Repository: wangheda/youtube-8m Branch: master Commit: 07e54b387ee0 Files: 581 Total size: 3.4 MB Directory structure: gitextract_1ok4slma/ ├── .gitignore ├── .gitmodules ├── LICENSE ├── README.md ├── eda/ │ ├── explore.ipynb │ └── vertical.tsv ├── youtube-8m-ensemble/ │ ├── .vimrc │ ├── CONTRIBUTING.md │ ├── LICENSE │ ├── README.md │ ├── __init__.py │ ├── all_ensemble_models/ │ │ ├── .vimrc │ │ ├── __init__.py │ │ ├── attention_linear_model.py │ │ ├── attention_linmatrix_model.py │ │ ├── attention_matrix_model.py │ │ ├── attention_moe_matrix_model.py │ │ ├── attention_moe_model.py │ │ ├── attention_rectified_linear_model.py │ │ ├── deep_combine_chain_model.py │ │ ├── input_moe_model.py │ │ ├── linear_regression_model.py │ │ ├── logistic_model.py │ │ ├── matrix_regression_model.py │ │ ├── mean_model.py │ │ ├── moe_model.py │ │ └── nonunit_matrix_regression_model.py │ ├── average_precision_calculator.py │ ├── check_distillation.py │ ├── check_video_id.py │ ├── check_video_id_match.py │ ├── cloudml-gpu-distributed.yaml │ ├── cloudml-gpu.yaml │ ├── data_augmentation.py │ ├── ensemble_command.example │ ├── ensemble_level_models.py │ ├── ensemble_scripts/ │ │ ├── .vimrc │ │ ├── after_submission_no1.conf │ │ ├── after_submission_no2.conf │ │ ├── after_submission_no3.conf │ │ ├── after_submission_no4.conf │ │ ├── auto-preensemble-deep_combine_chain_model.sh │ │ ├── auto-preensemble-matrix_model.sh │ │ ├── check-video_id.sh │ │ ├── check-video_id_match.sh │ │ ├── combine-tfrecords-frame-v2.sh │ │ ├── combine-tfrecords-frame.sh │ │ ├── combine-tfrecords-video-v2.sh │ │ ├── combine-tfrecords-video.sh │ │ ├── ensemble_no1.conf │ │ ├── ensemble_no10.conf │ │ ├── ensemble_no11.conf │ │ ├── ensemble_no12.conf │ │ ├── ensemble_no13.conf │ │ ├── ensemble_no14.conf │ │ ├── ensemble_no15.conf │ │ ├── ensemble_no16.conf │ │ ├── ensemble_no17.conf │ │ ├── ensemble_no18.conf │ │ ├── ensemble_no19.conf │ │ ├── ensemble_no2.conf │ │ ├── ensemble_no20.conf │ │ ├── ensemble_no21.conf │ │ ├── ensemble_no3.conf │ │ ├── ensemble_no4.conf │ │ ├── ensemble_no5.conf │ │ ├── ensemble_no6.conf │ │ ├── ensemble_no7.conf │ │ ├── ensemble_no8.conf │ │ ├── ensemble_no9.conf │ │ ├── eval-attention_linear_model.sh │ │ ├── eval-attention_linmatrix_model.sh │ │ ├── eval-attention_matrix_model.sh │ │ ├── eval-attention_moe_matrix_model.sh │ │ ├── eval-attention_moe_model.sh │ │ ├── eval-attention_rectified_linear_model.sh │ │ ├── eval-deep_combine_chain_model.sh │ │ ├── eval-input_moe_model.sh │ │ ├── eval-linear_model.sh │ │ ├── eval-matrix_model.sh │ │ ├── eval-mean_model.sh │ │ ├── eval-moe_model.sh │ │ ├── eval-nonunit_matrix_model.sh │ │ ├── explore-mean_model.log │ │ ├── explore-mean_model.sh │ │ ├── final_submission.conf │ │ ├── infer-attention_linear_model.sh │ │ ├── infer-attention_linmatrix_model.sh │ │ ├── infer-attention_matrix_model.sh │ │ ├── infer-attention_moe_matrix_model.sh │ │ ├── infer-attention_moe_model.sh │ │ ├── infer-attention_rectified_linear_model.sh │ │ ├── infer-linear_model.sh │ │ ├── infer-matrix_model.sh │ │ ├── infer-mean_model.sh │ │ ├── infer-moe_model.sh │ │ ├── make-bagging-of-ensembles.sh │ │ ├── make-virtual-groups.sh │ │ ├── preensemble-attention_matrix_model.sh │ │ ├── preensemble-matrix_model.sh │ │ ├── preensemble-mean_model.sh │ │ ├── train-attention_linear_model.sh │ │ ├── train-attention_linmatrix_model.sh │ │ ├── train-attention_matrix_model.sh │ │ ├── train-attention_moe_matrix_model.sh │ │ ├── train-attention_moe_model.sh │ │ ├── train-attention_rectified_linear_model.sh │ │ ├── train-deep_combine_chain_model.sh │ │ ├── train-input_moe_model.sh │ │ ├── train-linear_model.sh │ │ ├── train-matrix_model.sh │ │ ├── train-matrix_model_lr.sh │ │ ├── train-mean_model.sh │ │ ├── train-moe_model.sh │ │ └── train-nonunit_matrix_model.sh │ ├── eval.py │ ├── eval_util.py │ ├── feature_transform.py │ ├── inference-combine-tfrecords-frame.py │ ├── inference-combine-tfrecords-video.py │ ├── inference-pre-ensemble.py │ ├── inference.py │ ├── losses.py │ ├── mean_average_precision_calculator.py │ ├── model_selection_scripts/ │ │ ├── .vimrc │ │ ├── extend-step-mean_model.sh │ │ ├── get_extend_candidates.py │ │ ├── get_patterns.py │ │ ├── get_top_k.py │ │ └── greedy-selection-mean_model.sh │ ├── model_utils.py │ ├── models.py │ ├── readers.py │ ├── top_k_scripts/ │ │ ├── eval-attention_matrix_model.sh │ │ ├── infer-attention_matrix_model.sh │ │ ├── preensemble-attention_matrix_model.sh │ │ ├── run_top_k.sh │ │ └── train-attention_matrix_model.sh │ ├── train.py │ ├── training_utils/ │ │ ├── del.py │ │ ├── sample_conf.py │ │ ├── sample_freq.py │ │ └── select.py │ └── utils.py ├── youtube-8m-wangheda/ │ ├── .vimrc │ ├── CONTRIBUTING.md │ ├── LICENSE │ ├── README.md │ ├── __init__.py │ ├── all_data_augmentation/ │ │ ├── __init__.py │ │ ├── clipping_augmenter.py │ │ ├── default_augmenter.py │ │ ├── half_augmenter.py │ │ ├── half_video_augmenter.py │ │ └── noise_augmenter.py │ ├── all_feature_transform/ │ │ ├── __init__.py │ │ ├── avg_transformer.py │ │ ├── default_transformer.py │ │ ├── engineer_transformer.py │ │ ├── identical_transformer.py │ │ └── resolution_transformer.py │ ├── all_frame_models/ │ │ ├── .vimrc │ │ ├── __init__.py │ │ ├── bilstm_model.py │ │ ├── biunilstm_model.py │ │ ├── cnn_deep_combine_chain_model.py │ │ ├── cnn_kmax_model.py │ │ ├── cnn_lstm_memory_model.py │ │ ├── cnn_lstm_memory_multitask_model.py │ │ ├── cnn_lstm_memory_normalization_model.py │ │ ├── cnn_model.py │ │ ├── dbof_model.py │ │ ├── deep_cnn_deep_combine_chain_model.py │ │ ├── deep_lstm_model.py │ │ ├── distillchain_cnn_deep_combine_chain_model.py │ │ ├── distillchain_lstm_attention_max_pooling_model.py │ │ ├── distillchain_lstm_cnn_deep_combine_chain_model.py │ │ ├── distillchain_lstm_memory_deep_combine_chain_model.py │ │ ├── distillchain_lstm_parallel_finaloutput_model.py │ │ ├── distillchain_multiscale_cnn_lstm_model.py │ │ ├── frame_seg_model.py │ │ ├── framehop_lstm_memory_deep_combine_chain_model.py │ │ ├── framehop_lstm_memory_model.py │ │ ├── gru_pooling_model.py │ │ ├── gru_with_pooling_model.py │ │ ├── layernorm_lstm_memory_model.py │ │ ├── logistic_model.py │ │ ├── lstm_advanced_model.py │ │ ├── lstm_attention_lstm_model.py │ │ ├── lstm_attention_max_pooling_model.py │ │ ├── lstm_attention_model.py │ │ ├── lstm_auxloss_deep_combine_chain_model.py │ │ ├── lstm_cnn_deep_combine_chain_model.py │ │ ├── lstm_divided_model.py │ │ ├── lstm_look_back_model.py │ │ ├── lstm_memory_chain_model.py │ │ ├── lstm_memory_deep_chain_model.py │ │ ├── lstm_memory_input_chain_model.py │ │ ├── lstm_memory_model.py │ │ ├── lstm_memory_multitask_model.py │ │ ├── lstm_memory_normalization_model.py │ │ ├── lstm_memory_parallel_chain_model.py │ │ ├── lstm_model.py │ │ ├── lstm_multi_attention_model.py │ │ ├── lstm_multi_pooling_model.py │ │ ├── lstm_parallel_finaloutput_model.py │ │ ├── lstm_parallel_memory_model.py │ │ ├── lstm_parallel_model.py │ │ ├── lstm_pooling_model.py │ │ ├── lstm_positional_attention_max_pooling_model.py │ │ ├── lstm_with_mean_input_model.py │ │ ├── lstm_with_pooling_model.py │ │ ├── mm_lstm_memory_model.py │ │ ├── multi_view_cnn_deep_combine_chain_model.py │ │ ├── multires_lstm_memory_deep_combine_chain_model.py │ │ ├── multiscale_cnn_lstm_model.py │ │ ├── positional_cnn_deep_combine_chain_model.py │ │ ├── progressive_attention_lstm_model.py │ │ └── wide_and_deep_model.py │ ├── all_video_models/ │ │ ├── .vimrc │ │ ├── __init__.py │ │ ├── chain_main_relu_moe_model.py │ │ ├── chain_moe_model.py │ │ ├── chain_support_relu_moe_model.py │ │ ├── deep_chain_model.py │ │ ├── deep_combine_chain_model.py │ │ ├── distillchain_deep_combine_chain_model.py │ │ ├── hidden_chain_model.py │ │ ├── hidden_combine_chain_model.py │ │ ├── logistic_model.py │ │ ├── mlp_moe_model.py │ │ ├── moe_model.py │ │ ├── multitask_divergence_deep_combine_chain_model.py │ │ ├── multitask_divergence_moe_model.py │ │ ├── multitask_moe_model.py │ │ ├── shortcut_chain_support_relu_moe_model.py │ │ └── stage2_logistic_model.py │ ├── average_precision_calculator.py │ ├── bagging_scripts/ │ │ ├── cnn-deep-combine-chain-bagging.sh │ │ ├── distillation-video-dcc-bagging.sh │ │ ├── lstmattention8max-bagging.sh │ │ ├── lstmparalleloutput-bagging.sh │ │ └── video-deep-combine-chain-bagging.sh │ ├── boosting_scripts/ │ │ ├── cnn-deep-combine-chain-boosting.sh │ │ ├── distillation-cnn-dcc-boosting.sh │ │ ├── distillation-lstmcnn-dcc-boosting.sh │ │ ├── distillation-lstmparalleloutput-boosting.sh │ │ ├── distillation-multilstm-dcc-boosting.sh │ │ ├── distillation-multiscale-cnn-lstm-boosting.sh │ │ ├── distillation-positional-lstmattention8max-boosting.sh │ │ ├── distillation-video-dcc-boosting.sh │ │ ├── lstmattention8max-boosting-weightclip.sh │ │ ├── lstmparalleloutput-boosting-weightclip.sh │ │ ├── video-deep-combine-chain-boosting-discardhopeless.sh │ │ ├── video-deep-combine-chain-boosting-weightclip.sh │ │ └── video-deep-combine-chain-boosting.sh │ ├── cascade_scripts/ │ │ ├── distillchain-v2-hybridchain.sh │ │ ├── distillchain-v2-hybridchain2.sh │ │ └── distillchain-v2-videochain.sh │ ├── cloudml-gpu-distributed.yaml │ ├── cloudml-gpu.yaml │ ├── data_augmentation.py │ ├── data_augmentation_scripts/ │ │ ├── eval-chaining-video.sh │ │ ├── run-chaining-cnn.sh │ │ ├── run-chaining-lstm.sh │ │ ├── run-chaining-video.sh │ │ ├── run-multiple-attention-pooling-positional-embedding.sh │ │ ├── run-multiscale-cnn-lstm-model.sh │ │ └── run-parallel-lstm-memory.sh │ ├── eval.py │ ├── eval.sh │ ├── eval_scripts/ │ │ ├── eval-att-lstm.sh │ │ ├── eval-att.sh │ │ ├── eval-bi-uni-lstm.sh │ │ ├── eval-chain-model-relu.sh │ │ ├── eval-chain-model-suprelu.sh │ │ ├── eval-chain-moe-0.4.sh │ │ ├── eval-chain-moe-freq.sh │ │ ├── eval-chain-moe-suprelu-vert+freq.sh │ │ ├── eval-chain-moe.sh │ │ ├── eval-cnn-deep-combine-chain.sh │ │ ├── eval-cnn-model.sh │ │ ├── eval-dbof.sh │ │ ├── eval-deep-cnn-deep-combine-chain.sh │ │ ├── eval-distill-video-dcc-noise-scene1.sh │ │ ├── eval-distill-video-dcc-noise-scene2.sh │ │ ├── eval-distillchain-cnn-dcc.sh │ │ ├── eval-distillchain-lstmcnn.sh │ │ ├── eval-distillchain-lstmparalleloutput.sh │ │ ├── eval-distillchain-multilstm.sh │ │ ├── eval-distillchain-v2-boostinglstmparalleloutput.sh │ │ ├── eval-distillchain-v2-lstmattention8max.sh │ │ ├── eval-distillchain-v2-lstmcnn.sh │ │ ├── eval-distillchain-v2-lstmparalleloutput.sh │ │ ├── eval-distillchain-v2-multilstm.sh │ │ ├── eval-distillchain-v2-multiscale-cnnlstm.sh │ │ ├── eval-distillchain-v2-video-dcc.sh │ │ ├── eval-distillchain-video-dcc.sh │ │ ├── eval-frame-seg.sh │ │ ├── eval-framehop-lstmmem.sh │ │ ├── eval-layer-chain-moe8-freq.sh │ │ ├── eval-layer-moe-vert.sh │ │ ├── eval-lstm-attention-8max.sh │ │ ├── eval-lstm-cnn-deep-combine-chain.sh │ │ ├── eval-lstm-look-back.sh │ │ ├── eval-lstm-positional-attention-8max.sh │ │ ├── eval-lstmmem-augmenter.sh │ │ ├── eval-lstmmem-chain-freq.sh │ │ ├── eval-lstmmem-chain.sh │ │ ├── eval-lstmmem-cnnlstm.sh │ │ ├── eval-lstmmem-deep-chain.sh │ │ ├── eval-lstmmem-deep-combine-chain-length.sh │ │ ├── eval-lstmmem-dropout.sh │ │ ├── eval-lstmmem-feature.sh │ │ ├── eval-lstmmem-input-chain.sh │ │ ├── eval-lstmmem-input-noise.sh │ │ ├── eval-lstmmem-l2norm.sh │ │ ├── eval-lstmmem-layernorm.sh │ │ ├── eval-lstmmem-lowres.sh │ │ ├── eval-lstmmem-no-transform.sh │ │ ├── eval-lstmmem-noise.sh │ │ ├── eval-lstmmem-parallel.sh │ │ ├── eval-lstmmem-shortcut-chain-freq.sh │ │ ├── eval-lstmmem2048.sh │ │ ├── eval-lstmmemory-audio.sh │ │ ├── eval-lstmmemory-layer1.sh │ │ ├── eval-lstmmemory.sh │ │ ├── eval-lstmoutput-parallel.sh │ │ ├── eval-mem.sh │ │ ├── eval-mm-lstm.sh │ │ ├── eval-moe-baseline.sh │ │ ├── eval-moe-batchagreement1.sh │ │ ├── eval-moe-batchagreement2.sh │ │ ├── eval-moe-batchagreement3.sh │ │ ├── eval-moe-model.sh │ │ ├── eval-moe-topk-batchagreement1.sh │ │ ├── eval-moe-topk-batchagreement2.sh │ │ ├── eval-moe-topk-batchagreement3.sh │ │ ├── eval-multi-lstmmem-deep-chain.sh │ │ ├── eval-multi-view-cnn-deep-combine-chain.sh │ │ ├── eval-multires-lstm-deep-combine-chain.sh │ │ ├── eval-multitask-ce.sh │ │ ├── eval-multitask.sh │ │ ├── eval-positional-cnn-dcc.sh │ │ ├── eval-stage2-logistic.sh │ │ ├── eval-stage2-moe.sh │ │ ├── eval-video-deep-chain.sh │ │ ├── eval-video-deep-combine-addnoise.sh │ │ ├── eval-video-deep-combine-chain-dropout.sh │ │ ├── eval-video-deep-combine-chain-labelsmoothing.sh │ │ ├── eval-video-deep-combine-chain-noise.sh │ │ ├── eval-video-deep-combine-chain.sh │ │ ├── eval-video-distillchain-video-dcc.sh │ │ ├── eval-video-divergence-chain-model.sh │ │ ├── eval-video-divergence-moe-model.sh │ │ ├── eval-video-hidden-chain.sh │ │ ├── eval-video-hidden-combine-chain.sh │ │ ├── eval-video-logistic.sh │ │ ├── eval-video-moe.sh │ │ ├── eval-video-pairwise.sh │ │ └── eval-video-verydeep-combine-chain.sh │ ├── eval_util.py │ ├── feature_transform.py │ ├── frame_level_models.py │ ├── infer_scripts/ │ │ ├── infer-attentionlstm_moe4.sh │ │ ├── infer-biunilstm1024_moe8.sh │ │ ├── infer-cnn_deep_combine_chain.sh │ │ ├── infer-cnn_lstmmemory1024_moe8.sh │ │ ├── infer-cnn_model.sh │ │ ├── infer-dbof.sh │ │ ├── infer-deep_cnn_deep_combine.sh │ │ ├── infer-deeplstm1024_layer6_moe4.sh │ │ ├── infer-distill_video_dcc.sh │ │ ├── infer-distillation-cnn-dcc.sh │ │ ├── infer-distillation-lstmattention8max.sh │ │ ├── infer-distillation-lstmgate.sh │ │ ├── infer-distillation-video-dcc.sh │ │ ├── infer-distillation.sh │ │ ├── infer-distillchain-cnn-dcc.sh │ │ ├── infer-distillchain-lstmcnn.sh │ │ ├── infer-distillchain-lstmparalleloutput.sh │ │ ├── infer-distillchain-v2-boost-lstmparalleloutput.sh │ │ ├── infer-distillchain-v2-lstmattention8max.sh │ │ ├── infer-distillchain-v2-lstmcnn.sh │ │ ├── infer-distillchain-v2-lstmparalleloutput.sh │ │ ├── infer-distillchain-v2-multilstm.sh │ │ ├── infer-distillchain-v2-multiscal-cnnlstm.sh │ │ ├── infer-distillchain-v2-video-dcc.sh │ │ ├── infer-frame_seg.sh │ │ ├── infer-framehop_lstm.sh │ │ ├── infer-lstm_attention8_max.sh │ │ ├── infer-lstm_cnn_deep_combine_chain.sh │ │ ├── infer-lstmattlstm1024_moe8.sh │ │ ├── infer-lstmmemory-audio.sh │ │ ├── infer-lstmmemory-layer1.sh │ │ ├── infer-lstmmemory1024_deep_combine_chain_add_length.sh │ │ ├── infer-lstmmemory1024_moe8.sh │ │ ├── infer-lstmmemory2048_moe4.sh │ │ ├── infer-lstmparallelmemory1024_moe8.sh │ │ ├── infer-lstmparalleloutput1024_moe8.sh │ │ ├── infer-model_input.sh │ │ ├── infer-multilstmmemory1024_moe4_deep_chain.sh │ │ ├── infer-multires_lstm_deep_combine_chain.sh │ │ ├── infer-positional-lstmattention8max.sh │ │ ├── infer-video-distillchain-video-dcc.sh │ │ ├── infer-video_group_moe4_noise0.2_layer4_elu.sh │ │ ├── infer-video_logistic.sh │ │ ├── infer-video_moe16_model.sh │ │ └── infer-video_very_deep_combine_chain.sh │ ├── inference-layer.py │ ├── inference-pre-ensemble-get-input.py │ ├── inference-pre-ensemble-with-predictions.py │ ├── inference-pre-ensemble.py │ ├── inference-sample-error-analysis.py │ ├── inference-sample-error.py │ ├── inference-stage1.py │ ├── inference.py │ ├── losses.py │ ├── mean_average_precision_calculator.py │ ├── model_utils.py │ ├── models.py │ ├── readers.py │ ├── train-with-predictions.py │ ├── train-with-rebuild.py │ ├── train.py │ ├── training_scripts/ │ │ ├── run-cascade-75-chaining-cnn.sh │ │ ├── run-cascade-75-chaining-lstm-cnn.sh │ │ ├── run-cascade-75-chaining-lstm.sh │ │ ├── run-cascade-75-chaining-parallel-lstm.sh │ │ ├── run-cascade-75-chaining-video.sh │ │ ├── run-cascade-75-multiple-attention-pooling.sh │ │ ├── run-cascade-76-chaining-cnn.sh │ │ ├── run-cascade-76-chaining-lstm-cnn.sh │ │ ├── run-cascade-76-chaining-lstm.sh │ │ ├── run-cascade-76-chaining-video.sh │ │ ├── run-cascade-76-multiple-attention-pooling.sh │ │ ├── run-cascade-76-multiscale-cnn-lstm.sh │ │ ├── run-cascade-76-parallel-lstm-boosting.sh │ │ ├── run-cascade-76-parallel-lstm.sh │ │ ├── run-chaining-cnn.sh │ │ ├── run-chaining-deep-cnn.sh │ │ ├── run-chaining-lstm-cnn.sh │ │ ├── run-chaining-lstm.sh │ │ ├── run-chaining-multi-resolution-lstm.sh │ │ ├── run-chaining-shared-lstm.sh │ │ ├── run-chaining-video.sh │ │ ├── run-cnn-lstm.sh │ │ ├── run-cnn-model.sh │ │ ├── run-lstm-memory-cell1024.sh │ │ ├── run-lstm-memory-cell2048.sh │ │ ├── run-multiple-attention-pooling-positional-embedding.sh │ │ ├── run-multiscale-cnn-lstm-model.sh │ │ ├── run-parallel-lstm-memory.sh │ │ ├── run-parallel-lstm-output.sh │ │ └── run-temporal-pooling-lstm.sh │ ├── training_utils/ │ │ ├── del.py │ │ ├── human_readable_error_analysis.py │ │ ├── reweight_sample_freq.py │ │ ├── sample_freq.py │ │ ├── select.py │ │ └── video_original_boosting_error_analysis.py │ ├── utils.py │ └── video_level_models.py └── youtube-8m-zhangteng/ ├── CONTRIBUTING.md ├── LICENSE ├── README.md ├── YM_framemean.py ├── YM_labels_matrix.py ├── YM_labels_model.py ├── YM_labels_vocab.py ├── YM_readframefeature.py ├── __init__.py ├── average_precision_calculator.py ├── cloudml-gpu-distributed.yaml ├── cloudml-gpu.yaml ├── eval.py ├── eval_autoencoder.py ├── eval_distill.py ├── eval_embedding.py ├── eval_scripts/ │ ├── eval-distillchain_cnndcc_layer2moe4.sh │ ├── eval-distillchain_lstm_extend_moe8.sh │ ├── eval-distillchain_lstm_gate_moe8.sh │ ├── eval-distillchain_lstm_gate_moe8_v2.sh │ ├── eval-distillchain_lstm_glu2_moe8_v2.sh │ ├── eval-distillchain_lstm_moe8.sh │ ├── eval-distillchain_lstm_moe8_v2.sh │ ├── eval-distillchain_lstm_multiscale2layer_moe8.sh │ ├── eval-distillchain_lstm_multiscale4layer_moe8.sh │ ├── eval-distillchain_video_norm_moe8_local.sh │ ├── eval-distillsplit_lstm_gate_moe8.sh │ ├── eval-lstm2_attention8_max.sh │ ├── eval-lstm_attention8_max.sh │ ├── eval-lstm_gate_multiscale4_moe4.sh │ ├── eval-lstm_multiscale4_moe4.sh │ ├── eval-lstm_random_moe8.sh │ ├── eval-lstm_shortlayers_moe8.sh │ ├── eval-lstmbiglu_1024_moe8.sh │ ├── eval-lstmgate1024_moe8.sh │ ├── eval-lstmglu2_1024_moe8.sh │ ├── eval-video_knowledge_combine_chain.sh │ ├── eval-video_notzero_combine_chain.sh │ ├── eval-video_relabel_combine_chain.sh │ └── eval-video_softmax_combine_chain.sh ├── eval_util.py ├── frame_level_models.py ├── infer_scripts/ │ ├── infer-distillchain_cnndcc_layer2moe4.sh │ ├── infer-distillchain_cnndcc_layer2moe4_ensemble.sh │ ├── infer-distillchain_lstm_extend_moe8.sh │ ├── infer-distillchain_lstm_gate_moe8.sh │ ├── infer-distillchain_lstm_gate_moe8_v2.sh │ ├── infer-distillchain_lstm_glu2_moe8_v2.sh │ ├── infer-distillchain_lstm_moe8.sh │ ├── infer-distillchain_lstm_moe8_v2.sh │ ├── infer-distillchain_lstm_multiscale2layer_moe8.sh │ ├── infer-distillchain_lstm_multiscale4layer_moe8.sh │ ├── infer-distillchain_video_norm_moe8.sh │ ├── infer-distillchain_video_norm_moe8_local.sh │ ├── infer-distillsplit_lstm_gate_moe8.sh │ ├── infer-lstm2_attention8_max.sh │ ├── infer-lstm_attention8_max.sh │ ├── infer-lstm_attention_max_mean.sh │ ├── infer-lstm_gate_multiscale4_moe4.sh │ ├── infer-lstm_multiscale4_moe4.sh │ ├── infer-lstm_random_mean_moe8.sh │ ├── infer-lstm_shortlayers_moe8.sh │ ├── infer-lstmbiglu_1024_moe8.sh │ ├── infer-lstmgate1024_moe8.sh │ ├── infer-lstmglu2_1024_moe8.sh │ ├── infer-video_notzero_combine_chain.sh │ └── infer-video_relabel_combine_chain.sh ├── inference-pre-ensemble-distill.py ├── inference-pre-ensemble.py ├── inference.py ├── inference_autoencoder.py ├── inference_embedding.py ├── inference_test.py ├── inference_with_rebuild.py ├── labels_autoencoder.py ├── labels_embedding.py ├── labels_rbm.py ├── losses.py ├── losses_embedding.py ├── mean_average_precision_calculator.py ├── model_utils.py ├── models.py ├── readers.py ├── rnn_residual.py ├── train-with-rebuild.py ├── train.py ├── train_autoencoder.py ├── train_embedding.py ├── train_ensemble.py ├── train_scripts/ │ ├── run-attention-pooling-lstm-a.sh │ ├── run-attention-pooling-lstm-s.sh │ ├── run-attention-pooling-lstm2lstm.sh │ ├── run-attention-pooling.sh │ ├── run-bilstm-a.sh │ ├── run-cascade-76-lstm-a.sh │ ├── run-cascade-76-lstm-s.sh │ ├── run-cascade-76-lstm.sh │ ├── run-cascade-attention-pooling.sh │ ├── run-cascade-chaining-cnn-layer2.sh │ ├── run-cascade-chaining-video-normalize.sh │ ├── run-cascade-lstm-s-split.sh │ ├── run-cascade-lstm-s.sh │ ├── run-cascade-lstm.sh │ ├── run-cascade-multiscale-cnn-lstm-laery4.sh │ ├── run-cascade-multiscale-cnn-lstm-layer2.sh │ ├── run-chaining-video-add-confident.sh │ ├── run-chaining-video-infrequent-softmax.sh │ ├── run-chaining-video-normal.sh │ ├── run-chaining-video-vertical.sh │ ├── run-lstm-random-augmentation.sh │ ├── run-lstm-s.sh │ ├── run-multiscale-cnn-lstm.sh │ └── run-temporal-segment-lstm.sh ├── training_utils/ │ ├── del.py │ └── select.py ├── utils.py ├── video_level_models.py └── writers.py ================================================ FILE CONTENTS ================================================ ================================================ FILE: .gitignore ================================================ model/ prediction/ *.pdf */__pycache__/ *.out *.csv *.tfrecord *.pyc .*.swp eda/data/ ================================================ FILE: .gitmodules ================================================ [submodule "3rd_party/annoy"] path = 3rd_party/annoy url = git@github.com:spotify/annoy.git [submodule "tensorflow"] path = tensorflow url = git@github.com:tensorflow/tensorflow.git ================================================ FILE: LICENSE ================================================ Apache License Version 2.0, January 2004 http://www.apache.org/licenses/ TERMS AND CONDITIONS FOR USE, REPRODUCTION, AND DISTRIBUTION 1. Definitions. "License" shall mean the terms and conditions for use, reproduction, and distribution as defined by Sections 1 through 9 of this document. "Licensor" shall mean the copyright owner or entity authorized by the copyright owner that is granting the License. "Legal Entity" shall mean the union of the acting entity and all other entities that control, are controlled by, or are under common control with that entity. For the purposes of this definition, "control" means (i) the power, direct or indirect, to cause the direction or management of such entity, whether by contract or otherwise, or (ii) ownership of fifty percent (50%) or more of the outstanding shares, or (iii) beneficial ownership of such entity. "You" (or "Your") shall mean an individual or Legal Entity exercising permissions granted by this License. "Source" form shall mean the preferred form for making modifications, including but not limited to software source code, documentation source, and configuration files. "Object" form shall mean any form resulting from mechanical transformation or translation of a Source form, including but not limited to compiled object code, generated documentation, and conversions to other media types. "Work" shall mean the work of authorship, whether in Source or Object form, made available under the License, as indicated by a copyright notice that is included in or attached to the work (an example is provided in the Appendix below). "Derivative Works" shall mean any work, whether in Source or Object form, that is based on (or derived from) the Work and for which the editorial revisions, annotations, elaborations, or other modifications represent, as a whole, an original work of authorship. For the purposes of this License, Derivative Works shall not include works that remain separable from, or merely link (or bind by name) to the interfaces of, the Work and Derivative Works thereof. "Contribution" shall mean any work of authorship, including the original version of the Work and any modifications or additions to that Work or Derivative Works thereof, that is intentionally submitted to Licensor for inclusion in the Work by the copyright owner or by an individual or Legal Entity authorized to submit on behalf of the copyright owner. For the purposes of this definition, "submitted" means any form of electronic, verbal, or written communication sent to the Licensor or its representatives, including but not limited to communication on electronic mailing lists, source code control systems, and issue tracking systems that are managed by, or on behalf of, the Licensor for the purpose of discussing and improving the Work, but excluding communication that is conspicuously marked or otherwise designated in writing by the copyright owner as "Not a Contribution." "Contributor" shall mean Licensor and any individual or Legal Entity on behalf of whom a Contribution has been received by Licensor and subsequently incorporated within the Work. 2. Grant of Copyright License. Subject to the terms and conditions of this License, each Contributor hereby grants to You a perpetual, worldwide, non-exclusive, no-charge, royalty-free, irrevocable copyright license to reproduce, prepare Derivative Works of, publicly display, publicly perform, sublicense, and distribute the Work and such Derivative Works in Source or Object form. 3. Grant of Patent License. Subject to the terms and conditions of this License, each Contributor hereby grants to You a perpetual, worldwide, non-exclusive, no-charge, royalty-free, irrevocable (except as stated in this section) patent license to make, have made, use, offer to sell, sell, import, and otherwise transfer the Work, where such license applies only to those patent claims licensable by such Contributor that are necessarily infringed by their Contribution(s) alone or by combination of their Contribution(s) with the Work to which such Contribution(s) was submitted. If You institute patent litigation against any entity (including a cross-claim or counterclaim in a lawsuit) alleging that the Work or a Contribution incorporated within the Work constitutes direct or contributory patent infringement, then any patent licenses granted to You under this License for that Work shall terminate as of the date such litigation is filed. 4. Redistribution. You may reproduce and distribute copies of the Work or Derivative Works thereof in any medium, with or without modifications, and in Source or Object form, provided that You meet the following conditions: (a) You must give any other recipients of the Work or Derivative Works a copy of this License; and (b) You must cause any modified files to carry prominent notices stating that You changed the files; and (c) You must retain, in the Source form of any Derivative Works that You distribute, all copyright, patent, trademark, and attribution notices from the Source form of the Work, excluding those notices that do not pertain to any part of the Derivative Works; and (d) If the Work includes a "NOTICE" text file as part of its distribution, then any Derivative Works that You distribute must include a readable copy of the attribution notices contained within such NOTICE file, excluding those notices that do not pertain to any part of the Derivative Works, in at least one of the following places: within a NOTICE text file distributed as part of the Derivative Works; within the Source form or documentation, if provided along with the Derivative Works; or, within a display generated by the Derivative Works, if and wherever such third-party notices normally appear. The contents of the NOTICE file are for informational purposes only and do not modify the License. You may add Your own attribution notices within Derivative Works that You distribute, alongside or as an addendum to the NOTICE text from the Work, provided that such additional attribution notices cannot be construed as modifying the License. You may add Your own copyright statement to Your modifications and may provide additional or different license terms and conditions for use, reproduction, or distribution of Your modifications, or for any such Derivative Works as a whole, provided Your use, reproduction, and distribution of the Work otherwise complies with the conditions stated in this License. 5. Submission of Contributions. Unless You explicitly state otherwise, any Contribution intentionally submitted for inclusion in the Work by You to the Licensor shall be under the terms and conditions of this License, without any additional terms or conditions. Notwithstanding the above, nothing herein shall supersede or modify the terms of any separate license agreement you may have executed with Licensor regarding such Contributions. 6. Trademarks. This License does not grant permission to use the trade names, trademarks, service marks, or product names of the Licensor, except as required for reasonable and customary use in describing the origin of the Work and reproducing the content of the NOTICE file. 7. Disclaimer of Warranty. Unless required by applicable law or agreed to in writing, Licensor provides the Work (and each Contributor provides its Contributions) on an "AS IS" BASIS, WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied, including, without limitation, any warranties or conditions of TITLE, NON-INFRINGEMENT, MERCHANTABILITY, or FITNESS FOR A PARTICULAR PURPOSE. You are solely responsible for determining the appropriateness of using or redistributing the Work and assume any risks associated with Your exercise of permissions under this License. 8. Limitation of Liability. In no event and under no legal theory, whether in tort (including negligence), contract, or otherwise, unless required by applicable law (such as deliberate and grossly negligent acts) or agreed to in writing, shall any Contributor be liable to You for damages, including any direct, indirect, special, incidental, or consequential damages of any character arising as a result of this License or out of the use or inability to use the Work (including but not limited to damages for loss of goodwill, work stoppage, computer failure or malfunction, or any and all other commercial damages or losses), even if such Contributor has been advised of the possibility of such damages. 9. Accepting Warranty or Additional Liability. While redistributing the Work or Derivative Works thereof, You may choose to offer, and charge a fee for, acceptance of support, warranty, indemnity, or other liability obligations and/or rights consistent with this License. However, in accepting such obligations, You may act only on Your own behalf and on Your sole responsibility, not on behalf of any other Contributor, and only if You agree to indemnify, defend, and hold each Contributor harmless for any liability incurred by, or claims asserted against, such Contributor by reason of your accepting any such warranty or additional liability. END OF TERMS AND CONDITIONS APPENDIX: How to apply the Apache License to your work. To apply the Apache License to your work, attach the following boilerplate notice, with the fields enclosed by brackets "{}" replaced with your own identifying information. (Don't include the brackets!) The text should be enclosed in the appropriate comment syntax for the file format. We also recommend that a file or class name and description of purpose be included on the same "printed page" as the copyright notice for easier identification within third-party archives. Copyright 2017 Heda Wang and Teng Zhang Licensed under the Apache License, Version 2.0 (the "License"); you may not use this file except in compliance with the License. You may obtain a copy of the License at http://www.apache.org/licenses/LICENSE-2.0 Unless required by applicable law or agreed to in writing, software distributed under the License is distributed on an "AS IS" BASIS, WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied. See the License for the specific language governing permissions and limitations under the License. ================================================ FILE: README.md ================================================ # The Monkeytyping Solution to the Youtube-8M Video Understanding Challenge This is the solution repository of the 2nd place team monkeytyping, licensed under the Apache License 2.0. ## Dependencies Python 2.7 Tensorflow 1.0 Numpy 1.12 GNU Bash ## Resources For an understanding of our system, read the report of our solution: > https://arxiv.org/abs/1706.05150 Our source code: > https://github.com/wangheda/youtube-8m ## Useful scripts Training scripts (training a model may take 3-5 days) are in > youtube-8m-wangheda/training_scripts > youtube-8m-zhangteng/train_scripts Eval scripts for selecting best performing checkpoints > youtube-8m-wangheda/eval_scripts > youtube-8m-zhangteng/eval_scripts Infer scripts for generating intermediate files used by ensemble scripts > youtube-8m-wangheda/infer_scripts > youtube-8m-zhangteng/infer_scripts Ensemble scripts > youtube-8m-ensemble/ensemble_scripts ## Paths of models and data There are some conventions that we use in our code: models are saved in > ./model train1 data is saved in > /Youtube-8M/data/frame/train > /Youtube-8M/data/video/train validate1 data is saved in > /Youtube-8M/data/frame/validate > /Youtube-8M/data/video/validate test data is saved in > /Youtube-8M/data/frame/test > /Youtube-8M/data/video/test train2 data is saved in > /Youtube-8M/data/frame/ensemble_train > /Youtube-8M/data/video/ensemble_train validate2 data is saved in > /Youtube-8M/data/frame/ensemble_validate > /Youtube-8M/data/video/ensemble_validate intermediate results are stored in > /Youtube-8M/model_predictions/ensemble_train/[method] > /Youtube-8M/model_predictions/ensemble_validate/[method] > /Youtube-8M/model_predictions/test/[method] ## How to generate a solution ### Single model 1. Train a single model 2. evaluate the checkpoints to get the best one 3. infer the checkpoint to get intermediate result. ### Ensemble model 1. Write a configuration file 2. train a stacking model 3. evaluate the stacking model and pick the best checkpoint 4. infer the checkpoint to get a submission file ## Note Some of the single models are developed by Heda and some by Teng, so they are distributed in two folders. Bagging models are in `youtube-8m-wangheda/bagging_scripts`. Boosting and distillation models are in `youtube-8m-wangheda/bagging_scripts`. Cascade models are in `youtube-8m-wangheda/cascade_scripts`. Stacking models are in `youtube-8m-ensemble/ensemble_scripts`. ================================================ FILE: eda/explore.ipynb ================================================ { "cells": [ { "cell_type": "code", "execution_count": 16, "metadata": { "collapsed": true }, "outputs": [], "source": [ "import pandas as pd\n", "import matplotlib.pyplot as plt" ] }, { "cell_type": "code", "execution_count": 3, "metadata": { "collapsed": false }, "outputs": [], "source": [ "label_names = pd.read_csv(\"data/label_names.csv\")\n", "vocabulary = pd.read_csv(\"data/vocabulary.csv\")\n", "train_labels = pd.read_csv(\"data/train_labels.csv\", header=None, names=[\"id\", \"labels\"])\n", "validate_labels = pd.read_csv(\"data/validate_labels.csv\", header=None, names=[\"id\", \"labels\"])" ] }, { "cell_type": "markdown", "metadata": {}, "source": [ "## Lable names" ] }, { "cell_type": "code", "execution_count": 4, "metadata": { "collapsed": false }, "outputs": [ { "data": { "text/html": [ "
\n", "\n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", "
label_idlabel_name
00Games
11Vehicle
22Video game
33Concert
44Car
55Dance
66Animation
77Musician
88Football
99Music video
1010Animal
1111Motorsport
1212Food
1313Musical ensemble
1414Guitar
1515Cartoon
1616Performance art
1717Racing
1818Outdoor recreation
1919PC game
2020Trailer
2121Stadium
2222Nature
2323Mobile phone
2424String instrument
2525Toy
2626Cooking
2727Motorcycle
2828Fashion
2929Smartphone
.........
46864686Nisekoi
46874687Hollywood Palladium
46884688Skullgirls
46894689Al-Masjid an-Nabawi
46904690Shadow Fighter
46914691Reptile
46924692Abercrombie & Fitch
46934693Daffy Duck
46944694Cruelty to animals
46954695Strikeout
46964696Arjuna
46974697Scarlet Weather Rhapsody
46984698Voetbal International
46994699Death Row Records
47004700Shashlik
47014701Scarecrow
47024702Akame ga Kill!
47034703DJ Hero 2
47044704Gong
47054705SCP – Containment Breach
47064706Friday the 13th: Day of Death
47074707Poncho
47084708Visual novel
47094709Luge
47104710Mammal
47114711Aston Martin V8 Vantage (2005)
47124712Russian pyramid
47134713PBS Kids
47144714Air Gear
47154715Pina Records
\n", "

4716 rows × 2 columns

\n", "
" ], "text/plain": [ " label_id label_name\n", "0 0 Games\n", "1 1 Vehicle\n", "2 2 Video game\n", "3 3 Concert\n", "4 4 Car\n", "5 5 Dance\n", "6 6 Animation\n", "7 7 Musician\n", "8 8 Football\n", "9 9 Music video\n", "10 10 Animal\n", "11 11 Motorsport\n", "12 12 Food\n", "13 13 Musical ensemble\n", "14 14 Guitar\n", "15 15 Cartoon\n", "16 16 Performance art\n", "17 17 Racing\n", "18 18 Outdoor recreation\n", "19 19 PC game\n", "20 20 Trailer\n", "21 21 Stadium\n", "22 22 Nature\n", "23 23 Mobile phone\n", "24 24 String instrument\n", "25 25 Toy\n", "26 26 Cooking\n", "27 27 Motorcycle\n", "28 28 Fashion\n", "29 29 Smartphone\n", "... ... ...\n", "4686 4686 Nisekoi\n", "4687 4687 Hollywood Palladium\n", "4688 4688 Skullgirls\n", "4689 4689 Al-Masjid an-Nabawi\n", "4690 4690 Shadow Fighter\n", "4691 4691 Reptile\n", "4692 4692 Abercrombie & Fitch\n", "4693 4693 Daffy Duck\n", "4694 4694 Cruelty to animals\n", "4695 4695 Strikeout\n", "4696 4696 Arjuna\n", "4697 4697 Scarlet Weather Rhapsody\n", "4698 4698 Voetbal International\n", "4699 4699 Death Row Records\n", "4700 4700 Shashlik\n", "4701 4701 Scarecrow\n", "4702 4702 Akame ga Kill!\n", "4703 4703 DJ Hero 2\n", "4704 4704 Gong\n", "4705 4705 SCP – Containment Breach\n", "4706 4706 Friday the 13th: Day of Death\n", "4707 4707 Poncho\n", "4708 4708 Visual novel\n", "4709 4709 Luge\n", "4710 4710 Mammal\n", "4711 4711 Aston Martin V8 Vantage (2005)\n", "4712 4712 Russian pyramid\n", "4713 4713 PBS Kids\n", "4714 4714 Air Gear\n", "4715 4715 Pina Records\n", "\n", "[4716 rows x 2 columns]" ] }, "execution_count": 4, "metadata": {}, "output_type": "execute_result" } ], "source": [ "label_names" ] }, { "cell_type": "code", "execution_count": 5, "metadata": { "collapsed": false }, "outputs": [], "source": [ "label = {}\n", "for item in label_names.iterrows():\n", " label_id = item[1].label_id\n", " label_name = item[1].label_name\n", " label[label_id] = label_name" ] }, { "cell_type": "markdown", "metadata": {}, "source": [ "## Label details" ] }, { "cell_type": "code", "execution_count": 6, "metadata": { "collapsed": false }, "outputs": [ { "data": { "text/html": [ "
\n", "\n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", "
IndexTrainVideoCountKnowledgeGraphIdNameWikiUrlVertical1Vertical2Vertical3WikiDescription
00865174/m/03bt1ghGamehttps://en.wikipedia.org/wiki/GameGamesNaNNaNA game is structured form of play, usually und...
11683166/m/07yv9Vehiclehttps://en.wikipedia.org/wiki/VehicleAutos & VehiclesNaNNaNA vehicle is a mobile machine that transports ...
22522427/m/01mw1Video gamehttps://en.wikipedia.org/wiki/Video_gameGamesNaNNaNA video game is an electronic game that involv...
33497487/m/01jddzConcerthttps://en.wikipedia.org/wiki/ConcertArts & EntertainmentNaNNaNA concert is a live music performance in front...
44373952/m/0k4jCarhttps://en.wikipedia.org/wiki/CarAutos & VehiclesNaNNaNA car is a wheeled, self-powered motor vehicle...
55305143/m/026bkDancehttps://en.wikipedia.org/wiki/DanceArts & EntertainmentNaNNaNDance is a performance art form consisting of ...
66299054/m/0hcrAnimationhttps://en.wikipedia.org/wiki/AnimationArts & EntertainmentNaNNaNAnimation is the process of making the illusio...
77296577/m/09jwlMusicianhttps://en.wikipedia.org/wiki/MusicianArts & EntertainmentNaNNaNA musician is a person who plays a musical ins...
88219506/m/02vx4Association footballhttps://en.wikipedia.org/wiki/Association_foot...SportsNaNNaNAssociation football, more commonly known as f...
99217037/m/0mdxdMusic videohttps://en.wikipedia.org/wiki/Music_videoArts & EntertainmentNaNNaNA music video is a short film integrating a so...
1010203700/m/0jbkAnimalhttps://en.wikipedia.org/wiki/AnimalPets & AnimalsNaNNaNAnimals are multicellular, eukaryotic organism...
1111198630/m/0410tthMotorsporthttps://en.wikipedia.org/wiki/MotorsportSportsNaNNaNMotorsport or motorsports is a global term use...
1212196108/m/02wbmFoodhttps://en.wikipedia.org/wiki/FoodFood & DrinkNaNNaNFood is any substance consumed to provide nutr...
1313170351/m/05229Musical ensemblehttps://en.wikipedia.org/wiki/Musical_ensembleArts & EntertainmentNaNNaNA musical ensemble, also known as a music grou...
1414162864/m/0342hGuitarhttps://en.wikipedia.org/wiki/GuitarArts & EntertainmentNaNNaNThe guitar is a musical instrument classified ...
1515162157/m/0215nCartoonhttps://en.wikipedia.org/wiki/CartoonArts & EntertainmentNaNNaNA cartoon is a type of two-dimensional illustr...
1616160383/m/01350rPerformance arthttps://en.wikipedia.org/wiki/Performance_artArts & EntertainmentNaNNaNPerformance art is a performance presented to ...
1717121318/m/0dfbwRacinghttps://en.wikipedia.org/wiki/RacingSportsNaNNaNIn sport, racing is a competition of speed, ag...
1818110787/m/05b0n7kOutdoor recreationhttps://en.wikipedia.org/wiki/Outdoor_recreationHobbies & LeisureNaNNaNOutdoor recreation or outdoor activity refers ...
1919107038/m/04tr8sPC gamehttps://en.wikipedia.org/wiki/PC_gameGamesNaNNaNPC games, also known as computer games or pers...
202097912/m/03hdbfTrailer (promotion)https://en.wikipedia.org/wiki/Trailer_(promotion)Arts & EntertainmentNaNNaNA trailer is an advertisement or a commercial ...
212194966/m/019cfyStadiumhttps://en.wikipedia.org/wiki/StadiumArts & EntertainmentNaNNaNA stadium is a place or venue for outdoor spor...
222292866/m/05h0nNaturehttps://en.wikipedia.org/wiki/NatureScienceNaNNaNNature, in the broadest sense, is the natural,...
232390233/m/050k8Mobile phonehttps://en.wikipedia.org/wiki/Mobile_phoneInternet & TelecomNaNNaNA mobile phone is a portable telephone that ca...
242485888/m/0d8lmString instrumenthttps://en.wikipedia.org/wiki/String_instrumentArts & EntertainmentNaNNaNString instruments, stringed instruments, or c...
252584833/m/0138tlToyhttps://en.wikipedia.org/wiki/ToyShoppingNaNNaNA toy is an item that is generally used for ch...
262684600/m/01mtbCookinghttps://en.wikipedia.org/wiki/CookingFood & DrinkNaNNaNCooking or cookery is the art, technology and ...
272784366/m/04_svMotorcyclehttps://en.wikipedia.org/wiki/MotorcycleAutos & VehiclesNaNNaNA motorcycle is a two- or three-wheeled motor ...
282884355/m/032tlFashionhttps://en.wikipedia.org/wiki/FashionBeauty & FitnessNaNNaNFashion is a popular style or practice, especi...
292982217/m/0169zhSmartphonehttps://en.wikipedia.org/wiki/SmartphoneInternet & TelecomNaNNaNA smartphone is a mobile phone with an advance...
..............................
46864686113/m/0k3hjq7Nisekoihttps://en.wikipedia.org/wiki/Nisekoi(Unknown)NaNNaNNisekoi, released in English as Nisekoi: False...
46874687113/m/048hm5Hollywood Palladiumhttps://en.wikipedia.org/wiki/Hollywood_PalladiumArts & EntertainmentNaNNaNHollywood Palladium is a theater located at 62...
46884688112/m/0gtw3vhSkullgirlshttps://en.wikipedia.org/wiki/SkullgirlsGamesNaNNaNSkullgirls is a 2D fighting game developed by ...
46894689112/m/024902Al-Masjid an-Nabawihttps://en.wikipedia.org/wiki/Al-Masjid_an-NabawiTravelPeople & SocietyNaNAl-Masjid an-Nabawī is a mosque established an...
46904690111/m/0fv8j0Shadow Fighter (video game)https://en.wikipedia.org/wiki/Shadow_Fighter_(...GamesNaNNaNShadow Fighter is a computer game for the Comm...
46914691111/m/044p09Reptile (Mortal Kombat)https://en.wikipedia.org/wiki/Reptile_(Mortal_...GamesNaNNaNReptile is a video game character from the Mor...
46924692111/m/02z2m_Abercrombie & Fitchhttps://en.wikipedia.org/wiki/Abercrombie_%26_...ShoppingNaNNaNAbercrombie & Fitch is an American retailer th...
46934693110/m/0dng4Daffy Duckhttps://en.wikipedia.org/wiki/Daffy_DuckArts & EntertainmentNaNNaNDaffy Duck is an animated cartoon character pr...
46944694110/m/032nchCruelty to animalshttps://en.wikipedia.org/wiki/Cruelty_to_animalsPets & AnimalsNaNNaNCruelty to animals, also called animal abuse o...
46954695110/m/01k796Strikeouthttps://en.wikipedia.org/wiki/StrikeoutSportsNaNNaNIn baseball or softball, a strikeout occurs wh...
46964696109/m/0pjz3Arjunahttps://en.wikipedia.org/wiki/ArjunaPeople & SocietyReferenceNaNArjuna along with Krishna is the protagonist o...
46974697108/m/0661mn1Scarlet Weather Rhapsodyhttps://en.wikipedia.org/wiki/Scarlet_Weather_...Arts & EntertainmentNaNNaNTouhou Hisouten ~ Scarlet Weather Rhapsody. is...
46984698108/m/02z1c71Voetbal Internationalhttps://en.wikipedia.org/wiki/Voetbal_Internat...Books & LiteratureSportsNaNVoetbal International is a Dutch football maga...
46994699108/m/01n2m6Death Row Recordshttps://en.wikipedia.org/wiki/Death_Row_RecordsArts & EntertainmentNaNNaNDeath Row Records is a record company founded ...
47004700108/m/02p1p9mShashlikhttps://en.wikipedia.org/wiki/ShashlikFood & DrinkNaNNaNShashlik or shashlyk, is a dish of skewered an...
47014701107/m/01tcs0Scarecrow (DC Comics)https://en.wikipedia.org/wiki/Scarecrow_(DC_Co...Arts & EntertainmentNaNNaNThe Scarecrow is a fictional supervillain appe...
47024702107/m/0_83vt5Akame ga Kill!https://en.wikipedia.org/wiki/Akame_ga_Kill!(Unknown)NaNNaNAkame ga Kill! is a Japanese shōnen manga seri...
47034703107/m/0c3_50sDJ Hero 2https://en.wikipedia.org/wiki/DJ_Hero_2GamesNaNNaNDJ Hero 2 is a rhythm video game and a sequel ...
47044704106/m/0mbctGonghttps://en.wikipedia.org/wiki/GongArts & EntertainmentNaNNaNA gong is an African, East and South East Asia...
47054705105/m/0lq1lznSCP – Containment Breachhttps://en.wikipedia.org/wiki/SCP_%E2%80%93_Co...(Unknown)NaNNaNSCP – Containment Breach is an indie supernatu...
47064706104/m/0875jrFriday the 13th (franchise)https://en.wikipedia.org/wiki/Friday_the_13th_...Arts & EntertainmentNaNNaNFriday the 13th is an American horror franchis...
47074707104/m/03qtgnPonchohttps://en.wikipedia.org/wiki/PonchoShoppingNaNNaNA poncho is an outer garment designed to keep ...
47084708103/m/02stb2Visual novelhttps://en.wikipedia.org/wiki/Visual_novelGamesNaNNaNA visual novel is an interactive game, featuri...
47094709103/m/09f6bLugehttps://en.wikipedia.org/wiki/LugeSportsNaNNaNA luge /ˈluːʒ/ is a small one- or two-person s...
47104710103/m/04rkyMammalhttps://en.wikipedia.org/wiki/MammalScienceNaNNaNMammals are any vertebrates within the class M...
47114711103/m/06mk92Aston Martin Vantage (2005)https://en.wikipedia.org/wiki/Aston_Martin_Van...Autos & VehiclesNaNNaNThe Aston Martin Vantage is series of hand-bui...
47124712102/m/048ytxRussian pyramidhttps://en.wikipedia.org/wiki/Russian_pyramidGamesNaNNaNRussian pyramid, also known as Russian billiar...
47134713101/m/06_zh7PBS Kidshttps://en.wikipedia.org/wiki/PBS_KidsArts & EntertainmentPeople & SocietyNaNPBS Kids, stylized as PBS KIDS and formerly PT...
47144714101/m/089w_2Air Gearhttps://en.wikipedia.org/wiki/Air_GearArts & EntertainmentNaNNaNAir Gear is a shōnen manga written and illustr...
47154715101/m/02qg86lPina Recordshttps://en.wikipedia.org/wiki/Pina_RecordsArts & EntertainmentNaNNaNPina Records is a Puerto Rican record label fo...
\n", "

4716 rows × 9 columns

\n", "
" ], "text/plain": [ " Index TrainVideoCount KnowledgeGraphId Name \\\n", "0 0 865174 /m/03bt1gh Game \n", "1 1 683166 /m/07yv9 Vehicle \n", "2 2 522427 /m/01mw1 Video game \n", "3 3 497487 /m/01jddz Concert \n", "4 4 373952 /m/0k4j Car \n", "5 5 305143 /m/026bk Dance \n", "6 6 299054 /m/0hcr Animation \n", "7 7 296577 /m/09jwl Musician \n", "8 8 219506 /m/02vx4 Association football \n", "9 9 217037 /m/0mdxd Music video \n", "10 10 203700 /m/0jbk Animal \n", "11 11 198630 /m/0410tth Motorsport \n", "12 12 196108 /m/02wbm Food \n", "13 13 170351 /m/05229 Musical ensemble \n", "14 14 162864 /m/0342h Guitar \n", "15 15 162157 /m/0215n Cartoon \n", "16 16 160383 /m/01350r Performance art \n", "17 17 121318 /m/0dfbw Racing \n", "18 18 110787 /m/05b0n7k Outdoor recreation \n", "19 19 107038 /m/04tr8s PC game \n", "20 20 97912 /m/03hdbf Trailer (promotion) \n", "21 21 94966 /m/019cfy Stadium \n", "22 22 92866 /m/05h0n Nature \n", "23 23 90233 /m/050k8 Mobile phone \n", "24 24 85888 /m/0d8lm String instrument \n", "25 25 84833 /m/0138tl Toy \n", "26 26 84600 /m/01mtb Cooking \n", "27 27 84366 /m/04_sv Motorcycle \n", "28 28 84355 /m/032tl Fashion \n", "29 29 82217 /m/0169zh Smartphone \n", "... ... ... ... ... \n", "4686 4686 113 /m/0k3hjq7 Nisekoi \n", "4687 4687 113 /m/048hm5 Hollywood Palladium \n", "4688 4688 112 /m/0gtw3vh Skullgirls \n", "4689 4689 112 /m/024902 Al-Masjid an-Nabawi \n", "4690 4690 111 /m/0fv8j0 Shadow Fighter (video game) \n", "4691 4691 111 /m/044p09 Reptile (Mortal Kombat) \n", "4692 4692 111 /m/02z2m_ Abercrombie & Fitch \n", "4693 4693 110 /m/0dng4 Daffy Duck \n", "4694 4694 110 /m/032nch Cruelty to animals \n", "4695 4695 110 /m/01k796 Strikeout \n", "4696 4696 109 /m/0pjz3 Arjuna \n", "4697 4697 108 /m/0661mn1 Scarlet Weather Rhapsody \n", "4698 4698 108 /m/02z1c71 Voetbal International \n", "4699 4699 108 /m/01n2m6 Death Row Records \n", "4700 4700 108 /m/02p1p9m Shashlik \n", "4701 4701 107 /m/01tcs0 Scarecrow (DC Comics) \n", "4702 4702 107 /m/0_83vt5 Akame ga Kill! \n", "4703 4703 107 /m/0c3_50s DJ Hero 2 \n", "4704 4704 106 /m/0mbct Gong \n", "4705 4705 105 /m/0lq1lzn SCP – Containment Breach \n", "4706 4706 104 /m/0875jr Friday the 13th (franchise) \n", "4707 4707 104 /m/03qtgn Poncho \n", "4708 4708 103 /m/02stb2 Visual novel \n", "4709 4709 103 /m/09f6b Luge \n", "4710 4710 103 /m/04rky Mammal \n", "4711 4711 103 /m/06mk92 Aston Martin Vantage (2005) \n", "4712 4712 102 /m/048ytx Russian pyramid \n", "4713 4713 101 /m/06_zh7 PBS Kids \n", "4714 4714 101 /m/089w_2 Air Gear \n", "4715 4715 101 /m/02qg86l Pina Records \n", "\n", " WikiUrl Vertical1 \\\n", "0 https://en.wikipedia.org/wiki/Game Games \n", "1 https://en.wikipedia.org/wiki/Vehicle Autos & Vehicles \n", "2 https://en.wikipedia.org/wiki/Video_game Games \n", "3 https://en.wikipedia.org/wiki/Concert Arts & Entertainment \n", "4 https://en.wikipedia.org/wiki/Car Autos & Vehicles \n", "5 https://en.wikipedia.org/wiki/Dance Arts & Entertainment \n", "6 https://en.wikipedia.org/wiki/Animation Arts & Entertainment \n", "7 https://en.wikipedia.org/wiki/Musician Arts & Entertainment \n", "8 https://en.wikipedia.org/wiki/Association_foot... Sports \n", "9 https://en.wikipedia.org/wiki/Music_video Arts & Entertainment \n", "10 https://en.wikipedia.org/wiki/Animal Pets & Animals \n", "11 https://en.wikipedia.org/wiki/Motorsport Sports \n", "12 https://en.wikipedia.org/wiki/Food Food & Drink \n", "13 https://en.wikipedia.org/wiki/Musical_ensemble Arts & Entertainment \n", "14 https://en.wikipedia.org/wiki/Guitar Arts & Entertainment \n", "15 https://en.wikipedia.org/wiki/Cartoon Arts & Entertainment \n", "16 https://en.wikipedia.org/wiki/Performance_art Arts & Entertainment \n", "17 https://en.wikipedia.org/wiki/Racing Sports \n", "18 https://en.wikipedia.org/wiki/Outdoor_recreation Hobbies & Leisure \n", "19 https://en.wikipedia.org/wiki/PC_game Games \n", "20 https://en.wikipedia.org/wiki/Trailer_(promotion) Arts & Entertainment \n", "21 https://en.wikipedia.org/wiki/Stadium Arts & Entertainment \n", "22 https://en.wikipedia.org/wiki/Nature Science \n", "23 https://en.wikipedia.org/wiki/Mobile_phone Internet & Telecom \n", "24 https://en.wikipedia.org/wiki/String_instrument Arts & Entertainment \n", "25 https://en.wikipedia.org/wiki/Toy Shopping \n", "26 https://en.wikipedia.org/wiki/Cooking Food & Drink \n", "27 https://en.wikipedia.org/wiki/Motorcycle Autos & Vehicles \n", "28 https://en.wikipedia.org/wiki/Fashion Beauty & Fitness \n", "29 https://en.wikipedia.org/wiki/Smartphone Internet & Telecom \n", "... ... ... \n", "4686 https://en.wikipedia.org/wiki/Nisekoi (Unknown) \n", "4687 https://en.wikipedia.org/wiki/Hollywood_Palladium Arts & Entertainment \n", "4688 https://en.wikipedia.org/wiki/Skullgirls Games \n", "4689 https://en.wikipedia.org/wiki/Al-Masjid_an-Nabawi Travel \n", "4690 https://en.wikipedia.org/wiki/Shadow_Fighter_(... Games \n", "4691 https://en.wikipedia.org/wiki/Reptile_(Mortal_... Games \n", "4692 https://en.wikipedia.org/wiki/Abercrombie_%26_... Shopping \n", "4693 https://en.wikipedia.org/wiki/Daffy_Duck Arts & Entertainment \n", "4694 https://en.wikipedia.org/wiki/Cruelty_to_animals Pets & Animals \n", "4695 https://en.wikipedia.org/wiki/Strikeout Sports \n", "4696 https://en.wikipedia.org/wiki/Arjuna People & Society \n", "4697 https://en.wikipedia.org/wiki/Scarlet_Weather_... Arts & Entertainment \n", "4698 https://en.wikipedia.org/wiki/Voetbal_Internat... Books & Literature \n", "4699 https://en.wikipedia.org/wiki/Death_Row_Records Arts & Entertainment \n", "4700 https://en.wikipedia.org/wiki/Shashlik Food & Drink \n", "4701 https://en.wikipedia.org/wiki/Scarecrow_(DC_Co... Arts & Entertainment \n", "4702 https://en.wikipedia.org/wiki/Akame_ga_Kill! (Unknown) \n", "4703 https://en.wikipedia.org/wiki/DJ_Hero_2 Games \n", "4704 https://en.wikipedia.org/wiki/Gong Arts & Entertainment \n", "4705 https://en.wikipedia.org/wiki/SCP_%E2%80%93_Co... (Unknown) \n", "4706 https://en.wikipedia.org/wiki/Friday_the_13th_... Arts & Entertainment \n", "4707 https://en.wikipedia.org/wiki/Poncho Shopping \n", "4708 https://en.wikipedia.org/wiki/Visual_novel Games \n", "4709 https://en.wikipedia.org/wiki/Luge Sports \n", "4710 https://en.wikipedia.org/wiki/Mammal Science \n", "4711 https://en.wikipedia.org/wiki/Aston_Martin_Van... Autos & Vehicles \n", "4712 https://en.wikipedia.org/wiki/Russian_pyramid Games \n", "4713 https://en.wikipedia.org/wiki/PBS_Kids Arts & Entertainment \n", "4714 https://en.wikipedia.org/wiki/Air_Gear Arts & Entertainment \n", "4715 https://en.wikipedia.org/wiki/Pina_Records Arts & Entertainment \n", "\n", " Vertical2 Vertical3 \\\n", "0 NaN NaN \n", "1 NaN NaN \n", "2 NaN NaN \n", "3 NaN NaN \n", "4 NaN NaN \n", "5 NaN NaN \n", "6 NaN NaN \n", "7 NaN NaN \n", "8 NaN NaN \n", "9 NaN NaN \n", "10 NaN NaN \n", "11 NaN NaN \n", "12 NaN NaN \n", "13 NaN NaN \n", "14 NaN NaN \n", "15 NaN NaN \n", "16 NaN NaN \n", "17 NaN NaN \n", "18 NaN NaN \n", "19 NaN NaN \n", "20 NaN NaN \n", "21 NaN NaN \n", "22 NaN NaN \n", "23 NaN NaN \n", "24 NaN NaN \n", "25 NaN NaN \n", "26 NaN NaN \n", "27 NaN NaN \n", "28 NaN NaN \n", "29 NaN NaN \n", "... ... ... \n", "4686 NaN NaN \n", "4687 NaN NaN \n", "4688 NaN NaN \n", "4689 People & Society NaN \n", "4690 NaN NaN \n", "4691 NaN NaN \n", "4692 NaN NaN \n", "4693 NaN NaN \n", "4694 NaN NaN \n", "4695 NaN NaN \n", "4696 Reference NaN \n", "4697 NaN NaN \n", "4698 Sports NaN \n", "4699 NaN NaN \n", "4700 NaN NaN \n", "4701 NaN NaN \n", "4702 NaN NaN \n", "4703 NaN NaN \n", "4704 NaN NaN \n", "4705 NaN NaN \n", "4706 NaN NaN \n", "4707 NaN NaN \n", "4708 NaN NaN \n", "4709 NaN NaN \n", "4710 NaN NaN \n", "4711 NaN NaN \n", "4712 NaN NaN \n", "4713 People & Society NaN \n", "4714 NaN NaN \n", "4715 NaN NaN \n", "\n", " WikiDescription \n", "0 A game is structured form of play, usually und... \n", "1 A vehicle is a mobile machine that transports ... \n", "2 A video game is an electronic game that involv... \n", "3 A concert is a live music performance in front... \n", "4 A car is a wheeled, self-powered motor vehicle... \n", "5 Dance is a performance art form consisting of ... \n", "6 Animation is the process of making the illusio... \n", "7 A musician is a person who plays a musical ins... \n", "8 Association football, more commonly known as f... \n", "9 A music video is a short film integrating a so... \n", "10 Animals are multicellular, eukaryotic organism... \n", "11 Motorsport or motorsports is a global term use... \n", "12 Food is any substance consumed to provide nutr... \n", "13 A musical ensemble, also known as a music grou... \n", "14 The guitar is a musical instrument classified ... \n", "15 A cartoon is a type of two-dimensional illustr... \n", "16 Performance art is a performance presented to ... \n", "17 In sport, racing is a competition of speed, ag... \n", "18 Outdoor recreation or outdoor activity refers ... \n", "19 PC games, also known as computer games or pers... \n", "20 A trailer is an advertisement or a commercial ... \n", "21 A stadium is a place or venue for outdoor spor... \n", "22 Nature, in the broadest sense, is the natural,... \n", "23 A mobile phone is a portable telephone that ca... \n", "24 String instruments, stringed instruments, or c... \n", "25 A toy is an item that is generally used for ch... \n", "26 Cooking or cookery is the art, technology and ... \n", "27 A motorcycle is a two- or three-wheeled motor ... \n", "28 Fashion is a popular style or practice, especi... \n", "29 A smartphone is a mobile phone with an advance... \n", "... ... \n", "4686 Nisekoi, released in English as Nisekoi: False... \n", "4687 Hollywood Palladium is a theater located at 62... \n", "4688 Skullgirls is a 2D fighting game developed by ... \n", "4689 Al-Masjid an-Nabawī is a mosque established an... \n", "4690 Shadow Fighter is a computer game for the Comm... \n", "4691 Reptile is a video game character from the Mor... \n", "4692 Abercrombie & Fitch is an American retailer th... \n", "4693 Daffy Duck is an animated cartoon character pr... \n", "4694 Cruelty to animals, also called animal abuse o... \n", "4695 In baseball or softball, a strikeout occurs wh... \n", "4696 Arjuna along with Krishna is the protagonist o... \n", "4697 Touhou Hisouten ~ Scarlet Weather Rhapsody. is... \n", "4698 Voetbal International is a Dutch football maga... \n", "4699 Death Row Records is a record company founded ... \n", "4700 Shashlik or shashlyk, is a dish of skewered an... \n", "4701 The Scarecrow is a fictional supervillain appe... \n", "4702 Akame ga Kill! is a Japanese shōnen manga seri... \n", "4703 DJ Hero 2 is a rhythm video game and a sequel ... \n", "4704 A gong is an African, East and South East Asia... \n", "4705 SCP – Containment Breach is an indie supernatu... \n", "4706 Friday the 13th is an American horror franchis... \n", "4707 A poncho is an outer garment designed to keep ... \n", "4708 A visual novel is an interactive game, featuri... \n", "4709 A luge /ˈluːʒ/ is a small one- or two-person s... \n", "4710 Mammals are any vertebrates within the class M... \n", "4711 The Aston Martin Vantage is series of hand-bui... \n", "4712 Russian pyramid, also known as Russian billiar... \n", "4713 PBS Kids, stylized as PBS KIDS and formerly PT... \n", "4714 Air Gear is a shōnen manga written and illustr... \n", "4715 Pina Records is a Puerto Rican record label fo... \n", "\n", "[4716 rows x 9 columns]" ] }, "execution_count": 6, "metadata": {}, "output_type": "execute_result" } ], "source": [ "vocabulary" ] }, { "cell_type": "code", "execution_count": 7, "metadata": { "collapsed": false }, "outputs": [], "source": [ "vertical_dict = {}\n", "with open(\"vertical.tsv\", \"w\") as F:\n", " lines = []\n", " for num, item in vocabulary.iterrows():\n", " index = item[\"Index\"]\n", " vertical = item[\"Vertical1\"]\n", " if vertical not in vertical_dict:\n", " vertical_dict[vertical] = len(vertical_dict)\n", " lines.append(\"%d\\t%d\\n\" % (index, vertical_dict[vertical]))\n", " F.writelines(lines)" ] }, { "cell_type": "code", "execution_count": 8, "metadata": { "collapsed": false }, "outputs": [ { "name": "stdout", "output_type": "stream", "text": [ "25\n" ] } ], "source": [ "print len(vertical_dict)" ] }, { "cell_type": "markdown", "metadata": {}, "source": [ "## Training labels" ] }, { "cell_type": "code", "execution_count": 9, "metadata": { "collapsed": false }, "outputs": [ { "data": { "text/html": [ "
\n", "\n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", "
idlabels
0--DwgB78t-c16 5 430
1--NZRkXBV7k128 3 39 7 44 13 16 30
2--hoQ2sGG4M694
3--ABhs9ik7c2232 1 4 517 270
4--sBoaqBlzA45 125
5--7h1S4neDM0 48 10 356
6--F672jfCMo488 1067 1078
7--ezS5q-mZg1 962 12 625
8--XT8O4T3Wc379 6
9--Qgwg7mGZY0 33 2 2379 3423
10--_I8vffnIw768 37 72 330 331 29 373 23 189
11--VDzNzHHic20
12--pQCGgGjE80 2 836 432 665 2779
13--amuSGpp94118
14--mkDBHaFIs6
15--DyrrdZbUk1 114 1395 4 3409
16--cPdta0k4Y276
17--5iMsHNuDE1206 6 967
18--f5fEIOW1Q9
19--jcB8BTawk5 31
20--0soWyo5gQ1 403 4 1511 975
21--8MxffdH9k1 4
22--j9Z0JKrjE193 436 45 173
23--o719vqu6Q0 49 168 60 54
24--p02WMptcA49
25--0pQywa6OM1139
26--sW2Fk-LmY0 2 35
27--iEJ1NTTHQ8 813 79
28--39-Ml5Ylw99 3581
29--udooZCtNE200 25
.........
4906630zzCGK4T43sI16 3
4906631zz9AvyGAFhg0 2 35
4906632zzAhVFy3o5s233 222
4906633zzhKCCjBtN0208 170 830 246
4906634zzQfq7MBrfk80 10 438 247
4906635zz5H1l4fTRs0 33 2 1382
4906636zztyezN9m801002
4906637zz9urhmAd6M16 3 5
4906638zzieB3E90oc49 3 13 38 7
4906639zz-AlFFQAbI354 713 12 2482 466 1047
4906640zzYbFq5qF8w400 2017 972
4906641zzyTPLaoWxo3 38 39 7 13 49
4906642zzp9Q2bQhck1 4 202 11 17 164 92 447
4906643zz4L8mQ0Zr8112 59 316
4906644zzxF2j31aBE1 1323 11 172 17 278 728 57 27 669
4906645zzTw-YSIp1g163
4906646zznsbYnqvyU0 2 34 1069 19 118
4906647zz9z1DR2RXA24 14 63
4906648zzk5fGf0nMw8
4906649zzrCAKdNth4744 1794 3182
4906650zzX28lrE9Fc46
4906651zzMFAaqNGX4208
4906652zzhc0nTbC4Q49 5 95
4906653zztFMPeS4KU20
4906654zz5R0c6JHBk56 106 28
4906655zzRKzJ5iCRs416 113 652 339 580
4906656zzDg9v6733A1 4
4906657zzZl6-AxXv80 60
4906658zzoW1afkGKk2033 1083 3959
4906659zzfubODk_8E10 6 15
\n", "

4906660 rows × 2 columns

\n", "
" ], "text/plain": [ " id labels\n", "0 --DwgB78t-c 16 5 430\n", "1 --NZRkXBV7k 128 3 39 7 44 13 16 30\n", "2 --hoQ2sGG4M 694\n", "3 --ABhs9ik7c 2232 1 4 517 270\n", "4 --sBoaqBlzA 45 125\n", "5 --7h1S4neDM 0 48 10 356\n", "6 --F672jfCMo 488 1067 1078\n", "7 --ezS5q-mZg 1 962 12 625\n", "8 --XT8O4T3Wc 379 6\n", "9 --Qgwg7mGZY 0 33 2 2379 3423\n", "10 --_I8vffnIw 768 37 72 330 331 29 373 23 189\n", "11 --VDzNzHHic 20\n", "12 --pQCGgGjE8 0 2 836 432 665 2779\n", "13 --amuSGpp94 118\n", "14 --mkDBHaFIs 6\n", "15 --DyrrdZbUk 1 114 1395 4 3409\n", "16 --cPdta0k4Y 276\n", "17 --5iMsHNuDE 1206 6 967\n", "18 --f5fEIOW1Q 9\n", "19 --jcB8BTawk 5 31\n", "20 --0soWyo5gQ 1 403 4 1511 975\n", "21 --8MxffdH9k 1 4\n", "22 --j9Z0JKrjE 193 436 45 173\n", "23 --o719vqu6Q 0 49 168 60 54\n", "24 --p02WMptcA 49\n", "25 --0pQywa6OM 1139\n", "26 --sW2Fk-LmY 0 2 35\n", "27 --iEJ1NTTHQ 8 813 79\n", "28 --39-Ml5Ylw 99 3581\n", "29 --udooZCtNE 200 25\n", "... ... ...\n", "4906630 zzCGK4T43sI 16 3\n", "4906631 zz9AvyGAFhg 0 2 35\n", "4906632 zzAhVFy3o5s 233 222\n", "4906633 zzhKCCjBtN0 208 170 830 246\n", "4906634 zzQfq7MBrfk 80 10 438 247\n", "4906635 zz5H1l4fTRs 0 33 2 1382\n", "4906636 zztyezN9m80 1002\n", "4906637 zz9urhmAd6M 16 3 5\n", "4906638 zzieB3E90oc 49 3 13 38 7\n", "4906639 zz-AlFFQAbI 354 713 12 2482 466 1047\n", "4906640 zzYbFq5qF8w 400 2017 972\n", "4906641 zzyTPLaoWxo 3 38 39 7 13 49\n", "4906642 zzp9Q2bQhck 1 4 202 11 17 164 92 447\n", "4906643 zz4L8mQ0Zr8 112 59 316\n", "4906644 zzxF2j31aBE 1 1323 11 172 17 278 728 57 27 669\n", "4906645 zzTw-YSIp1g 163\n", "4906646 zznsbYnqvyU 0 2 34 1069 19 118\n", "4906647 zz9z1DR2RXA 24 14 63\n", "4906648 zzk5fGf0nMw 8\n", "4906649 zzrCAKdNth4 744 1794 3182\n", "4906650 zzX28lrE9Fc 46\n", "4906651 zzMFAaqNGX4 208\n", "4906652 zzhc0nTbC4Q 49 5 95\n", "4906653 zztFMPeS4KU 20\n", "4906654 zz5R0c6JHBk 56 106 28\n", "4906655 zzRKzJ5iCRs 416 113 652 339 580\n", "4906656 zzDg9v6733A 1 4\n", "4906657 zzZl6-AxXv8 0 60\n", "4906658 zzoW1afkGKk 2033 1083 3959\n", "4906659 zzfubODk_8E 10 6 15\n", "\n", "[4906660 rows x 2 columns]" ] }, "execution_count": 9, "metadata": {}, "output_type": "execute_result" } ], "source": [ "train_labels" ] }, { "cell_type": "markdown", "metadata": {}, "source": [ "### show some of the label combinations" ] }, { "cell_type": "code", "execution_count": 10, "metadata": { "collapsed": false, "scrolled": true }, "outputs": [ { "name": "stdout", "output_type": "stream", "text": [ "5-Dance / 16-Performance art / 430-Ballroom dance / \n", "3-Concert / 7-Musician / 13-Musical ensemble / 16-Performance art / 30-Drummer / 39-Drums / 44-Drums / 128-Cymbal / \n", "694-Tai chi / \n", "1-Vehicle / 4-Car / 270-Sedan / 517-Audi / 2232-Audi A3 / \n", "45-Cosmetics / 125-Television / \n", "0-Games / 10-Animal / 48-Pet / 356-Star / \n", "488-Door / 1067-Lock / 1078-Installation art / \n", "1-Vehicle / 12-Food / 625-Factory / 962-McDonald's / \n", "6-Animation / 379-Touhou Project / \n", "0-Games / 2-Video game / 33-Weapon / 2379-Tom Clancy's Ghost Recon / 3423-Tom Clancy's Ghost Recon: Future Soldier / \n", "23-Mobile phone / 29-Smartphone / 37-Gadget / 72-iPhone / 189-iPad / 330-Money / 331-iPod touch / 373-Clash of Clans / 768-Diamond / \n", "20-Trailer / \n", "0-Games / 2-Video game / 432-Kingdom Hearts / 665-Kingdom Hearts / 836-Sora / 2779-Universe of Kingdom Hearts / \n", "118-Grand Theft Auto V / \n", "6-Animation / \n", "1-Vehicle / 4-Car / 114-Engine / 1395-Valve / 3409-Nissan Altima / \n", "276-Room / \n", "6-Animation / 967-Text / 1206-Cinema 4D / \n", "9-Music video / \n", "5-Dance / 31-Disc jockey / \n", "1-Vehicle / 4-Car / 403-Toyota / 975-Dune / 1511-Toyota Land Cruiser / \n", "1-Vehicle / 4-Car / \n", "45-Cosmetics / 173-Eye shadow / 193-Eye liner / 436-Concealer / \n", "0-Games / 49-School / 54-Highlight film / 60-Basketball / 168-Basketball moves / \n", "49-School / \n", "1139-Belle / \n", "0-Games / 2-Video game / 35-Minecraft / \n", "8-Football / 79-American football / 813-Running back / \n", "99-Christmas / 3581-Rudolph the Red-Nosed Reindeer / \n", "25-Toy / 200-Doll / \n", "1427-TalesRunner / \n", "983-Lexus / \n", "59-Winter sport / 84-Snow / 177-Skiing / \n", "3-Concert / 5-Dance / 16-Performance art / \n", "3-Concert / 5-Dance / 7-Musician / 13-Musical ensemble / 38-Orchestra / 46-Choir / \n", "3-Concert / 5-Dance / 68-Lighting / \n", "1-Vehicle / 4-Car / 11-Motorsport / 17-Racing / 40-Road / 41-Sports car / 57-Race track / 135-Drifting / 212-Highway / 231-Supercar / 705-Forza Motorsport / 1530-Forza Motorsport 4 / 2261-Nissan Silvia / \n", "5-Dance / \n", "12-Food / 26-Cooking / 32-Recipe / 226-Dessert / 232-Cake / 236-Baking / 295-Chocolate / 534-Milk / 759-Icing / 985-Strawberry / 1728-Chocolate cake / \n", "2-Video game / 20-Trailer / \n", "42-Fishing / 902-Fishing bait / 1018-Salmon / \n", "3-Concert / \n", "0-Games / 2-Video game / 1027-Terraria / 2821-Cactus / \n", "1-Vehicle / 4-Car / 11-Motorsport / 17-Racing / 25-Toy / 57-Race track / 92-Radio-controlled model / 167-Four-wheel drive / 174-Farm / 202-Radio-controlled car / 718-Duck / 926-Traxxas / \n", "1-Vehicle / 17-Racing / 27-Motorcycle / 114-Engine / 1036-Moped / \n", "2-Video game / 47-Personal computer / 216-Call of Duty: Modern Warfare 3 / \n", "36-Piano / 53-Keyboard / 107-Musical keyboard / 265-Electronic keyboard / 397-Organ / 1265-Kyle Kingson / \n", "3-Concert / 5-Dance / 68-Lighting / 133-Nightclub / \n", "42-Fishing / 315-Knife / 510-Fishing lure / 3745-Eel / \n", "7-Musician / 36-Piano / \n", "337-Advertising / \n", "0-Games / 447-Sand / \n", "6-Animation / 2176-Fist of the North Star / 3349-Kenshiro / \n", "6-Animation / 15-Cartoon / 911-Studio / \n", "1-Vehicle / 4-Car / 11-Motorsport / 17-Racing / 25-Toy / 92-Radio-controlled model / 135-Drifting / 202-Radio-controlled car / \n", "1-Vehicle / 4-Car / 11-Motorsport / 40-Road / 70-Driving / 952-Chevrolet Corvette / 2885-Chevrolet Corvette (C6) / \n", "10-Animal / 91-Fish / 225-Lake / 258-Aquarium / \n", "1-Vehicle / 4-Car / 11-Motorsport / 124-Tractor / 249-Heavy equipment / 252-Tractor pulling / \n", "6-Animation / 25-Toy / 269-Comic book / 1049-Collectible card game / \n", "6-Animation / 15-Cartoon / 723-Pig / \n", "86-Plant / 136-Gardening / 279-Garden / 363-Soil / \n", "12-Food / \n", "0-Games / 1-Vehicle / 2-Video game / 4-Car / 11-Motorsport / 17-Racing / 41-Sports car / 57-Race track / 70-Driving / 135-Drifting / \n", "0-Games / 19-PC game / 51-Strategy video game / 165-League of Legends / \n", "1-Vehicle / 4-Car / 125-Television / 950-Blu-ray disc / \n", "1-Vehicle / 4-Car / 74-Truck / 158-Tire / 1429-Dump truck / \n", "28-Fashion / \n", "253-Carnival / 641-Mask / \n", "12-Food / \n", "174-Farm / \n", "5-Dance / 16-Performance art / \n", "23-Mobile phone / 29-Smartphone / 37-Gadget / 47-Personal computer / 145-Tablet computer / \n", "131-Computer / 497-Macintosh / \n", "1-Vehicle / 4-Car / 11-Motorsport / 17-Racing / 41-Sports car / 57-Race track / 711-Gran Turismo / 790-Nissan GT-R / 1519-Gran Turismo 6 / \n", "1-Vehicle / 4-Car / 11-Motorsport / 41-Sports car / 231-Supercar / 569-Lamborghini / \n", "1650-The Idolmaster / 1755-The Idolmaster / \n", "0-Games / 2-Video game / 55-Video game console / 3217-Neo Geo / \n", "23-Mobile phone / 29-Smartphone / \n", "712-Outer space / 992-Rocket / 1508-Spacecraft / 1992-Space Shuttle / \n", "350-Circus / 788-Juggling / \n", "1-Vehicle / 50-Aircraft / 82-Airplane / 102-Building / 129-Aviation / 139-Airport / 151-Landing / 157-Jet aircraft / 175-Airliner / 187-Takeoff / 188-Airline / 254-Runway / 1736-Microsoft Flight Simulator / 2136-Boeing 757 / 2391-American Airlines / \n", "3-Concert / 16-Performance art / \n", "125-Television / 1172-VHS / \n", "0-Games / 211-RuneScape / \n", "46-Choir / 318-Church / \n", "36-Piano / 53-Keyboard / \n", "130-Gymnastics / \n", "45-Cosmetics / 436-Concealer / \n", "1-Vehicle / 11-Motorsport / 124-Tractor / 156-Agriculture / 174-Farm / 252-Tractor pulling / 617-Maize / \n", "6-Animation / 15-Cartoon / 4244-Rozen Maiden / \n", "0-Games / 2-Video game / 19-PC game / 3235-Monopoly / \n", "7-Musician / 13-Musical ensemble / 46-Choir / 150-Violin / \n", "1-Vehicle / 50-Aircraft / 129-Aviation / 151-Landing / 451-Cockpit / \n", "1-Vehicle / 1409-Missile / 1643-Submarine / 3842-Radio-controlled submarine / \n", "66-Bollywood / \n", "10-Animal / 80-Horse / \n", "0-Games / 2-Video game / 19-PC game / 33-Weapon / 96-Soldier / 257-Counter-Strike / \n", "3-Concert / 7-Musician / 30-Drummer / 44-Drums / \n", "345-Loudspeaker / 484-Subwoofer / \n", "772-Bowling / 930-Ten-pin bowling / 1217-Bowling ball / \n", "45-Cosmetics / 238-Nail / 300-Nail art / 304-Nail polish / 307-Hand / 324-Finger / 338-Manicure / \n", "0-Games / 2-Video game / 43-Call of Duty / 116-Pokémon / 134-Call of Duty: Black Ops / 138-Call of Duty: Black Ops II / 3050-Mewtwo / \n", "10-Animal / 48-Pet / 71-Dog / 182-Puppy / 2626-Shih Tzu / \n", "102-Building / 812-Plastic / 1384-Greenhouse / \n", "23-Mobile phone / 29-Smartphone / 37-Gadget / 72-iPhone / 189-iPad / 272-iPod / 331-iPod touch / 508-iPhone 5 / \n", "10-Animal / 71-Dog / 1784-Goblin / \n", "2071-Indiana Jones / \n", "0-Games / 489-Board game / 2267-Geometry Dash / 2833-Go / \n", "1-Vehicle / 62-Train / 64-Transport / 224-Rapid transit / \n", "1-Vehicle / 4-Car / 706-Deer / 955-Police officer / \n", "14-Guitar / 24-String instrument / 401-Cello / \n", "3-Concert / 7-Musician / 13-Musical ensemble / 38-Orchestra / 230-Brass instrument / 545-Quartet (ensemble) / 3185-Bronze / \n", "261-Tool / 521-Metal / 1033-Drill / 1073-Chainsaw / \n", "9-Music video / \n", "6-Animation / \n", "1-Vehicle / 4-Car / 11-Motorsport / 17-Racing / 57-Race track / \n", "6-Animation / \n", "28-Fashion / 657-Autumn / \n", "3-Concert / 7-Musician / 1684-Hillsong Church / \n", "85-Combat / \n", "67-Cycling / 69-Bicycle / 469-Bag / 835-Backpack / \n", "3-Concert / 7-Musician / 13-Musical ensemble / 30-Drummer / 39-Drums / \n", "6-Animation / 15-Cartoon / 116-Pokémon / 186-Pokémon / \n", "6-Animation / 15-Cartoon / 25-Toy / \n", "1-Vehicle / 4-Car / 11-Motorsport / 167-Four-wheel drive / 198-Off-road vehicle / 327-Mud / 360-Jeep / 565-Mud bogging / \n", "0-Games / 2-Video game / 19-PC game / 43-Call of Duty / 138-Call of Duty: Black Ops II / \n", "1-Vehicle / 4-Car / 263-Light / \n", "113-House / 416-Apartment / \n", "9-Music video / \n", "1-Vehicle / 4-Car / 20-Trailer / \n", "1-Vehicle / 11-Motorsport / \n", "1-Vehicle / 4-Car / 519-Recreational vehicle / \n", "1-Vehicle / 25-Toy / 92-Radio-controlled model / 140-LEGO / 202-Radio-controlled car / \n", "1-Vehicle / 4-Car / 114-Engine / 234-Ford / 473-Ford Mustang / 509-Classic car / \n", "10-Animal / 208-Wood / 540-Insect / 775-Bee / 784-Honey / 1121-Beehive / 2501-Picture frame / \n", "0-Games / 211-RuneScape / 1178-Quest / \n", "5-Dance / 16-Performance art / 115-Ballet / \n", "7-Musician / 36-Piano / 53-Keyboard / 104-Pianist / 107-Musical keyboard / \n", "0-Games / 8-Football / 21-Stadium / 73-Ball / \n", "1-Vehicle / 18-Outdoor recreation / 22-Nature / 67-Cycling / 69-Bicycle / 197-Trail / 220-Mountain bike / 284-Mountain biking / \n", "8-Football / \n", "3-Concert / 16-Performance art / \n", "0-Games / 2-Video game / 111-PlayStation 3 / \n", "1-Vehicle / 4-Car / 27-Motorcycle / 114-Engine / \n", "8-Football / 21-Stadium / \n", "9-Music video / \n", "5-Dance / 46-Choir / \n", "1-Vehicle / \n", "14-Guitar / 24-String instrument / 100-Electric guitar / \n", "0-Games / 1-Vehicle / 2-Video game / 34-Action-adventure game / 118-Grand Theft Auto V / 176-PlayStation 4 / 181-Xbox / 314-Xbox One / \n", "1-Vehicle / 50-Aircraft / 82-Airplane / 139-Airport / \n", "0-Games / 2-Video game / 34-Action-adventure game / 85-Combat / 656-Metal Gear / 2698-Metal Gear Rising: Revengeance / \n", "20-Trailer / \n", "0-Games / 6-Animation / 51-Strategy video game / 460-Yu-Gi-Oh! Trading Card Game / \n", "3-Concert / 16-Performance art / \n", "9-Music video / \n", "0-Games / 21-Stadium / 54-Highlight film / 81-Athlete / 562-Pitcher / 1111-Home run / \n", "6-Animation / \n", "6-Animation / 15-Cartoon / 89-Comics / 294-Batman / 803-DC Comics / 2082-Catwoman / \n", "36-Piano / 53-Keyboard / 107-Musical keyboard / 1909-Analog synthesizer / 2521-Jupiter / \n", "1-Vehicle / 4-Car / 11-Motorsport / 17-Racing / 40-Road / 41-Sports car / 70-Driving / 231-Supercar / 344-Coupé / 1056-Porsche 911 / 1965-Porsche Carrera / 2208-Porsche 911 GT3 / \n", "6-Animation / 15-Cartoon / 801-Inuyasha / \n", "12-Food / 26-Cooking / 32-Recipe / 232-Cake / 236-Baking / 3020-Raspberry / \n", "3-Concert / 38-Orchestra / 1431-Madison Square Garden / \n", "1-Vehicle / 50-Aircraft / 129-Aviation / 139-Airport / 151-Landing / 157-Jet aircraft / 175-Airliner / 187-Takeoff / 188-Airline / 254-Runway / 471-Wing / 1316-Airbus A330 / \n", "28-Fashion / 56-Hair / 106-Hairstyle / 583-Wig / \n", "0-Games / 2-Video game / 229-Halo / 470-Halo 3 / \n", "56-Hair / 106-Hairstyle / \n", "6-Animation / 15-Cartoon / 1866-Katara / \n", "0-Games / 2-Video game / 51-Strategy video game / 791-Magic: The Gathering / \n", "6-Animation / 25-Toy / 140-LEGO / 477-Transformers / 990-Optimus Prime / \n", "2-Video game / 23-Mobile phone / 29-Smartphone / 37-Gadget / 72-iPhone / 189-iPad / \n", "0-Games / 8-Football / 21-Stadium / 65-Kick / 73-Ball / \n", "14-Guitar / 63-Acoustic guitar / \n", "3674-Barcode / 3780-QR code / \n", "3-Concert / 5-Dance / 21-Stadium / 342-Windows Media Video / \n", "8-Football / \n", "9-Music video / \n", "0-Games / 1-Vehicle / 2-Video game / 74-Truck / 328-Simulation video game / 1099-Euro Truck Simulator 2 / \n", "23-Mobile phone / 29-Smartphone / 37-Gadget / 101-Telephone / 2217-Moto G / \n", "0-Games / 2-Video game / 33-Weapon / 43-Call of Duty / 422-Call of Duty: Ghosts / \n", "75-Wrestling / 85-Combat / \n", "0-Games / 2-Video game / 75-Wrestling / 127-Xbox 360 / 777-WWE 2K / 2629-WWE SmackDown vs. Raw 2010 / \n", "3-Concert / 16-Performance art / \n", "468-Biology / 1110-Cell / \n", "3-Concert / 5-Dance / \n", "3-Concert / 7-Musician / \n", "5-Dance / 115-Ballet / 980-Solo dance / \n", "28-Fashion / \n", "3-Concert / 5-Dance / 16-Performance art / 95-Talent show / \n", "227-Water / \n", "0-Games / 2-Video game / 474-Destiny / \n", "2-Video game / 20-Trailer / 34-Action-adventure game / 1144-Castlevania / \n", "110-Album / \n", "23-Mobile phone / 29-Smartphone / 101-Telephone / 932-Microsoft Lumia / 3376-Memory card / \n", "2627-Little Red Riding Hood / \n", "6-Animation / 123-Naruto / \n", "1-Vehicle / 18-Outdoor recreation / 67-Cycling / 69-Bicycle / 220-Mountain bike / 284-Mountain biking / 520-Bicycle frame / \n", "0-Games / 60-Basketball / 81-Athlete / 255-Slam dunk / \n", "961-Sega Genesis / \n", "5-Dance / 16-Performance art / 98-Festival / \n", "0-Games / 2-Video game / 241-Fighting game / 764-The King of Fighters / 2714-Iori Yagami / 3160-Kyo Kusanagi / \n", "0-Games / 2-Video game / 6-Animation / 241-Fighting game / 486-M.U.G.E.N / \n", "345-Loudspeaker / \n", "695-Balloon / \n", "8-Football / \n", "18-Outdoor recreation / 42-Fishing / 120-Winter / 262-Ice / 698-Fisherman / 896-Largemouth bass / 2836-Walleye / \n", "1-Vehicle / 50-Aircraft / 923-Boeing 777 / \n", "1-Vehicle / 4-Car / 349-Chevrolet / 406-Chevrolet / 509-Classic car / 2446-Used car / \n", "6-Animation / \n", "0-Games / 8-Football / 79-American football / \n", "287-Human swimming / 329-Swimming pool / \n", "0-Games / 2-Video game / 505-Mega Man / 2602-Mega Man X / \n", "18-Outdoor recreation / 59-Winter sport / 177-Skiing / 305-Ski / 789-Alpine skiing / \n", "14-Guitar / 63-Acoustic guitar / \n", "1-Vehicle / \n", "9-Music video / \n", "536-Glass / 640-Bottle / \n", "1-Vehicle / 4-Car / 27-Motorcycle / 105-Motorcycling / 114-Engine / 336-Scooter / 1036-Moped / 1448-Vespa / \n", "130-Gymnastics / 204-Gym / \n", "60-Basketball / \n", "223-Weather / \n", "3-Concert / \n", "546-Stitch / 607-Lace / 1057-Embroidery / 1804-Sewing needle / \n", "0-Games / 2-Video game / 1419-GameSpot / \n", "10-Animal / 48-Pet / 71-Dog / 182-Puppy / 1278-Labrador Retriever / \n", "3-Concert / 7-Musician / \n", "3-Concert / \n", "3-Concert / \n", "6-Animation / 15-Cartoon / 97-Drawing / \n", "0-Games / 2-Video game / 159-World of Warcraft / 171-Warcraft / 849-Tales / \n", "0-Games / 8-Football / \n", "0-Games / 2-Video game / 3296-Transistor / \n", "8-Football / 76-Ball / \n", "23-Mobile phone / 29-Smartphone / 72-iPhone / 101-Telephone / 272-iPod / \n", "1-Vehicle / 4-Car / 1324-Roadster / \n", "56-Hair / 106-Hairstyle / 340-Afro-textured hair / \n", "8-Football / \n", "86-Plant / 136-Gardening / 184-Vegetable / 1209-Bean / \n", "3-Concert / 16-Performance art / \n", "3-Concert / \n", "66-Bollywood / \n", "1-Vehicle / 62-Train / 64-Transport / 103-Rail transport / 119-Locomotive / 143-Railroad car / 152-Track / \n", "28-Fashion / 206-Shoe / 333-Nike; Inc. / 368-Sneakers / 1423-Nike Air Max / 1999-Eggplant / \n", "6-Animation / 15-Cartoon / \n", "6-Animation / \n", "102-Building / 574-Elevator / \n", "3-Concert / 7-Musician / 13-Musical ensemble / 24-String instrument / 36-Piano / 38-Orchestra / 150-Violin / 376-Flute / 450-Viola / \n", "3-Concert / 13-Musical ensemble / 16-Performance art / \n", "61-Art / 97-Drawing / 118-Grand Theft Auto V / 201-Grand Theft Auto: San Andreas / \n", "410-Teacher / 776-Kindergarten / \n", "59-Winter sport / 84-Snow / 177-Skiing / 281-Snowboarding / 305-Ski / \n", "3-Concert / 7-Musician / 13-Musical ensemble / 38-Orchestra / \n", "25-Toy / 89-Comics / 435-Action figure / 2121-Marvel Legends / 3109-Clint Barton / \n", "337-Advertising / \n", "6-Animation / 123-Naruto / \n", "3-Concert / 7-Musician / 13-Musical ensemble / 14-Guitar / 38-Orchestra / 39-Drums / \n", "1578-Red Dead Redemption / 1891-Red Dead / \n", "12-Food / 232-Cake / 295-Chocolate / 1102-Cake decorating / 1189-Torte / 2941-Wedding cake / \n", "6-Animation / 15-Cartoon / \n", "260-Prayer / \n", "5-Dance / 49-School / 1019-Christ / \n", "1-Vehicle / 114-Engine / \n", "33-Weapon / \n", "6-Animation / 15-Cartoon / 1130-Armour / 2398-Personal armor / 3185-Bronze / \n", "18-Outdoor recreation / 22-Nature / \n", "45-Cosmetics / \n", "14-Guitar / 24-String instrument / 63-Acoustic guitar / \n", "13-Musical ensemble / 21-Stadium / 79-American football / 203-Marching band / 209-University / \n", "1-Vehicle / 4-Car / 2396-Coyote / \n", "45-Cosmetics / 402-Jewellery / \n", "83-Skateboarding / 108-Skateboard / 556-Skateboarding trick / \n", "3-Concert / 5-Dance / 16-Performance art / \n", "6-Animation / 15-Cartoon / 294-Batman / \n", "1584-Kettlebell / \n", "0-Games / 8-Football / 65-Kick / 73-Ball / 90-Sports game / 180-FIFA 15 / \n", "125-Television / 223-Weather / \n", "45-Cosmetics / 193-Eye liner / 195-Mascara / 196-Eye / \n", "1058-Volcano / \n", "6-Animation / 15-Cartoon / 61-Art / 97-Drawing / 379-Touhou Project / 408-Sketch / 2162-Marker pen / \n", "3-Concert / \n", "3-Concert / 5-Dance / 7-Musician / 13-Musical ensemble / 30-Drummer / 39-Drums / \n", "3-Concert / 7-Musician / \n", "1254-Cave / \n", "3-Concert / 133-Nightclub / \n", "131-Computer / 141-Microsoft Windows / \n", "6-Animation / 15-Cartoon / 123-Naruto / 183-Manga / \n", "10-Animal / 1330-Bull riding / \n", "1-Vehicle / 11-Motorsport / 17-Racing / 27-Motorcycle / 172-Motocross / 253-Carnival / \n", "18-Outdoor recreation / 59-Winter sport / 84-Snow / 120-Winter / 177-Skiing / 281-Snowboarding / 305-Ski / 789-Alpine skiing / \n", "5-Dance / 31-Disc jockey / \n", "5-Dance / 49-School / \n", "36-Piano / 265-Electronic keyboard / \n", "36-Piano / 53-Keyboard / 107-Musical keyboard / \n", "9-Music video / \n", "5-Dance / \n", "6-Animation / 269-Comic book / 460-Yu-Gi-Oh! Trading Card Game / 494-Card game / \n", "774-Galaxy / \n", "3-Concert / 7-Musician / \n", "0-Games / 2-Video game / 19-PC game / 192-Battlefield / 374-Battlefield 3 / 464-First-person Shooter / \n", "1-Vehicle / 17-Racing / 50-Aircraft / \n", "0-Games / 2-Video game / 6-Animation / 2237-Kantai Collection / \n", "36-Piano / 46-Choir / 53-Keyboard / 104-Pianist / 107-Musical keyboard / \n", "12-Food / \n", "0-Games / 2-Video game / 35-Minecraft / \n", "33-Weapon / 315-Knife / \n", "22-Nature / 86-Plant / 136-Gardening / 279-Garden / 1023-Rose / \n", "36-Piano / \n", "0-Games / 60-Basketball / \n", "9-Music video / \n", "121-Photography / \n", "5-Dance / 78-Wedding / 3693-Groomsman / \n", "59-Winter sport / 84-Snow / 120-Winter / 177-Skiing / 281-Snowboarding / 789-Alpine skiing / 1830-Downhill / \n", "3478-Meme / \n", "0-Games / 2-Video game / 43-Call of Duty / 194-Call of Duty: Modern Warfare 2 / 2074-Killzone / \n", "49-School / \n", "9-Music video / \n", "6-Animation / 61-Art / 981-Sculpture / 1420-Gear / \n", "223-Weather / 419-Earth / 448-Planet / 593-Moon / 654-Sun / 2643-Asteroid / 4478-Sunspot / \n", "8-Football / \n", "0-Games / 2-Video game / 19-PC game / 34-Action-adventure game / 201-Grand Theft Auto: San Andreas / 858-Carl Johnson / \n", "6-Animation / 4129-Crayon / \n", "3-Concert / \n", "2-Video game / 425-City / 2307-SimCity / 3290-SimCity / \n", "6-Animation / 15-Cartoon / \n", "242-Paper / \n", "3-Concert / \n", "110-Album / \n", "23-Mobile phone / 29-Smartphone / 37-Gadget / 72-iPhone / 272-iPod / 331-iPod touch / 554-iPhone 4 / 1155-iPhone 3GS / \n", "1-Vehicle / 4-Car / 11-Motorsport / 17-Racing / 27-Motorcycle / 57-Race track / \n", "128-Cymbal / \n", "1-Vehicle / 27-Motorcycle / 158-Tire / \n", "10-Animal / \n", "0-Games / 8-Football / \n", "10-Animal / 48-Pet / 71-Dog / 182-Puppy / 1084-Bulldog / 1845-Pug / \n", "6-Animation / 31-Disc jockey / \n", "1-Vehicle / 82-Airplane / 92-Radio-controlled model / 149-Model aircraft / 160-Radio-controlled aircraft / 366-Dragon / 471-Wing / \n", "1-Vehicle / 4-Car / 245-BMW / 344-Coupé / \n", "1-Vehicle / 27-Motorcycle / 67-Cycling / 69-Bicycle / 114-Engine / 806-Chopper / \n", "22-Nature / 86-Plant / \n", "2-Video game / 33-Weapon / 96-Soldier / 736-ARMA / 1034-ARMA 2 / \n", "1-Vehicle / 4-Car / 11-Motorsport / 40-Road / 64-Transport / 74-Truck / \n", "222-Surfing / 672-Bodyboarding / \n", "22-Nature / 147-River / 283-Rock / \n", "5-Dance / 16-Performance art / 95-Talent show / \n", "0-Games / 2-Video game / 581-PlayStation / 3161-Tarzan / \n", "7-Musician / 13-Musical ensemble / 44-Drums / 122-Weight training / 203-Marching band / 250-Parade / 820-Bagpipes / 1646-Pipe band / \n", "30-Drummer / 39-Drums / 44-Drums / 128-Cymbal / 146-Snare drum / \n", "121-Photography / 155-Camera / 1855-Video camera / \n", "25-Toy / 89-Comics / 319-Robot / 435-Action figure / \n", "886-James Bond / \n", "1-Vehicle / 11-Motorsport / 17-Racing / 27-Motorcycle / 172-Motocross / \n", "14-Guitar / 4101-Lap steel guitar / \n", "0-Games / 2-Video game / 35-Minecraft / \n", "1-Vehicle / 4-Car / 11-Motorsport / 17-Racing / 41-Sports car / 57-Race track / 135-Drifting / 1301-Nissan Skyline / \n", "9-Music video / 96-Soldier / \n", "0-Games / 23-Mobile phone / 29-Smartphone / 37-Gadget / \n", "37-Gadget / 47-Personal computer / 228-Laptop / 513-Computer keyboard / \n", "3-Concert / 7-Musician / 14-Guitar / \n", "1-Vehicle / \n", "130-Gymnastics / \n", "10-Animal / 540-Insect / 775-Bee / 1296-Nest / 2867-Wasp / \n", "110-Album / \n", "3-Concert / 13-Musical ensemble / \n", "0-Games / 1-Vehicle / 2-Video game / 4-Car / 34-Action-adventure game / 118-Grand Theft Auto V / 330-Money / \n", "10-Animal / 48-Pet / 71-Dog / 182-Puppy / 3046-Maltese / \n", "0-Games / 2-Video game / 241-Fighting game / 1961-Virtua Fighter 5 / 3077-Virtua Fighter / \n", "5-Dance / 16-Performance art / \n", "1-Vehicle / 2635-Water well / \n", "0-Games / 2-Video game / 19-PC game / 33-Weapon / 43-Call of Duty / 111-PlayStation 3 / 127-Xbox 360 / 134-Call of Duty: Black Ops / 194-Call of Duty: Modern Warfare 2 / \n", "10-Animal / 80-Horse / \n", "28-Fashion / 206-Shoe / 333-Nike; Inc. / 368-Sneakers / 4171-Air Force 1 / \n", "6-Animation / 1206-Cinema 4D / \n", "86-Plant / 136-Gardening / 156-Agriculture / 174-Farm / 279-Garden / 1720-Grape / 2473-Vineyard / 4644-Vine / \n", "1-Vehicle / 4-Car / 17-Racing / 41-Sports car / 234-Ford / 352-Need for Speed / 1361-Ford GT / 1600-Need for Speed: World / \n", "0-Games / 2-Video game / 35-Minecraft / 55-Video game console / 127-Xbox 360 / 181-Xbox / 810-Village / \n", "148-Painting / 153-Dress / 303-Paint / 476-Textile / \n", "3-Concert / 7-Musician / 13-Musical ensemble / 46-Choir / \n", "12-Food / 26-Cooking / 32-Recipe / 52-Dish / 58-Cuisine / 215-Eating / 274-Meat / 442-Bread / 2313-Bánh / \n", "6-Animation / 15-Cartoon / 89-Comics / 123-Naruto / 183-Manga / \n", "28-Fashion / 206-Shoe / 333-Nike; Inc. / 368-Sneakers / \n", "3-Concert / \n", "3-Concert / 39-Drums / \n", "0-Games / 1-Vehicle / 2-Video game / 4-Car / 34-Action-adventure game / 40-Road / 118-Grand Theft Auto V / 127-Xbox 360 / 796-Fire engine / \n", "36-Piano / 53-Keyboard / 107-Musical keyboard / 397-Organ / 1666-Hammond organ / \n", "9-Music video / \n", "0-Games / 8-Football / 21-Stadium / 65-Kick / 76-Ball / \n", "9-Music video / \n", "15-Cartoon / 123-Naruto / 285-Sasuke Uchiha / \n", "0-Games / 2-Video game / 237-Final Fantasy / 1658-Dissidia Final Fantasy / \n", "98-Festival / 163-Running / \n", "0-Games / 8-Football / 76-Ball / \n", "75-Wrestling / 85-Combat / \n", "842-Harp / \n", "1-Vehicle / 87-Boat / 498-Motorboat / \n", "2297-Elf / \n", "5-Dance / 16-Performance art / 130-Gymnastics / 350-Circus / 2358-Contortion / \n", "0-Games / 8-Football / 21-Stadium / 65-Kick / 76-Ball / 81-Athlete / \n", "0-Games / 745-Writing / 971-Alphabet / \n", "0-Games / 199-Wii / 960-Badminton / \n", "10-Animal / 1116-Goat / \n", "0-Games / 5-Dance / 60-Basketball / \n", "0-Games / 130-Gymnastics / 1030-Trampoline / 1335-Trampolining / \n", "1-Vehicle / 4-Car / 1237-Baby carriage / 1932-Child safety seat / \n", "465-Microphone / \n", "88-Machine / 261-Tool / 400-Woodturning / 972-Lathe / 2256-Bearing / \n", "20-Trailer / \n", "12-Food / 86-Plant / 136-Gardening / 184-Vegetable / 279-Garden / 671-Tomato / 1022-Chili pepper / 1384-Greenhouse / \n", "117-Boxing / \n", "1-Vehicle / 4-Car / 74-Truck / 158-Tire / \n", "3-Concert / 7-Musician / 9-Music video / 13-Musical ensemble / \n", "6-Animation / 123-Naruto / \n", "1-Vehicle / 4-Car / 11-Motorsport / 17-Racing / 92-Radio-controlled model / 167-Four-wheel drive / 202-Radio-controlled car / 926-Traxxas / \n", "6-Animation / 15-Cartoon / 126-The Walt Disney Company / 1106-Character / 1266-Death Note / \n", "6-Animation / 15-Cartoon / 359-Harry Potter (Literary Series) / 3605-Puppetry / \n", "339-Architecture / 394-Map / 425-City / \n", "7-Musician / 14-Guitar / 24-String instrument / 100-Electric guitar / \n", "324-Finger / 444-Family / 1599-Bone / \n", "282-Skin / 2108-Hair removal / \n", "1-Vehicle / 6-Animation / 449-Honda / 610-Snowmobile / \n", "18-Outdoor recreation / 59-Winter sport / 112-Ice skating / 316-Figure skating / 632-Ice dancing / \n", "78-Wedding / \n", "0-Games / 59-Winter sport / 112-Ice skating / 142-Hockey / \n", "1-Vehicle / 50-Aircraft / 82-Airplane / 129-Aviation / 929-Aerobatics / \n", "0-Games / 256-Tennis / 974-Racket / 1337-Serve / 1626-Forehand / \n", "612-Beer / \n", "9-Music video / 13-Musical ensemble / \n", "49-School / \n", "0-Games / 326-Metin2 / \n", "3-Concert / \n", "0-Games / 8-Football / 73-Ball / \n", "3-Concert / \n", "5-Dance / 2943-KFC / \n", "402-Jewellery / 829-Bead / 859-Beadwork / 907-Necklace / 1001-Gemstone / 1300-Earrings / \n", "1172-VHS / \n", "1000-AdventureQuest Worlds / \n", "1-Vehicle / 64-Transport / 156-Agriculture / 174-Farm / 660-Rural area / 931-Combine Harvester / \n", "85-Combat / 693-Sword / 694-Tai chi / \n", "1-Vehicle / 4-Car / 11-Motorsport / 17-Racing / 270-Sedan / 680-Subaru / 1041-Subaru Impreza / 1286-Subaru / 1897-Subaru Impreza WRX / \n", "93-Comedy / 355-Comedian / \n", "97-Drawing / 148-Painting / 717-Darth Vader / 4331-Darth Maul / \n", "93-Comedy / \n", "1-Vehicle / 4-Car / 114-Engine / 610-Snowmobile / \n", "122-Weight training / 164-Beach / 214-Muscle / \n", "0-Games / 8-Football / 21-Stadium / 65-Kick / 73-Ball / \n", "12-Food / 26-Cooking / 32-Recipe / 52-Dish / 58-Cuisine / 109-Cooking show / 184-Vegetable / 210-Cookware and bakeware / 301-Meal / \n", "1-Vehicle / 2457-Organism / \n", "8-Football / \n", "42-Fishing / 423-Coast / 1008-Kite / 4380-Sport kite / \n", "359-Harry Potter (Literary Series) / \n", "131-Computer / 141-Microsoft Windows / \n", "28-Fashion / 78-Wedding / 190-Bride / \n", "0-Games / 8-Football / \n", "1-Vehicle / 62-Train / 64-Transport / 152-Track / 3015-Monorail / \n", "8-Football / 90-Sports game / 180-FIFA 15 / \n", "3-Concert / \n", "6-Animation / 15-Cartoon / 816-Gundam / \n", "0-Games / 35-Minecraft / 488-Door / \n", "12-Food / 86-Plant / 136-Gardening / 184-Vegetable / 215-Eating / 348-Fruit / 1060-Raw food / \n", "96-Soldier / \n", "3-Concert / 31-Disc jockey / \n", "7-Musician / 46-Choir / 397-Organ / 870-Chapel / \n", "1-Vehicle / 4-Car / 277-Sport utility vehicle / 515-Dodge / \n", "1-Vehicle / 4-Car / 167-Four-wheel drive / 198-Off-road vehicle / 360-Jeep / 1205-Jeep Wrangler / \n", "425-City / \n", "110-Album / \n", "18-Outdoor recreation / 67-Cycling / 69-Bicycle / 221-Skatepark / 377-BMX bike / \n", "12-Food / 26-Cooking / 32-Recipe / 236-Baking / 485-Dough / 560-Sugar / 600-Flour / 1284-Frying / 1576-Doughnut / \n", "0-Games / 2-Video game / 51-Strategy video game / 2182-Mario Party / \n", "56-Hair / \n", "10-Animal / 42-Fishing / 71-Dog / \n", "1318-Flashlight / \n", "56-Hair / 106-Hairstyle / 340-Afro-textured hair / \n", "0-Games / 60-Basketball / \n", "248-Hotel / 369-Resort / \n", "6-Animation / 701-Beyblade / \n", "1-Vehicle / 4-Car / 40-Road / 64-Transport / 74-Truck / 1020-Emergency vehicle / 1168-Neighbourhood / \n", "0-Games / 434-Ninja / 2169-Ninja World / \n", "12-Food / \n", "18-Outdoor recreation / 163-Running / \n", "246-Furniture / 580-Bedroom / 830-Bed / 1281-Couch / \n", "31-Disc jockey / \n", "0-Games / 2-Video game / 34-Action-adventure game / 243-Sonic the Hedgehog / 251-Sonic the Hedgehog / 1512-Sonic Generations / \n", "9-Music video / \n", "4499-Fable II / \n", "4584-Street Legal Racing: Redline / \n", "116-Pokémon / 186-Pokémon / \n", "0-Games / 60-Basketball / 168-Basketball moves / \n", "1-Vehicle / 4-Car / 11-Motorsport / 17-Racing / 41-Sports car / 70-Driving / \n", "56-Hair / \n", "0-Games / 1000-AdventureQuest Worlds / \n", "9-Music video / \n", "2625-Gabrielle / \n", "694-Tai chi / \n", "222-Surfing / \n", "14-Guitar / 1231-Mandolin / \n", "1-Vehicle / 27-Motorcycle / 105-Motorcycling / \n", "18-Outdoor recreation / 83-Skateboarding / 108-Skateboard / \n", "3-Concert / 5-Dance / 13-Musical ensemble / \n", "3-Concert / \n", "22-Nature / 42-Fishing / 275-Diving / 288-Underwater / 381-Underwater diving / 407-Scuba diving / \n", "9-Music video / \n", "6-Animation / 15-Cartoon / 396-One Piece / \n", "0-Games / 2-Video game / 90-Sports game / 180-FIFA 15 / \n", "12-Food / 99-Christmas / \n", "23-Mobile phone / 29-Smartphone / 37-Gadget / 101-Telephone / 178-Samsung Galaxy / \n", "222-Surfing / \n", "3-Concert / \n", "36-Piano / 53-Keyboard / 107-Musical keyboard / \n", "1-Vehicle / 22-Nature / 87-Boat / 92-Radio-controlled model / 225-Lake / \n", "3-Concert / 16-Performance art / \n", "56-Hair / 106-Hairstyle / 583-Wig / \n", "102-Building / 113-House / 170-Home improvement / 276-Room / 339-Architecture / 416-Apartment / \n", "8-Football / 21-Stadium / 77-Arena / \n", "0-Games / 1576-Doughnut / 3277-The Simpsons: Tapped Out / \n", "3-Concert / 16-Performance art / \n", "0-Games / 2-Video game / 311-Street Fighter / 461-Street Fighter IV / 496-Super Street Fighter IV / 563-Ryu / 2468-M. Bison / 2526-Zangief / \n", "23-Mobile phone / 29-Smartphone / 37-Gadget / 553-Sony Xperia / 4635-Sony Ericsson Xperia X10 Mini / \n", "10-Animal / 48-Pet / 71-Dog / 182-Puppy / \n", "3-Concert / 7-Musician / 13-Musical ensemble / 30-Drummer / 39-Drums / 1108-Mushroom / \n", "78-Wedding / \n", "1-Vehicle / 27-Motorcycle / 105-Motorcycling / 114-Engine / 154-Wheel / 158-Tire / 207-Exhaust system / 806-Chopper / 2927-Shovel / \n", "3-Concert / 7-Musician / 13-Musical ensemble / 38-Orchestra / \n", "163-Running / 206-Shoe / 368-Sneakers / \n", "2-Video game / 380-Assassin's Creed / 1048-Assassin's Creed / \n", "5-Dance / \n", "66-Bollywood / \n", "1-Vehicle / 4-Car / 517-Audi / 531-Hatchback / 945-MINI Cooper / 1221-MINI / 4496-Citroën DS3 / \n", "1-Vehicle / 4-Car / \n", "0-Games / 2-Video game / 111-PlayStation 3 / 216-Call of Duty: Modern Warfare 3 / \n", "10-Animal / 18-Outdoor recreation / 80-Horse / 247-Stallion / 289-Dressage / \n", "3-Concert / \n", "10-Animal / 12-Food / 132-Bird / \n", "1-Vehicle / 2-Video game / 4-Car / 11-Motorsport / 17-Racing / 27-Motorcycle / 2712-Just Cause / \n", "0-Games / 372-Playing card / 791-Magic: The Gathering / \n", "7-Musician / 13-Musical ensemble / 46-Choir / \n", "6-Animation / 15-Cartoon / 123-Naruto / 183-Manga / 285-Sasuke Uchiha / 622-Sakura Haruno / \n", "93-Comedy / 205-Newscaster / \n", "7-Musician / 14-Guitar / 36-Piano / 44-Drums / \n", "10-Animal / 86-Plant / 91-Fish / 258-Aquarium / 2029-Filter / \n", "83-Skateboarding / 108-Skateboard / 221-Skatepark / 336-Scooter / 595-Champion / \n", "3-Concert / \n", "1964-TV4 / \n", "0-Games / 8-Football / \n", "20-Trailer / \n", "0-Games / 1-Vehicle / 74-Truck / 124-Tractor / 174-Farm / 715-Farming Simulator / \n", "467-Chipmunk / \n", "8-Football / 79-American football / \n", "15-Cartoon / 183-Manga / 396-One Piece / \n", "0-Games / 199-Wii / \n", "3-Concert / 7-Musician / 13-Musical ensemble / 30-Drummer / 39-Drums / \n", "0-Games / 2-Video game / 445-Guitar Hero / 724-Rock Band / \n", "1-Vehicle / 4-Car / 299-Volkswagen Passenger Cars / 827-Pump / \n", "7-Musician / 9-Music video / 14-Guitar / \n", "776-Kindergarten / \n", "0-Games / 8-Football / 54-Highlight film / \n", "0-Games / 489-Board game / 2833-Go / 4510-Shogi / \n", "0-Games / 2-Video game / 55-Video game console / 320-Super Smash Bros. / 692-Link / \n", "783-Zee TV / \n", "3-Concert / 5-Dance / 16-Performance art / \n", "14-Guitar / 842-Harp / 3172-Lyre / \n", "259-Fire / \n", "7-Musician / 14-Guitar / 24-String instrument / 63-Acoustic guitar / \n", "2324-Hang / \n", "9-Music video / \n", "1-Vehicle / 62-Train / 64-Transport / 84-Snow / 103-Rail transport / 120-Winter / \n", "23-Mobile phone / 29-Smartphone / 72-iPhone / 189-iPad / 272-iPod / 331-iPod touch / \n", "61-Art / 83-Skateboarding / 108-Skateboard / \n", "0-Games / 2-Video game / 264-Star Wars / \n", "1-Vehicle / 4-Car / 154-Wheel / 158-Tire / 493-Rim / \n", "3-Concert / \n", "3-Concert / 7-Musician / 389-Restaurant / \n", "0-Games / 2-Video game / 6-Animation / 506-Garry's Mod / \n", "6-Animation / 15-Cartoon / 89-Comics / \n", "0-Games / 463-Cricket / 1495-Twenty20 / \n", "1-Vehicle / 4-Car / 25-Toy / 84-Snow / 92-Radio-controlled model / 202-Radio-controlled car / 449-Honda / 3086-Honda CR-V / \n", "6-Animation / 4527-South Park: The Stick of Truth / \n", "12-Food / 26-Cooking / 32-Recipe / 58-Cuisine / \n", "1-Vehicle / 92-Radio-controlled model / 113-House / 149-Model aircraft / 155-Camera / 160-Radio-controlled aircraft / 393-Unmanned aerial vehicle / 604-Quadcopter / \n", "6-Animation / 15-Cartoon / 185-Dragon Ball / 244-Goku / 418-Gohan / \n", "1-Vehicle / 62-Train / 119-Locomotive / 143-Railroad car / 428-Rail transport modelling / \n", "6-Animation / 15-Cartoon / 89-Comics / 3595-Shaman King / \n", "12-Food / 86-Plant / 865-Turtle / 1500-Cherry blossom / \n", "6-Animation / 213-Forest / 1555-Sword Art Online / \n", "0-Games / 2-Video game / 192-Battlefield / 374-Battlefield 3 / 383-Battlefield 4 / \n", "1-Vehicle / 4-Car / 87-Boat / 498-Motorboat / 2338-Inflatable boat / \n", "3-Concert / 30-Drummer / \n", "0-Games / 2-Video game / 19-PC game / \n", "3162-Postage stamp / \n", "1-Vehicle / 4-Car / 114-Engine / 509-Classic car / 1301-Nissan Skyline / \n", "1-Vehicle / 62-Train / 64-Transport / 161-Train station / \n", "23-Mobile phone / 101-Telephone / \n", "47-Personal computer / 131-Computer / 141-Microsoft Windows / \n", "66-Bollywood / \n", "14-Guitar / \n", "3-Concert / 7-Musician / 98-Festival / \n", "12-Food / 26-Cooking / 32-Recipe / 52-Dish / 58-Cuisine / 274-Meat / 542-Potato / 968-Pie / 1108-Mushroom / 2650-Frying pan / \n", "405-Braid / 1249-Rainbow Loom / \n", "0-Games / 49-School / 81-Athlete / 562-Pitcher / \n", "22-Nature / 164-Beach / \n", "2-Video game / 127-Xbox 360 / \n", "20-Trailer / 42-Fishing / 222-Surfing / 917-Sail / \n", "10-Animal / 61-Art / 97-Drawing / 148-Painting / 936-Watercolor paint / 1063-Ink / 2630-Doodle / \n", "0-Games / 2-Video game / 19-PC game / 20-Trailer / 359-Harry Potter (Literary Series) / 4489-Harry Potter and the Philosopher's Stone / \n", "3-Concert / \n", "5-Dance / \n", "28-Fashion / 206-Shoe / 4151-Flip-flops / \n", "72-iPhone / 83-Skateboarding / 108-Skateboard / 221-Skatepark / \n", "6-Animation / 15-Cartoon / 61-Art / 97-Drawing / 183-Manga / 408-Sketch / \n", "14-Guitar / 63-Acoustic guitar / \n", "497-Macintosh / 3974-Mac Mini / \n", "5-Dance / \n", "1-Vehicle / 4-Car / 20-Trailer / \n", "0-Games / 2-Video game / 55-Video game console / 317-PlayStation / \n", "1-Vehicle / 87-Boat / 308-Ship / 579-Yacht / 1957-Luxury yacht / \n", "1-Vehicle / 4-Car / 11-Motorsport / \n", "0-Games / 8-Football / 21-Stadium / \n", "12-Food / 52-Dish / 58-Cuisine / 184-Vegetable / 301-Meal / 391-Chicken meat / 1569-Steamed rice / 3049-Convenience store / \n", "121-Photography / \n", "31-Disc jockey / \n", "122-Weight training / 214-Muscle / \n", "0-Games / 2-Video game / 55-Video game console / 111-PlayStation 3 / 127-Xbox 360 / 181-Xbox / 317-PlayStation / 386-Xbox / 1482-PlayStation Network / \n", "130-Gymnastics / \n", "0-Games / 2-Video game / 55-Video game console / 199-Wii / 774-Galaxy / 1397-Super Mario Galaxy / 1759-Super Mario Galaxy 2 / \n", "49-School / \n", "1-Vehicle / 4-Car / 114-Engine / 3399-Timing belt / 3939-Nissan Maxima / \n", "9-Music video / \n", "339-Architecture / \n", "925-Brain / \n", "61-Art / 97-Drawing / 408-Sketch / 479-Graffiti / 538-Illustration / \n", "0-Games / 2-Video game / 20-Trailer / 34-Action-adventure game / 111-PlayStation 3 / 176-PlayStation 4 / 656-Metal Gear / 2212-Metal Gear Solid V: The Phantom Pain / \n", "14-Guitar / 24-String instrument / 63-Acoustic guitar / \n", "0-Games / 2-Video game / 33-Weapon / 43-Call of Duty / 302-Call of Duty 4: Modern Warfare / \n", "8-Football / 358-Rugby football / \n", "122-Weight training / 529-Squat / 532-Barbell / \n", "7-Musician / 13-Musical ensemble / 46-Choir / \n", "7-Musician / 14-Guitar / 24-String instrument / 100-Electric guitar / \n", "9-Music video / 14-Guitar / \n", "66-Bollywood / \n", "137-Tree / 1198-Birth / \n", "1-Vehicle / 9-Music video / 62-Train / 428-Rail transport modelling / \n", "2642-CNBC / \n", "5-Dance / \n", "3-Concert / 5-Dance / \n", "9-Music video / 99-Christmas / 558-Santa Claus / 964-Christmas tree / \n", "3-Concert / 16-Performance art / \n", "3-Concert / \n", "7-Musician / 14-Guitar / 15-Cartoon / 24-String instrument / 63-Acoustic guitar / \n", "5-Dance / 31-Disc jockey / 133-Nightclub / \n", "263-Light / 1375-Laser lighting display / \n", "20-Trailer / 264-Star Wars / \n", "3-Concert / 7-Musician / 13-Musical ensemble / 38-Orchestra / 46-Choir / 376-Flute / \n", "23-Mobile phone / 29-Smartphone / 37-Gadget / 101-Telephone / \n", "9-Music video / \n", "0-Games / 2-Video game / 19-PC game / 34-Action-adventure game / \n", "0-Games / 2-Video game / 19-PC game / 33-Weapon / 43-Call of Duty / 96-Soldier / 422-Call of Duty: Ghosts / 4485-Warhawk / \n", "6-Animation / 15-Cartoon / \n", "7-Musician / 219-Accordion / \n", "3-Concert / 68-Lighting / 263-Light / \n", "35-Minecraft / \n", "59-Winter sport / 81-Athlete / 112-Ice skating / 142-Hockey / 312-Ice rink / \n", "49-School / \n", "0-Games / 2-Video game / 6-Animation / 1878-SD Gundam Capsule Fighter / \n", "0-Games / 2-Video game / 6-Animation / 15-Cartoon / 1668-Claw crane / \n", "3-Concert / 7-Musician / \n", "2-Video game / 429-Website / \n", "0-Games / 326-Metin2 / \n", "20-Trailer / 3226-Chucky / \n", "45-Cosmetics / \n", "1-Vehicle / 4-Car / 11-Motorsport / 17-Racing / 40-Road / 41-Sports car / 57-Race track / 70-Driving / 212-Highway / 231-Supercar / 711-Gran Turismo / 953-Gran Turismo 5 / 1521-Jaguar Cars / \n", "37-Gadget / 47-Personal computer / 145-Tablet computer / 189-iPad / 1464-iPad Mini / \n", "0-Games / 8-Football / 1247-Penalty kick / \n", "0-Games / 83-Skateboarding / 221-Skatepark / 336-Scooter / \n", "0-Games / 2-Video game / 47-Personal computer / 1028-Devil May Cry / 2344-DmC: Devil May Cry / \n", "18-Outdoor recreation / 33-Weapon / \n", "3-Concert / 7-Musician / 13-Musical ensemble / 16-Performance art / 30-Drummer / \n", "117-Boxing / \n", "3-Concert / 13-Musical ensemble / 30-Drummer / 133-Nightclub / \n", "0-Games / 2-Video game / 19-PC game / 51-Strategy video game / 382-Dota 2 / 549-Defense of the Ancients / \n", "0-Games / 59-Winter sport / 77-Arena / 112-Ice skating / 142-Hockey / 312-Ice rink / \n", "0-Games / 530-Cue sports / 568-Pool / 994-Cue stick / \n", "9-Music video / 3124-Kendama / \n", "1-Vehicle / 11-Motorsport / 27-Motorcycle / 84-Snow / 105-Motorcycling / 120-Winter / 262-Ice / 2070-Suzuki Hayabusa / \n", "49-School / 410-Teacher / \n", "13-Musical ensemble / \n", "5-Dance / 49-School / \n", "702-Counter-Strike: Source / \n", "9-Music video / 1221-MINI / \n", "0-Games / 2-Video game / 47-Personal computer / 118-Grand Theft Auto V / 464-First-person Shooter / \n", "10-Animal / 80-Horse / \n", "10-Animal / 208-Wood / 540-Insect / 775-Bee / 1296-Nest / \n", "0-Games / 49-School / 237-Final Fantasy / \n", "3-Concert / 68-Lighting / \n", "3-Concert / 7-Musician / 13-Musical ensemble / 38-Orchestra / 46-Choir / \n", "0-Games / 710-Handball / \n", "110-Album / \n", "0-Games / 2-Video game / 60-Basketball / 77-Arena / 90-Sports game / 168-Basketball moves / 255-Slam dunk / 1262-NBA 2K14 / 2156-Rookie / \n", "10-Animal / 48-Pet / 166-Cat / 469-Bag / 1142-Laundry / \n", "10-Animal / 48-Pet / 166-Cat / 346-Kitten / \n", "66-Bollywood / \n", "1-Vehicle / 74-Truck / 114-Engine / \n", "3-Concert / 5-Dance / 95-Talent show / 191-Cheerleading / \n", "5-Dance / 16-Performance art / 133-Nightclub / \n", "267-Mountain / \n", "5-Dance / \n", "616-The Bible / \n", "1-Vehicle / 4-Car / 261-Tool / 4066-Wrench / \n", "0-Games / 2-Video game / 116-Pokémon / 186-Pokémon / 1204-Pokémon Ruby and Sapphire / 1665-Pokémon Emerald / 2236-Emerald / \n", "512-Printing / 696-T-shirt / 1757-Screen printing / \n", "1-Vehicle / 4-Car / 40-Road / \n", "36-Piano / 53-Keyboard / 104-Pianist / 107-Musical keyboard / \n", "14-Guitar / \n", "1-Vehicle / 18-Outdoor recreation / 22-Nature / 67-Cycling / 69-Bicycle / 197-Trail / 213-Forest / 220-Mountain bike / 278-GoPro / 284-Mountain biking / \n", "2-Video game / 47-Personal computer / 1045-Computer case / 4515-Antec / \n", "12-Food / 26-Cooking / 32-Recipe / 226-Dessert / 232-Cake / 295-Chocolate / 1176-Pumpkin / 1707-Microwave oven / 2768-Mug / \n", "6-Animation / 15-Cartoon / 183-Manga / 185-Dragon Ball / 244-Goku / \n", "810-Village / \n", "78-Wedding / \n", "0-Games / 2-Video game / 11-Motorsport / 2815-F-Zero / \n", "0-Games / 2-Video game / 37-Gadget / 55-Video game console / 90-Sports game / 292-Handheld game console / 317-PlayStation / 480-FIFA 13 / 682-PlayStation Vita / \n", "1-Vehicle / 11-Motorsport / 17-Racing / 18-Outdoor recreation / 27-Motorcycle / 172-Motocross / \n", "0-Games / 67-Cycling / 69-Bicycle / 83-Skateboarding / 222-Surfing / 669-Extreme sport / 2014-Vans / \n", "25-Toy / 89-Comics / 185-Dragon Ball / 244-Goku / 435-Action figure / 1986-Piccolo / \n", "1-Vehicle / 4-Car / 11-Motorsport / 17-Racing / 40-Road / 41-Sports car / 57-Race track / 349-Chevrolet / 378-Drag racing / \n", "0-Games / 1-Vehicle / 2-Video game / 4-Car / 11-Motorsport / 17-Racing / 41-Sports car / 57-Race track / 70-Driving / \n", "1-Vehicle / 62-Train / 64-Transport / 103-Rail transport / 119-Locomotive / 143-Railroad car / 161-Train station / 224-Rapid transit / \n", "0-Games / 2-Video game / 19-PC game / 1566-Doom / \n", "3-Concert / 5-Dance / 133-Nightclub / 613-Bar / \n", "1-Vehicle / 4-Car / 40-Road / 120-Winter / 2016-Volvo Cars / \n", "0-Games / 2851-Fight Night / \n", "31-Disc jockey / 133-Nightclub / \n", "18-Outdoor recreation / 217-Hunting / 4035-Pheasant / \n", "1-Vehicle / 4-Car / 9-Music video / 70-Driving / \n", "5-Dance / \n", "10-Animal / 25-Toy / 1152-Plush / 1212-Stuffed toy / 1497-Wool / \n", "1-Vehicle / 62-Train / 110-Album / \n", "0-Games / 5-Dance / 49-School / 191-Cheerleading / \n", "0-Games / 2-Video game / 6-Animation / 15-Cartoon / 877-Hatsune Miku: Project DIVA / 1648-Hatsune Miku: Project DIVA F / 2682-Hatsune Miku: Project DIVA F 2nd / \n", "9-Music video / 14-Guitar / 15-Cartoon / 24-String instrument / \n", "66-Bollywood / \n", "766-President of the United States / \n", "0-Games / 2-Video game / 2431-Metal Slug / \n", "14-Guitar / 24-String instrument / 63-Acoustic guitar / \n", "6-Animation / 15-Cartoon / 243-Sonic the Hedgehog / 251-Sonic the Hedgehog / \n", "1-Vehicle / 3-Concert / 11-Motorsport / 27-Motorcycle / 67-Cycling / 69-Bicycle / 350-Circus / \n", "12-Food / 26-Cooking / 32-Recipe / 179-Kitchen / 184-Vegetable / 1439-Refrigerator / 2614-Lettuce / \n", "3-Concert / 5-Dance / 16-Performance art / \n", "30-Drummer / 39-Drums / 44-Drums / 128-Cymbal / 146-Snare drum / \n", "3-Concert / \n", "93-Comedy / \n", "5-Dance / 16-Performance art / \n", "5-Dance / \n", "1-Vehicle / 11-Motorsport / \n", "12-Food / 391-Chicken meat / 944-Hamburger / 2943-KFC / \n", "75-Wrestling / \n", "5-Dance / \n", "0-Games / 8-Football / 21-Stadium / \n", "0-Games / 2-Video game / 34-Action-adventure game / 1028-Devil May Cry / 2336-Devil May Cry 3: Dante's Awakening / \n", "0-Games / 2-Video game / 19-PC game / 34-Action-adventure game / 43-Call of Duty / 118-Grand Theft Auto V / \n", "12-Food / 26-Cooking / 32-Recipe / 52-Dish / 58-Cuisine / \n", "1-Vehicle / 11-Motorsport / 135-Drifting / 1887-Tricycle / \n", "1-Vehicle / 84-Snow / 120-Winter / 223-Weather / 1145-Winter storm / \n", "318-Church / \n", "121-Photography / 155-Camera / 566-Digital camera / 629-Camera lens / 721-Digital SLR / 3328-Canon EOS 7D / \n", "10-Animal / 12-Food / 274-Meat / 763-Snake / \n", "3-Concert / 5-Dance / \n", "3-Concert / 7-Musician / 14-Guitar / 30-Drummer / 39-Drums / 44-Drums / \n", "110-Album / \n", "0-Games / 1-Vehicle / 2-Video game / 4-Car / 34-Action-adventure game / 297-Grand Theft Auto IV / 658-Grand Theft Auto: The Lost and Damned / \n", "626-Puzzle / \n", "0-Games / 2-Video game / 23-Mobile phone / 29-Smartphone / 37-Gadget / \n", "0-Games / 1-Vehicle / 2-Video game / 124-Tractor / 127-Xbox 360 / 174-Farm / 314-Xbox One / 715-Farming Simulator / \n", "79-American football / \n", "1163-Carpet / 2547-Staples Center / \n", "2-Video game / 55-Video game console / 127-Xbox 360 / 181-Xbox / \n", "0-Games / 2-Video game / 43-Call of Duty / 111-PlayStation 3 / 127-Xbox 360 / 194-Call of Duty: Modern Warfare 2 / 302-Call of Duty 4: Modern Warfare / \n", "121-Photography / \n", "9-Music video / 549-Defense of the Ancients / \n", "12-Food / 26-Cooking / 32-Recipe / 58-Cuisine / 274-Meat / 620-Barbecue / 688-Beef / 1115-Steak / \n", "36-Piano / 110-Album / \n", "0-Games / 2-Video game / 37-Gadget / 55-Video game console / 292-Handheld game console / 317-PlayStation / 362-PlayStation Portable / 682-PlayStation Vita / 2260-Monster Hunter Portable 3rd / \n", "0-Games / 942-Silkroad Online / \n", "9-Music video / \n", "0-Games / 2-Video game / 19-PC game / 1861-Smite / 4525-Hi-Rez Studios / \n", "12-Food / 26-Cooking / 32-Recipe / 52-Dish / 58-Cuisine / 184-Vegetable / 236-Baking / 475-Cheese / 485-Dough / 662-Pizza / 671-Tomato / 752-Pasta / 1230-Italian food / 1381-Garlic / 3364-Basil / \n", "0-Games / 1-Vehicle / 2-Video game / 4-Car / 19-PC game / 34-Action-adventure game / 118-Grand Theft Auto V / \n", "10-Animal / 42-Fishing / 91-Fish / 147-River / 225-Lake / 268-Recreational fishing / 896-Largemouth bass / \n", "14-Guitar / 63-Acoustic guitar / \n", "1-Vehicle / 27-Motorcycle / 206-Shoe / 757-Boot / \n", "12-Food / 26-Cooking / 184-Vegetable / 348-Fruit / \n", "110-Album / 311-Street Fighter / 461-Street Fighter IV / 2239-Street Fighter II: The World Warrior / 2468-M. Bison / 4579-Street Fighter Alpha 3 / \n", "364-Sitcom / \n", "1-Vehicle / 11-Motorsport / 17-Racing / 27-Motorcycle / 172-Motocross / \n", "3-Concert / \n", "223-Weather / 259-Fire / 365-Fireworks / \n", "20-Trailer / 1710-Pirates of the Caribbean / \n", "257-Counter-Strike / \n", "273-News program / \n", "7-Musician / 14-Guitar / 24-String instrument / 63-Acoustic guitar / \n", "23-Mobile phone / 29-Smartphone / 452-Watch / 627-Clock / 1260-Smartwatch / 3727-Android Wear / \n", "3-Concert / 5-Dance / 31-Disc jockey / 68-Lighting / 133-Nightclub / \n", "98-Festival / \n", "15-Cartoon / 1696-Raven / \n", "6-Animation / \n", "0-Games / 779-Point Blank / \n", "1-Vehicle / 11-Motorsport / 17-Racing / 27-Motorcycle / 172-Motocross / \n", "1621-Ork / \n", "14-Guitar / \n", "31-Disc jockey / \n", "9-Music video / 28-Fashion / \n", "10-Animal / 18-Outdoor recreation / 22-Nature / 42-Fishing / 91-Fish / 147-River / 225-Lake / 268-Recreational fishing / 478-Fishing rod / \n", "3-Concert / 7-Musician / 14-Guitar / 39-Drums / \n", "6-Animation / 20-Trailer / \n", "23-Mobile phone / 29-Smartphone / 72-iPhone / \n", "938-Steel / \n", "3-Concert / 5-Dance / 16-Performance art / \n", "5-Dance / \n", "1-Vehicle / 4-Car / 74-Truck / 137-Tree / 1093-Loader / 1611-Garbage truck / \n", "8-Football / \n", "0-Games / 8-Football / 21-Stadium / 163-Running / \n", "22-Nature / 1531-Sunrise / \n", "25-Toy / \n", "9-Music video / \n", "0-Games / 2-Video game / 199-Wii / \n", "372-Playing card / 746-Card manipulation / \n", "3-Concert / \n", "23-Mobile phone / 29-Smartphone / 37-Gadget / 101-Telephone / \n", "0-Games / 2-Video game / 1007-Portal / 1274-Portal 2 / \n", "205-Newscaster / 223-Weather / \n", "1-Vehicle / 4-Car / 11-Motorsport / 17-Racing / 41-Sports car / 57-Race track / 135-Drifting / \n", "45-Cosmetics / 56-Hair / 193-Eye liner / 195-Mascara / 196-Eye / 218-Eyelash / 282-Skin / 291-Rouge / 385-Face / 436-Concealer / \n", "1-Vehicle / 88-Machine / 249-Heavy equipment / 817-Bulldozer / 822-Crane / \n", "93-Comedy / 355-Comedian / \n", "3-Concert / 5-Dance / 16-Performance art / \n", "12-Food / \n", "83-Skateboarding / 108-Skateboard / 351-Jumping / \n", "0-Games / 2-Video game / 34-Action-adventure game / 362-PlayStation Portable / 1125-God of War / 1432-Kratos / \n", "9-Music video / \n", "3-Concert / \n", "2-Video game / 319-Robot / \n", "1975-Amiga / \n", "1899-E-book / \n", "3442-Résumé / \n", "1-Vehicle / 25-Toy / 62-Train / 1279-Plarail / 1685-Toy train / \n", "3-Concert / 5-Dance / 16-Performance art / 95-Talent show / \n", "3-Concert / 7-Musician / 13-Musical ensemble / 14-Guitar / 24-String instrument / 30-Drummer / 44-Drums / \n", "0-Games / 2-Video game / 79-American football / 90-Sports game / 561-Madden NFL / \n", "1-Vehicle / 12-Food / 666-Concrete / \n", "18-Outdoor recreation / 59-Winter sport / 112-Ice skating / 703-Roller skating / 1937-Inline skates / 2747-Inline skating / \n", "30-Drummer / 39-Drums / 44-Drums / 128-Cymbal / 146-Snare drum / 2073-Sabian / \n", "3-Concert / 16-Performance art / 68-Lighting / \n", "12-Food / \n", "20-Trailer / \n", "0-Games / \n", "5-Dance / \n", "23-Mobile phone / 37-Gadget / 47-Personal computer / 145-Tablet computer / 189-iPad / 1459-Magazine / \n", "7-Musician / 13-Musical ensemble / 38-Orchestra / 250-Parade / 371-Firefighter / \n", "23-Mobile phone / 29-Smartphone / 37-Gadget / 72-iPhone / 272-iPod / \n", "5-Dance / 31-Disc jockey / \n", "14-Guitar / 36-Piano / 53-Keyboard / \n", "23-Mobile phone / 29-Smartphone / 37-Gadget / 72-iPhone / 101-Telephone / 155-Camera / 178-Samsung Galaxy / 899-Samsung Galaxy S III / 3648-Samsung Galaxy S III Mini / \n", "0-Games / 2-Video game / 116-Pokémon / 186-Pokémon / 1204-Pokémon Ruby and Sapphire / \n", "1-Vehicle / 4-Car / 11-Motorsport / 650-Convertible / 1455-Cadillac / \n", "9-Music video / \n", "5-Dance / 16-Performance art / \n", "60-Basketball / \n", "36-Piano / 53-Keyboard / 104-Pianist / 107-Musical keyboard / \n", "47-Personal computer / 145-Tablet computer / 228-Laptop / 2581-Microsoft Surface / \n", "468-Biology / 925-Brain / \n", "12-Food / 26-Cooking / 662-Pizza / \n", "1-Vehicle / 67-Cycling / 69-Bicycle / 520-Bicycle frame / 645-Road bicycle / \n", "1-Vehicle / 266-Helicopter / 3602-Mini-Z / \n", "3-Concert / 7-Musician / \n", "0-Games / 2-Video game / 1178-Quest / 1196-Dragon Age / 1735-Dragon Age: Inquisition / 2700-Dragon Age II / \n", "98-Festival / 253-Carnival / \n", "9-Music video / \n", "0-Games / 2-Video game / 34-Action-adventure game / 201-Grand Theft Auto: San Andreas / 1069-Easter egg / \n", "3-Concert / 68-Lighting / \n", "0-Games / 2-Video game / 831-Plants vs. Zombies / \n", "1-Vehicle / 4-Car / 11-Motorsport / 17-Racing / 25-Toy / 40-Road / 57-Race track / 92-Radio-controlled model / 202-Radio-controlled car / 270-Sedan / \n", "4463-Santana Lopez / \n", "0-Games / 88-Machine / 535-Slot machine / \n", "7-Musician / 14-Guitar / 24-String instrument / 63-Acoustic guitar / \n", "0-Games / 756-Tibia / \n", "23-Mobile phone / 29-Smartphone / 47-Personal computer / 72-iPhone / 145-Tablet computer / 189-iPad / 272-iPod / 921-Grand Theft Auto: Vice City / \n", "28-Fashion / 45-Cosmetics / 56-Hair / 583-Wig / 607-Lace / \n", "0-Games / 8-Football / 54-Highlight film / 79-American football / 813-Running back / \n", "6-Animation / 15-Cartoon / 123-Naruto / 801-Inuyasha / \n", "0-Games / 535-Slot machine / 3153-Freddy Krueger / \n", "1535-Prize / \n", "61-Art / 110-Album / \n", "28-Fashion / 45-Cosmetics / 56-Hair / \n", "1-Vehicle / 4-Car / 11-Motorsport / 57-Race track / \n", "12-Food / 26-Cooking / 52-Dish / 58-Cuisine / 215-Eating / 1218-Japanese Cuisine / \n", "0-Games / 2-Video game / 541-Black-and-white / 1446-Game Boy / \n", "222-Surfing / 815-Surfboard / \n", "1-Vehicle / 12-Food / 124-Tractor / 156-Agriculture / 174-Farm / 249-Heavy equipment / 617-Maize / 1473-Forage harvester / \n", "13-Musical ensemble / \n", "3-Concert / 7-Musician / 30-Drummer / 39-Drums / \n", "1-Vehicle / 4-Car / 11-Motorsport / 41-Sports car / 344-Coupé / 618-Fiat Automobiles / \n", "0-Games / 159-World of Warcraft / 171-Warcraft / \n", "3-Concert / 7-Musician / 13-Musical ensemble / 14-Guitar / 49-School / \n", "12-Food / 26-Cooking / 32-Recipe / 52-Dish / 58-Cuisine / \n", "0-Games / 2-Video game / 23-Mobile phone / 29-Smartphone / 809-Logo / \n", "155-Camera / \n", "2-Video game / 34-Action-adventure game / 1396-Lara Croft / \n", "30-Drummer / 39-Drums / 44-Drums / 128-Cymbal / 146-Snare drum / \n", "0-Games / 2-Video game / 35-Minecraft / \n", "1-Vehicle / 4-Car / 11-Motorsport / 41-Sports car / 344-Coupé / 2052-Infiniti / \n", "61-Art / \n", "1-Vehicle / 4-Car / 154-Wheel / 158-Tire / 493-Rim / \n", "0-Games / 2-Video game / 1631-PlanetSide 2 / 2026-PlanetSide / \n", "85-Combat / 117-Boxing / 699-Kick / \n", "8-Football / 595-Champion / \n", "5-Dance / \n", "5-Dance / 209-University / \n", "14-Guitar / 24-String instrument / \n", "7-Musician / 13-Musical ensemble / 46-Choir / \n", "0-Games / 2-Video game / 851-Digimon / \n", "3-Concert / 7-Musician / 13-Musical ensemble / \n", "66-Bollywood / \n", "0-Games / 211-RuneScape / 635-Achievement / \n", "59-Winter sport / 112-Ice skating / 262-Ice / 316-Figure skating / 833-Ice skate / \n", "3-Concert / 7-Musician / \n", "0-Games / 2-Video game / 35-Minecraft / 365-Fireworks / \n", "1-Vehicle / 4-Car / 11-Motorsport / 17-Racing / 41-Sports car / 57-Race track / 378-Drag racing / \n", "98-Festival / \n", "322-Mixtape / \n", "8-Football / \n", "12-Food / 26-Cooking / 32-Recipe / 274-Meat / \n", "22-Nature / 2493-The Hobbit / \n", "5-Dance / 115-Ballet / 776-Kindergarten / \n", "25-Toy / 140-LEGO / 264-Star Wars / 852-Lego Star Wars / \n", "5-Dance / 4336-Angkor Wat / \n", "0-Games / 2-Video game / 19-PC game / 43-Call of Duty / 315-Knife / 422-Call of Duty: Ghosts / 4485-Warhawk / \n", "85-Combat / 1832-Hand-to-hand combat / \n", "3-Concert / 7-Musician / \n", "6-Animation / 15-Cartoon / \n", "9-Music video / 125-Television / \n", "410-Teacher / 745-Writing / \n", "67-Cycling / 69-Bicycle / 220-Mountain bike / 284-Mountain biking / 351-Jumping / \n", "12-Food / 367-Drink / 612-Beer / 3126-Ale / \n", "3-Concert / 7-Musician / 30-Drummer / \n", "3-Concert / 7-Musician / \n", "28-Fashion / 45-Cosmetics / \n", "0-Games / 2-Video game / 20-Trailer / 413-PlayStation 2 / 4226-Neo / \n", "6-Animation / 15-Cartoon / 1732-Oni / \n", "0-Games / 33-Weapon / 3786-Killing Floor / \n", "557-Salad / \n", "1-Vehicle / 18-Outdoor recreation / 22-Nature / 67-Cycling / 69-Bicycle / 197-Trail / 220-Mountain bike / 284-Mountain biking / 3665-Giant Bicycles / \n", "9-Music video / \n", "14-Guitar / 63-Acoustic guitar / \n", "3-Concert / 5-Dance / \n", "2-Video game / 25-Toy / 229-Halo / \n", "0-Games / 8-Football / \n", "325-Vampire / 725-Human / \n", "0-Games / 162-Medicine / 468-Biology / \n", "3-Concert / \n", "1-Vehicle / 50-Aircraft / 82-Airplane / 139-Airport / \n", "28-Fashion / 56-Hair / 106-Hairstyle / 405-Braid / \n", "398-The Legend of Zelda / \n", "7-Musician / 36-Piano / 150-Violin / 376-Flute / \n", "2183-Stardoll / \n", "12-Food / 26-Cooking / 32-Recipe / 179-Kitchen / 367-Drink / 985-Strawberry / 1174-Smoothie / \n", "125-Television / \n", "20-Trailer / 1177-Spider / \n", "925-Brain / \n", "9-Music video / \n", "1-Vehicle / 575-Traffic / \n", "6-Animation / 15-Cartoon / 342-Windows Media Video / \n", "3-Concert / 7-Musician / 14-Guitar / \n", "14-Guitar / 24-String instrument / 63-Acoustic guitar / \n", "0-Games / 2-Video game / 257-Counter-Strike / 1109-Counter-Strike Online / \n", "0-Games / 51-Strategy video game / 373-Clash of Clans / 3067-Boom Beach / \n", "736-ARMA / 1034-ARMA 2 / \n", "66-Bollywood / \n", "7-Musician / 13-Musical ensemble / 38-Orchestra / \n", "22-Nature / 559-Night / 739-Guild Wars / 839-Guild Wars 2 / \n", "282-Skin / 385-Face / \n", "28-Fashion / 45-Cosmetics / 56-Hair / 106-Hairstyle / \n", "3242-Seeley Booth / \n", "8-Football / 12-Food / 58-Cuisine / 944-Hamburger / \n", "10-Animal / 22-Nature / 86-Plant / 136-Gardening / 540-Insect / 642-Leaf / \n", "0-Games / 8-Football / \n", "1-Vehicle / 4-Car / 11-Motorsport / 680-Subaru / 1286-Subaru / \n", "1-Vehicle / 4-Car / 118-Grand Theft Auto V / 201-Grand Theft Auto: San Andreas / \n", "0-Games / 8-Football / 54-Highlight film / \n", "83-Skateboarding / 108-Skateboard / 154-Wheel / \n", "2653-Wedding ring / \n", "6-Animation / 15-Cartoon / 1281-Couch / \n", "0-Games / 35-Minecraft / 775-Bee / \n", "0-Games / 1401-The Witcher 3: Wild Hunt / \n", "0-Games / 51-Strategy video game / 165-League of Legends / \n", "0-Games / 2-Video game / 19-PC game / 118-Grand Theft Auto V / 192-Battlefield / 383-Battlefield 4 / \n", "0-Games / 278-GoPro / 463-Cricket / \n", "110-Album / \n", "1-Vehicle / 4-Car / 88-Machine / 309-Computer hardware / 602-Central processing unit / \n", "1-Vehicle / 4-Car / 70-Driving / 94-Dashcam / 212-Highway / \n", "0-Games / 2-Video game / 34-Action-adventure game / 111-PlayStation 3 / 118-Grand Theft Auto V / \n", "1-Vehicle / 4-Car / 114-Engine / 154-Wheel / 234-Ford / 918-Brake / 2899-Ford Explorer / \n", "1-Vehicle / 50-Aircraft / 82-Airplane / 92-Radio-controlled model / 149-Model aircraft / 160-Radio-controlled aircraft / \n", "1-Vehicle / 4-Car / 11-Motorsport / 1221-MINI / 1418-Model car / \n", "5-Dance / 44-Drums / 98-Festival / 250-Parade / \n", "0-Games / 2-Video game / 89-Comics / 737-Joker / 1392-Injustice: Gods Among Us / \n", "29-Smartphone / 1366-LG Optimus series / \n", "12-Food / 26-Cooking / 32-Recipe / 52-Dish / 784-Honey / 1022-Chili pepper / \n", "1-Vehicle / 87-Boat / 92-Radio-controlled model / 342-Windows Media Video / 3311-Hovercraft / \n", "121-Photography / 155-Camera / 721-Digital SLR / \n", "6-Animation / 248-Hotel / 369-Resort / 468-Biology / \n", "3-Concert / 7-Musician / 24-String instrument / 842-Harp / \n", "5-Dance / 16-Performance art / 115-Ballet / 191-Cheerleading / \n", "6-Animation / \n", "593-Moon / 654-Sun / \n", "6-Animation / 15-Cartoon / \n", "1-Vehicle / 4-Car / 40-Road / 70-Driving / 94-Dashcam / 212-Highway / 2020-Toyota 86 / \n", "0-Games / 51-Strategy video game / 373-Clash of Clans / \n", "22-Nature / 164-Beach / \n", "512-Printing / \n", "6-Animation / 97-Drawing / 183-Manga / 269-Comic book / 396-One Piece / \n", "1-Vehicle / 88-Machine / 249-Heavy equipment / 283-Rock / 976-Demolition / 1995-Hammer / \n", "0-Games / 772-Bowling / 930-Ten-pin bowling / \n", "3-Concert / 68-Lighting / \n", "6-Animation / 10-Animal / 15-Cartoon / 80-Horse / 290-Pony / 343-My Little Pony / \n", "9-Music video / \n", "1-Vehicle / 4-Car / 1074-Hyundai / 4199-Hyundai Sonata / \n", "1-Vehicle / 27-Motorcycle / 154-Wheel / 158-Tire / 918-Brake / 4170-Honda Shadow / \n", "3-Concert / \n", "3-Concert / \n", "14-Guitar / 49-School / 63-Acoustic guitar / \n", "66-Bollywood / \n", "1507-Calligraphy / \n", "3676-Dance Central 3 / \n", "0-Games / 1-Vehicle / 2-Video game / 4-Car / 34-Action-adventure game / 201-Grand Theft Auto: San Andreas / 858-Carl Johnson / \n", "6-Animation / 97-Drawing / \n", "0-Games / 8-Football / 54-Highlight film / 79-American football / \n", "6-Animation / 15-Cartoon / 20-Trailer / 185-Dragon Ball / \n", "2582-Split / \n", "1-Vehicle / 4-Car / 11-Motorsport / 17-Racing / 728-Dirt track racing / \n", "144-Amusement park / 248-Hotel / 369-Resort / \n", "2001-Harry Potter and the Deathly Hallows / \n", "6-Animation / 15-Cartoon / \n", "10-Animal / 22-Nature / 132-Bird / 280-Wildlife / \n", "12-Food / 26-Cooking / 32-Recipe / 52-Dish / 58-Cuisine / 109-Cooking show / 184-Vegetable / 210-Cookware and bakeware / 239-Roasting / 1295-Stir frying / 2959-Wok / \n", "3-Concert / 68-Lighting / \n", "0-Games / 2-Video game / 1895-Werewolf / 2054-Diablo II / \n", "6-Animation / \n", "5-Dance / 59-Winter sport / 112-Ice skating / 126-The Walt Disney Company / 262-Ice / 316-Figure skating / \n", "46-Choir / \n", "110-Album / \n", "8-Football / \n", "0-Games / 2-Video game / 19-PC game / 788-Juggling / 4130-Rumble Fighter / \n", "1-Vehicle / 4-Car / 64-Transport / 371-Firefighter / 796-Fire engine / 1020-Emergency vehicle / \n", "1-Vehicle / 4-Car / 245-BMW / 4393-BMW 5 Series (F10) / \n", "23-Mobile phone / 29-Smartphone / 37-Gadget / 47-Personal computer / 72-iPhone / 145-Tablet computer / 189-iPad / 1150-iPad 2 / 1504-iPad 3 / \n", "23-Mobile phone / 29-Smartphone / \n", "4359-Anno / \n", "3-Concert / 7-Musician / 13-Musical ensemble / 16-Performance art / \n", "0-Games / 51-Strategy video game / 165-League of Legends / \n", "8-Football / \n", "1-Vehicle / 50-Aircraft / 82-Airplane / 151-Landing / 157-Jet aircraft / 388-Fighter aircraft / \n", "18-Outdoor recreation / 144-Amusement park / 321-Roller coaster / 334-Amusement ride / \n", "3-Concert / 7-Musician / 13-Musical ensemble / 2375-Ney / \n", "1-Vehicle / 23-Mobile phone / \n", "45-Cosmetics / 173-Eye shadow / 1534-H&M / \n", "1-Vehicle / 4-Car / 64-Transport / 286-Bus / \n", "3-Concert / 427-Trumpet / \n", "45-Cosmetics / 276-Room / 402-Jewellery / \n", "5-Dance / 296-Saxophone / \n", "5-Dance / 250-Parade / \n", "365-Fireworks / \n", "3-Concert / \n", "31-Disc jockey / \n", "5-Dance / \n", "1-Vehicle / 17-Racing / 18-Outdoor recreation / 67-Cycling / 69-Bicycle / 163-Running / 220-Mountain bike / 284-Mountain biking / 570-Road bicycle racing / 645-Road bicycle / 674-Marathon / \n", "18-Outdoor recreation / 22-Nature / 42-Fishing / 85-Combat / 137-Tree / 147-River / 351-Jumping / 1449-Rubber band / 1921-Bungee jumping / \n", "5-Dance / \n", "22-Nature / \n", "3-Concert / 7-Musician / 30-Drummer / \n", "5-Dance / 16-Performance art / 115-Ballet / 153-Dress / 1253-Ballet Dancer / \n", "0-Games / 2-Video game / 19-PC game / 241-Fighting game / 764-The King of Fighters / 2354-The King of Fighters 2002 / \n", "5-Dance / 78-Wedding / \n", "3-Concert / 68-Lighting / \n", "1-Vehicle / 4-Car / 11-Motorsport / 17-Racing / \n", "44-Drums / \n", "943-Stove / \n", "66-Bollywood / \n", "246-Furniture / 920-Chair / 2669-Leather crafting / 3329-Upholstery / \n", "1862-Corel / \n", "126-The Walt Disney Company / \n", "0-Games / 8-Football / 54-Highlight film / 79-American football / \n", "121-Photography / \n", "3-Concert / 7-Musician / \n", "6-Animation / 20-Trailer / \n", "14-Guitar / 63-Acoustic guitar / \n", "3-Concert / \n", "3-Concert / 7-Musician / 24-String instrument / \n", "0-Games / 2-Video game / 55-Video game console / 581-PlayStation / \n", "89-Comics / \n", "3-Concert / 7-Musician / 13-Musical ensemble / 16-Performance art / 38-Orchestra / 98-Festival / \n", "1-Vehicle / 12-Food / 124-Tractor / 156-Agriculture / 174-Farm / 249-Heavy equipment / 931-Combine Harvester / \n", "3-Concert / 133-Nightclub / 150-Violin / 988-Ibiza / \n", "1-Vehicle / 84-Snow / 120-Winter / 155-Camera / 522-Helmet / 1323-Helmet camera / \n", "260-Prayer / \n", "7-Musician / 24-String instrument / 150-Violin / \n", "0-Games / 2-Video game / 90-Sports game / 180-FIFA 15 / 561-Madden NFL / \n", "9-Music video / \n", "6-Animation / 15-Cartoon / \n", "110-Album / \n", "0-Games / 2-Video game / 19-PC game / 1861-Smite / \n", "3-Concert / 7-Musician / 36-Piano / 107-Musical keyboard / \n", "7-Musician / 14-Guitar / 24-String instrument / 361-Flamenco / 2380-Flamenco guitar / \n", "10-Animal / 537-Zoo / 1198-Birth / \n", "3-Concert / 16-Performance art / \n", "1-Vehicle / 4-Car / 41-Sports car / 231-Supercar / 569-Lamborghini / 1418-Model car / 2225-Lamborghini Murciélago / \n", "3-Concert / 7-Musician / \n", "9-Music video / \n", "1-Vehicle / 87-Boat / \n", "23-Mobile phone / 29-Smartphone / 72-iPhone / 516-Headphones / 832-Headset / \n", "96-Soldier / 99-Christmas / 466-Gift / 1107-United States Navy / \n", "283-Rock / 1001-Gemstone / \n", "1-Vehicle / 4-Car / 88-Machine / 114-Engine / \n", "1-Vehicle / 4-Car / 17-Racing / 57-Race track / 3602-Mini-Z / \n", "726-Clarinet / \n", "1-Vehicle / 4-Car / \n", "23-Mobile phone / 616-The Bible / \n", "56-Hair / 61-Art / 457-Knitting / 653-Thread / 1536-Woven fabric / \n", "10-Animal / 80-Horse / 247-Stallion / 289-Dressage / \n", "1078-Installation art / 1678-Foam / \n", "49-School / 476-Textile / 1479-Quilt / \n", "10-Animal / 80-Horse / 247-Stallion / \n", "1-Vehicle / 4-Car / 552-Honda Civic / \n", "1-Vehicle / 4-Car / 40-Road / 94-Dashcam / 234-Ford / 473-Ford Mustang / 1285-Parking / \n", "1-Vehicle / 4-Car / 41-Sports car / \n", "2-Video game / 192-Battlefield / 374-Battlefield 3 / 2948-Dogfight / 3450-Battlefield 1942 / \n", "3-Concert / \n", "45-Cosmetics / 173-Eye shadow / 193-Eye liner / 195-Mascara / 196-Eye / 218-Eyelash / 291-Rouge / \n", "155-Camera / 278-GoPro / \n", "0-Games / 2-Video game / 19-PC game / 43-Call of Duty / 134-Call of Duty: Black Ops / 138-Call of Duty: Black Ops II / 431-Call of Duty: Advanced Warfare / 1248-Call of Duty: Zombies / \n", "827-Pump / \n", "23-Mobile phone / 29-Smartphone / 37-Gadget / 345-Loudspeaker / 1221-MINI / 1565-Logitech / 3285-Boombox / \n", "3-Concert / 16-Performance art / 46-Choir / \n", "85-Combat / 117-Boxing / 399-Kickboxing / 1239-K-1 / \n", "59-Winter sport / 84-Snow / 120-Winter / 177-Skiing / \n", "18-Outdoor recreation / 163-Running / \n", "3-Concert / 5-Dance / 13-Musical ensemble / 16-Performance art / 38-Orchestra / \n", "3-Concert / 7-Musician / 36-Piano / 53-Keyboard / 104-Pianist / \n", "9-Music video / \n", "4171-Air Force 1 / \n", "467-Chipmunk / 846-Squirrel / \n", "13-Musical ensemble / 820-Bagpipes / 1646-Pipe band / \n", "3522-ProSieben / \n", "0-Games / 2-Video game / 43-Call of Duty / 431-Call of Duty: Advanced Warfare / \n", "3-Concert / \n", "1-Vehicle / 4-Car / 64-Transport / 74-Truck / 250-Parade / 371-Firefighter / 796-Fire engine / 1020-Emergency vehicle / \n", "0-Games / 2-Video game / 77-Arena / 211-RuneScape / \n", "0-Games / 2-Video game / 35-Minecraft / \n", "296-Saxophone / 1006-Tenor saxophone / \n", "0-Games / 2-Video game / 33-Weapon / 441-The Elder Scrolls / 3418-The Elder Scrolls III: Morrowind / \n", "3-Concert / 7-Musician / 13-Musical ensemble / 14-Guitar / 30-Drummer / 39-Drums / 44-Drums / \n", "339-Architecture / \n", "0-Games / 2-Video game / 6-Animation / 15-Cartoon / 723-Pig / 1426-Treasure / \n", "0-Games / 2-Video game / 229-Halo / 644-Halo: Reach / \n", "0-Games / 2-Video game / 34-Action-adventure game / 201-Grand Theft Auto: San Andreas / \n", "45-Cosmetics / 235-Lipstick / 1012-Lip gloss / \n", "414-Computer monitor / \n", "1-Vehicle / 4-Car / 277-Sport utility vehicle / 1806-GMC / 2683-GMC / \n", "6-Animation / \n", "20-Trailer / 2096-The Hobbit / 2493-The Hobbit / 3792-Dwarf / \n", "341-Sketch comedy / \n", "10-Animal / 48-Pet / 71-Dog / 1252-Dog agility / \n", "5-Dance / \n", "0-Games / 8-Football / 1338-Devil / \n", "0-Games / 8-Football / 21-Stadium / 65-Kick / 76-Ball / \n", "0-Games / 2-Video game / 19-PC game / 43-Call of Duty / 118-Grand Theft Auto V / 422-Call of Duty: Ghosts / \n", "0-Games / 35-Minecraft / 1151-Temple / \n", "3-Concert / \n", "28-Fashion / 757-Boot / 1481-Sunglasses / 2326-Royal Air Force / \n", "9-Music video / \n", "66-Bollywood / \n", "23-Mobile phone / 29-Smartphone / 37-Gadget / 101-Telephone / \n", "320-Super Smash Bros. / 335-Nintendo 3DS / 443-Wii U / 697-Super Smash Bros. for Nintendo 3DS and Wii U / 3050-Mewtwo / \n", "20-Trailer / 294-Batman / \n", "18-Outdoor recreation / 59-Winter sport / 84-Snow / 120-Winter / 177-Skiing / 305-Ski / \n", "20-Trailer / \n", "0-Games / 75-Wrestling / \n", "283-Rock / 353-Wall / \n", "12-Food / 26-Cooking / 184-Vegetable / 215-Eating / 944-Hamburger / \n", "3-Concert / \n", "6-Animation / 15-Cartoon / 243-Sonic the Hedgehog / 251-Sonic the Hedgehog / \n", "358-Rugby football / \n", "5-Dance / 16-Performance art / 95-Talent show / \n", "1-Vehicle / 4-Car / 11-Motorsport / 17-Racing / 57-Race track / \n", "3-Concert / 38-Orchestra / \n", "3-Concert / 7-Musician / \n", "1-Vehicle / 87-Boat / 225-Lake / 498-Motorboat / \n", "12-Food / 32-Recipe / 226-Dessert / 232-Cake / 236-Baking / 295-Chocolate / 578-Cookie / 1322-Biscuit / \n", "47-Personal computer / 131-Computer / 309-Computer hardware / 957-Desktop computer / 1045-Computer case / \n", "33-Weapon / 315-Knife / \n", "5-Dance / 16-Performance art / 115-Ballet / \n", "35-Minecraft / \n", "1-Vehicle / 83-Skateboarding / 108-Skateboard / \n", "9-Music video / \n", "131-Computer / 189-iPad / \n", "3-Concert / 5-Dance / 49-School / 130-Gymnastics / 191-Cheerleading / \n", "1-Vehicle / 250-Parade / \n", "9-Music video / 594-Music festival / \n", "25-Toy / 140-LEGO / \n", "9-Music video / \n", "0-Games / 8-Football / \n", "3-Concert / 13-Musical ensemble / 49-School / \n", "0-Games / 8-Football / 21-Stadium / 65-Kick / 73-Ball / \n", "155-Camera / \n", "0-Games / 8-Football / 21-Stadium / 65-Kick / 73-Ball / 76-Ball / 81-Athlete / 90-Sports game / \n", "5-Dance / 49-School / \n", "737-Joker / \n", "8-Football / 21-Stadium / 60-Basketball / 77-Arena / \n", "1-Vehicle / 62-Train / 103-Rail transport / \n", "3-Concert / 7-Musician / 13-Musical ensemble / 46-Choir / \n", "1-Vehicle / 62-Train / 74-Truck / 2655-Vehicle horn / \n", "75-Wrestling / 977-WWE '13 / \n", "67-Cycling / 69-Bicycle / 86-Plant / 136-Gardening / 520-Bicycle frame / \n", "110-Album / \n", "3-Concert / \n", "9-Music video / \n", "5-Dance / 16-Performance art / \n", "0-Games / 2-Video game / 19-PC game / 33-Weapon / 43-Call of Duty / 96-Soldier / 134-Call of Duty: Black Ops / 216-Call of Duty: Modern Warfare 3 / 302-Call of Duty 4: Modern Warfare / \n", "6-Animation / 15-Cartoon / \n", "0-Games / 2-Video game / 19-PC game / 1412-Half-Life / \n", "2383-White House / \n", "1-Vehicle / 4-Car / 11-Motorsport / \n", "23-Mobile phone / 29-Smartphone / 37-Gadget / 72-iPhone / 101-Telephone / 527-Battery / 1159-Battery charger / 2163-Rechargeable battery / \n", "5-Dance / 361-Flamenco / \n", "14-Guitar / 63-Acoustic guitar / \n", "1-Vehicle / 4-Car / 212-Highway / 1415-Kia Motors / \n", "10-Animal / 48-Pet / 91-Fish / 258-Aquarium / \n", "14-Guitar / 39-Drums / 44-Drums / 521-Metal / \n", "3-Concert / 68-Lighting / 98-Festival / 3285-Boombox / \n", "0-Games / 2-Video game / 6-Animation / 849-Tales / 3415-Tales of Vesperia / \n", "3808-Goggles / \n", "12-Food / 827-Pump / 1616-Electrical wiring / \n", "5-Dance / \n", "12-Food / 26-Cooking / 32-Recipe / 52-Dish / 58-Cuisine / 109-Cooking show / 184-Vegetable / \n", "0-Games / 159-World of Warcraft / 171-Warcraft / \n", "45-Cosmetics / 436-Concealer / \n", "3-Concert / \n", "3-Concert / \n", "9-Music video / \n", "1-Vehicle / 4-Car / 11-Motorsport / 17-Racing / 40-Road / 57-Race track / 378-Drag racing / \n", "1-Vehicle / 18-Outdoor recreation / 42-Fishing / 87-Boat / \n", "28-Fashion / 45-Cosmetics / 56-Hair / 106-Hairstyle / 195-Mascara / 218-Eyelash / 235-Lipstick / 291-Rouge / \n", "7-Musician / 36-Piano / 53-Keyboard / 104-Pianist / \n", "3-Concert / \n", "5-Dance / 16-Performance art / \n", "0-Games / 2-Video game / 1091-Saints Row / 1695-Saints Row: The Third / \n", "398-The Legend of Zelda / 541-Black-and-white / 1228-The Legend of Zelda: Ocarina of Time / \n", "6-Animation / 15-Cartoon / 4276-Toaster / \n", "2532-Pan flute / \n", "0-Games / 2-Video game / 19-PC game / 51-Strategy video game / 382-Dota 2 / 549-Defense of the Ancients / 1378-Multiplayer online battle arena / \n", "3-Concert / 7-Musician / 13-Musical ensemble / 30-Drummer / 230-Brass instrument / \n", "5-Dance / \n", "810-Village / \n", "0-Games / 211-RuneScape / \n", "98-Festival / 102-Building / 2341-Fine art / \n", "6-Animation / 15-Cartoon / \n", "1-Vehicle / 4-Car / 154-Wheel / 234-Ford / 918-Brake / 2256-Bearing / 3684-Ford Bronco / \n", "12-Food / 3117-Match / \n", "5-Dance / \n", "309-Computer hardware / \n", "0-Games / 35-Minecraft / \n", "14-Guitar / 63-Acoustic guitar / \n", "6-Animation / \n", "56-Hair / 99-Christmas / 106-Hairstyle / 405-Braid / \n", "2799-Ys / \n", "46-Choir / \n", "25-Toy / \n", "6-Animation / 34-Action-adventure game / 264-Star Wars / \n", "1880-Marinera / \n", "1249-Rainbow Loom / \n", "3-Concert / 7-Musician / 13-Musical ensemble / 46-Choir / 318-Church / \n", "0-Games / 211-RuneScape / 1178-Quest / \n", "3-Concert / 5-Dance / 153-Dress / \n", "1-Vehicle / 74-Truck / \n", "8-Football / \n", "12-Food / 26-Cooking / 32-Recipe / 179-Kitchen / \n", "14-Guitar / 24-String instrument / \n", "18-Outdoor recreation / 163-Running / 674-Marathon / \n", "0-Games / 2-Video game / 33-Weapon / \n", "0-Games / 2-Video game / 19-PC game / 33-Weapon / 43-Call of Duty / 96-Soldier / 216-Call of Duty: Modern Warfare 3 / \n", "3-Concert / 7-Musician / 30-Drummer / 39-Drums / 44-Drums / \n", "5-Dance / 16-Performance art / \n", "1-Vehicle / 11-Motorsport / 27-Motorcycle / 608-Wheelie / 927-Supermoto / \n", "0-Games / 772-Bowling / 930-Ten-pin bowling / \n", "0-Games / \n", "12-Food / 612-Beer / 720-Wine / \n", "6-Animation / 547-Bleach / \n", "1-Vehicle / 4-Car / 901-Mitsubishi / 1461-Electric car / 2304-Electric vehicle / \n", "144-Amusement park / 248-Hotel / 369-Resort / \n", "3-Concert / 5-Dance / 16-Performance art / 95-Talent show / \n", "6-Animation / 15-Cartoon / 89-Comics / 183-Manga / 185-Dragon Ball / 244-Goku / \n", "0-Games / 8-Football / 54-Highlight film / 79-American football / \n", "0-Games / 59-Winter sport / 112-Ice skating / 142-Hockey / 312-Ice rink / \n", "0-Games / 523-Cattle / 626-Puzzle / \n", "61-Art / 587-Clay / 970-Monkey / \n", "5-Dance / 44-Drums / \n", "0-Games / 8-Football / 21-Stadium / 77-Arena / \n", "86-Plant / \n", "3-Concert / 16-Performance art / 46-Choir / \n", "1-Vehicle / 4-Car / 207-Exhaust system / 299-Volkswagen Passenger Cars / 1761-Volkswagen Jetta / 3155-Volkswagen Golf Mk4 / \n", "25-Toy / 626-Puzzle / 631-Cube / 740-Rubik's Cube / \n", "0-Games / 19-PC game / 159-World of Warcraft / 171-Warcraft / \n", "6-Animation / 9-Music video / 97-Drawing / 408-Sketch / \n", "0-Games / 8-Football / 21-Stadium / 54-Highlight film / 65-Kick / 77-Arena / \n", "3-Concert / 7-Musician / 30-Drummer / \n", "0-Games / 8-Football / 110-Album / \n", "7-Musician / 13-Musical ensemble / 38-Orchestra / \n", "5-Dance / 36-Piano / 53-Keyboard / 763-Snake / \n", "0-Games / 1-Vehicle / 50-Aircraft / 92-Radio-controlled model / 149-Model aircraft / 160-Radio-controlled aircraft / 266-Helicopter / 392-Radio-controlled helicopter / \n", "6-Animation / \n", "3-Concert / 7-Musician / 13-Musical ensemble / 38-Orchestra / 230-Brass instrument / \n", "3-Concert / 46-Choir / 49-School / 776-Kindergarten / \n", "3-Concert / 7-Musician / 30-Drummer / \n", "31-Disc jockey / \n", "3-Concert / 7-Musician / 16-Performance art / 98-Festival / 594-Music festival / \n", "5-Dance / \n", "0-Games / 2-Video game / 1505-Ultimate Marvel vs. Capcom 3 / \n", "14-Guitar / 63-Acoustic guitar / \n", "22-Nature / \n", "28-Fashion / 153-Dress / 390-Model / 495-Runway / \n", "1-Vehicle / 4-Car / 41-Sports car / 231-Supercar / 569-Lamborghini / \n", "25-Toy / \n", "208-Wood / 981-Sculpture / 1346-Wood carving / 1480-Astronaut / \n", "222-Surfing / 233-Ocean / \n", "0-Games / 2-Video game / 51-Strategy video game / 356-Star / 373-Clash of Clans / \n", "56-Hair / 123-Naruto / 583-Wig / \n", "3-Concert / 16-Performance art / \n", "8-Football / 9-Music video / \n", "12-Food / 26-Cooking / 32-Recipe / 58-Cuisine / 3288-Mustard / \n", "5-Dance / \n", "12-Food / 786-Juice / 1081-Lemon / 3598-Lime / \n", "12-Food / 32-Recipe / 2588-Mayonnaise / \n", "122-Weight training / 204-Gym / 214-Muscle / \n", "15-Cartoon / 185-Dragon Ball / 244-Goku / 418-Gohan / \n", "0-Games / 2-Video game / 6-Animation / 514-Tekken / 797-Tekken / 1470-Tekken Tag Tournament 2 / \n", "30-Drummer / 39-Drums / 44-Drums / 1953-Roland V-Drums / \n", "9-Music video / 66-Bollywood / \n", "2-Video game / \n", "25-Toy / 75-Wrestling / 164-Beach / 435-Action figure / \n", "6-Animation / 439-Barbie / \n", "31-Disc jockey / \n", "3-Concert / 7-Musician / 16-Performance art / 30-Drummer / \n", "7-Musician / 14-Guitar / 24-String instrument / 100-Electric guitar / \n", "12-Food / 617-Maize / \n", "46-Choir / 99-Christmas / \n", "1-Vehicle / 4-Car / 154-Wheel / \n", "5-Dance / \n", "0-Games / 2-Video game / 1681-Mu Online / \n", "45-Cosmetics / 466-Gift / \n", "0-Games / 2-Video game / 19-PC game / 33-Weapon / 96-Soldier / 257-Counter-Strike / \n", "8-Football / 358-Rugby football / \n", "130-Gymnastics / 1417-Balance beam / 2745-Beam / \n", "1621-Ork / \n", "3-Concert / 7-Musician / 13-Musical ensemble / 14-Guitar / 24-String instrument / 30-Drummer / 39-Drums / \n", "0-Games / 379-Touhou Project / 3849-Perfect Cherry Blossom / \n", "6-Animation / 3588-Berserk / \n", "1-Vehicle / 283-Rock / 1409-Missile / \n", "1-Vehicle / 4-Car / 11-Motorsport / 41-Sports car / 70-Driving / 207-Exhaust system / 270-Sedan / 344-Coupé / 349-Chevrolet / 406-Chevrolet / 947-Muffler / 1752-Chevrolet SS / \n", "6-Animation / 123-Naruto / \n", "10-Animal / 18-Outdoor recreation / 80-Horse / 247-Stallion / 438-Mare / 3383-Filly / \n", "163-Running / \n", "89-Comics / \n", "12-Food / \n", "66-Bollywood / \n", "3-Concert / 68-Lighting / 77-Arena / \n", "12-Food / 26-Cooking / 32-Recipe / 52-Dish / 58-Cuisine / 274-Meat / \n", "12-Food / \n", "6-Animation / 185-Dragon Ball / 244-Goku / \n", "12-Food / 424-Chicken / 597-Soup / \n", "1-Vehicle / 4-Car / 1194-Automotive lighting / \n", "6-Animation / 15-Cartoon / 61-Art / 251-Sonic the Hedgehog / \n", "9-Music video / 3401-Aggro Berlin / \n", "23-Mobile phone / 29-Smartphone / 101-Telephone / \n", "0-Games / 2-Video game / 19-PC game / 33-Weapon / 43-Call of Duty / 96-Soldier / 194-Call of Duty: Modern Warfare 2 / \n", "5-Dance / 31-Disc jockey / \n", "36-Piano / 53-Keyboard / 104-Pianist / 107-Musical keyboard / 237-Final Fantasy / \n", "0-Games / 2-Video game / 34-Action-adventure game / 35-Minecraft / 1007-Portal / \n", "61-Art / 238-Nail / 300-Nail art / 338-Manicure / \n", "1-Vehicle / 4-Car / 1134-Taxicab / \n", "9-Music video / \n", "5-Dance / 66-Bollywood / \n", "3-Concert / 68-Lighting / \n", "1-Vehicle / 4-Car / 11-Motorsport / 41-Sports car / 70-Driving / 207-Exhaust system / 231-Supercar / \n", "7-Musician / 13-Musical ensemble / 98-Festival / 203-Marching band / \n", "18-Outdoor recreation / 137-Tree / 148-Painting / 303-Paint / 2124-Spray painting / 2853-Aerosol paint / 3188-Camouflage / \n", "0-Games / 1991-Rappelz / \n", "1-Vehicle / 2-Video game / 4-Car / 1333-Wangan Midnight / \n", "18-Outdoor recreation / 83-Skateboarding / 108-Skateboard / 556-Skateboarding trick / \n", "3-Concert / 31-Disc jockey / 133-Nightclub / \n", "12-Food / 184-Vegetable / 348-Fruit / 671-Tomato / 1193-Orange / 2676-Citrus / 3207-Peel / 3293-Vegetable carving / \n", "23-Mobile phone / 29-Smartphone / \n", "5-Dance / 6-Animation / 16-Performance art / 126-The Walt Disney Company / 576-Walt Disney World / 1139-Belle / 1265-Kyle Kingson / \n", "0-Games / 372-Playing card / 791-Magic: The Gathering / 1784-Goblin / \n", "9-Music video / \n", "1338-Devil / \n", "1-Vehicle / 4-Car / 70-Driving / 94-Dashcam / \n", "0-Games / 2-Video game / 453-Pro Evolution Soccer / 3526-Pro Evolution Soccer 2009 / \n", "20-Trailer / 66-Bollywood / \n", "5-Dance / 59-Winter sport / 112-Ice skating / \n", "7-Musician / 14-Guitar / 24-String instrument / 63-Acoustic guitar / 1936-Bouzouki / \n", "3-Concert / 7-Musician / \n", "0-Games / 2-Video game / 8-Football / 54-Highlight film / 90-Sports game / 111-PlayStation 3 / 453-Pro Evolution Soccer / \n", "31-Disc jockey / \n", "61-Art / 148-Painting / \n", "1-Vehicle / 4-Car / 40-Road / 94-Dashcam / 2850-Chauffeur / \n", "28-Fashion / 56-Hair / 106-Hairstyle / 340-Afro-textured hair / 919-Hair conditioner / \n", "56-Hair / \n", "0-Games / 8-Football / 90-Sports game / 453-Pro Evolution Soccer / \n", "0-Games / 2-Video game / 19-PC game / 33-Weapon / 43-Call of Duty / 216-Call of Duty: Modern Warfare 3 / \n", "3214-Black Widow / \n", "0-Games / 710-Handball / \n", "7-Musician / 14-Guitar / 24-String instrument / 63-Acoustic guitar / \n", "0-Games / 10-Animal / 51-Strategy video game / 550-Dinosaur / 1718-Spore / \n", "0-Games / 75-Wrestling / \n", "14-Guitar / 63-Acoustic guitar / \n", "14-Guitar / 63-Acoustic guitar / \n", "1-Vehicle / 4-Car / 70-Driving / 1321-Škoda Auto / 2690-Škoda Octavia / \n", "0-Games / 8-Football / 79-American football / \n", "5-Dance / \n", "31-Disc jockey / \n", "1-Vehicle / 4-Car / 11-Motorsport / 86-Plant / 137-Tree / 167-Four-wheel drive / 198-Off-road vehicle / 360-Jeep / 565-Mud bogging / \n", "18-Outdoor recreation / 83-Skateboarding / 108-Skateboard / 221-Skatepark / \n", "31-Disc jockey / \n", "1-Vehicle / 22-Nature / 50-Aircraft / 92-Radio-controlled model / 149-Model aircraft / 160-Radio-controlled aircraft / 266-Helicopter / 392-Radio-controlled helicopter / \n", "9-Music video / 14-Guitar / 24-String instrument / 63-Acoustic guitar / 4484-Nikon D7000 / \n", "1-Vehicle / 67-Cycling / 69-Bicycle / 220-Mountain bike / 3611-Folding bicycle / \n", "45-Cosmetics / \n", "37-Gadget / 832-Headset / \n", "10-Animal / 42-Fishing / 91-Fish / 275-Diving / 288-Underwater / 381-Underwater diving / \n", "0-Games / 2-Video game / 159-World of Warcraft / 171-Warcraft / \n", "3-Concert / 7-Musician / 38-Orchestra / 219-Accordion / \n", "12-Food / 612-Beer / \n", "3-Concert / 7-Musician / 13-Musical ensemble / \n", "1-Vehicle / 4-Car / 11-Motorsport / 17-Racing / 27-Motorcycle / 57-Race track / 853-Dune buggy / \n", "0-Games / 60-Basketball / 77-Arena / 168-Basketball moves / \n", "208-Wood / 303-Paint / 400-Woodturning / 462-Table / 673-Saw / 2548-Table saw / \n", "7-Musician / 13-Musical ensemble / 38-Orchestra / 150-Violin / 230-Brass instrument / 1574-French horn / \n", "3-Concert / 7-Musician / 13-Musical ensemble / 24-String instrument / 670-Banjo / \n", "0-Games / 2-Video game / 19-PC game / 43-Call of Duty / 96-Soldier / 194-Call of Duty: Modern Warfare 2 / \n", "141-Microsoft Windows / \n", "1-Vehicle / 6-Animation / 50-Aircraft / 82-Airplane / 129-Aviation / \n", "49-School / 99-Christmas / \n", "60-Basketball / 168-Basketball moves / \n", "5-Dance / 889-Swing / \n", "1-Vehicle / 62-Train / 64-Transport / 103-Rail transport / 119-Locomotive / 143-Railroad car / 161-Train station / \n", "6-Animation / 15-Cartoon / 477-Transformers / 2186-Megatron / 3398-Transformers: Generation 1 / \n", "10-Animal / 48-Pet / 166-Cat / 346-Kitten / \n", "0-Games / 35-Minecraft / 3331-Quarry / \n", "12-Food / 26-Cooking / 32-Recipe / 36-Piano / 53-Keyboard / 109-Cooking show / 215-Eating / 236-Baking / \n", "8-Football / \n", "61-Art / 447-Sand / 3282-Sand art and play / 3550-Sand animation / \n", "5-Dance / 78-Wedding / 259-Fire / \n", "8-Football / \n", "102-Building / 113-House / 170-Home improvement / 276-Room / 580-Bedroom / \n", "5-Dance / 16-Performance art / 115-Ballet / \n", "3-Concert / 7-Musician / 13-Musical ensemble / 46-Choir / 318-Church / 870-Chapel / \n", "18-Outdoor recreation / 42-Fishing / 548-Parachuting / 573-Paragliding / 769-Red Bull / \n", "3822-Mold / \n", "38-Orchestra / \n", "497-Macintosh / 819-World Wide Web / \n", "6-Animation / \n", "5-Dance / \n", "170-Home improvement / 652-Bathroom / \n", "6-Animation / 20-Trailer / \n", "1-Vehicle / 4-Car / 88-Machine / 114-Engine / \n", "1-Vehicle / 18-Outdoor recreation / \n", "1-Vehicle / 11-Motorsport / 172-Motocross / \n", "3-Concert / 708-Mermaid / \n", "0-Games / 2-Video game / 43-Call of Duty / 704-Call of Duty: World at War / \n", "0-Games / 8-Football / 21-Stadium / 65-Kick / 76-Ball / 77-Arena / 81-Athlete / \n", "8-Football / 696-T-shirt / 2693-Puma SE / \n", "0-Games / 2-Video game / 33-Weapon / 96-Soldier / 257-Counter-Strike / \n", "13-Musical ensemble / 46-Choir / \n", "28-Fashion / 45-Cosmetics / 676-Handbag / 785-Jeans / \n", "2-Video game / 141-Microsoft Windows / 2302-Dying Light / \n", "3-Concert / 7-Musician / 16-Performance art / \n", "840-Museum / \n", "6-Animation / \n", "3-Concert / 7-Musician / 30-Drummer / \n", "356-Star / \n", "37-Gadget / 47-Personal computer / 309-Computer hardware / 700-Hard disk drive / \n", "3-Concert / 7-Musician / 14-Guitar / \n", "47-Personal computer / 58-Cuisine / 145-Tablet computer / \n", "162-Medicine / \n", "162-Medicine / 209-University / \n", "163-Running / \n", "1-Vehicle / 59-Winter sport / 84-Snow / 120-Winter / 177-Skiing / 305-Ski / \n", "6-Animation / \n", "9-Music video / \n", "42-Fishing / 825-Stream / \n", "136-Gardening / 585-Lawn / 2437-String trimmer / \n", "1-Vehicle / 18-Outdoor recreation / 22-Nature / 40-Road / 67-Cycling / 69-Bicycle / 86-Plant / 137-Tree / 197-Trail / 213-Forest / 1887-Tricycle / 3713-Recumbent bicycle / \n", "3-Concert / 7-Musician / \n", "1264-Head / \n", "28-Fashion / 45-Cosmetics / 56-Hair / \n", "36-Piano / 150-Violin / 401-Cello / \n", "1-Vehicle / 27-Motorcycle / 2242-Pillow / \n", "2887-Swing / \n", "3-Concert / 5-Dance / 16-Performance art / 38-Orchestra / 68-Lighting / \n", "23-Mobile phone / 29-Smartphone / 37-Gadget / 155-Camera / 566-Digital camera / \n", "481-Kayak / 949-Rowing / 1090-Canoe / 1305-Canoeing / 1372-Paddle / 3858-Canoe Slalom / \n", "0-Games / 2-Video game / 47-Personal computer / 456-Go-kart / 518-Mario Kart / 2901-Mario Kart 64 / \n", "0-Games / 2-Video game / 19-PC game / 51-Strategy video game / 382-Dota 2 / 549-Defense of the Ancients / \n", "10-Animal / 18-Outdoor recreation / 80-Horse / 247-Stallion / 289-Dressage / 351-Jumping / 404-Livestock / \n", "83-Skateboarding / 1105-Fingerboard / \n", "28-Fashion / 153-Dress / 2392-Blouse / \n", "209-University / 3503-Doctor / \n", "3-Concert / \n", "22-Nature / \n", "0-Games / 2-Video game / 20-Trailer / 379-Touhou Project / 518-Mario Kart / \n", "1-Vehicle / 18-Outdoor recreation / 86-Plant / 124-Tractor / 156-Agriculture / 174-Farm / 252-Tractor pulling / \n", "0-Games / 6-Animation / 15-Cartoon / 323-Toddler / \n", "0-Games / 2-Video game / 19-PC game / 35-Minecraft / \n", "2-Video game / 243-Sonic the Hedgehog / \n", "9-Music video / \n", "9-Music video / \n", "5-Dance / 350-Circus / \n", "242-Paper / 370-Origami / 466-Gift / 469-Bag / \n", "2-Video game / 20-Trailer / 426-Resident Evil / \n", "419-Earth / \n", "18-Outdoor recreation / 163-Running / 674-Marathon / \n", "1-Vehicle / 4-Car / 2616-Chevrolet S-10 / \n", "812-Plastic / 2913-Drawer / \n", "0-Games / 2-Video game / 19-PC game / 47-Personal computer / 192-Battlefield / 374-Battlefield 3 / \n", "1-Vehicle / 4-Car / 11-Motorsport / 41-Sports car / 154-Wheel / 234-Ford / 349-Chevrolet / 406-Chevrolet / 473-Ford Mustang / 647-Muscle car / 761-Chevrolet Camaro / 952-Chevrolet Corvette / \n", "10-Animal / 71-Dog / \n", "28-Fashion / 45-Cosmetics / \n", "3-Concert / 7-Musician / 13-Musical ensemble / 38-Orchestra / \n", "3-Concert / 7-Musician / 13-Musical ensemble / 14-Guitar / 24-String instrument / 30-Drummer / 39-Drums / 44-Drums / \n", "159-World of Warcraft / \n", "10-Animal / 22-Nature / 155-Camera / 280-Wildlife / 846-Squirrel / 4388-Canon EOS 650D / \n", "122-Weight training / \n", "1-Vehicle / 4-Car / 11-Motorsport / 17-Racing / 27-Motorcycle / 57-Race track / 67-Cycling / 69-Bicycle / 2133-Road racing / \n", "12-Food / 18-Outdoor recreation / \n", "0-Games / 2-Video game / 19-PC game / 35-Minecraft / \n", "0-Games / 2-Video game / 4-Car / 34-Action-adventure game / 118-Grand Theft Auto V / \n", "5-Dance / \n", "293-Slide show / 444-Family / \n", "6-Animation / 97-Drawing / 148-Painting / 3411-Digital painting / \n", "1-Vehicle / 4-Car / 11-Motorsport / 135-Drifting / \n", "504-Web page / 819-World Wide Web / \n", "318-Church / \n", "1-Vehicle / 92-Radio-controlled model / 149-Model aircraft / 160-Radio-controlled aircraft / 266-Helicopter / 392-Radio-controlled helicopter / \n", "79-American football / 122-Weight training / 204-Gym / 532-Barbell / \n", "3-Concert / 68-Lighting / \n", "3-Concert / 16-Performance art / \n", "1-Vehicle / 4-Car / 11-Motorsport / 17-Racing / 70-Driving / 2518-Dirt 3 / \n", "342-Windows Media Video / 548-Parachuting / 4334-Paratrooper / \n", "9-Music video / \n", "0-Games / 2-Video game / 8-Football / 76-Ball / 90-Sports game / 453-Pro Evolution Soccer / 1719-Pro Evolution Soccer 2015 / \n", "1794-Curtain / \n", "0-Games / 8-Football / 79-American football / \n", "15-Cartoon / 123-Naruto / 183-Manga / \n", "10-Animal / 18-Outdoor recreation / 22-Nature / 42-Fishing / 91-Fish / 132-Bird / 147-River / 268-Recreational fishing / 478-Fishing rod / \n", "31-Disc jockey / \n", "110-Album / \n", "3-Concert / \n", "1-Vehicle / 4-Car / 11-Motorsport / 17-Racing / 41-Sports car / 1763-Lotus Cars / \n", "23-Mobile phone / 29-Smartphone / 47-Personal computer / 72-iPhone / 131-Computer / \n", "102-Building / 248-Hotel / \n", "1-Vehicle / 85-Combat / 259-Fire / 262-Ice / 371-Firefighter / 1245-Alarm device / \n", "1-Vehicle / 11-Motorsport / 17-Racing / 27-Motorcycle / 172-Motocross / \n", "7-Musician / 14-Guitar / 24-String instrument / \n", "3-Concert / 7-Musician / 30-Drummer / 39-Drums / \n", "661-Stretching / \n", "28-Fashion / 206-Shoe / 333-Nike; Inc. / \n", "273-News program / \n", "1-Vehicle / 92-Radio-controlled model / 149-Model aircraft / 160-Radio-controlled aircraft / 393-Unmanned aerial vehicle / 604-Quadcopter / \n", "5-Dance / 430-Ballroom dance / 678-Latin dance / 889-Swing / \n", "0-Games / 5-Dance / 79-American football / 191-Cheerleading / \n", "410-Teacher / \n", "117-Boxing / \n", "59-Winter sport / 112-Ice skating / 142-Hockey / \n", "9-Music video / \n", "9-Music video / \n", "689-Harmonica / 810-Village / \n", "0-Games / 8-Football / 21-Stadium / 54-Highlight film / 65-Kick / 76-Ball / 77-Arena / \n", "9-Music video / 347-Climbing / 2080-Cup / 3686-The North Face / \n", "1-Vehicle / 4-Car / 3679-Super GT / 4612-Sepang International Circuit / \n", "12-Food / 2397-Cabbage / \n", "1-Vehicle / 27-Motorcycle / 105-Motorcycling / \n", "8-Football / \n", "5-Dance / 16-Performance art / 769-Red Bull / \n", "2077-Chemical reaction / \n", "0-Games / 2-Video game / 159-World of Warcraft / 171-Warcraft / \n", "0-Games / 8-Football / \n", "1-Vehicle / 4-Car / 11-Motorsport / 74-Truck / 375-Pickup truck / 2683-GMC / \n", "0-Games / 2-Video game / 803-DC Comics / 1515-DC Universe Online / \n", "0-Games / 2-Video game / 19-PC game / 35-Minecraft / \n", "10-Animal / 48-Pet / 71-Dog / 1084-Bulldog / 2438-French Bulldog / \n", "3-Concert / 7-Musician / 16-Performance art / 38-Orchestra / \n", "89-Comics / 396-One Piece / 749-Figurine / \n", "0-Games / 2-Video game / 217-Hunting / 596-Monster Hunter / 2181-Monster Hunter 4 / \n", "8-Football / \n", "30-Drummer / 39-Drums / 44-Drums / 128-Cymbal / 146-Snare drum / \n", "3-Concert / 7-Musician / 13-Musical ensemble / 38-Orchestra / \n", "66-Bollywood / \n", "5-Dance / 49-School / 191-Cheerleading / \n", "0-Games / 2-Video game / 229-Halo / 470-Halo 3 / 2002-Halo 3: ODST / \n", "0-Games / 2-Video game / 33-Weapon / 96-Soldier / 192-Battlefield / 383-Battlefield 4 / \n", "0-Games / 2-Video game / 33-Weapon / 1394-Alliance of Valiant Arms / \n", "5-Dance / \n", "123-Naruto / \n", "3-Concert / 7-Musician / \n", "0-Games / 2-Video game / 19-PC game / 43-Call of Duty / 134-Call of Duty: Black Ops / 216-Call of Duty: Modern Warfare 3 / \n", "1-Vehicle / \n", "0-Games / 1-Vehicle / 2-Video game / 19-PC game / 33-Weapon / 96-Soldier / 734-DayZ / 736-ARMA / 1034-ARMA 2 / \n", "0-Games / 54-Highlight film / 60-Basketball / 168-Basketball moves / \n", "372-Playing card / 746-Card manipulation / \n", "533-Oil / 536-Glass / 1510-Liquid / 4325-Detergent / \n", "9-Music video / \n", "0-Games / 379-Touhou Project / \n", "12-Food / 26-Cooking / 32-Recipe / 52-Dish / 58-Cuisine / 184-Vegetable / 688-Beef / \n", "5-Dance / 98-Festival / \n", "10-Animal / 166-Cat / 970-Monkey / \n", "1004-Samurai / \n", "0-Games / 19-PC game / 51-Strategy video game / 165-League of Legends / \n", "3-Concert / \n", "0-Games / 2-Video game / 237-Final Fantasy / \n", "6-Animation / 15-Cartoon / 251-Sonic the Hedgehog / \n", "4175-8 mm film / \n", "23-Mobile phone / 29-Smartphone / 932-Microsoft Lumia / \n", "196-Eye / 402-Jewellery / 2167-Pendant / \n", "6-Animation / 15-Cartoon / 379-Touhou Project / \n", "12-Food / 720-Wine / 1720-Grape / 4470-Winery / \n", "36-Piano / 53-Keyboard / 107-Musical keyboard / 265-Electronic keyboard / 1909-Analog synthesizer / \n", "75-Wrestling / 1042-Jumbotron / \n", "31-Disc jockey / \n", "31-Disc jockey / \n", "133-Nightclub / \n", "10-Animal / 20-Trailer / 48-Pet / 71-Dog / 298-Television advertisement / \n", "457-Knitting / 546-Stitch / 653-Thread / \n", "10-Animal / 25-Toy / 71-Dog / 319-Robot / \n", "5-Dance / 16-Performance art / 3831-Supreme / \n", "83-Skateboarding / 108-Skateboard / 221-Skatepark / \n", "56-Hair / 607-Lace / \n", "47-Personal computer / 309-Computer hardware / 513-Computer keyboard / \n", "0-Games / 2-Video game / 34-Action-adventure game / 201-Grand Theft Auto: San Andreas / \n", "530-Cue sports / 568-Pool / \n", "1-Vehicle / 4-Car / 74-Truck / 259-Fire / 371-Firefighter / 2414-Firefighting / \n", "45-Cosmetics / \n", "1601-Social media / \n", "2-Video game / 961-Sega Genesis / 1135-Dolphin / 2638-Master System / 3980-Sega CD / \n", "1-Vehicle / 2-Video game / 20-Trailer / 34-Action-adventure game / 118-Grand Theft Auto V / 297-Grand Theft Auto IV / 658-Grand Theft Auto: The Lost and Damned / \n", "12-Food / 26-Cooking / 32-Recipe / 52-Dish / 58-Cuisine / 597-Soup / 2234-Avocado / \n", "8-Football / \n", "23-Mobile phone / 29-Smartphone / 37-Gadget / 101-Telephone / 4128-Nokia N97 / \n", "5-Dance / 1240-Mambo / \n", "0-Games / 326-Metin2 / \n", "33-Weapon / \n", "3-Concert / 7-Musician / 30-Drummer / \n", "22-Nature / \n", "905-Flash Video / 2808-Printed circuit board / \n", "37-Gadget / 47-Personal computer / 228-Laptop / 414-Computer monitor / \n", "20-Trailer / 298-Television advertisement / 1059-Movie theater / \n", "8-Football / 21-Stadium / 77-Arena / \n", "3-Concert / \n", "22-Nature / \n", "246-Furniture / 251-Sonic the Hedgehog / 276-Room / \n", "0-Games / 10-Animal / 384-Horse racing / \n", "2532-Pan flute / \n", "8-Football / 9-Music video / 76-Ball / \n", "5-Dance / 892-Aerobics / \n", "8-Football / \n", "5-Dance / \n", "1-Vehicle / 4-Car / 11-Motorsport / 41-Sports car / 761-Chevrolet Camaro / 947-Muffler / \n", "5-Dance / 95-Talent show / \n", "68-Lighting / 121-Photography / 721-Digital SLR / \n", "31-Disc jockey / 2193-CDJ / \n", "93-Comedy / \n", "0-Games / 8-Football / 21-Stadium / 81-Athlete / \n", "7-Musician / 13-Musical ensemble / 46-Choir / \n", "6-Animation / 15-Cartoon / 851-Digimon / \n", "3-Concert / \n", "1-Vehicle / 88-Machine / 363-Soil / \n", "78-Wedding / \n", "0-Games / 2-Video game / 19-PC game / 34-Action-adventure game / 35-Minecraft / \n", "7-Musician / 14-Guitar / 24-String instrument / 100-Electric guitar / \n", "7-Musician / 13-Musical ensemble / 46-Choir / 689-Harmonica / \n", "10-Animal / 86-Plant / 91-Fish / 184-Vegetable / 258-Aquarium / 1258-Goldfish / \n", "8-Football / 110-Album / \n", "9-Music video / \n", "0-Games / 19-PC game / 35-Minecraft / \n", "3-Concert / 16-Performance art / \n", "9-Music video / \n", "23-Mobile phone / 29-Smartphone / 37-Gadget / 42-Fishing / 72-iPhone / 189-iPad / 730-Final Fantasy VII / \n", "1222-Fountain / \n", "1-Vehicle / 4-Car / 70-Driving / \n", "5-Dance / 16-Performance art / \n", "5-Dance / 16-Performance art / 95-Talent show / \n", "122-Weight training / 204-Gym / \n", "111-PlayStation 3 / 243-Sonic the Hedgehog / 4019-Sonic the Hedgehog CD / \n", "12-Food / 122-Weight training / 214-Muscle / 215-Eating / \n", "10-Animal / 48-Pet / 132-Bird / 598-Parrot / \n", "6-Animation / \n", "242-Paper / 370-Origami / \n", "0-Games / 47-Personal computer / 504-Web page / \n", "75-Wrestling / \n", "0-Games / 20-Trailer / 33-Weapon / 3708-Dead Space 3 / \n", "0-Games / 2-Video game / 686-Diablo III / 872-Diablo / \n", "14-Guitar / 24-String instrument / 1063-Ink / \n", "271-Tank / \n", "946-Baseball park / \n", "3-Concert / 5-Dance / 7-Musician / 13-Musical ensemble / 46-Choir / 98-Festival / 318-Church / \n", "6-Animation / \n", "1-Vehicle / 4-Car / 27-Motorcycle / 114-Engine / 1036-Moped / \n", "3-Concert / 13-Musical ensemble / \n", "0-Games / 2-Video game / 192-Battlefield / 383-Battlefield 4 / \n", "150-Violin / 545-Quartet (ensemble) / 1002-String quartet / \n", "93-Comedy / \n", "1-Vehicle / 4-Car / 40-Road / 70-Driving / 94-Dashcam / 155-Camera / 212-Highway / \n", "9-Music video / \n", "1-Vehicle / 4-Car / 124-Tractor / 363-Soil / 459-Plough / \n", "6-Animation / 123-Naruto / \n", "110-Album / \n", "0-Games / 2-Video game / 464-First-person Shooter / \n", "0-Games / 2-Video game / 2662-Brave Frontier / \n", "3-Concert / \n", "429-Website / 504-Web page / \n", "3-Concert / 7-Musician / 16-Performance art / 465-Microphone / \n", "3-Concert / 133-Nightclub / \n", "3-Concert / 13-Musical ensemble / \n", "0-Games / 2-Video game / 1027-Terraria / 1149-Castle / \n", "227-Water / 287-Human swimming / 329-Swimming pool / \n", "3-Concert / 68-Lighting / 365-Fireworks / \n", "28-Fashion / 206-Shoe / \n", "0-Games / 5-Dance / \n", "1-Vehicle / 64-Transport / 124-Tractor / 174-Farm / 278-GoPro / 2218-Manure / \n", "0-Games / 2-Video game / 3664-Super Mario 3D Land / \n", "18-Outdoor recreation / 83-Skateboarding / 108-Skateboard / \n", "0-Games / 1-Vehicle / 2-Video game / 4-Car / 34-Action-adventure game / 118-Grand Theft Auto V / \n", "3-Concert / 7-Musician / 13-Musical ensemble / 16-Performance art / 30-Drummer / 39-Drums / \n", "1-Vehicle / 4-Car / 245-BMW / 492-Floor / \n", "1-Vehicle / 666-Concrete / 1869-Cement / \n", "1-Vehicle / 33-Weapon / 96-Soldier / 271-Tank / \n", "3-Concert / 7-Musician / \n", "6-Animation / 25-Toy / 269-Comic book / 701-Beyblade / \n", "1-Vehicle / 4-Car / 11-Motorsport / 17-Racing / 41-Sports car / 57-Race track / 135-Drifting / 169-Rallying / \n", "3-Concert / \n", "0-Games / 2-Video game / 477-Transformers / 990-Optimus Prime / 2186-Megatron / 2364-Cybertron / 3759-Transformers: Fall of Cybertron / \n", "1-Vehicle / 4-Car / 41-Sports car / 449-Honda / 552-Honda Civic / \n", "3-Concert / 2032-Pepsi / \n", "3-Concert / \n", "1-Vehicle / 22-Nature / 573-Paragliding / 1013-Walking / \n", "60-Basketball / \n", "9-Music video / \n", "1-Vehicle / 62-Train / 64-Transport / 103-Rail transport / 119-Locomotive / 143-Railroad car / 428-Rail transport modelling / 2935-Garden railway / \n", "1791-Stargate / \n", "0-Games / 2-Video game / 2348-Chrono Trigger / \n", "0-Games / 774-Galaxy / \n", "85-Combat / 117-Boxing / \n", "0-Games / 8-Football / \n", "3-Concert / 5-Dance / 153-Dress / \n", "0-Games / 60-Basketball / \n", "3-Concert / 7-Musician / 16-Performance art / \n", "72-iPhone / \n", "12-Food / 582-Candy / \n", "117-Boxing / \n", "1-Vehicle / 42-Fishing / 50-Aircraft / 82-Airplane / 129-Aviation / 139-Airport / 151-Landing / 157-Jet aircraft / 175-Airliner / 187-Takeoff / 188-Airline / 254-Runway / 471-Wing / 856-Airbus A320 family / \n", "8-Football / 21-Stadium / \n", "1-Vehicle / 4-Car / 406-Chevrolet / \n", "1-Vehicle / 11-Motorsport / 27-Motorcycle / 105-Motorcycling / \n", "1-Vehicle / 4-Car / 11-Motorsport / 17-Racing / 57-Race track / 74-Truck / 124-Tractor / 252-Tractor pulling / \n", "1-Vehicle / 4-Car / 11-Motorsport / 17-Racing / 70-Driving / 169-Rallying / \n", "1-Vehicle / 62-Train / 64-Transport / 103-Rail transport / 119-Locomotive / 143-Railroad car / 161-Train station / 224-Rapid transit / \n", "1312-Dynasty Warriors / 3721-Dynasty Warriors 7 / 3813-Dynasty Warriors: Gundam / 3897-Koei / \n", "131-Computer / \n", "1-Vehicle / 74-Truck / \n", "0-Games / 6-Animation / 15-Cartoon / \n", "20-Trailer / 3792-Dwarf / \n", "1854-Centella / \n", "6-Animation / 126-The Walt Disney Company / 144-Amusement park / 576-Walt Disney World / 679-Mickey Mouse / 1234-Mouse / 1327-Minnie Mouse / \n", "3-Concert / 7-Musician / \n", "0-Games / 2-Video game / 216-Call of Duty: Modern Warfare 3 / \n", "3-Concert / \n", "330-Money / 731-Coin / 2390-Dollar / 2937-Banknote / \n", "9-Music video / 28-Fashion / 153-Dress / 483-Costume / \n", "75-Wrestling / \n", "3-Concert / \n", "627-Clock / 3312-Seiko / 3750-It's a Small World / \n", "75-Wrestling / \n", "0-Games / 1-Vehicle / 92-Radio-controlled model / 149-Model aircraft / 160-Radio-controlled aircraft / 393-Unmanned aerial vehicle / \n", "3-Concert / 5-Dance / 16-Performance art / \n", "0-Games / 8-Football / 79-American football / \n", "0-Games / 8-Football / 21-Stadium / 65-Kick / 73-Ball / 77-Arena / 81-Athlete / \n", "12-Food / 242-Paper / \n", "93-Comedy / \n", "93-Comedy / \n", "12-Food / 26-Cooking / 58-Cuisine / 232-Cake / \n", "219-Accordion / 1994-Garmon / \n", "28-Fashion / 894-Jacket / \n", "1-Vehicle / 11-Motorsport / 17-Racing / 27-Motorcycle / 172-Motocross / \n", "102-Building / 113-House / 339-Architecture / \n", "0-Games / 6-Animation / 15-Cartoon / 25-Toy / 126-The Walt Disney Company / \n", "1-Vehicle / 4-Car / 74-Truck / 158-Tire / 167-Four-wheel drive / 375-Pickup truck / 515-Dodge / \n", "307-Hand / \n", "0-Games / 2-Video game / 34-Action-adventure game / 380-Assassin's Creed / \n", "49-School / 444-Family / \n", "8-Football / \n", "40-Road / 70-Driving / \n", "5-Dance / 16-Performance art / 95-Talent show / \n", "59-Winter sport / 84-Snow / 177-Skiing / 305-Ski / 789-Alpine skiing / 1830-Downhill / \n", "61-Art / 148-Painting / 641-Mask / \n", "1-Vehicle / 4-Car / 18-Outdoor recreation / 22-Nature / 25-Toy / 74-Truck / 92-Radio-controlled model / 147-River / 202-Radio-controlled car / 787-Monster truck / \n", "20-Trailer / \n", "1-Vehicle / 4-Car / 11-Motorsport / 17-Racing / 40-Road / 41-Sports car / 57-Race track / 70-Driving / 169-Rallying / 482-Asphalt / 1634-Ford Fiesta / \n", "36-Piano / \n", "1-Vehicle / \n", "219-Accordion / 655-Diatonic button accordion / \n", "0-Games / 8-Football / 21-Stadium / 81-Athlete / \n", "0-Games / 8-Football / \n", "10-Animal / 48-Pet / 132-Bird / \n", "0-Games / 326-Metin2 / \n", "6-Animation / 15-Cartoon / \n", "0-Games / 51-Strategy video game / 127-Xbox 360 / 194-Call of Duty: Modern Warfare 2 / \n", "3-Concert / \n", "0-Games / 2-Video game / 19-PC game / 33-Weapon / 96-Soldier / 464-First-person Shooter / 1543-Crysis / \n", "0-Games / 60-Basketball / 255-Slam dunk / \n", "0-Games / 35-Minecraft / \n", "1-Vehicle / 4-Car / 158-Tire / \n", "7-Musician / 13-Musical ensemble / 39-Drums / 44-Drums / 146-Snare drum / \n", "5-Dance / 59-Winter sport / 112-Ice skating / 262-Ice / 316-Figure skating / 833-Ice skate / \n", "0-Games / 8-Football / 21-Stadium / 65-Kick / 73-Ball / 77-Arena / 81-Athlete / \n", "23-Mobile phone / 29-Smartphone / 242-Paper / 370-Origami / 631-Cube / \n", "47-Personal computer / 131-Computer / 309-Computer hardware / 602-Central processing unit / 957-Desktop computer / 1045-Computer case / \n", "985-Strawberry / \n", "102-Building / 113-House / 339-Architecture / \n", "89-Comics / \n", "3-Concert / \n", "0-Games / 2-Video game / 20-Trailer / 35-Minecraft / 394-Map / \n", "8-Football / 595-Champion / \n", "31-Disc jockey / \n", "18-Outdoor recreation / 83-Skateboarding / 108-Skateboard / 221-Skatepark / \n", "3-Concert / 7-Musician / 24-String instrument / \n", "45-Cosmetics / 56-Hair / 871-Hair coloring / 1269-Human hair color / 3540-Violet / \n", "6-Animation / 15-Cartoon / \n", "8-Football / 54-Highlight film / 73-Ball / 595-Champion / \n", "14-Guitar / 24-String instrument / \n", "36-Piano / 53-Keyboard / 107-Musical keyboard / 265-Electronic keyboard / 592-Synthesia / \n", "6-Animation / 15-Cartoon / 60-Basketball / \n", "113-House / \n", "21-Stadium / 463-Cricket / 1495-Twenty20 / \n", "37-Gadget / \n", "1-Vehicle / 4-Car / 11-Motorsport / 41-Sports car / 70-Driving / 1324-Roadster / 1907-Ferrari F430 / 4020-Ferrari 360 / \n", "134-Call of Duty: Black Ops / 163-Running / \n", "10-Animal / 25-Toy / 97-Drawing / \n", "12-Food / 1044-Coca-Cola / \n", "98-Festival / \n", "1-Vehicle / 4-Car / 94-Dashcam / 155-Camera / \n", "783-Zee TV / \n", "12-Food / 136-Gardening / 227-Water / 259-Fire / \n", "121-Photography / \n", "0-Games / 240-Arcade game / 1802-Jubeat / \n", "0-Games / 8-Football / 21-Stadium / 65-Kick / 76-Ball / 77-Arena / 81-Athlete / \n", "10-Animal / 42-Fishing / 91-Fish / 268-Recreational fishing / 510-Fishing lure / \n", "0-Games / 159-World of Warcraft / 171-Warcraft / \n", "7-Musician / 14-Guitar / 24-String instrument / \n", "0-Games / 8-Football / 54-Highlight film / 79-American football / \n", "3-Concert / 16-Performance art / \n", "345-Loudspeaker / 484-Subwoofer / \n", "1-Vehicle / 11-Motorsport / 17-Racing / 27-Motorcycle / 172-Motocross / \n", "1035-Bank / \n", "3-Concert / 7-Musician / 13-Musical ensemble / 38-Orchestra / 219-Accordion / \n", "93-Comedy / \n", "12-Food / 26-Cooking / 32-Recipe / 786-Juice / 1329-Blender / \n", "3-Concert / \n", "31-Disc jockey / \n", "2-Video game / 90-Sports game / 180-FIFA 15 / 480-FIFA 13 / \n", "93-Comedy / 341-Sketch comedy / \n", "12-Food / 86-Plant / 156-Agriculture / 1158-Wheat / 2920-Food grain / 3260-Barley / \n", "20-Trailer / \n", "4545-Arwen / \n", "0-Games / 25-Toy / 455-Unidentified flying object / 1498-Key / 1668-Claw crane / \n", "256-Tennis / 974-Racket / 1337-Serve / \n", "122-Weight training / 214-Muscle / 1250-Biceps / 2045-Biceps curl / \n", "20-Trailer / 325-Vampire / \n", "7-Musician / 13-Musical ensemble / 46-Choir / \n", "12-Food / 26-Cooking / 32-Recipe / 52-Dish / 58-Cuisine / 109-Cooking show / 179-Kitchen / 215-Eating / 295-Chocolate / 301-Meal / 1024-Culinary art / \n", "3-Concert / 296-Saxophone / \n", "174-Farm / 348-Fruit / 366-Dragon / \n", "3-Concert / \n", "0-Games / 2-Video game / 910-Dragon Quest / 4172-Final Fantasy III / \n", "0-Games / 2-Video game / 34-Action-adventure game / 3263-Shadow of the Colossus / \n", "452-Watch / \n", "205-Newscaster / 273-News program / \n", "66-Bollywood / \n", "12-Food / 26-Cooking / 32-Recipe / 52-Dish / 109-Cooking show / 236-Baking / 391-Chicken meat / 1081-Lemon / 2309-Low-carbohydrate diet / \n", "66-Bollywood / \n", "3-Concert / \n", "28-Fashion / 56-Hair / \n", "0-Games / 8-Football / 21-Stadium / 65-Kick / \n", "0-Games / 1-Vehicle / 2-Video game / 4-Car / 6-Animation / 19-PC game / 1077-Lightning McQueen / \n", "1-Vehicle / 18-Outdoor recreation / 67-Cycling / 69-Bicycle / 197-Trail / 220-Mountain bike / 284-Mountain biking / \n", "3-Concert / 5-Dance / 7-Musician / 16-Performance art / \n", "248-Hotel / 288-Underwater / \n", "2-Video game / 189-iPad / 1150-iPad 2 / \n", "5-Dance / 85-Combat / \n", "0-Games / 2-Video game / 34-Action-adventure game / 1144-Castlevania / \n", "0-Games / 2-Video game / 90-Sports game / 180-FIFA 15 / 480-FIFA 13 / \n", "222-Surfing / 225-Lake / 233-Ocean / \n", "6-Animation / \n", "20-Trailer / \n", "12-Food / 88-Machine / 208-Wood / 812-Plastic / \n", "42-Fishing / \n", "18-Outdoor recreation / 84-Snow / 120-Winter / 262-Ice / 2058-Snowman / \n", "12-Food / 85-Combat / 259-Fire / 620-Barbecue / \n", "28-Fashion / 56-Hair / 106-Hairstyle / \n", "5-Dance / 25-Toy / 130-Gymnastics / \n", "5-Dance / 117-Boxing / \n", "9-Music video / \n", "31-Disc jockey / \n", "28-Fashion / 45-Cosmetics / 1811-Forever 21 / 2963-Converse / \n", "6-Animation / 15-Cartoon / 25-Toy / 290-Pony / 343-My Little Pony / \n", "0-Games / 2-Video game / 37-Gadget / 55-Video game console / 292-Handheld game console / 317-PlayStation / 362-PlayStation Portable / 387-Spider-Man / 682-PlayStation Vita / 3561-The Amazing Spider-Man / \n", "0-Games / 2-Video game / 15-Cartoon / 241-Fighting game / 486-M.U.G.E.N / \n", "97-Drawing / 502-Sky / 559-Night / \n", "7-Musician / 38-Orchestra / 150-Violin / \n", "7-Musician / 38-Orchestra / \n", "7-Musician / 13-Musical ensemble / 46-Choir / 318-Church / 870-Chapel / \n", "164-Beach / 222-Surfing / 233-Ocean / \n", "30-Drummer / 39-Drums / 44-Drums / 128-Cymbal / 146-Snare drum / 1127-Drum stick / \n", "10-Animal / 48-Pet / 132-Bird / \n", "28-Fashion / \n", "2-Video game / 55-Video game console / 1083-Role-playing game / 2142-Sega Saturn / 2806-SaGa / \n", "0-Games / 1-Vehicle / 2-Video game / 27-Motorcycle / 34-Action-adventure game / 47-Personal computer / 201-Grand Theft Auto: San Andreas / 858-Carl Johnson / \n", "1-Vehicle / 25-Toy / 62-Train / 64-Transport / 119-Locomotive / 152-Track / 428-Rail transport modelling / \n", "10-Animal / 14-Guitar / 61-Art / 981-Sculpture / \n", "8-Football / 21-Stadium / 77-Arena / \n", "219-Accordion / 655-Diatonic button accordion / \n", "75-Wrestling / \n", "0-Games / 2-Video game / 241-Fighting game / \n", "0-Games / 159-World of Warcraft / 171-Warcraft / \n", "7-Musician / 14-Guitar / 100-Electric guitar / \n", "8-Football / 358-Rugby football / \n", "3151-Kermit the Frog / \n", "6-Animation / 15-Cartoon / 809-Logo / \n", "1-Vehicle / 11-Motorsport / 17-Racing / 59-Winter sport / 84-Snow / 120-Winter / 610-Snowmobile / \n", "49-School / 209-University / \n", "574-Elevator / 691-Cargo / 1746-Otis Elevator Company / \n", "1-Vehicle / 4-Car / 41-Sports car / 154-Wheel / 158-Tire / 2868-Nissan 240SX / \n", "31-Disc jockey / \n", "631-Cube / 740-Rubik's Cube / \n", "10-Animal / 18-Outdoor recreation / 217-Hunting / \n", "3-Concert / 93-Comedy / \n", "2147-Fiber / 2958-Electrical connector / \n", "10-Animal / 80-Horse / \n", "6-Animation / \n", "3-Concert / 14-Guitar / 16-Performance art / 39-Drums / \n", "222-Surfing / 672-Bodyboarding / 815-Surfboard / \n", "5-Dance / \n", "12-Food / 26-Cooking / 32-Recipe / 58-Cuisine / 210-Cookware and bakeware / 239-Roasting / 1284-Frying / 2483-Deep frying / \n", "0-Games / 2-Video game / 34-Action-adventure game / 199-Wii / \n", "2-Video game / \n", "98-Festival / 365-Fireworks / \n", "0-Games / 2-Video game / 33-Weapon / 43-Call of Duty / 96-Soldier / 111-PlayStation 3 / 194-Call of Duty: Modern Warfare 2 / 216-Call of Duty: Modern Warfare 3 / \n", "45-Cosmetics / 385-Face / \n", "0-Games / 2-Video game / 900-MacBook / 1657-MacBook Air / 1928-BioShock / 2645-BioShock Infinite / \n", "93-Comedy / \n", "1-Vehicle / 4-Car / 349-Chevrolet / 406-Chevrolet / 3793-Chevrolet Chevy II / Nova / \n", "0-Games / 8-Football / 21-Stadium / \n", "1119-Sensor / \n", "7-Musician / 14-Guitar / 24-String instrument / \n", "93-Comedy / \n", "20-Trailer / \n", "176-PlayStation 4 / \n", "31-Disc jockey / \n", "9-Music video / 78-Wedding / \n", "10-Animal / 18-Outdoor recreation / 22-Nature / \n", "10-Animal / 48-Pet / 166-Cat / 346-Kitten / \n", "12-Food / 18-Outdoor recreation / 22-Nature / 124-Tractor / 156-Agriculture / 174-Farm / 617-Maize / \n", "3-Concert / \n", "502-Sky / \n", "12-Food / 88-Machine / \n", "22-Nature / 213-Forest / 423-Coast / \n", "1-Vehicle / 4-Car / 11-Motorsport / 17-Racing / 41-Sports car / 70-Driving / \n", "1-Vehicle / 62-Train / 103-Rail transport / 119-Locomotive / 143-Railroad car / 3348-BVE Trainsim / \n", "85-Combat / 117-Boxing / 1016-Glove / \n", "6-Animation / 123-Naruto / \n", "7-Musician / 14-Guitar / 24-String instrument / 100-Electric guitar / 2656-Guitar Center / \n", "8-Football / \n", "88-Machine / 536-Glass / 619-Manufacturing / \n", "20-Trailer / \n", "1-Vehicle / 124-Tractor / 136-Gardening / 249-Heavy equipment / 363-Soil / 2138-Compost / \n", "6-Animation / 15-Cartoon / 2047-Popeye / \n", "1-Vehicle / 4-Car / 70-Driving / 94-Dashcam / 299-Volkswagen Passenger Cars / \n", "0-Games / 211-RuneScape / \n", "9-Music video / \n", "10-Animal / 48-Pet / 132-Bird / 1767-Cockatiel / \n", "12-Food / 26-Cooking / 32-Recipe / 52-Dish / \n", "5-Dance / \n", "61-Art / 97-Drawing / 242-Paper / 768-Diamond / \n", "10-Animal / 1160-Giant panda / \n", "8-Football / \n", "0-Games / 1-Vehicle / 50-Aircraft / \n", "0-Games / \n", "130-Gymnastics / 920-Chair / \n", "5-Dance / 16-Performance art / \n", "6-Animation / 15-Cartoon / 2408-Lucky Star / \n", "3-Concert / \n", "3-Concert / 13-Musical ensemble / \n", "0-Games / 6-Animation / 10-Animal / 15-Cartoon / 55-Video game console / \n", "3-Concert / \n", "0-Games / 256-Tennis / \n", "56-Hair / 340-Afro-textured hair / 405-Braid / 919-Hair conditioner / \n", "0-Games / 2-Video game / 60-Basketball / 90-Sports game / \n", "246-Furniture / 307-Hand / 462-Table / 1057-Embroidery / 3071-Towel / \n", "2-Video game / 25-Toy / 43-Call of Duty / \n", "1-Vehicle / 4-Car / 11-Motorsport / 541-Black-and-white / \n", "9-Music video / \n", "1-Vehicle / 4-Car / 11-Motorsport / 17-Racing / 40-Road / 41-Sports car / 57-Race track / 70-Driving / 135-Drifting / 1583-Toyota Corolla / 4584-Street Legal Racing: Redline / \n", "125-Television / 3934-Radiator / \n", "22-Nature / 164-Beach / 233-Ocean / 275-Diving / 423-Coast / \n", "3-Concert / 7-Musician / 13-Musical ensemble / 38-Orchestra / \n", "5-Dance / 1623-Shaolin Kung Fu / \n", "196-Eye / \n", "5-Dance / 78-Wedding / 153-Dress / 190-Bride / 421-Gown / \n", "117-Boxing / \n", "88-Machine / 1226-Metalworking / 3073-Log splitter / \n", "3-Concert / 7-Musician / 13-Musical ensemble / 14-Guitar / 36-Piano / 44-Drums / 150-Violin / \n", "97-Drawing / 408-Sketch / 528-Pencil / 3491-Stained glass / \n", "14-Guitar / 63-Acoustic guitar / \n", "9-Music video / 110-Album / \n", "130-Gymnastics / 1835-Majorette / \n", "1-Vehicle / 4-Car / \n", "0-Games / 2-Video game / 43-Call of Duty / 216-Call of Duty: Modern Warfare 3 / \n", "3-Concert / 46-Choir / 3707-Yale University / \n", "3-Concert / 126-The Walt Disney Company / 144-Amusement park / 365-Fireworks / 576-Walt Disney World / 1463-Magic Kingdom / \n", "16-Performance art / 78-Wedding / 115-Ballet / \n", "9-Music video / \n", "1-Vehicle / 4-Car / 22-Nature / 40-Road / 94-Dashcam / 800-Mountain pass / \n", "0-Games / 2-Video game / 23-Mobile phone / 29-Smartphone / 37-Gadget / 72-iPhone / 189-iPad / 272-iPod / 331-iPod touch / \n", "616-The Bible / \n", "6-Animation / 15-Cartoon / 1674-Ouran High School Host Club / \n", "56-Hair / 919-Hair conditioner / \n", "0-Games / 8-Football / 21-Stadium / 65-Kick / 73-Ball / 76-Ball / 77-Arena / 81-Athlete / \n", "0-Games / 8-Football / 54-Highlight film / 79-American football / \n", "5-Dance / 16-Performance art / 95-Talent show / 115-Ballet / \n", "5-Dance / \n", "23-Mobile phone / 29-Smartphone / 37-Gadget / \n", "14-Guitar / 24-String instrument / 63-Acoustic guitar / 361-Flamenco / \n", "912-Silver / \n", "0-Games / 49-School / 60-Basketball / 142-Hockey / \n", "222-Surfing / 233-Ocean / 672-Bodyboarding / 815-Surfboard / \n", "23-Mobile phone / 29-Smartphone / 178-Samsung Galaxy / 504-Web page / 1287-Samsung Galaxy Tab series / 1390-Samsung Galaxy S II / \n", "986-Suit / \n", "1-Vehicle / 9-Music video / 87-Boat / 579-Yacht / \n", "15-Cartoon / 4525-Hi-Rez Studios / \n", "10-Animal / 80-Horse / 247-Stallion / 438-Mare / 1871-Foal / 2018-Arabian horse / 2553-Colt / \n", "9-Music video / \n", "3-Concert / 5-Dance / 68-Lighting / 144-Amusement park / 366-Dragon / 559-Night / \n", "8-Football / 595-Champion / \n", "14-Guitar / 63-Acoustic guitar / \n", "3-Concert / 7-Musician / 13-Musical ensemble / 30-Drummer / 39-Drums / 44-Drums / 56-Hair / \n", "20-Trailer / 387-Spider-Man / \n", "67-Cycling / 215-Eating / 3101-Indoor cycling / \n", "1862-Corel / \n", "6-Animation / 25-Toy / 200-Doll / \n", "79-American football / \n", "12-Food / 26-Cooking / 32-Recipe / 52-Dish / 58-Cuisine / 109-Cooking show / 184-Vegetable / 944-Hamburger / 1022-Chili pepper / 1332-Onion / 1355-Vegetarian food / 1381-Garlic / 3100-Chickpea / \n", "3-Concert / 7-Musician / 13-Musical ensemble / 14-Guitar / 38-Orchestra / \n", "1-Vehicle / 87-Boat / 308-Ship / 1107-United States Navy / \n", "1-Vehicle / 27-Motorcycle / 40-Road / 283-Rock / 1358-Canyon / 1808-National park / \n", "2852-Burger King / \n", "3051-Sarah Walker / \n", "12-Food / 26-Cooking / 32-Recipe / 52-Dish / 58-Cuisine / 109-Cooking show / 210-Cookware and bakeware / 893-Fish as food / 3379-Prawn / \n", "9-Music video / \n", "1-Vehicle / 4-Car / 11-Motorsport / 17-Racing / 40-Road / 378-Drag racing / \n", "6-Animation / 15-Cartoon / 1563-Stick figure / \n", "22-Nature / 86-Plant / \n", "9-Music video / \n", "0-Games / 2-Video game / 89-Comics / 185-Dragon Ball / 1462-Dragon Ball XenoVerse / \n", "1-Vehicle / 4-Car / 124-Tractor / \n", "2429-India TV / \n", "0-Games / 2-Video game / 6-Animation / 111-PlayStation 3 / 547-Bleach / \n", "3-Concert / 16-Performance art / \n", "208-Wood / 246-Furniture / \n", "444-Family / \n", "0-Games / 582-Candy / 1112-Candy Crush Saga / \n", "3-Concert / \n", "5-Dance / 49-School / 191-Cheerleading / \n", "56-Hair / 106-Hairstyle / \n", "12-Food / 232-Cake / 1102-Cake decorating / 1365-Fondant icing / \n", "93-Comedy / 341-Sketch comedy / 364-Sitcom / \n", "1-Vehicle / 50-Aircraft / 188-Airline / \n", "10-Animal / \n", "5-Dance / 49-School / \n", "31-Disc jockey / 49-School / \n", "7-Musician / 14-Guitar / 24-String instrument / 63-Acoustic guitar / \n", "131-Computer / 141-Microsoft Windows / \n", "3-Concert / 7-Musician / 13-Musical ensemble / \n", "25-Toy / 1105-Fingerboard / \n", "9-Music video / \n", "3-Concert / 7-Musician / 14-Guitar / \n", "1465-FOX / \n", "0-Games / 2-Video game / 237-Final Fantasy / 4661-Final Fantasy II / \n", "85-Combat / 117-Boxing / 399-Kickboxing / 699-Kick / \n", "47-Personal computer / 131-Computer / 228-Laptop / 513-Computer keyboard / 811-Mouse / \n", "3-Concert / 7-Musician / 13-Musical ensemble / 14-Guitar / 30-Drummer / 39-Drums / 44-Drums / 1044-Coca-Cola / \n", "3-Concert / 7-Musician / 13-Musical ensemble / 24-String instrument / \n", "3-Concert / 7-Musician / 13-Musical ensemble / 38-Orchestra / 150-Violin / \n", "710-Handball / \n", "0-Games / 257-Counter-Strike / \n", "1-Vehicle / 4-Car / 167-Four-wheel drive / 234-Ford / 2899-Ford Explorer / \n", "110-Album / 2681-Marshmallow / \n", "3-Concert / 7-Musician / 13-Musical ensemble / 46-Choir / \n", "1226-Metalworking / 2035-Forge / \n", "1-Vehicle / 4-Car / 11-Motorsport / 17-Racing / \n", "5-Dance / 130-Gymnastics / \n", "356-Star / 654-Sun / 712-Outer space / 774-Galaxy / 2938-Nebula / \n", "12-Food / 367-Drink / 612-Beer / \n", "3-Concert / 1684-Hillsong Church / \n", "130-Gymnastics / 1255-Ribbon / \n", "2884-Unturned / \n", "219-Accordion / 655-Diatonic button accordion / \n", "61-Art / 415-Sewing / 476-Textile / 1398-Triangle / 1479-Quilt / 2373-Square / 2966-Patchwork / \n", "3-Concert / 7-Musician / 13-Musical ensemble / 46-Choir / \n", "1-Vehicle / 4-Car / 11-Motorsport / 17-Racing / 27-Motorcycle / 40-Road / 57-Race track / 278-GoPro / 456-Go-kart / \n", "22-Nature / \n", "14-Guitar / \n", "12-Food / 52-Dish / 278-GoPro / \n", "23-Mobile phone / 29-Smartphone / 37-Gadget / 72-iPhone / 1009-iPhone 3G / \n", "1-Vehicle / 4-Car / 154-Wheel / \n", "1-Vehicle / 4-Car / 270-Sedan / 449-Honda / 552-Honda Civic / \n", "3-Concert / \n", "12-Food / 1162-Dietary supplement / \n", "13-Musical ensemble / 38-Orchestra / 98-Festival / 253-Carnival / \n", "5-Dance / 115-Ballet / \n", "1-Vehicle / 4-Car / 27-Motorcycle / 114-Engine / 3004-Softail / \n", "1-Vehicle / 4-Car / 1391-Kia / 1415-Kia Motors / \n", "0-Games / 1-Vehicle / 2-Video game / 4-Car / 11-Motorsport / 17-Racing / 41-Sports car / 57-Race track / 127-Xbox 360 / 1419-GameSpot / \n", "0-Games / 2-Video game / 19-PC game / 33-Weapon / 3977-Lost Planet / \n", "144-Amusement park / 321-Roller coaster / 334-Amusement ride / 1059-Movie theater / 2786-3D television / \n", "0-Games / 2-Video game / 121-Photography / 125-Television / 1348-Panasonic Corporation / \n", "6-Animation / 15-Cartoon / \n", "3-Concert / 7-Musician / 13-Musical ensemble / 16-Performance art / 39-Drums / \n", "0-Games / 2-Video game / 75-Wrestling / 1582-WWE 2K14 / \n", "8-Football / 67-Cycling / 69-Bicycle / 570-Road bicycle racing / \n", "8-Football / \n", "5-Dance / 115-Ballet / \n", "7-Musician / 14-Guitar / 266-Helicopter / \n", "694-Tai chi / \n", "42-Fishing / \n", "5-Dance / \n", "5-Dance / 66-Bollywood / \n", "10-Animal / 763-Snake / \n", "12-Food / 3544-Rubber stamp / \n", "0-Games / 116-Pokémon / 186-Pokémon / 3347-Misty / \n", "0-Games / 2-Video game / 6-Animation / 33-Weapon / 229-Halo / \n", "5-Dance / \n", "9-Music video / \n", "3-Concert / \n", "3-Concert / 30-Drummer / \n", "9-Music video / \n", "10-Animal / 22-Nature / 91-Fish / 2829-Dictionary / \n", "0-Games / 8-Football / \n", "0-Games / 242-Paper / 370-Origami / \n", "0-Games / 725-Human / 1698-Darts / \n", "3-Concert / \n", "5-Dance / 130-Gymnastics / 1474-Acrobatic gymnastics / \n", "66-Bollywood / \n", "1-Vehicle / 62-Train / 64-Transport / 94-Dashcam / 103-Rail transport / 119-Locomotive / 143-Railroad car / 152-Track / 161-Train station / 224-Rapid transit / \n", "1-Vehicle / 11-Motorsport / 124-Tractor / 156-Agriculture / 174-Farm / 252-Tractor pulling / 363-Soil / 459-Plough / \n", "3-Concert / 7-Musician / 16-Performance art / 30-Drummer / \n", "9-Music video / \n", "7-Musician / 13-Musical ensemble / 14-Guitar / 44-Drums / \n", "28-Fashion / 56-Hair / 106-Hairstyle / 405-Braid / \n", "0-Games / 117-Boxing / \n", "3-Concert / 7-Musician / 13-Musical ensemble / 36-Piano / 230-Brass instrument / 427-Trumpet / \n", "10-Animal / 48-Pet / 132-Bird / 598-Parrot / 1767-Cockatiel / 2024-Cockatoo / \n", "1-Vehicle / 33-Weapon / 271-Tank / \n", "33-Weapon / \n", "25-Toy / 200-Doll / 1828-Campervan / 1842-Dollhouse / \n", "3-Concert / 7-Musician / 13-Musical ensemble / 46-Choir / 49-School / \n", "3-Concert / \n", "144-Amusement park / 321-Roller coaster / 334-Amusement ride / \n", "6-Animation / 15-Cartoon / 877-Hatsune Miku: Project DIVA / 1648-Hatsune Miku: Project DIVA F / 2682-Hatsune Miku: Project DIVA F 2nd / \n", "0-Games / 2-Video game / 6-Animation / 15-Cartoon / 89-Comics / 123-Naruto / 285-Sasuke Uchiha / 591-Naruto: Ultimate Ninja / 2299-Naruto Shippuden: Ultimate Ninja Storm 2 / \n", "46-Choir / \n", "9-Music video / \n", "28-Fashion / 45-Cosmetics / 429-Website / \n", "3-Concert / 7-Musician / 36-Piano / 53-Keyboard / 104-Pianist / 107-Musical keyboard / 878-Composer / \n", "0-Games / 8-Football / 21-Stadium / 65-Kick / 73-Ball / \n", "14-Guitar / 63-Acoustic guitar / \n", "1-Vehicle / 4-Car / 11-Motorsport / 17-Racing / 975-Dune / \n", "122-Weight training / 162-Medicine / 214-Muscle / 1162-Dietary supplement / \n", "18-Outdoor recreation / 42-Fishing / \n", "18-Outdoor recreation / 83-Skateboarding / 108-Skateboard / 221-Skatepark / 1031-Fisheye lens / \n", "1-Vehicle / 4-Car / 87-Boat / 246-Furniture / 579-Yacht / \n", "1-Vehicle / 11-Motorsport / 22-Nature / 27-Motorcycle / 197-Trail / \n", "36-Piano / 53-Keyboard / 104-Pianist / 107-Musical keyboard / \n", "6-Animation / 15-Cartoon / 99-Christmas / 243-Sonic the Hedgehog / 251-Sonic the Hedgehog / \n", "12-Food / 26-Cooking / 179-Kitchen / 215-Eating / \n", "3-Concert / \n", "0-Games / 8-Football / 21-Stadium / 54-Highlight film / \n", "0-Games / 60-Basketball / 168-Basketball moves / 4466-Korfball / \n", "3-Concert / 98-Festival / 594-Music festival / \n", "56-Hair / 106-Hairstyle / \n", "0-Games / 2-Video game / 229-Halo / 470-Halo 3 / \n", "6-Animation / 15-Cartoon / \n", "6-Animation / 15-Cartoon / 20-Trailer / 294-Batman / 2069-Flash / 3473-Nightwing / \n", "45-Cosmetics / 56-Hair / 235-Lipstick / 282-Skin / 385-Face / \n", "3-Concert / 31-Disc jockey / 68-Lighting / 133-Nightclub / 263-Light / \n", "2-Video game / 1985-The Last of Us / \n", "843-The Doctor / \n", "0-Games / 8-Football / \n", "0-Games / 2-Video game / 34-Action-adventure game / 243-Sonic the Hedgehog / 251-Sonic the Hedgehog / \n", "6-Animation / 15-Cartoon / 20-Trailer / 116-Pokémon / 183-Manga / \n", "3-Concert / 7-Musician / \n", "0-Games / 8-Football / 79-American football / \n", "23-Mobile phone / 101-Telephone / \n", "5-Dance / \n", "6-Animation / 330-Money / 402-Jewellery / \n", "6-Animation / 243-Sonic the Hedgehog / 251-Sonic the Hedgehog / 1197-Hedgehog / 1299-Shadow the Hedgehog / 3321-Shadow the Hedgehog / \n", "0-Games / 8-Football / 54-Highlight film / 79-American football / \n", "78-Wedding / 136-Gardening / \n", "5-Dance / 16-Performance art / 95-Talent show / \n", "1-Vehicle / 50-Aircraft / 778-Gliding / 1263-Glider / 1317-Glider / \n", "162-Medicine / \n", "3-Concert / \n", "1-Vehicle / 4-Car / 11-Motorsport / 27-Motorcycle / 105-Motorcycling / 2345-Kawasaki Ninja 250R / \n", "1-Vehicle / 4-Car / 1242-Car wash / \n", "0-Games / 2-Video game / 19-PC game / 257-Counter-Strike / 702-Counter-Strike: Source / \n", "0-Games / 2-Video game / 264-Star Wars / \n", "125-Television / \n", "4214-Spike / \n", "59-Winter sport / 84-Snow / 120-Winter / 177-Skiing / 305-Ski / \n", "61-Art / 148-Painting / \n", "0-Games / 2-Video game / 34-Action-adventure game / 243-Sonic the Hedgehog / 251-Sonic the Hedgehog / 1512-Sonic Generations / \n", "0-Games / 2-Video game / 33-Weapon / 43-Call of Duty / 85-Combat / 111-PlayStation 3 / 134-Call of Duty: Black Ops / 677-Game controller / \n", "0-Games / 2-Video game / 19-PC game / 43-Call of Duty / 56-Hair / 194-Call of Duty: Modern Warfare 2 / \n", "22-Nature / 86-Plant / 136-Gardening / 279-Garden / \n", "0-Games / 8-Football / 90-Sports game / \n", "25-Toy / 200-Doll / 412-Monster / 503-Monster High / \n", "7-Musician / 13-Musical ensemble / 46-Choir / \n", "0-Games / 2-Video game / 19-PC game / 34-Action-adventure game / 118-Grand Theft Auto V / 1069-Easter egg / \n", "23-Mobile phone / 29-Smartphone / 72-iPhone / \n", "75-Wrestling / 85-Combat / \n", "3-Concert / 7-Musician / 13-Musical ensemble / 38-Orchestra / 203-Marching band / 230-Brass instrument / \n", "0-Games / 2-Video game / 33-Weapon / 134-Call of Duty: Black Ops / \n", "1-Vehicle / 25-Toy / 33-Weapon / 140-LEGO / 271-Tank / \n", "3-Concert / 46-Choir / 318-Church / \n", "3-Concert / 545-Quartet (ensemble) / \n", "1-Vehicle / 67-Cycling / 69-Bicycle / 197-Trail / 220-Mountain bike / 284-Mountain biking / \n", "10-Animal / 86-Plant / 91-Fish / 136-Gardening / 258-Aquarium / 1258-Goldfish / \n", "8-Football / \n", "0-Games / 8-Football / 54-Highlight film / 79-American football / 81-Athlete / \n", "6-Animation / 15-Cartoon / 379-Touhou Project / 2170-The Embodiment of Scarlet Devil / \n", "10-Animal / 48-Pet / 71-Dog / 2059-Doberman Pinscher / \n", "9-Music video / \n", "1-Vehicle / 4-Car / 17-Racing / 27-Motorcycle / 40-Road / 212-Highway / \n", "3-Concert / 7-Musician / 16-Performance art / \n", "9-Music video / \n", "463-Cricket / 1476-Bowling (cricket) / \n", "9-Music video / \n", "10-Animal / 18-Outdoor recreation / 80-Horse / 247-Stallion / \n", "28-Fashion / \n", "36-Piano / 53-Keyboard / 107-Musical keyboard / \n", "59-Winter sport / 177-Skiing / 373-Clash of Clans / 1764-Ski jumping / \n", "3-Concert / \n", "9-Music video / \n", "1-Vehicle / 4-Car / 1614-Honda Accord / \n", "283-Rock / \n", "0-Games / 2-Video game / 55-Video game console / 292-Handheld game console / 320-Super Smash Bros. / \n", "5-Dance / 306-High school / \n", "3-Concert / 7-Musician / 13-Musical ensemble / 46-Choir / 860-Christian Church / \n", "44-Drums / 246-Furniture / 1825-Cajón / \n", "10-Animal / 18-Outdoor recreation / 80-Horse / 247-Stallion / 289-Dressage / 351-Jumping / \n", "12-Food / 26-Cooking / 179-Kitchen / \n", "47-Personal computer / 131-Computer / 497-Macintosh / 504-Web page / \n", "39-Drums / 44-Drums / 146-Snare drum / \n", "1-Vehicle / 2-Video game / 50-Aircraft / \n", "5-Dance / 66-Bollywood / 4017-Television film / \n", "5-Dance / \n", "0-Games / 8-Football / 21-Stadium / 54-Highlight film / 65-Kick / 73-Ball / 76-Ball / 77-Arena / \n", "0-Games / 2-Video game / 596-Monster Hunter / 1935-Monster Hunter Tri / \n", "10-Animal / 18-Outdoor recreation / 22-Nature / 42-Fishing / 91-Fish / 217-Hunting / 275-Diving / 288-Underwater / 381-Underwater diving / 407-Scuba diving / \n", "6-Animation / 15-Cartoon / 577-Penguin / \n", "8-Football / \n", "325-Vampire / \n", "897-Avengers / 1469-X-Men / \n", "1-Vehicle / 4-Car / 124-Tractor / 156-Agriculture / 174-Farm / 363-Soil / 459-Plough / 906-Sowing / \n", "223-Weather / \n", "10-Animal / 80-Horse / 247-Stallion / 1434-Saddle / \n", "61-Art / 457-Knitting / 546-Stitch / \n", "5-Dance / 76-Ball / 2389-Exercise ball / \n", "141-Microsoft Windows / \n", "79-American football / 303-Paint / 522-Helmet / \n", "6-Animation / \n", "2867-Wasp / \n", "8-Football / \n", "1-Vehicle / 4-Car / 40-Road / 259-Fire / 371-Firefighter / 1633-Fire station / \n", "5-Dance / 16-Performance art / 98-Festival / \n", "2196-Boiler / \n", "0-Games / 35-Minecraft / \n", "0-Games / 8-Football / 21-Stadium / 65-Kick / 81-Athlete / \n", "6-Animation / 185-Dragon Ball / 200-Doll / 242-Paper / \n", "23-Mobile phone / 29-Smartphone / 204-Gym / 2488-Fitness Center / \n", "7-Musician / 14-Guitar / \n", "0-Games / 1047-Easter egg / \n", "0-Games / 2-Video game / 85-Combat / 803-DC Comics / 1515-DC Universe Online / 3068-Green Arrow / \n", "0-Games / 1-Vehicle / 2-Video game / 4-Car / 11-Motorsport / 17-Racing / \n", "6-Animation / 15-Cartoon / \n", "5-Dance / 16-Performance art / \n", "30-Drummer / 39-Drums / 44-Drums / 123-Naruto / 128-Cymbal / 146-Snare drum / \n", "20-Trailer / \n", "0-Games / 25-Toy / 76-Ball / 323-Toddler / \n", "46-Choir / \n", "5-Dance / \n", "18-Outdoor recreation / 22-Nature / 144-Amusement park / 321-Roller coaster / 334-Amusement ride / 3395-Asterix / \n", "1-Vehicle / 4-Car / 11-Motorsport / 27-Motorcycle / 105-Motorcycling / 158-Tire / 207-Exhaust system / 2617-Kawasaki Ninja ZX-10R / \n", "0-Games / 1-Vehicle / 2-Video game / 4-Car / 19-PC game / 34-Action-adventure game / 118-Grand Theft Auto V / \n", "0-Games / 211-RuneScape / 2685-Giant / \n", "1-Vehicle / 4-Car / 11-Motorsport / 17-Racing / 27-Motorcycle / 40-Road / \n", "48-Pet / 159-World of Warcraft / 171-Warcraft / \n", "0-Games / 2-Video game / 43-Call of Duty / 118-Grand Theft Auto V / 134-Call of Duty: Black Ops / 138-Call of Duty: Black Ops II / 1543-Crysis / \n", "5-Dance / \n", "28-Fashion / 56-Hair / 106-Hairstyle / 919-Hair conditioner / 1359-Hair iron / 3584-Hair roller / \n", "1-Vehicle / 18-Outdoor recreation / 59-Winter sport / 84-Snow / 120-Winter / 610-Snowmobile / \n", "10-Animal / 48-Pet / 71-Dog / 84-Snow / \n", "9-Music video / \n", "170-Home improvement / 246-Furniture / 830-Bed / \n", "121-Photography / 155-Camera / 721-Digital SLR / \n", "1-Vehicle / 11-Motorsport / 17-Racing / 27-Motorcycle / \n", "0-Games / 8-Football / \n", "0-Games / 2-Video game / 1144-Castlevania / \n", "89-Comics / \n", "3540-Violet / \n", "61-Art / 89-Comics / 294-Batman / 435-Action figure / 737-Joker / \n", "66-Bollywood / \n", "1-Vehicle / 11-Motorsport / 18-Outdoor recreation / 27-Motorcycle / \n", "170-Home improvement / 303-Paint / 353-Wall / 1179-Brick / \n", "85-Combat / 1832-Hand-to-hand combat / \n", "2620-Federal Reserve System / \n", "23-Mobile phone / 29-Smartphone / 37-Gadget / 345-Loudspeaker / \n", "1-Vehicle / 4-Car / 509-Classic car / \n", "0-Games / 1-Vehicle / 2-Video game / 4-Car / 34-Action-adventure game / 40-Road / 74-Truck / 118-Grand Theft Auto V / 886-James Bond / \n", "3-Concert / 5-Dance / 16-Performance art / \n", "9-Music video / \n", "0-Games / 12-Food / 88-Machine / 822-Crane / 1314-Garage / 1668-Claw crane / \n", "0-Games / 2-Video game / 240-Arcade game / 1467-Pump It Up / 2651-Pump It Up Fiesta / \n", "0-Games / 2-Video game / 20-Trailer / 35-Minecraft / \n", "1-Vehicle / 715-Farming Simulator / \n", "3-Concert / 5-Dance / 13-Musical ensemble / \n", "6-Animation / 15-Cartoon / 1338-Devil / \n", "1-Vehicle / 27-Motorcycle / 105-Motorcycling / 114-Engine / 806-Chopper / \n", "6-Animation / 1192-Hello Kitty / \n", "5-Dance / 703-Roller skating / \n", "0-Games / 2-Video game / 192-Battlefield / 374-Battlefield 3 / \n", "376-Flute / 3344-Tin whistle / \n", "465-Microphone / \n", "2-Video game / 6-Animation / 15-Cartoon / 111-PlayStation 3 / 123-Naruto / 413-PlayStation 2 / 591-Naruto: Ultimate Ninja / \n", "523-Cattle / \n", "10-Animal / 48-Pet / 71-Dog / 182-Puppy / 709-Bull / 824-Terrier / 2388-Bull Terrier / \n", "0-Games / 2-Video game / 19-PC game / 33-Weapon / 43-Call of Duty / 134-Call of Duty: Black Ops / 138-Call of Duty: Black Ops II / 422-Call of Duty: Ghosts / \n", "6-Animation / 15-Cartoon / \n", "20-Trailer / \n", "3-Concert / 5-Dance / 16-Performance art / \n", "102-Building / 113-House / 339-Architecture / 416-Apartment / 939-Condominium / \n", "434-Ninja / 865-Turtle / 951-Teenage Mutant Ninja Turtles / \n", "30-Drummer / 39-Drums / 44-Drums / 128-Cymbal / 146-Snare drum / \n", "7-Musician / 46-Choir / 3463-Nun / \n", "31-Disc jockey / \n", "3-Concert / \n", "31-Disc jockey / \n", "0-Games / 8-Football / 21-Stadium / 54-Highlight film / 79-American football / 81-Athlete / \n", "6-Animation / \n", "9-Music video / \n", "170-Home improvement / 492-Floor / 934-Flooring / 3359-Laminate flooring / \n", "125-Television / 209-University / \n", "1-Vehicle / 4-Car / 11-Motorsport / 3058-BMW 7 Series / \n", "0-Games / 2-Video game / 192-Battlefield / 374-Battlefield 3 / 383-Battlefield 4 / \n", "1-Vehicle / 11-Motorsport / \n", "60-Basketball / \n", "6-Animation / 15-Cartoon / \n", "1-Vehicle / 11-Motorsport / 27-Motorcycle / 169-Rallying / \n", "10-Animal / 136-Gardening / 458-Gold / 521-Metal / 731-Coin / 959-Metal detector / \n", "121-Photography / 155-Camera / 721-Digital SLR / 1119-Sensor / \n", "1-Vehicle / 4-Car / 270-Sedan / 827-Pump / 2509-Mazda3 / \n", "122-Weight training / 204-Gym / \n", "0-Games / 2-Video game / 19-PC game / 33-Weapon / 2594-Mount & Blade: Warband / 2622-Mount & Blade / \n", "45-Cosmetics / 173-Eye shadow / \n", "5-Dance / \n", "1-Vehicle / 4-Car / 88-Machine / 114-Engine / 741-Mower / \n", "66-Bollywood / \n", "10-Animal / 48-Pet / 71-Dog / 293-Slide show / \n", "156-Agriculture / 174-Farm / 328-Simulation video game / 715-Farming Simulator / \n", "36-Piano / 53-Keyboard / 104-Pianist / \n", "0-Games / 2-Video game / 269-Comic book / 489-Board game / \n", "28-Fashion / 93-Comedy / \n", "1-Vehicle / 11-Motorsport / 17-Racing / 27-Motorcycle / 105-Motorcycling / 172-Motocross / 491-Enduro / \n", "260-Prayer / \n", "0-Games / 1-Vehicle / 2-Video game / 4-Car / 11-Motorsport / 17-Racing / 41-Sports car / 57-Race track / 344-Coupé / 569-Lamborghini / 1763-Lotus Cars / 2400-Asphalt / 4481-Pagani Huayra / \n", "93-Comedy / \n", "12-Food / 26-Cooking / 32-Recipe / 109-Cooking show / 1174-Smoothie / \n", "1-Vehicle / 83-Skateboarding / 108-Skateboard / \n", "0-Games / 2-Video game / 240-Arcade game / 241-Fighting game / 311-Street Fighter / 461-Street Fighter IV / 496-Super Street Fighter IV / 563-Ryu / \n", "6-Animation / 15-Cartoon / \n", "28-Fashion / \n", "30-Drummer / 39-Drums / 44-Drums / 128-Cymbal / 146-Snare drum / \n", "30-Drummer / 39-Drums / 44-Drums / 128-Cymbal / 146-Snare drum / \n", "6-Animation / 434-Ninja / 951-Teenage Mutant Ninja Turtles / 3508-Leonardo / 3667-Raphael / \n", "0-Games / 2-Video game / 33-Weapon / 43-Call of Duty / 216-Call of Duty: Modern Warfare 3 / \n", "3-Concert / 5-Dance / 16-Performance art / \n", "1-Vehicle / 4-Car / 18-Outdoor recreation / 74-Truck / 137-Tree / 158-Tire / 167-Four-wheel drive / 197-Trail / 198-Off-road vehicle / 234-Ford / 327-Mud / 360-Jeep / \n", "0-Games / 1-Vehicle / 2-Video game / 4-Car / 34-Action-adventure game / 118-Grand Theft Auto V / 176-PlayStation 4 / \n", "1-Vehicle / 249-Heavy equipment / 433-Construction / \n", "3-Concert / 16-Performance art / \n", "8-Football / \n", "5-Dance / 259-Fire / 350-Circus / 788-Juggling / \n", "0-Games / 1-Vehicle / 156-Agriculture / 174-Farm / 715-Farming Simulator / \n", "3-Concert / 7-Musician / 13-Musical ensemble / 14-Guitar / 30-Drummer / 39-Drums / 44-Drums / \n", "0-Games / 2-Video game / 19-PC game / 33-Weapon / 43-Call of Duty / 96-Soldier / 134-Call of Duty: Black Ops / \n", "1-Vehicle / 4-Car / 40-Road / 94-Dashcam / 212-Highway / \n", "0-Games / 34-Action-adventure game / 55-Video game console / 118-Grand Theft Auto V / 127-Xbox 360 / 386-Xbox / \n", "10-Animal / 12-Food / 26-Cooking / 32-Recipe / 52-Dish / 58-Cuisine / 91-Fish / 893-Fish as food / \n", "5-Dance / \n", "3-Concert / 7-Musician / 13-Musical ensemble / 38-Orchestra / 46-Choir / \n", "0-Games / 489-Board game / 3863-Temple Run / \n", "8-Football / \n", "9-Music video / \n", "319-Robot / 1885-Robotics / \n", "410-Teacher / \n", "23-Mobile phone / 29-Smartphone / 37-Gadget / 72-iPhone / 101-Telephone / \n", "10-Animal / 12-Food / 91-Fish / 3103-Green tea / \n", "1-Vehicle / 4-Car / 299-Volkswagen Passenger Cars / 531-Hatchback / 668-Volkswagen Golf / 2214-Multi-valve / 2286-Grand Tourer Injection / \n", "0-Games / 2-Video game / 33-Weapon / 43-Call of Duty / 302-Call of Duty 4: Modern Warfare / \n", "46-Choir / 716-Coral / \n", "67-Cycling / 69-Bicycle / 1067-Lock / 2516-Lock picking / \n", "0-Games / 2-Video game / 34-Action-adventure game / 118-Grand Theft Auto V / \n", "9-Music video / \n", "0-Games / 2-Video game / 6-Animation / 111-PlayStation 3 / 2984-.hack / \n", "3-Concert / 16-Performance art / \n", "31-Disc jockey / 322-Mixtape / \n", "0-Games / 8-Football / 73-Ball / \n", "131-Computer / 141-Microsoft Windows / \n", "0-Games / 8-Football / 65-Kick / \n", "0-Games / 2-Video game / 34-Action-adventure game / 127-Xbox 360 / 2205-Bully / \n", "310-The Sims / 543-The Sims 2 / \n", "88-Machine / 400-Woodturning / \n", "1-Vehicle / 27-Motorcycle / 105-Motorcycling / 207-Exhaust system / \n", "18-Outdoor recreation / 59-Winter sport / 84-Snow / 177-Skiing / 305-Ski / \n", "5-Dance / 115-Ballet / \n", "12-Food / 184-Vegetable / 348-Fruit / \n", "0-Games / 8-Football / 73-Ball / \n", "0-Games / 8-Football / \n", "110-Album / \n", "3-Concert / 5-Dance / 16-Performance art / 95-Talent show / \n", "0-Games / 2-Video game / 35-Minecraft / \n", "1-Vehicle / 4-Car / 11-Motorsport / 17-Racing / 41-Sports car / 135-Drifting / 231-Supercar / 711-Gran Turismo / 953-Gran Turismo 5 / \n", "0-Games / 21-Stadium / 54-Highlight film / 79-American football / \n", "7-Musician / 36-Piano / 53-Keyboard / 104-Pianist / \n", "365-Fireworks / \n", "3-Concert / 7-Musician / 13-Musical ensemble / \n", "59-Winter sport / 81-Athlete / 112-Ice skating / 142-Hockey / 262-Ice / 312-Ice rink / 833-Ice skate / 1973-Hockey puck / \n", "1-Vehicle / 11-Motorsport / 27-Motorcycle / 447-Sand / 751-Desert / 975-Dune / 2875-Quadracycle / \n", "6-Animation / 15-Cartoon / 183-Manga / 701-Beyblade / \n", "9-Music video / 13-Musical ensemble / 14-Guitar / 39-Drums / 44-Drums / \n", "75-Wrestling / 435-Action figure / 1441-Mattel / \n", "10-Animal / 227-Water / \n", "1-Vehicle / 4-Car / 11-Motorsport / 27-Motorcycle / 105-Motorcycling / \n", "0-Games / 287-Human swimming / \n", "1-Vehicle / 8-Football / \n", "3-Concert / 7-Musician / 13-Musical ensemble / 38-Orchestra / 46-Choir / \n", "1107-United States Navy / \n", "8-Football / 21-Stadium / 77-Arena / \n", "5-Dance / 333-Nike; Inc. / \n", "23-Mobile phone / 29-Smartphone / 125-Television / \n", "955-Police officer / \n", "223-Weather / \n", "93-Comedy / \n", "38-Orchestra / \n", "0-Games / 2-Video game / \n", "1-Vehicle / 4-Car / \n", "3-Concert / 5-Dance / 98-Festival / \n", "22-Nature / 67-Cycling / 69-Bicycle / 179-Kitchen / 197-Trail / 213-Forest / 220-Mountain bike / 284-Mountain biking / \n", "1-Vehicle / 4-Car / 64-Transport / 74-Truck / 259-Fire / 371-Firefighter / 1633-Fire station / \n", "1-Vehicle / 18-Outdoor recreation / \n", "3-Concert / 7-Musician / \n", "0-Games / 2-Video game / 4468-Breath of Fire / \n", "1-Vehicle / 50-Aircraft / 82-Airplane / 92-Radio-controlled model / 149-Model aircraft / 160-Radio-controlled aircraft / \n", "2-Video game / 1028-Devil May Cry / 2336-Devil May Cry 3: Dante's Awakening / \n", "28-Fashion / 45-Cosmetics / 56-Hair / 106-Hairstyle / \n", "47-Personal computer / 145-Tablet computer / 189-iPad / \n", "2614-Lettuce / \n", "1-Vehicle / 4-Car / 11-Motorsport / 234-Ford / 406-Chevrolet / 473-Ford Mustang / 509-Classic car / 761-Chevrolet Camaro / 1148-Hot rod / 2061-Chevrolet Chevelle / 2078-Dodge Challenger / \n", "122-Weight training / 214-Muscle / \n", "3-Concert / 7-Musician / 16-Performance art / \n", "648-Clown / \n", "0-Games / 6-Animation / 25-Toy / \n", "0-Games / 60-Basketball / 168-Basketball moves / 255-Slam dunk / \n", "14-Guitar / \n", "3-Concert / 7-Musician / 13-Musical ensemble / 24-String instrument / 150-Violin / 401-Cello / 450-Viola / 545-Quartet (ensemble) / 639-Fiddle / 1002-String quartet / \n", "0-Games / 23-Mobile phone / 29-Smartphone / 492-Floor / \n", "0-Games / 2-Video game / 6-Animation / 15-Cartoon / 584-Super Mario Bros. / \n", "298-Television advertisement / 337-Advertising / \n", "376-Flute / 1326-Recorder / \n", "12-Food / \n", "3-Concert / 31-Disc jockey / 98-Festival / \n", "0-Games / 2-Video game / 60-Basketball / 90-Sports game / 1262-NBA 2K14 / \n", "0-Games / 54-Highlight film / 60-Basketball / 77-Arena / 81-Athlete / 168-Basketball moves / 255-Slam dunk / \n", "3-Concert / 30-Drummer / \n", "53-Keyboard / 1909-Analog synthesizer / \n", "1-Vehicle / 4-Car / 11-Motorsport / 270-Sedan / 680-Subaru / 1286-Subaru / \n", "6-Animation / 483-Costume / 596-Monster Hunter / \n", "0-Games / 2-Video game / 903-Mass Effect / \n", "1-Vehicle / 27-Motorcycle / 3410-Paper model / \n", "20-Trailer / \n", "0-Games / 455-Unidentified flying object / 1668-Claw crane / \n", "0-Games / 756-Tibia / \n", "0-Games / 781-Pachinko / \n", "0-Games / 2-Video game / 19-PC game / 33-Weapon / 96-Soldier / 192-Battlefield / 383-Battlefield 4 / \n", "8-Football / 76-Ball / \n", "5-Dance / 444-Family / \n", "1-Vehicle / 4-Car / 259-Fire / \n", "3-Concert / \n", "12-Food / 26-Cooking / 32-Recipe / 52-Dish / 58-Cuisine / 232-Cake / \n", "3-Concert / 7-Musician / 13-Musical ensemble / \n", "0-Games / 8-Football / 21-Stadium / 54-Highlight film / 81-Athlete / \n", "163-Running / \n", "22-Nature / 136-Gardening / 137-Tree / 642-Leaf / 657-Autumn / \n", "3-Concert / 16-Performance art / \n", "313-Book / 1640-Library / \n", "3-Concert / \n", "1057-Embroidery / 1451-Ribbon / \n", "5-Dance / 46-Choir / 49-School / \n", "18-Outdoor recreation / 83-Skateboarding / 108-Skateboard / \n", "0-Games / 6-Animation / 15-Cartoon / 89-Comics / 185-Dragon Ball / 244-Goku / 418-Gohan / 869-Dragon Ball Z: Budokai Tenkaichi / \n", "10-Animal / 1591-Leopard / 3055-Gecko / \n", "356-Star / 419-Earth / 712-Outer space / 1702-Telescope / 2643-Asteroid / 3511-Meteoroid / 3942-Meteor shower / \n", "75-Wrestling / 85-Combat / 3305-Tag team / 3912-Face / \n", "412-Monster / \n", "3-Concert / 7-Musician / \n", "640-Bottle / 1510-Liquid / 1544-Laboratory / \n", "3-Concert / \n", "1-Vehicle / 4-Car / 11-Motorsport / 17-Racing / 40-Road / 41-Sports car / \n", "6-Animation / 15-Cartoon / 1527-Pretty Cure / \n", "47-Personal computer / 131-Computer / 228-Laptop / \n", "42-Fishing / 370-Origami / 510-Fishing lure / \n", "12-Food / 58-Cuisine / \n", "36-Piano / 110-Album / 585-Lawn / \n", "0-Games / 2-Video game / 20-Trailer / 55-Video game console / 111-PlayStation 3 / \n", "0-Games / 1-Vehicle / 2-Video game / 4-Car / 55-Video game console / 2249-Driver / \n", "3-Concert / 16-Performance art / \n", "0-Games / 127-Xbox 360 / 386-Xbox / \n", "0-Games / 60-Basketball / 4355-Netball / \n", "425-City / 1528-Euro / \n", "3-Concert / 7-Musician / 13-Musical ensemble / 46-Choir / \n", "0-Games / 2-Video game / 116-Pokémon / 186-Pokémon / 1387-Pokémon Battle Revolution / \n", "1-Vehicle / 18-Outdoor recreation / 27-Motorcycle / 67-Cycling / 69-Bicycle / 327-Mud / \n", "7-Musician / 14-Guitar / 24-String instrument / 63-Acoustic guitar / 376-Flute / \n", "0-Games / 2-Video game / 51-Strategy video game / \n", "25-Toy / 4670-Footbag / \n", "9-Music video / \n", "96-Soldier / 205-Newscaster / 444-Family / \n", "0-Games / 2-Video game / 432-Kingdom Hearts / 665-Kingdom Hearts / 836-Sora / 941-Kingdom Hearts II / 1730-Roxas / \n", "47-Personal computer / 131-Computer / 228-Laptop / 309-Computer hardware / \n", "10-Animal / 537-Zoo / 571-Bear / 1160-Giant panda / \n", "0-Games / 60-Basketball / 75-Wrestling / 595-Champion / \n", "20-Trailer / \n", "0-Games / 59-Winter sport / 112-Ice skating / 142-Hockey / 262-Ice / 312-Ice rink / \n", "365-Fireworks / \n", "3-Concert / \n", "3-Concert / \n", "110-Album / \n", "1232-Perfect World / 2394-Perfect World / \n", "12-Food / 2926-Barista / \n", "5-Dance / \n", "9-Music video / \n", "3-Concert / 5-Dance / 16-Performance art / 98-Festival / \n", "122-Weight training / 532-Barbell / 1250-Biceps / \n", "1-Vehicle / 4-Car / 158-Tire / 2742-Mini 4WD / \n", "1-Vehicle / 50-Aircraft / 129-Aviation / 151-Landing / 157-Jet aircraft / 175-Airliner / \n", "1-Vehicle / 4-Car / 449-Honda / 552-Honda Civic / \n", "0-Games / 1-Vehicle / 2-Video game / 4-Car / 34-Action-adventure game / 118-Grand Theft Auto V / \n", "31-Disc jockey / \n", "1-Vehicle / 4-Car / 11-Motorsport / 84-Snow / 120-Winter / 135-Drifting / 344-Coupé / 683-Renault / 2455-Renault Mégane / \n", "2093-Coating / \n", "5-Dance / 396-One Piece / \n", "6-Animation / \n", "0-Games / 2-Video game / 637-Five Nights at Freddy's / \n", "110-Album / \n", "769-Red Bull / \n", "3-Concert / \n", "412-Monster / \n", "78-Wedding / \n", "0-Games / 33-Weapon / 229-Halo / 470-Halo 3 / \n", "31-Disc jockey / \n", "10-Animal / 80-Horse / 351-Jumping / \n", "0-Games / 2-Video game / 34-Action-adventure game / 84-Snow / 118-Grand Theft Auto V / 176-PlayStation 4 / \n", "121-Photography / 512-Printing / 525-Printer / 2634-Inkjet printing / \n", "12-Food / 26-Cooking / 389-Restaurant / 504-Web page / \n", "3-Concert / \n", "3-Concert / 7-Musician / 16-Performance art / 465-Microphone / \n", "5-Dance / 16-Performance art / \n", "6-Animation / 15-Cartoon / \n", "3-Concert / \n", "0-Games / 489-Board game / \n", "1-Vehicle / 4-Car / 11-Motorsport / 17-Racing / 135-Drifting / \n", "9-Music video / \n", "23-Mobile phone / 29-Smartphone / 37-Gadget / 47-Personal computer / 145-Tablet computer / 621-Asus / \n", "3-Concert / 16-Performance art / \n", "1-Vehicle / 4-Car / 17-Racing / 67-Cycling / 69-Bicycle / 92-Radio-controlled model / 202-Radio-controlled car / \n", "3-Concert / 5-Dance / \n", "22-Nature / \n", "12-Food / 25-Toy / 295-Chocolate / 582-Candy / 1408-Snack / \n", "148-Painting / 998-Landscape / 3030-Landscape painting / \n", "1-Vehicle / 4-Car / 1078-Installation art / 1119-Sensor / 1285-Parking / \n", "3-Concert / \n", "45-Cosmetics / 173-Eye shadow / 193-Eye liner / 195-Mascara / 196-Eye / 218-Eyelash / \n", "18-Outdoor recreation / 59-Winter sport / 112-Ice skating / 316-Figure skating / 632-Ice dancing / \n", "18-Outdoor recreation / 96-Soldier / \n", "36-Piano / 53-Keyboard / 107-Musical keyboard / 592-Synthesia / \n", "21-Stadium / 191-Cheerleading / 306-High school / 1662-Ōendan / \n", "365-Fireworks / 1293-RollerCoaster Tycoon 3 / \n", "9-Music video / \n", "1-Vehicle / 4-Car / 74-Truck / 945-MINI Cooper / \n", "1-Vehicle / 4-Car / 84-Snow / 120-Winter / 198-Off-road vehicle / 406-Chevrolet / \n", "0-Games / 2-Video game / 6-Animation / \n", "3-Concert / 16-Performance art / 263-Light / \n", "2-Video game / 5-Dance / 199-Wii / \n", "0-Games / 1852-Final Fantasy XI / \n", "2-Video game / 47-Personal computer / 131-Computer / 309-Computer hardware / 602-Central processing unit / \n", "0-Games / 35-Minecraft / \n", "117-Boxing / \n", "0-Games / 240-Arcade game / 535-Slot machine / \n", "46-Choir / 318-Church / 397-Organ / 870-Chapel / \n", "0-Games / 530-Cue sports / \n", "12-Food / 25-Toy / 200-Doll / 232-Cake / 1365-Fondant icing / \n", "102-Building / \n", "3-Concert / 7-Musician / 14-Guitar / \n", "3-Concert / 1684-Hillsong Church / \n", "147-River / 164-Beach / 722-Waterfall / \n", "6-Animation / 15-Cartoon / 20-Trailer / 126-The Walt Disney Company / \n", "56-Hair / 106-Hairstyle / \n", "325-Vampire / \n", "36-Piano / 53-Keyboard / 104-Pianist / 107-Musical keyboard / 265-Electronic keyboard / \n", "0-Games / 2-Video game / 511-Mortal Kombat / 1291-Mortal Kombat / 4684-Kung Lao / \n", "28-Fashion / 45-Cosmetics / 785-Jeans / \n", "6-Animation / 2110-D.Gray-man / \n", "110-Album / \n", "18-Outdoor recreation / 59-Winter sport / 112-Ice skating / 316-Figure skating / \n", "15-Cartoon / 477-Transformers / 3398-Transformers: Generation 1 / 3795-Starscream / \n", "102-Building / 113-House / 310-The Sims / 328-Simulation video game / 564-The Sims 3 / \n", "0-Games / 19-PC game / 51-Strategy video game / 165-League of Legends / \n", "170-Home improvement / 276-Room / \n", "113-House / \n", "219-Accordion / \n", "12-Food / 2080-Cup / \n", "61-Art / 402-Jewellery / 829-Bead / 859-Beadwork / \n", "9-Music video / \n", "2-Video game / 33-Weapon / 55-Video game console / 2034-Dreamcast / \n", "516-Headphones / \n", "10-Animal / 424-Chicken / \n", "6-Animation / 15-Cartoon / 183-Manga / 801-Inuyasha / \n", "125-Television / 1348-Panasonic Corporation / 3984-Plasma display / \n", "18-Outdoor recreation / 83-Skateboarding / 108-Skateboard / \n", "419-Earth / 448-Planet / 593-Moon / 616-The Bible / 1480-Astronaut / 1686-Orbit / \n", "9-Music video / \n", "20-Trailer / \n", "3-Concert / 7-Musician / 13-Musical ensemble / 16-Performance art / 46-Choir / \n", "12-Food / \n", "3-Concert / \n", "1-Vehicle / 4-Car / 18-Outdoor recreation / 74-Truck / 167-Four-wheel drive / 198-Off-road vehicle / 277-Sport utility vehicle / 375-Pickup truck / 403-Toyota / 1840-Toyota Hilux / \n", "5-Dance / 16-Performance art / 115-Ballet / \n", "1-Vehicle / \n", "121-Photography / 155-Camera / 566-Digital camera / 629-Camera lens / 721-Digital SLR / 3258-Leica Camera / \n", "1-Vehicle / 4-Car / \n", "0-Games / 8-Football / 21-Stadium / 73-Ball / \n", "110-Album / \n", "121-Photography / 155-Camera / 2248-Zoom lens / \n", "93-Comedy / \n", "2-Video game / 6-Animation / 15-Cartoon / 320-Super Smash Bros. / 398-The Legend of Zelda / 692-Link / 697-Super Smash Bros. for Nintendo 3DS and Wii U / \n", "31-Disc jockey / \n", "31-Disc jockey / \n", "15-Cartoon / 116-Pokémon / 186-Pokémon / \n", "6-Animation / 20-Trailer / 264-Star Wars / 714-Jedi / 717-Darth Vader / 2576-Clone trooper / \n", "0-Games / 1721-Tetris / \n", "12-Food / 26-Cooking / 32-Recipe / 122-Weight training / 542-Potato / 780-Fat / 1283-Pancake / 2575-Sweet potato / 2995-Blueberry / \n", "3-Concert / \n", "1-Vehicle / 4-Car / 299-Volkswagen Passenger Cars / 1498-Key / \n", "2-Video game / 229-Halo / 470-Halo 3 / 644-Halo: Reach / 866-Halo 4 / 2889-343 Industries / \n", "3-Concert / \n", "5-Dance / 13-Musical ensemble / 16-Performance art / \n", "3-Concert / 5-Dance / 16-Performance art / 889-Swing / \n", "1-Vehicle / 490-Rain / \n", "7-Musician / 14-Guitar / 24-String instrument / 376-Flute / \n", "6-Animation / 15-Cartoon / \n", "12-Food / 720-Wine / \n", "12-Food / \n", "0-Games / 2-Video game / \n", "30-Drummer / 39-Drums / 44-Drums / 123-Naruto / 128-Cymbal / 146-Snare drum / \n", "237-Final Fantasy / 2204-Final Fantasy VI / \n", "66-Bollywood / \n", "10-Animal / 22-Nature / 48-Pet / 71-Dog / \n", "22-Nature / 213-Forest / 423-Coast / \n", "3-Concert / 5-Dance / 16-Performance art / \n", "9-Music video / \n", "5-Dance / \n", "42-Fishing / 1306-Carp fishing / \n", "18-Outdoor recreation / 4695-Strikeout / \n", "47-Personal computer / 1045-Computer case / \n", "0-Games / 88-Machine / 240-Arcade game / 535-Slot machine / \n", "20-Trailer / 3395-Asterix / \n", "35-Minecraft / \n", "10-Animal / 80-Horse / \n", "3-Concert / 7-Musician / 13-Musical ensemble / 24-String instrument / 38-Orchestra / 150-Violin / 401-Cello / 450-Viola / \n", "0-Games / 51-Strategy video game / 373-Clash of Clans / 1001-Gemstone / \n", "23-Mobile phone / 29-Smartphone / 37-Gadget / 101-Telephone / 141-Microsoft Windows / \n", "6-Animation / 15-Cartoon / \n", "98-Festival / \n", "1-Vehicle / 4-Car / 11-Motorsport / 17-Racing / 41-Sports car / 57-Race track / 1286-Subaru / 3679-Super GT / \n", "60-Basketball / \n", "5-Dance / 16-Performance art / 678-Latin dance / \n", "6-Animation / 15-Cartoon / \n", "293-Slide show / \n", "0-Games / 1604-Satan / \n", "3-Concert / 16-Performance art / \n", "6-Animation / 708-Mermaid / \n", "1-Vehicle / 4-Car / 74-Truck / 158-Tire / \n", "3-Concert / \n", "6-Animation / \n", "2952-Ventriloquism / \n", "3-Concert / 13-Musical ensemble / 597-Soup / \n", "8-Football / \n", "2-Video game / 47-Personal computer / 908-Video card / \n", "1-Vehicle / 62-Train / 64-Transport / \n", "18-Outdoor recreation / 83-Skateboarding / 108-Skateboard / 221-Skatepark / 1781-Ollie / \n", "3-Concert / 7-Musician / 13-Musical ensemble / 46-Choir / \n", "10-Animal / 18-Outdoor recreation / 22-Nature / 42-Fishing / 91-Fish / 1306-Carp fishing / \n", "3-Concert / \n", "522-Helmet / 748-Iron Man / \n", "8-Football / \n", "162-Medicine / \n", "0-Games / 2-Video game / 51-Strategy video game / 165-League of Legends / \n", "260-Prayer / \n", "466-Gift / \n", "1-Vehicle / 4-Car / 114-Engine / 154-Wheel / 403-Toyota / 2256-Bearing / 3399-Timing belt / 3685-Pulley / \n", "7-Musician / 13-Musical ensemble / 38-Orchestra / \n", "10-Animal / 537-Zoo / 970-Monkey / \n", "1-Vehicle / 62-Train / 64-Transport / 103-Rail transport / 119-Locomotive / 143-Railroad car / 152-Track / \n", "0-Games / 2-Video game / 19-PC game / 43-Call of Duty / 96-Soldier / 194-Call of Duty: Modern Warfare 2 / 276-Room / 1892-Dormitory / 2879-Medal of Honor: Warfighter / \n", "2-Video game / 237-Final Fantasy / 1131-Final Fantasy XIII / \n", "10-Animal / 48-Pet / 71-Dog / 1278-Labrador Retriever / \n", "66-Bollywood / 113-House / 339-Architecture / 888-Living room / \n", "2-Video game / 55-Video game console / \n", "6-Animation / 290-Pony / 343-My Little Pony / \n", "5-Dance / 342-Windows Media Video / \n", "5-Dance / 98-Festival / 250-Parade / 253-Carnival / \n", "429-Website / 819-World Wide Web / \n", "1584-Kettlebell / \n", "3-Concert / \n", "8-Football / \n", "217-Hunting / 882-Pen / 1873-Crossbow / \n", "28-Fashion / \n", "66-Bollywood / \n", "3-Concert / 16-Performance art / 30-Drummer / \n", "5-Dance / 8-Football / 21-Stadium / 4106-Moonwalk / \n", "3-Concert / 13-Musical ensemble / \n", "5-Dance / 16-Performance art / 95-Talent show / \n", "2996-Caterpillar / \n", "10-Animal / 48-Pet / 1147-Hamster / 1234-Mouse / 1716-Rat / \n", "31-Disc jockey / \n", "10-Animal / 18-Outdoor recreation / 22-Nature / 42-Fishing / 275-Diving / 287-Human swimming / 288-Underwater / 1135-Dolphin / \n", "0-Games / 2-Video game / 35-Minecraft / \n", "9-Music video / \n", "415-Sewing / \n", "10-Animal / 80-Horse / 247-Stallion / 438-Mare / \n", "1222-Fountain / \n", "0-Games / 2-Video game / 55-Video game console / 1118-Kirby / 1133-Kirby / \n", "1-Vehicle / 4-Car / \n", "5-Dance / 9-Music video / 31-Disc jockey / \n", "1-Vehicle / 4-Car / 395-Camping / \n", "1-Vehicle / 4-Car / \n", "783-Zee TV / \n", "0-Games / 2-Video game / 243-Sonic the Hedgehog / 251-Sonic the Hedgehog / 1860-Sonic Adventure / 1890-Dr. Eggman / 2034-Dreamcast / \n", "0-Games / 35-Minecraft / \n", "0-Games / 54-Highlight film / 60-Basketball / 168-Basketball moves / \n", "1-Vehicle / 87-Boat / \n", "3-Concert / \n", "1-Vehicle / 4-Car / 114-Engine / 332-Mercedes-Benz / 854-Diesel engine / \n", "10-Animal / 571-Bear / \n", "1-Vehicle / 50-Aircraft / 82-Airplane / 92-Radio-controlled model / 149-Model aircraft / 160-Radio-controlled aircraft / \n", "0-Games / 8-Football / 65-Kick / \n", "25-Toy / 113-House / 200-Doll / 323-Toddler / 652-Bathroom / 1113-Toilet / 2153-Sink / 2861-Baby Alive / \n", "0-Games / 2-Video game / 1251-The Witcher / 2497-The Witcher 2: Assassins of Kings / \n", "3-Concert / 16-Performance art / \n", "7-Musician / 13-Musical ensemble / 21-Stadium / 79-American football / 203-Marching band / 209-University / 250-Parade / \n", "1-Vehicle / 50-Aircraft / 139-Airport / 151-Landing / 175-Airliner / 188-Airline / 254-Runway / 472-Boeing 737 / 1736-Microsoft Flight Simulator / \n", "1-Vehicle / 4-Car / 11-Motorsport / 27-Motorcycle / \n", "36-Piano / \n", "8-Football / \n", "0-Games / 2-Video game / 34-Action-adventure game / 118-Grand Theft Auto V / 176-PlayStation 4 / 297-Grand Theft Auto IV / 317-PlayStation / \n", "88-Machine / 521-Metal / 943-Stove / 2331-Melting / \n", "1-Vehicle / 4-Car / 11-Motorsport / 40-Road / 212-Highway / 947-Muffler / \n", "0-Games / 2-Video game / 19-PC game / 55-Video game console / 292-Handheld game console / 320-Super Smash Bros. / \n", "1-Vehicle / 4-Car / 158-Tire / 299-Volkswagen Passenger Cars / 668-Volkswagen Golf / \n", "12-Food / 26-Cooking / 32-Recipe / 52-Dish / 58-Cuisine / \n", "20-Trailer / \n", "1090-Canoe / 1305-Canoeing / 2273-Polo / \n", "2848-12-inch single / \n", "3-Concert / 7-Musician / 13-Musical ensemble / 30-Drummer / 44-Drums / \n", "0-Games / 2-Video game / 111-PlayStation 3 / 127-Xbox 360 / 243-Sonic the Hedgehog / 251-Sonic the Hedgehog / 1512-Sonic Generations / \n", "45-Cosmetics / 56-Hair / 173-Eye shadow / 195-Mascara / 291-Rouge / \n", "0-Games / 35-Minecraft / \n", "75-Wrestling / \n", "28-Fashion / \n", "1-Vehicle / 27-Motorcycle / 493-Rim / \n", "0-Games / 326-Metin2 / \n", "3-Concert / 46-Choir / \n", "5-Dance / \n", "93-Comedy / \n", "9-Music video / \n", "22-Nature / 42-Fishing / 233-Ocean / 1075-Whale / 1135-Dolphin / 3797-Humpback whale / \n", "1-Vehicle / 11-Motorsport / 124-Tractor / 174-Farm / 249-Heavy equipment / 252-Tractor pulling / 817-Bulldozer / \n", "7-Musician / 9-Music video / 14-Guitar / \n", "840-Museum / \n", "0-Games / 2-Video game / 229-Halo / 470-Halo 3 / 644-Halo: Reach / \n", "4076-Ultima / \n", "28-Fashion / 56-Hair / 106-Hairstyle / 1359-Hair iron / \n", "61-Art / 148-Painting / 269-Comic book / \n", "0-Games / 1-Vehicle / 2-Video game / 4-Car / 34-Action-adventure game / 201-Grand Theft Auto: San Andreas / 387-Spider-Man / \n", "3-Concert / 7-Musician / 13-Musical ensemble / 38-Orchestra / 46-Choir / \n", "6-Animation / \n", "12-Food / 26-Cooking / 32-Recipe / 52-Dish / 226-Dessert / 662-Pizza / \n", "5-Dance / \n", "0-Games / 79-American football / \n", "10-Animal / 48-Pet / 71-Dog / 182-Puppy / \n", "1-Vehicle / 62-Train / 64-Transport / 103-Rail transport / 119-Locomotive / 143-Railroad car / 152-Track / 161-Train station / \n", "1-Vehicle / 62-Train / 64-Transport / 103-Rail transport / 119-Locomotive / 143-Railroad car / 428-Rail transport modelling / \n", "6-Animation / 15-Cartoon / \n", "75-Wrestling / 85-Combat / \n", "3-Concert / 7-Musician / 4089-Maracas / \n", "6-Animation / 25-Toy / 1850-Playmobil / \n", "9-Music video / \n", "2-Video game / 20-Trailer / \n", "5-Dance / 66-Bollywood / \n", "23-Mobile phone / 29-Smartphone / 101-Telephone / \n", "14-Guitar / 24-String instrument / 63-Acoustic guitar / \n", "85-Combat / \n", "1-Vehicle / 4-Car / 11-Motorsport / 135-Drifting / \n", "130-Gymnastics / \n", "20-Trailer / \n", "12-Food / 2329-Felt / \n", "20-Trailer / \n", "10-Animal / 48-Pet / 71-Dog / \n", "0-Games / 2-Video game / 1566-Doom / 2352-Doom II: Hell on Earth / \n", "0-Games / 2-Video game / 20-Trailer / 43-Call of Duty / 422-Call of Duty: Ghosts / \n", "10-Animal / 86-Plant / 2284-Bamboo / \n", "0-Games / 35-Minecraft / 102-Building / 113-House / \n", "3-Concert / 7-Musician / 13-Musical ensemble / 14-Guitar / 30-Drummer / \n", "0-Games / 2-Video game / 51-Strategy video game / \n", "0-Games / 2-Video game / 910-Dragon Quest / \n", "0-Games / 6-Animation / 460-Yu-Gi-Oh! Trading Card Game / \n", "356-Star / 502-Sky / \n", "18-Outdoor recreation / 83-Skateboarding / 108-Skateboard / 221-Skatepark / \n", "122-Weight training / 162-Medicine / 214-Muscle / \n", "18-Outdoor recreation / 59-Winter sport / 112-Ice skating / 262-Ice / 316-Figure skating / \n", "3-Concert / 7-Musician / 16-Performance art / \n", "10-Animal / 846-Squirrel / \n", "219-Accordion / 655-Diatonic button accordion / \n", "1-Vehicle / 4-Car / 40-Road / 70-Driving / \n", "240-Arcade game / \n", "7-Musician / 13-Musical ensemble / 46-Choir / \n", "3-Concert / 5-Dance / 16-Performance art / 99-Christmas / \n", "56-Hair / 106-Hairstyle / \n", "6-Animation / 242-Paper / 295-Chocolate / 466-Gift / \n", "0-Games / 8-Football / 21-Stadium / 73-Ball / \n", "45-Cosmetics / 49-School / \n", "3-Concert / 13-Musical ensemble / \n", "7-Musician / 14-Guitar / 24-String instrument / 63-Acoustic guitar / 1971-Epiphone / \n", "14-Guitar / 24-String instrument / \n", "1-Vehicle / 4-Car / \n", "764-The King of Fighters / \n", "10-Animal / 48-Pet / 71-Dog / 182-Puppy / \n", "1-Vehicle / 11-Motorsport / 27-Motorcycle / 105-Motorcycling / 336-Scooter / \n", "0-Games / 75-Wrestling / \n", "102-Building / \n", "28-Fashion / 153-Dress / 495-Runway / \n", "61-Art / 97-Drawing / 538-Illustration / \n", "110-Album / \n", "164-Beach / \n", "1751-Image scanner / \n", "425-City / \n", "227-Water / 827-Pump / 2635-Water well / \n", "0-Games / 23-Mobile phone / 29-Smartphone / 178-Samsung Galaxy / \n", "5-Dance / 16-Performance art / 769-Red Bull / \n", "14-Guitar / 63-Acoustic guitar / \n", "0-Games / 2-Video game / 19-PC game / 35-Minecraft / \n", "0-Games / 47-Personal computer / 141-Microsoft Windows / 228-Laptop / \n", "1-Vehicle / 4-Car / 11-Motorsport / \n", "3-Concert / 7-Musician / 13-Musical ensemble / 24-String instrument / 38-Orchestra / 150-Violin / 401-Cello / 450-Viola / 545-Quartet (ensemble) / 726-Clarinet / 1002-String quartet / \n", "0-Games / 2-Video game / 19-PC game / 47-Personal computer / \n", "1-Vehicle / 18-Outdoor recreation / 87-Boat / 366-Dragon / 949-Rowing / 2678-Dragon boat / \n", "0-Games / 2-Video game / 6-Animation / 10-Animal / 290-Pony / 343-My Little Pony / \n", "3-Concert / 7-Musician / 13-Musical ensemble / 14-Guitar / 24-String instrument / 1669-Vodka / \n", "14-Guitar / 2120-Tears / \n", "1-Vehicle / 4-Car / 17-Racing / \n", "3-Concert / 5-Dance / 16-Performance art / 95-Talent show / \n", "3-Concert / 7-Musician / 13-Musical ensemble / 38-Orchestra / \n", "0-Games / 59-Winter sport / 81-Athlete / 112-Ice skating / 142-Hockey / 312-Ice rink / \n", "0-Games / 2-Video game / 6-Animation / 15-Cartoon / 19-PC game / 123-Naruto / 285-Sasuke Uchiha / 591-Naruto: Ultimate Ninja / \n", "9-Music video / \n", "36-Piano / \n", "3-Concert / 16-Performance art / 46-Choir / \n", "4400-Rain gutter / \n", "23-Mobile phone / 29-Smartphone / 37-Gadget / 72-iPhone / 452-Watch / 1260-Smartwatch / 3727-Android Wear / \n", "23-Mobile phone / 29-Smartphone / 345-Loudspeaker / \n", "0-Games / 2-Video game / 237-Final Fantasy / 730-Final Fantasy VII / \n", "9-Music video / \n", "0-Games / 211-RuneScape / \n", "0-Games / 2-Video game / 243-Sonic the Hedgehog / 251-Sonic the Hedgehog / 2869-Bomberman / \n", "359-Harry Potter (Literary Series) / \n", "0-Games / 8-Football / 54-Highlight film / 79-American football / \n", "5-Dance / \n", "31-Disc jockey / \n", "0-Games / 8-Football / 54-Highlight film / 79-American football / \n", "3-Concert / \n", "6-Animation / 15-Cartoon / 440-Lion / \n", "0-Games / 1182-Softball / \n", "133-Nightclub / 613-Bar / \n", "455-Unidentified flying object / \n", "3-Concert / 16-Performance art / \n", "1-Vehicle / 4-Car / 27-Motorcycle / 40-Road / 105-Motorcycling / \n", "0-Games / 2-Video game / 2944-La Tale / \n", "1222-Fountain / \n", "107-Musical keyboard / \n", "1035-Bank / 1528-Euro / \n", "1-Vehicle / 4-Car / \n", "1-Vehicle / \n", "2531-Dojo / \n", "0-Games / 2-Video game / 8-Football / 90-Sports game / 180-FIFA 15 / \n", "1-Vehicle / 4-Car / 11-Motorsport / 17-Racing / \n", "18-Outdoor recreation / 22-Nature / 147-River / 722-Waterfall / \n", "1-Vehicle / 4-Car / 114-Engine / 1575-Clutch / \n", "2-Video game / \n", "22-Nature / 36-Piano / \n", "93-Comedy / 341-Sketch comedy / 364-Sitcom / \n", "5-Dance / \n", "1-Vehicle / 11-Motorsport / 17-Racing / 27-Motorcycle / 105-Motorcycling / 197-Trail / \n", "0-Games / 90-Sports game / 180-FIFA 15 / 848-FIFA 12 / \n", "9-Music video / \n", "122-Weight training / 214-Muscle / \n", "0-Games / 2-Video game / 432-Kingdom Hearts / 4250-Kingdom Hearts 3D: Dream Drop Distance / \n", "0-Games / 8-Football / 21-Stadium / 65-Kick / 73-Ball / 76-Ball / \n", "66-Bollywood / 905-Flash Video / \n", "3-Concert / 7-Musician / \n", "2-Video game / 89-Comics / 2172-Deadpool / \n", "5-Dance / 98-Festival / 203-Marching band / \n", "9-Music video / \n", "365-Fireworks / 942-Silkroad Online / \n", "10-Animal / 18-Outdoor recreation / 217-Hunting / 3022-Hare / \n", "0-Games / 60-Basketball / \n", "2042-Nut / \n", "466-Gift / 627-Clock / \n", "102-Building / 113-House / \n", "3-Concert / \n", "9-Music video / \n", "3063-Friendship bracelet / \n", "1572-Husband / \n", "9-Music video / 22-Nature / 283-Rock / \n", "3-Concert / 7-Musician / 30-Drummer / 39-Drums / 44-Drums / \n", "1-Vehicle / 62-Train / \n", "14-Guitar / 24-String instrument / 100-Electric guitar / 1259-Fender Telecaster / \n", "0-Games / 2-Video game / 19-PC game / 43-Call of Duty / 47-Personal computer / 127-Xbox 360 / 181-Xbox / 383-Battlefield 4 / 677-Game controller / 704-Call of Duty: World at War / 1248-Call of Duty: Zombies / 1872-Xbox 360 Controller / \n", "12-Food / 367-Drink / 613-Bar / 904-Cocktail / 1065-Bartender / \n", "9-Music video / \n", "8-Football / 924-Trophy / \n", "1-Vehicle / 11-Motorsport / 27-Motorcycle / 197-Trail / 213-Forest / \n", "3-Concert / 3044-Sesame / \n", "93-Comedy / 1369-Necktie / \n", "7-Musician / 30-Drummer / 39-Drums / 44-Drums / 128-Cymbal / 146-Snare drum / 4131-Mapex Drums / \n", "5-Dance / 98-Festival / \n", "5-Dance / 16-Performance art / 95-Talent show / 2198-Irish dance / \n", "1-Vehicle / 4-Car / \n", "12-Food / 32-Recipe / 236-Baking / 1408-Snack / 2046-Popcorn / \n", "6-Animation / 15-Cartoon / 411-MapleStory / \n", "319-Robot / \n", "0-Games / 8-Football / \n", "309-Computer hardware / \n", "0-Games / 2-Video game / 116-Pokémon / 186-Pokémon / 1387-Pokémon Battle Revolution / \n", "9-Music video / \n", "0-Games / 2-Video game / \n", "367-Drink / 612-Beer / \n", "238-Nail / 300-Nail art / 303-Paint / 304-Nail polish / 307-Hand / 324-Finger / 338-Manicure / 873-Acrylic paint / 936-Watercolor paint / \n", "937-Black belt / \n", "7-Musician / 9-Music video / 13-Musical ensemble / \n", "0-Games / 2-Video game / 19-PC game / 33-Weapon / 43-Call of Duty / 302-Call of Duty 4: Modern Warfare / \n", "2-Video game / 23-Mobile phone / 47-Personal computer / 145-Tablet computer / \n", "20-Trailer / \n", "1-Vehicle / 4-Car / 277-Sport utility vehicle / 298-Television advertisement / 4493-Nissan X-Trail / \n", "10-Animal / 280-Wildlife / 537-Zoo / 999-Elephant / \n", "59-Winter sport / 84-Snow / 120-Winter / 177-Skiing / 281-Snowboarding / \n", "0-Games / 2-Video game / 51-Strategy video game / 165-League of Legends / \n", "1-Vehicle / 4-Car / 40-Road / 94-Dashcam / \n", "0-Games / 544-Table tennis / \n", "8-Football / \n", "804-Shopping mall / \n", "3-Concert / \n", "0-Games / 2-Video game / 635-Achievement / 948-Far Cry / \n", "0-Games / 1680-Seafight / \n", "1-Vehicle / 10-Animal / 87-Boat / 225-Lake / \n", "5-Dance / 31-Disc jockey / 78-Wedding / 121-Photography / 2269-Wedding photography / \n", "51-Strategy video game / 321-Roller coaster / 334-Amusement ride / 1293-RollerCoaster Tycoon 3 / 3620-RollerCoaster Tycoon / \n", "1-Vehicle / 27-Motorcycle / 207-Exhaust system / \n", "1-Vehicle / 27-Motorcycle / 197-Trail / 213-Forest / \n", "460-Yu-Gi-Oh! Trading Card Game / \n", "12-Food / 3250-Chubby Bunny / \n", "0-Games / 2-Video game / 34-Action-adventure game / \n", "14-Guitar / 100-Electric guitar / \n", "10-Animal / 12-Food / 26-Cooking / 52-Dish / 58-Cuisine / 215-Eating / 424-Chicken / 1438-Street food / \n", "10-Animal / 18-Outdoor recreation / 80-Horse / 247-Stallion / 289-Dressage / 404-Livestock / 4322-Dutch Warmblood / \n", "3-Concert / 16-Performance art / 253-Carnival / \n", "3-Concert / 14-Guitar / 502-Sky / \n", "31-Disc jockey / \n", "0-Games / 1-Vehicle / 4-Car / 11-Motorsport / 17-Racing / 169-Rallying / 711-Gran Turismo / 2476-Maze / \n", "20-Trailer / \n", "117-Boxing / \n", "1-Vehicle / 50-Aircraft / 82-Airplane / 92-Radio-controlled model / 149-Model aircraft / 160-Radio-controlled aircraft / 266-Helicopter / 392-Radio-controlled helicopter / \n", "20-Trailer / \n", "0-Games / 2-Video game / 19-PC game / 33-Weapon / 192-Battlefield / 374-Battlefield 3 / 383-Battlefield 4 / \n", "273-News program / \n", "689-Harmonica / \n", "0-Games / 211-RuneScape / \n", "170-Home improvement / 248-Hotel / \n", "743-Edward Cullen / \n", "7-Musician / 36-Piano / 150-Violin / 401-Cello / \n", "14-Guitar / 24-String instrument / 63-Acoustic guitar / 2435-Tokyo Ghoul / \n", "2118-Tongue / \n", "3-Concert / \n", "1-Vehicle / 92-Radio-controlled model / 149-Model aircraft / 160-Radio-controlled aircraft / 233-Ocean / 278-GoPro / 393-Unmanned aerial vehicle / \n", "3-Concert / 30-Drummer / 39-Drums / \n", "9-Music video / 330-Money / \n", "0-Games / 2-Video game / 511-Mortal Kombat / 635-Achievement / \n", "23-Mobile phone / 29-Smartphone / 37-Gadget / 72-iPhone / 189-iPad / 272-iPod / 331-iPod touch / \n", "28-Fashion / 45-Cosmetics / 246-Furniture / 276-Room / 580-Bedroom / \n", "282-Skin / 3861-Toner / \n", "0-Games / 51-Strategy video game / 165-League of Legends / \n", "0-Games / 1-Vehicle / 2-Video game / 34-Action-adventure game / 50-Aircraft / 201-Grand Theft Auto: San Andreas / \n", "0-Games / 1-Vehicle / 2-Video game / 11-Motorsport / 1817-Initial D Arcade Stage / \n", "1-Vehicle / 4-Car / 27-Motorcycle / \n", "61-Art / \n", "23-Mobile phone / 29-Smartphone / 37-Gadget / 72-iPhone / 189-iPad / 272-iPod / 1899-E-book / \n", "30-Drummer / 39-Drums / 44-Drums / 128-Cymbal / 146-Snare drum / \n", "10-Animal / 1234-Mouse / \n", "463-Cricket / 1476-Bowling (cricket) / \n", "5-Dance / 889-Swing / \n", "238-Nail / 1325-Gel / \n", "10-Animal / 22-Nature / 166-Cat / \n", "12-Food / 1208-Coconut / \n", "125-Television / \n", "5-Dance / 16-Performance art / 4089-Maracas / \n", "36-Piano / 164-Beach / 978-Love song / \n", "22-Nature / 61-Art / 147-River / 148-Painting / 227-Water / \n", "0-Games / 2-Video game / 19-PC game / 47-Personal computer / 131-Computer / 192-Battlefield / 383-Battlefield 4 / \n", "31-Disc jockey / \n", "819-World Wide Web / \n", "3-Concert / 68-Lighting / \n", "7-Musician / 36-Piano / 104-Pianist / \n", "5-Dance / 49-School / \n", "9-Music video / \n", "0-Games / 159-World of Warcraft / 171-Warcraft / 1356-Paladin / 1882-Warrior / \n", "25-Toy / 140-LEGO / 884-Lego minifigure / \n", "67-Cycling / 69-Bicycle / 377-BMX bike / \n", "1-Vehicle / 50-Aircraft / 82-Airplane / 129-Aviation / 157-Jet aircraft / 388-Fighter aircraft / 1483-United States Air Force / 2954-McDonnell Douglas F-15 Eagle / \n", "0-Games / 2-Video game / 19-PC game / 928-Borderlands 2 / \n", "293-Slide show / \n", "1-Vehicle / 4-Car / 344-Coupé / 1314-Garage / 2639-Automobile repair shop / \n", "47-Personal computer / 228-Laptop / \n", "6-Animation / 547-Bleach / 3562-Hollow / \n", "0-Games / 49-School / 54-Highlight film / 60-Basketball / \n", "9-Music video / \n", "98-Festival / 110-Album / \n", "1-Vehicle / 4-Car / 11-Motorsport / \n", "0-Games / \n", "624-Number / 1546-Line / \n", "0-Games / 2-Video game / 6-Animation / 15-Cartoon / 89-Comics / 185-Dragon Ball / 244-Goku / 418-Gohan / 869-Dragon Ball Z: Budokai Tenkaichi / 1244-Trunks / 4335-Dragon Ball Z: Burst Limit / \n", "0-Games / 2-Video game / 6-Animation / 20-Trailer / 55-Video game console / 111-PlayStation 3 / 176-PlayStation 4 / 396-One Piece / \n", "6-Animation / 1199-Kinect / \n", "1-Vehicle / 4-Car / 1046-Remote control / 1245-Alarm device / \n", "45-Cosmetics / 235-Lipstick / \n", "7-Musician / 14-Guitar / 24-String instrument / 100-Electric guitar / \n", "150-Violin / \n", "3-Concert / 7-Musician / \n", "28-Fashion / 206-Shoe / 757-Boot / \n", "22-Nature / 84-Snow / 800-Mountain pass / \n", "1-Vehicle / 11-Motorsport / 27-Motorcycle / 105-Motorcycling / 207-Exhaust system / 806-Chopper / 4170-Honda Shadow / \n", "12-Food / 26-Cooking / 32-Recipe / 52-Dish / 542-Potato / 2175-Lamb and mutton / 3384-Cutlet / \n", "36-Piano / 1555-Sword Art Online / \n", "3-Concert / 5-Dance / 671-Tomato / \n", "1-Vehicle / 62-Train / 1279-Plarail / \n", "121-Photography / \n", "0-Games / 49-School / 60-Basketball / \n", "122-Weight training / \n", "20-Trailer / \n", "1113-Toilet / \n", "22-Nature / 164-Beach / \n", "0-Games / 8-Football / 21-Stadium / 65-Kick / 76-Ball / 81-Athlete / \n", "3-Concert / 7-Musician / 13-Musical ensemble / 30-Drummer / 39-Drums / 44-Drums / 128-Cymbal / \n", "0-Games / 2-Video game / 19-PC game / 1187-Ragnarok Online / \n", "12-Food / \n", "1-Vehicle / 4-Car / 74-Truck / 2007-Intermodal container / \n", "1-Vehicle / 4-Car / \n", "5-Dance / \n", "20-Trailer / 222-Surfing / 357-DVD / 815-Surfboard / \n", "3-Concert / 16-Performance art / 3389-West Side Story / \n", "8-Football / \n", "14-Guitar / 100-Electric guitar / \n", "3-Concert / 16-Performance art / \n", "10-Animal / 987-Sheep / \n", "59-Winter sport / 84-Snow / 177-Skiing / 305-Ski / \n", "1-Vehicle / 18-Outdoor recreation / 22-Nature / 40-Road / 283-Rock / \n", "3-Concert / 7-Musician / 13-Musical ensemble / 36-Piano / 38-Orchestra / 230-Brass instrument / 427-Trumpet / 634-Trombone / 1574-French horn / \n", "10-Animal / 17-Racing / 384-Horse racing / \n", "1-Vehicle / 4-Car / 1742-Lowrider / \n", "0-Games / 1-Vehicle / 2-Video game / 11-Motorsport / 17-Racing / 27-Motorcycle / 172-Motocross / \n", "0-Games / 1-Vehicle / 2-Video game / 34-Action-adventure game / 1275-Watch Dogs / \n", "1786-Dam / \n", "12-Food / 26-Cooking / 32-Recipe / 52-Dish / 58-Cuisine / 109-Cooking show / 179-Kitchen / 184-Vegetable / 301-Meal / 557-Salad / 742-Countertop / 1748-Carrot / \n", "467-Chipmunk / \n", "0-Games / 51-Strategy video game / 356-Star / 373-Clash of Clans / \n", "59-Winter sport / 84-Snow / 120-Winter / 281-Snowboarding / 501-Snowboard / \n", "10-Animal / 174-Farm / 723-Pig / 1123-Domestic pig / \n", "2162-Marker pen / \n", "12-Food / 86-Plant / 136-Gardening / 156-Agriculture / 174-Farm / 184-Vegetable / 279-Garden / 671-Tomato / 1384-Greenhouse / \n", "260-Prayer / \n", "13-Musical ensemble / 46-Choir / \n", "14-Guitar / 24-String instrument / 63-Acoustic guitar / \n", "1363-Candle / \n", "88-Machine / 208-Wood / 261-Tool / 400-Woodturning / 640-Bottle / 972-Lathe / \n", "6-Animation / 701-Beyblade / 1004-Samurai / 3868-Beyblade: Shogun Steel / \n", "36-Piano / 53-Keyboard / 397-Organ / 646-Pipe organ / \n", "25-Toy / 140-LEGO / 313-Book / 434-Ninja / 884-Lego minifigure / 1478-Lego Ninjago / \n", "1-Vehicle / 11-Motorsport / 17-Racing / 27-Motorcycle / 57-Race track / 172-Motocross / \n", "36-Piano / 53-Keyboard / 104-Pianist / \n", "298-Television advertisement / 337-Advertising / \n", "3-Concert / 7-Musician / 30-Drummer / 44-Drums / \n", "0-Games / 6-Animation / 460-Yu-Gi-Oh! Trading Card Game / \n", "0-Games / 47-Personal computer / 55-Video game console / 111-PlayStation 3 / \n", "1-Vehicle / 11-Motorsport / 22-Nature / 198-Off-road vehicle / 213-Forest / 327-Mud / 565-Mud bogging / \n", "0-Games / 8-Football / 21-Stadium / 65-Kick / 77-Arena / \n", "125-Television / \n", "102-Building / 113-House / 170-Home improvement / 208-Wood / 707-Roof / 879-Ceiling / \n", "126-The Walt Disney Company / 144-Amusement park / 576-Walt Disney World / \n", "0-Games / 35-Minecraft / \n", "59-Winter sport / 83-Skateboarding / 108-Skateboard / 112-Ice skating / \n", "25-Toy / 200-Doll / 786-Juice / 2861-Baby Alive / \n", "25-Toy / 476-Textile / 2914-Wreath / \n", "414-Computer monitor / 621-Asus / \n", "0-Games / 2-Video game / 43-Call of Duty / 134-Call of Duty: Black Ops / 194-Call of Duty: Modern Warfare 2 / \n", "10-Animal / 18-Outdoor recreation / 80-Horse / 247-Stallion / 289-Dressage / 404-Livestock / 438-Mare / 1434-Saddle / \n", "0-Games / 85-Combat / 117-Boxing / \n", "452-Watch / \n", "1-Vehicle / 4-Car / 154-Wheel / \n", "12-Food / 26-Cooking / 32-Recipe / 58-Cuisine / 210-Cookware and bakeware / 236-Baking / 442-Bread / 485-Dough / 578-Cookie / 1283-Pancake / 2995-Blueberry / \n", "1-Vehicle / 4-Car / \n", "3-Concert / 5-Dance / 7-Musician / 30-Drummer / 133-Nightclub / \n", "9-Music video / 78-Wedding / 190-Bride / 318-Church / \n", "3071-Towel / \n", "99-Christmas / 242-Paper / 370-Origami / \n", "22-Nature / 499-Cloud / \n", "0-Games / 1-Vehicle / 2-Video game / 4-Car / 11-Motorsport / 17-Racing / 57-Race track / 3586-F1 2011 / 3731-F1 2010 / \n", "725-Human / \n", "23-Mobile phone / 29-Smartphone / 37-Gadget / 101-Telephone / 178-Samsung Galaxy / 487-Samsung Electronics / 1015-Samsung Galaxy Note series / 1776-Samsung Galaxy Note II / \n", "1-Vehicle / 4-Car / 11-Motorsport / 17-Racing / 31-Disc jockey / 135-Drifting / \n", "0-Games / 2-Video game / 19-PC game / 33-Weapon / 96-Soldier / 192-Battlefield / 3450-Battlefield 1942 / \n", "1-Vehicle / 205-Newscaster / 273-News program / \n", "439-Barbie / \n", "12-Food / 26-Cooking / 32-Recipe / 52-Dish / 58-Cuisine / 301-Meal / \n", "6-Animation / 123-Naruto / 434-Ninja / 2169-Ninja World / \n", "1-Vehicle / 2-Video game / 4-Car / 70-Driving / 1758-Test Drive / 1766-Test Drive Unlimited / 2113-Test Drive Unlimited 2 / \n", "5-Dance / 16-Performance art / 93-Comedy / \n", "2-Video game / 47-Personal computer / 294-Batman / 464-First-person Shooter / 1261-Batman: Arkham City / \n", "6-Animation / 25-Toy / 435-Action figure / \n", "6-Animation / \n", "0-Games / 2-Video game / 55-Video game console / 2101-Ace Combat / \n", "23-Mobile phone / 29-Smartphone / 101-Telephone / 465-Microphone / 802-Google Nexus / \n", "7-Musician / 13-Musical ensemble / 30-Drummer / 39-Drums / \n", "0-Games / 8-Football / 21-Stadium / 54-Highlight film / 65-Kick / \n", "1459-Magazine / \n", "122-Weight training / 214-Muscle / 1055-Back / \n", "18-Outdoor recreation / 83-Skateboarding / 108-Skateboard / \n", "5-Dance / 115-Ballet / 3092-Cinderella / \n", "0-Games / 8-Football / \n", "9-Music video / 256-Tennis / \n", "1-Vehicle / 4-Car / 164-Beach / \n", "20-Trailer / \n", "6-Animation / 15-Cartoon / 126-The Walt Disney Company / 2076-Goofy / \n", "36-Piano / 53-Keyboard / 104-Pianist / 2555-Harpsichord / \n", "6-Animation / \n", "3-Concert / 7-Musician / \n", "28-Fashion / 49-School / \n", "0-Games / 2-Video game / 35-Minecraft / \n", "88-Machine / \n", "1-Vehicle / 50-Aircraft / 82-Airplane / 92-Radio-controlled model / 129-Aviation / 149-Model aircraft / 160-Radio-controlled aircraft / \n", "12-Food / 26-Cooking / 32-Recipe / 52-Dish / 58-Cuisine / 109-Cooking show / 179-Kitchen / 226-Dessert / 232-Cake / 295-Chocolate / 684-Ice cream / \n", "0-Games / 10-Animal / 15-Cartoon / \n", "18-Outdoor recreation / 83-Skateboarding / 108-Skateboard / 221-Skatepark / 1031-Fisheye lens / \n", "0-Games / 8-Football / \n", "0-Games / 2-Video game / 96-Soldier / 192-Battlefield / 374-Battlefield 3 / 383-Battlefield 4 / \n", "0-Games / 2-Video game / \n", "1-Vehicle / 11-Motorsport / 124-Tractor / 156-Agriculture / 252-Tractor pulling / 1129-Hay / \n", "5-Dance / 2449-Chuck E. Cheese's / \n", "49-School / \n", "139-Airport / 1013-Walking / \n", "5-Dance / 16-Performance art / \n", "293-Slide show / \n", "18-Outdoor recreation / 347-Climbing / 828-Rock climbing / \n", "0-Games / 194-Call of Duty: Modern Warfare 2 / \n", "3-Concert / \n", "0-Games / 2-Video game / \n", "5-Dance / \n", "20-Trailer / \n", "260-Prayer / 1604-Satan / \n", "0-Games / 2-Video game / 586-Nintendo Entertainment System / \n", "2-Video game / 23-Mobile phone / 29-Smartphone / 37-Gadget / 101-Telephone / \n", "3-Concert / \n", "3031-Aragorn / \n", "5-Dance / \n", "0-Games / 2-Video game / 19-PC game / 192-Battlefield / 374-Battlefield 3 / \n", "3-Concert / 7-Musician / \n", "3-Concert / 5-Dance / 49-School / 95-Talent show / 191-Cheerleading / \n", "1-Vehicle / 11-Motorsport / 27-Motorcycle / 67-Cycling / 69-Bicycle / 377-BMX bike / \n", "12-Food / 26-Cooking / 32-Recipe / 58-Cuisine / 109-Cooking show / 1022-Chili pepper / 2609-Canning / 4266-Jalapeño / \n", "1-Vehicle / 10-Animal / 42-Fishing / 87-Boat / 91-Fish / \n", "3-Concert / 7-Musician / 36-Piano / 53-Keyboard / 104-Pianist / 107-Musical keyboard / \n", "20-Trailer / \n", "10-Animal / 84-Snow / 217-Hunting / 706-Deer / \n", "9-Music video / 20-Trailer / 110-Album / \n", "0-Games / 1029-Aion: The Tower of Eternity / \n", "0-Games / 2-Video game / 464-First-person Shooter / 1749-Titanfall / \n", "693-Sword / \n", "452-Watch / 619-Manufacturing / 627-Clock / \n", "1-Vehicle / 4-Car / 3197-Citroën C4 / \n", "86-Plant / 279-Garden / \n", "7-Musician / 14-Guitar / 24-String instrument / 63-Acoustic guitar / \n", "0-Games / 310-The Sims / 328-Simulation video game / 543-The Sims 2 / \n", "8-Football / \n", "6-Animation / 25-Toy / 435-Action figure / \n", "2816-Billboard / \n", "15-Cartoon / 165-League of Legends / \n", "0-Games / 8-Football / 21-Stadium / 54-Highlight film / 79-American football / \n", "5-Dance / \n", "99-Christmas / \n", "46-Choir / \n", "5-Dance / 16-Performance art / 39-Drums / \n", "37-Gadget / 47-Personal computer / 145-Tablet computer / 189-iPad / 1150-iPad 2 / \n", "162-Medicine / \n", "66-Bollywood / \n", "15-Cartoon / 903-Mass Effect / \n", "0-Games / 1-Vehicle / 2-Video game / 4-Car / 11-Motorsport / 17-Racing / 199-Wii / 3920-Maximum Destruction / \n", "6-Animation / 15-Cartoon / 412-Monster / 503-Monster High / \n", "0-Games / 8-Football / 21-Stadium / 65-Kick / 73-Ball / 77-Arena / 81-Athlete / \n", "10-Animal / 80-Horse / 247-Stallion / 289-Dressage / 290-Pony / \n", "1-Vehicle / 4-Car / 64-Transport / 74-Truck / 249-Heavy equipment / 1611-Garbage truck / 3856-Landfill / \n", "6-Animation / 577-Penguin / \n", "3-Concert / 7-Musician / 13-Musical ensemble / 39-Drums / 230-Brass instrument / \n", "23-Mobile phone / 29-Smartphone / 37-Gadget / 178-Samsung Galaxy / 1629-Samsung Galaxy Note / 1727-Samsung Galaxy Note 3 / \n", "3-Concert / \n", "2-Video game / 55-Video game console / 127-Xbox 360 / \n", "5-Dance / 16-Performance art / \n", "3-Concert / 7-Musician / 13-Musical ensemble / 38-Orchestra / \n", "12-Food / 26-Cooking / 52-Dish / 58-Cuisine / 2313-Bánh / \n", "10-Animal / 22-Nature / 91-Fish / 258-Aquarium / \n", "298-Television advertisement / 356-Star / \n", "56-Hair / 106-Hairstyle / 607-Lace / 1292-Ponytail / \n", "20-Trailer / 434-Ninja / \n", "1-Vehicle / 62-Train / 2904-Express train / \n", "14-Guitar / 24-String instrument / 100-Electric guitar / \n", "3-Concert / 38-Orchestra / 68-Lighting / \n", "2-Video game / 43-Call of Duty / 134-Call of Duty: Black Ops / 138-Call of Duty: Black Ops II / \n", "3-Concert / 68-Lighting / \n", "8-Football / 73-Ball / 76-Ball / \n", "3-Concert / 13-Musical ensemble / \n", "1-Vehicle / 4-Car / 270-Sedan / \n", "36-Piano / 53-Keyboard / \n", "9-Music video / \n", "10-Animal / 42-Fishing / 91-Fish / 258-Aquarium / \n", "6-Animation / 61-Art / 97-Drawing / 148-Painting / 183-Manga / 528-Pencil / 628-Portrait / \n", "0-Games / 21-Stadium / 54-Highlight film / 81-Athlete / 562-Pitcher / 946-Baseball park / \n", "36-Piano / 53-Keyboard / 104-Pianist / 107-Musical keyboard / 265-Electronic keyboard / \n", "9-Music video / \n", "1-Vehicle / 4-Car / 371-Firefighter / 663-Street / 796-Fire engine / 1633-Fire station / \n", "5-Dance / 95-Talent show / \n", "125-Television / \n", "5-Dance / \n", "9-Music video / \n", "167-Four-wheel drive / 626-Puzzle / 631-Cube / 740-Rubik's Cube / \n", "15-Cartoon / 31-Disc jockey / \n", "5-Dance / 16-Performance art / \n", "0-Games / 8-Football / 21-Stadium / 54-Highlight film / 65-Kick / 76-Ball / 81-Athlete / 767-Goalkeeper / 1247-Penalty kick / \n", "3-Concert / 30-Drummer / 68-Lighting / \n", "10-Animal / 48-Pet / 132-Bird / 424-Chicken / 1136-Poultry / \n", "162-Medicine / 1599-Bone / \n", "22-Nature / 147-River / 283-Rock / 751-Desert / \n", "0-Games / 8-Football / \n", "1-Vehicle / 249-Heavy equipment / 822-Crane / 2194-Wind power / \n", "1-Vehicle / 4-Car / 2826-Suzuki Swift / \n", "6-Animation / 61-Art / 97-Drawing / 148-Painting / 1620-Pearl / \n", "260-Prayer / \n", "116-Pokémon / 330-Money / \n", "5-Dance / 9-Music video / \n", "0-Games / 8-Football / 54-Highlight film / 79-American football / 813-Running back / \n", "122-Weight training / 214-Muscle / 532-Barbell / 1055-Back / \n", "0-Games / 6-Animation / 995-Attack on Titan / \n", "1-Vehicle / 8-Football / 21-Stadium / 77-Arena / \n", "278-GoPro / \n", "1-Vehicle / 62-Train / 64-Transport / 103-Rail transport / 119-Locomotive / 143-Railroad car / 152-Track / 161-Train station / \n", "150-Violin / 639-Fiddle / \n", "12-Food / 226-Dessert / 232-Cake / 560-Sugar / 578-Cookie / 759-Icing / 795-Cupcake / 3641-Royal icing / \n", "67-Cycling / 69-Bicycle / 221-Skatepark / 377-BMX bike / \n", "23-Mobile phone / 29-Smartphone / 37-Gadget / 101-Telephone / \n", "1-Vehicle / 4-Car / 1321-Škoda Auto / 1540-Station Wagon / \n", "0-Games / 2-Video game / 55-Video game console / \n", "23-Mobile phone / 29-Smartphone / 121-Photography / 155-Camera / \n", "971-Alphabet / \n", "0-Games / 2-Video game / 6-Animation / 116-Pokémon / 186-Pokémon / 690-Pokémon X and Y / 1005-Pokémon Omega Ruby and Alpha Sapphire / \n", "1-Vehicle / 25-Toy / 62-Train / 144-Amusement park / 507-Thomas the Tank Engine / \n", "1-Vehicle / 27-Motorcycle / 207-Exhaust system / \n", "0-Games / 2-Video game / 19-PC game / 33-Weapon / 635-Achievement / 924-Trophy / \n", "96-Soldier / \n", "6-Animation / 15-Cartoon / 99-Christmas / 467-Chipmunk / \n", "1-Vehicle / 11-Motorsport / 124-Tractor / 156-Agriculture / 174-Farm / 249-Heavy equipment / 252-Tractor pulling / 459-Plough / 660-Rural area / 3820-Harrow / \n", "1-Vehicle / 82-Airplane / 114-Engine / 1263-Glider / \n", "1-Vehicle / 4-Car / 11-Motorsport / 17-Racing / 57-Race track / 172-Motocross / 252-Tractor pulling / \n", "1-Vehicle / 4-Car / 11-Motorsport / 169-Rallying / \n", "0-Games / 2-Video game / 20-Trailer / \n", "3-Concert / 5-Dance / 16-Performance art / \n", "12-Food / 26-Cooking / 58-Cuisine / \n", "25-Toy / 1152-Plush / 2689-Furby / \n", "1-Vehicle / 4-Car / 114-Engine / 449-Honda / 1614-Honda Accord / 3086-Honda CR-V / \n", "5-Dance / 115-Ballet / \n", "126-The Walt Disney Company / \n", "0-Games / 2-Video game / 89-Comics / 241-Fighting game / 311-Street Fighter / \n", "75-Wrestling / \n", "97-Drawing / \n", "9-Music video / \n", "6-Animation / 15-Cartoon / \n", "1-Vehicle / 87-Boat / 164-Beach / \n", "23-Mobile phone / 29-Smartphone / 101-Telephone / 2929-Galaxy Nexus / \n", "1-Vehicle / 5-Dance / \n", "230-Brass instrument / 296-Saxophone / 1006-Tenor saxophone / \n", "0-Games / 2-Video game / 75-Wrestling / 777-WWE 2K / 2629-WWE SmackDown vs. Raw 2010 / \n", "6-Animation / 15-Cartoon / 1724-Homer Simpson / 3367-Bart Simpson / \n", "303-Paint / 353-Wall / 488-Door / \n", "6-Animation / 15-Cartoon / \n", "0-Games / 60-Basketball / 168-Basketball moves / \n", "10-Animal / 12-Food / 523-Cattle / \n", "0-Games / 14-Guitar / 445-Guitar Hero / 1144-Castlevania / \n", "5-Dance / 16-Performance art / \n", "3-Concert / 38-Orchestra / \n", "126-The Walt Disney Company / 144-Amusement park / 576-Walt Disney World / \n", "12-Food / 1363-Candle / \n", "3-Concert / \n", "9-Music video / \n", "45-Cosmetics / 238-Nail / 300-Nail art / 304-Nail polish / 307-Hand / 324-Finger / 338-Manicure / 823-Artificial nails / \n", "1-Vehicle / 12-Food / 124-Tractor / 156-Agriculture / 174-Farm / 984-Silage / \n", "2-Video game / 43-Call of Duty / 394-Map / 431-Call of Duty: Advanced Warfare / \n", "14-Guitar / 63-Acoustic guitar / \n", "0-Games / 19-PC game / 1187-Ragnarok Online / \n", "1-Vehicle / 62-Train / 64-Transport / 103-Rail transport / 161-Train station / 224-Rapid transit / \n", "3-Concert / 250-Parade / \n", "402-Jewellery / 476-Textile / \n", "222-Surfing / 672-Bodyboarding / \n", "419-Earth / \n", "67-Cycling / 69-Bicycle / 259-Fire / \n", "309-Computer hardware / 908-Video card / \n", "6-Animation / \n", "8-Football / \n", "1-Vehicle / 62-Train / 64-Transport / 103-Rail transport / 119-Locomotive / 143-Railroad car / 152-Track / 691-Cargo / 821-Rail freight transport / \n", "402-Jewellery / 863-Body piercing / \n", "0-Games / 2-Video game / 37-Gadget / 55-Video game console / 199-Wii / 292-Handheld game console / 443-Wii U / \n", "0-Games / 1-Vehicle / 124-Tractor / 156-Agriculture / 174-Farm / 458-Gold / 715-Farming Simulator / \n", "325-Vampire / 1950-The CW / \n", "1-Vehicle / 62-Train / 64-Transport / 94-Dashcam / 152-Track / 161-Train station / 224-Rapid transit / \n", "6-Animation / 15-Cartoon / \n", "12-Food / 86-Plant / 348-Fruit / 2500-Pear / \n", "9-Music video / \n", "6-Animation / 15-Cartoon / 1884-Inazuma Eleven / 2450-Inazuma Eleven / \n", "0-Games / 2-Video game / 19-PC game / 33-Weapon / 51-Strategy video game / 271-Tank / 409-World of Tanks / \n", "88-Machine / 1057-Embroidery / \n", "205-Newscaster / 223-Weather / \n", "9-Music video / \n", "0-Games / 2-Video game / 19-PC game / 33-Weapon / 43-Call of Duty / 96-Soldier / 111-PlayStation 3 / 194-Call of Duty: Modern Warfare 2 / \n", "0-Games / 1232-Perfect World / 2394-Perfect World / \n", "25-Toy / 35-Minecraft / 140-LEGO / 3899-Lego Minecraft / \n", "0-Games / 159-World of Warcraft / 171-Warcraft / 1213-World of Warcraft: Wrath of the Lich King / \n", "1-Vehicle / 4-Car / 94-Dashcam / 449-Honda / \n", "12-Food / 26-Cooking / 32-Recipe / 52-Dish / 58-Cuisine / 226-Dessert / 232-Cake / 236-Baking / 534-Milk / 968-Pie / \n", "60-Basketball / \n", "3-Concert / 7-Musician / 13-Musical ensemble / 14-Guitar / 44-Drums / 689-Harmonica / \n", "1-Vehicle / 4-Car / \n", "0-Games / 51-Strategy video game / 382-Dota 2 / 549-Defense of the Ancients / \n", "122-Weight training / 204-Gym / 214-Muscle / 529-Squat / 532-Barbell / \n", "5-Dance / 986-Suit / \n", "3-Concert / 7-Musician / 13-Musical ensemble / 30-Drummer / 38-Orchestra / 39-Drums / 44-Drums / 46-Choir / \n", "5-Dance / \n", "2-Video game / 47-Personal computer / 1996-Intel Core i7 / \n", "1-Vehicle / 74-Truck / 249-Heavy equipment / 1429-Dump truck / \n", "31-Disc jockey / \n", "1-Vehicle / 4-Car / 11-Motorsport / 70-Driving / 349-Chevrolet / 406-Chevrolet / 3793-Chevrolet Chevy II / Nova / \n", "5-Dance / 66-Bollywood / \n", "1-Vehicle / 286-Bus / \n", "0-Games / 2-Video game / 34-Action-adventure game / 163-Running / 201-Grand Theft Auto: San Andreas / 858-Carl Johnson / \n", "49-School / \n", "0-Games / 2-Video game / 33-Weapon / \n", "7-Musician / 13-Musical ensemble / 49-School / 203-Marching band / 306-High school / \n", "1-Vehicle / 10-Animal / 18-Outdoor recreation / 42-Fishing / 87-Boat / 91-Fish / 268-Recreational fishing / \n", "31-Disc jockey / \n", "1605-Second Life / \n", "1-Vehicle / 648-Clown / \n", "1-Vehicle / 11-Motorsport / 124-Tractor / 156-Agriculture / 249-Heavy equipment / 327-Mud / \n", "3-Concert / 16-Performance art / 38-Orchestra / \n", "1-Vehicle / 4-Car / 11-Motorsport / 124-Tractor / 252-Tractor pulling / \n", "8-Football / 21-Stadium / 77-Arena / \n", "5-Dance / 115-Ballet / \n", "0-Games / 35-Minecraft / \n", "31-Disc jockey / \n", "1-Vehicle / 18-Outdoor recreation / 83-Skateboarding / 108-Skateboard / 221-Skatepark / 336-Scooter / \n", "164-Beach / 222-Surfing / \n", "7-Musician / 13-Musical ensemble / 389-Restaurant / \n", "0-Games / 2541-Allods Online / \n", "516-Headphones / \n", "3-Concert / 7-Musician / 13-Musical ensemble / 38-Orchestra / 133-Nightclub / \n", "9-Music video / \n", "0-Games / 8-Football / 21-Stadium / 54-Highlight film / 79-American football / \n", "0-Games / 8-Football / 21-Stadium / 65-Kick / 73-Ball / 76-Ball / 77-Arena / 81-Athlete / \n", "0-Games / 8-Football / 21-Stadium / 65-Kick / 73-Ball / \n", "250-Parade / \n", "0-Games / 2-Video game / 1681-Mu Online / 4288-Carnage / \n", "45-Cosmetics / 238-Nail / 300-Nail art / 304-Nail polish / 307-Hand / 324-Finger / 338-Manicure / 1128-Glitter / \n", "85-Combat / 694-Tai chi / \n", "39-Drums / 44-Drums / 1458-Electronic drum / \n", "0-Games / 1-Vehicle / 2-Video game / 4-Car / 11-Motorsport / 17-Racing / 41-Sports car / 57-Race track / 635-Achievement / 924-Trophy / \n", "12-Food / 26-Cooking / 32-Recipe / 58-Cuisine / 1408-Snack / 2046-Popcorn / \n", "3-Concert / 16-Performance art / \n", "0-Games / 2-Video game / 6-Animation / 241-Fighting game / 514-Tekken / 797-Tekken / 1470-Tekken Tag Tournament 2 / 2812-Kazuya Mishima / \n", "9-Music video / \n", "30-Drummer / 39-Drums / 44-Drums / 128-Cymbal / 146-Snare drum / \n", "12-Food / 26-Cooking / 32-Recipe / 52-Dish / 179-Kitchen / 354-Egg / \n", "1-Vehicle / 198-Off-road vehicle / 741-Mower / \n", "3-Concert / 16-Performance art / 99-Christmas / \n", "1-Vehicle / 27-Motorcycle / \n", "10-Animal / 48-Pet / 132-Bird / 598-Parrot / 1341-Parakeet / \n", "0-Games / 8-Football / 21-Stadium / 65-Kick / 73-Ball / 81-Athlete / \n", "0-Games / 1-Vehicle / 2-Video game / 4-Car / 34-Action-adventure game / 82-Airplane / 118-Grand Theft Auto V / \n", "0-Games / 8-Football / 21-Stadium / 65-Kick / 73-Ball / 76-Ball / \n", "1019-Christ / \n", "0-Games / 19-PC game / 51-Strategy video game / 165-League of Legends / 1494-Jungle / \n", "455-Unidentified flying object / \n", "8-Football / 79-American football / \n", "99-Christmas / \n", "28-Fashion / 785-Jeans / 2315-Top / \n", "1-Vehicle / 50-Aircraft / 139-Airport / 151-Landing / 157-Jet aircraft / 175-Airliner / 187-Takeoff / 188-Airline / 254-Runway / 472-Boeing 737 / 1529-Jet engine / 2064-Boeing 737 Next Generation / \n", "0-Games / 33-Weapon / \n", "1-Vehicle / 4-Car / 11-Motorsport / 41-Sports car / 70-Driving / 234-Ford / 473-Ford Mustang / 647-Muscle car / 1361-Ford GT / \n", "117-Boxing / \n", "624-Number / \n", "1-Vehicle / 4-Car / 11-Motorsport / 41-Sports car / 234-Ford / 270-Sedan / 332-Mercedes-Benz / 473-Ford Mustang / 2753-SEAT León / \n", "10-Animal / 22-Nature / 86-Plant / 723-Pig / 1215-Wild boar / \n", "3-Concert / \n", "10-Animal / 22-Nature / 588-Reptile / 763-Snake / \n", "960-Badminton / \n", "1-Vehicle / 27-Motorcycle / \n", "14-Guitar / \n", "9-Music video / \n", "0-Games / 34-Action-adventure game / 201-Grand Theft Auto: San Andreas / \n", "0-Games / 2140-Tony Hawk's / \n", "23-Mobile phone / 29-Smartphone / 37-Gadget / 72-iPhone / 452-Watch / 1260-Smartwatch / \n", "2610-Rosary / \n", "9-Music video / \n", "1-Vehicle / 25-Toy / 369-Resort / \n", "13-Musical ensemble / \n", "3-Concert / 36-Piano / 46-Choir / \n", "6-Animation / 1303-Fullmetal Alchemist / \n", "0-Games / 2-Video game / 1496-LittleBigPlanet 2 / 1672-LittleBigPlanet / \n", "31-Disc jockey / \n", "0-Games / 2-Video game / 34-Action-adventure game / 387-Spider-Man / 3561-The Amazing Spider-Man / \n", "1-Vehicle / 62-Train / 224-Rapid transit / 337-Advertising / 3348-BVE Trainsim / \n", "208-Wood / 246-Furniture / \n", "6-Animation / 15-Cartoon / \n", "28-Fashion / 45-Cosmetics / 676-Handbag / \n", "5-Dance / 31-Disc jockey / \n", "3-Concert / \n", "1-Vehicle / 4-Car / 733-Washing / 1242-Car wash / \n", "8-Football / 13-Musical ensemble / 203-Marching band / \n", "3-Concert / 7-Musician / \n", "25-Toy / 323-Toddler / 447-Sand / \n", "12-Food / 476-Textile / \n", "0-Games / 2-Video game / 731-Coin / \n", "227-Water / 442-Bread / \n", "36-Piano / 150-Violin / \n", "0-Games / 8-Football / 21-Stadium / \n", "0-Games / 8-Football / 21-Stadium / \n", "0-Games / 8-Football / 358-Rugby football / \n", "7-Musician / 13-Musical ensemble / 21-Stadium / 49-School / 77-Arena / 79-American football / 203-Marching band / \n", "6-Animation / 31-Disc jockey / \n", "5-Dance / 49-School / 191-Cheerleading / 306-High school / \n", "6-Animation / 2965-A Certain Magical Index / \n", "369-Resort / \n", "5-Dance / \n", "14-Guitar / 24-String instrument / 100-Electric guitar / 1259-Fender Telecaster / \n", "693-Sword / \n", "7-Musician / 13-Musical ensemble / 21-Stadium / 49-School / 79-American football / 203-Marching band / 306-High school / \n", "0-Games / 2-Video game / 43-Call of Duty / 194-Call of Duty: Modern Warfare 2 / \n", "6-Animation / 15-Cartoon / \n", "75-Wrestling / \n", "5-Dance / 678-Latin dance / \n", "1-Vehicle / 4-Car / 11-Motorsport / 17-Racing / 41-Sports car / 234-Ford / 473-Ford Mustang / 1877-Shelby Mustang / \n", "0-Games / 211-RuneScape / \n", "3-Concert / 720-Wine / \n", "1-Vehicle / 4-Car / 27-Motorcycle / 1380-Kawasaki motorcycles / 4455-Kawasaki KLR650 / \n", "3-Concert / 7-Musician / 13-Musical ensemble / 46-Choir / \n", "5-Dance / \n", "6-Animation / 85-Combat / \n", "1-Vehicle / 4-Car / 270-Sedan / 299-Volkswagen Passenger Cars / 342-Windows Media Video / 526-Volkswagen / \n", "1-Vehicle / 11-Motorsport / 27-Motorcycle / 363-Soil / 660-Rural area / \n", "25-Toy / 99-Christmas / 200-Doll / 439-Barbie / 558-Santa Claus / \n", "1-Vehicle / 4-Car / 11-Motorsport / 17-Racing / 41-Sports car / 169-Rallying / \n", "0-Games / 2-Video game / 34-Action-adventure game / 127-Xbox 360 / 2205-Bully / \n", "3-Concert / 7-Musician / 24-String instrument / 150-Violin / 639-Fiddle / \n", "1-Vehicle / 10-Animal / 17-Racing / 57-Race track / 80-Horse / 384-Horse racing / 1694-Harness racing / \n", "14-Guitar / 110-Album / \n", "0-Games / 79-American football / \n", "49-School / \n", "2233-Derek Shepherd / \n", "155-Camera / 566-Digital camera / \n", "66-Bollywood / \n", "3-Concert / \n", "0-Games / 60-Basketball / 168-Basketball moves / 255-Slam dunk / \n", "164-Beach / 248-Hotel / 369-Resort / \n", "0-Games / 2-Video game / 34-Action-adventure game / 111-PlayStation 3 / 3832-Beyond: Two Souls / \n", "10-Animal / 384-Horse racing / \n", "0-Games / 8-Football / 21-Stadium / \n", "1-Vehicle / 50-Aircraft / 82-Airplane / 92-Radio-controlled model / 149-Model aircraft / 160-Radio-controlled aircraft / \n", "330-Money / 731-Coin / 912-Silver / 4247-Bullion / \n", "1-Vehicle / 50-Aircraft / 188-Airline / \n", "246-Furniture / 462-Table / 920-Chair / \n", "6-Animation / 121-Photography / 278-GoPro / \n", "25-Toy / 126-The Walt Disney Company / 200-Doll / 354-Egg / 439-Barbie / 996-Princess / \n", "20-Trailer / \n", "2-Video game / 25-Toy / 140-LEGO / 229-Halo / 264-Star Wars / 852-Lego Star Wars / \n", "12-Food / \n", "9-Music video / \n", "912-Silver / \n", "0-Games / 2-Video game / 35-Minecraft / \n", "6-Animation / 61-Art / 97-Drawing / 148-Painting / 353-Wall / \n", "229-Halo / 470-Halo 3 / \n", "162-Medicine / \n", "5-Dance / 115-Ballet / \n", "0-Games / 6-Animation / 1878-SD Gundam Capsule Fighter / \n", "125-Television / \n", "1-Vehicle / 62-Train / \n", "1-Vehicle / 4-Car / 74-Truck / 88-Machine / 249-Heavy equipment / 1093-Loader / 2445-Backhoe / \n", "7-Musician / 13-Musical ensemble / 38-Orchestra / 230-Brass instrument / 1003-Tuba / \n", "1-Vehicle / 4-Car / 11-Motorsport / 17-Racing / 403-Toyota / \n", "1704-Antena 3 / \n", "6-Animation / 243-Sonic the Hedgehog / 251-Sonic the Hedgehog / 1299-Shadow the Hedgehog / \n", "1-Vehicle / 4-Car / 11-Motorsport / 207-Exhaust system / 349-Chevrolet / 406-Chevrolet / 761-Chevrolet Camaro / \n", "3-Concert / \n", "36-Piano / 53-Keyboard / 104-Pianist / 107-Musical keyboard / 397-Organ / 646-Pipe organ / \n", "3-Concert / 7-Musician / 30-Drummer / \n", "30-Drummer / 39-Drums / 44-Drums / 128-Cymbal / \n", "1-Vehicle / 50-Aircraft / 82-Airplane / 139-Airport / 151-Landing / 175-Airliner / 472-Boeing 737 / \n", "7-Musician / 13-Musical ensemble / 98-Festival / 203-Marching band / 250-Parade / 663-Street / \n", "1-Vehicle / 62-Train / 64-Transport / 103-Rail transport / 152-Track / 428-Rail transport modelling / \n", "1-Vehicle / 11-Motorsport / 124-Tractor / 156-Agriculture / 174-Farm / 249-Heavy equipment / 252-Tractor pulling / 363-Soil / 459-Plough / \n", "0-Games / 2-Video game / 948-Far Cry / \n", "641-Mask / \n", "0-Games / 8-Football / 54-Highlight film / \n", "12-Food / 215-Eating / 232-Cake / 600-Flour / \n", "9-Music video / \n", "5-Dance / \n", "3-Concert / 7-Musician / 13-Musical ensemble / 16-Performance art / 3355-Bingo / \n", "23-Mobile phone / 47-Personal computer / 131-Computer / \n", "49-School / 306-High school / \n", "1-Vehicle / 4-Car / 40-Road / 64-Transport / 74-Truck / 259-Fire / 371-Firefighter / 796-Fire engine / 1633-Fire station / \n", "59-Winter sport / 84-Snow / 177-Skiing / 278-GoPro / 281-Snowboarding / 305-Ski / 501-Snowboard / \n", "6-Animation / 15-Cartoon / 311-Street Fighter / 461-Street Fighter IV / 496-Super Street Fighter IV / 563-Ryu / 3936-Gouken / \n", "0-Games / 2-Video game / 8-Football / 90-Sports game / 180-FIFA 15 / 480-FIFA 13 / \n", "5-Dance / \n", "15-Cartoon / 440-Lion / \n", "22-Nature / 121-Photography / 155-Camera / 283-Rock / 629-Camera lens / 2248-Zoom lens / 3922-Wide-angle lens / \n", "93-Comedy / \n", "9-Music video / \n", "0-Games / 8-Football / 21-Stadium / \n", "1-Vehicle / 4-Car / 298-Television advertisement / 683-Renault / \n", "12-Food / 200-Doll / 1487-Reborn doll / \n", "35-Minecraft / 242-Paper / 370-Origami / 385-Face / \n", "0-Games / 2-Video game / 116-Pokémon / 186-Pokémon / 1204-Pokémon Ruby and Sapphire / 1665-Pokémon Emerald / 2236-Emerald / \n", "313-Book / 2464-Bookcase / \n", "12-Food / 26-Cooking / 32-Recipe / \n", "1-Vehicle / 4-Car / 11-Motorsport / 27-Motorcycle / 40-Road / 105-Motorcycling / \n", "28-Fashion / \n", "75-Wrestling / \n", "6-Animation / 25-Toy / 140-LEGO / \n", "1-Vehicle / 4-Car / 11-Motorsport / 41-Sports car / 517-Audi / \n", "0-Games / 51-Strategy video game / 1718-Spore / \n", "0-Games / 8-Football / 21-Stadium / 65-Kick / 76-Ball / 81-Athlete / \n", "9-Music video / 31-Disc jockey / \n", "1-Vehicle / 25-Toy / 62-Train / 1279-Plarail / \n", "1-Vehicle / 4-Car / 11-Motorsport / 17-Racing / 27-Motorcycle / 158-Tire / 378-Drag racing / 1609-Auto Race / \n", "0-Games / 2-Video game / 19-PC game / 51-Strategy video game / 165-League of Legends / \n", "9-Music video / \n", "1-Vehicle / 18-Outdoor recreation / 67-Cycling / 69-Bicycle / 197-Trail / 220-Mountain bike / 284-Mountain biking / 645-Road bicycle / \n", "163-Running / \n", "0-Games / 21-Stadium / 54-Highlight film / 81-Athlete / 562-Pitcher / 946-Baseball park / \n", "113-House / \n", "205-Newscaster / 2642-CNBC / \n", "59-Winter sport / 1829-Sledding / 4446-Bobsleigh / \n", "3-Concert / \n", "1-Vehicle / 4-Car / 27-Motorcycle / 40-Road / 105-Motorcycling / 278-GoPro / \n", "7-Musician / 13-Musical ensemble / 46-Choir / 860-Christian Church / \n", "0-Games / 2-Video game / 19-PC game / 23-Mobile phone / 35-Minecraft / 3253-Mobile game / \n", "12-Food / 1424-Coupon / \n", "47-Personal computer / 141-Microsoft Windows / 1043-Router / \n", "66-Bollywood / \n", "9-Music video / \n", "23-Mobile phone / 29-Smartphone / 72-iPhone / 272-iPod / 331-iPod touch / \n", "743-Edward Cullen / \n", "116-Pokémon / \n", "0-Games / 2-Video game / 1251-The Witcher / 2497-The Witcher 2: Assassins of Kings / \n", "61-Art / 97-Drawing / 512-Printing / 1346-Wood carving / 2967-Printmaking / \n", "10-Animal / 22-Nature / 132-Bird / \n", "121-Photography / 293-Slide show / \n", "3-Concert / 7-Musician / 38-Orchestra / \n", "1-Vehicle / 4-Car / 11-Motorsport / 135-Drifting / \n", "1-Vehicle / 50-Aircraft / 92-Radio-controlled model / 149-Model aircraft / 160-Radio-controlled aircraft / 266-Helicopter / 392-Radio-controlled helicopter / \n", "1-Vehicle / 4-Car / 11-Motorsport / 17-Racing / 41-Sports car / 57-Race track / 135-Drifting / \n", "1-Vehicle / 4-Car / 403-Toyota / 3219-Daihatsu / \n", "83-Skateboarding / 121-Photography / 336-Scooter / 3866-Panasonic Lumix DMC-GH4 / \n", "98-Festival / \n", "0-Games / 2-Video game / 237-Final Fantasy / 1658-Dissidia Final Fantasy / 1857-Sephiroth / 2293-Cloud Strife / \n", "0-Games / 8-Football / 21-Stadium / 65-Kick / 73-Ball / 76-Ball / 77-Arena / \n", "1-Vehicle / 11-Motorsport / 17-Racing / 27-Motorcycle / 105-Motorcycling / 172-Motocross / \n", "232-Cake / 1007-Portal / \n", "8-Football / \n", "0-Games / 2-Video game / 19-PC game / 506-Garry's Mod / 2290-Trouble in Terrorist Town / \n", "1-Vehicle / 4-Car / 11-Motorsport / 27-Motorcycle / 172-Motocross / 491-Enduro / 1520-Minibike / \n", "18-Outdoor recreation / 83-Skateboarding / 108-Skateboard / 221-Skatepark / \n", "0-Games / 25-Toy / 1154-Anpanman / \n", "2-Video game / 1099-Euro Truck Simulator 2 / \n", "0-Games / 35-Minecraft / \n", "0-Games / 8-Football / 21-Stadium / 65-Kick / 73-Ball / \n", "0-Games / 89-Comics / 285-Sasuke Uchiha / 486-M.U.G.E.N / \n", "117-Boxing / \n", "6-Animation / 25-Toy / 126-The Walt Disney Company / 200-Doll / \n", "0-Games / 2-Video game / 20-Trailer / 111-PlayStation 3 / \n", "6-Animation / 15-Cartoon / 1445-Donald Duck / \n", "222-Surfing / \n", "1-Vehicle / 4-Car / 11-Motorsport / 299-Volkswagen Passenger Cars / 864-Speedometer / 1661-Volkswagen Gol / \n", "260-Prayer / \n", "10-Animal / 22-Nature / 280-Wildlife / \n", "7-Musician / 150-Violin / \n", "313-Book / \n", "5-Dance / 616-The Bible / \n", "1-Vehicle / 4-Car / 11-Motorsport / 17-Racing / 40-Road / 212-Highway / 473-Ford Mustang / \n", "162-Medicine / \n", "6-Animation / 15-Cartoon / 126-The Walt Disney Company / \n", "98-Festival / \n", "0-Games / 60-Basketball / \n", "28-Fashion / \n", "10-Animal / 22-Nature / 132-Bird / 217-Hunting / 585-Lawn / 1037-Grass / 4640-Decoy / \n", "1-Vehicle / 84-Snow / 120-Winter / 459-Plough / 1145-Winter storm / 1548-Snowplow / \n", "1-Vehicle / 50-Aircraft / 82-Airplane / 102-Building / 129-Aviation / 139-Airport / 151-Landing / 157-Jet aircraft / 175-Airliner / 187-Takeoff / 188-Airline / 223-Weather / 254-Runway / 499-Cloud / \n", "10-Animal / 12-Food / 48-Pet / 132-Bird / \n", "9-Music video / \n", "14-Guitar / 63-Acoustic guitar / 670-Banjo / \n", "3-Concert / \n", "325-Vampire / \n", "75-Wrestling / \n", "9-Music video / \n", "10-Animal / 18-Outdoor recreation / 80-Horse / 247-Stallion / 289-Dressage / 466-Gift / \n", "209-University / \n", "5-Dance / 6-Animation / 15-Cartoon / 1527-Pretty Cure / \n", "3-Concert / 7-Musician / 16-Performance art / 38-Orchestra / \n", "6-Animation / \n", "3-Concert / \n", "7-Musician / 14-Guitar / 24-String instrument / 63-Acoustic guitar / \n", "130-Gymnastics / \n", "3-Concert / 7-Musician / 30-Drummer / \n", "6-Animation / 2489-Sheriff Woody / \n", "3-Concert / \n", "3-Concert / 978-Love song / \n", "708-Mermaid / \n", "20-Trailer / 229-Halo / \n", "0-Games / 8-Football / 21-Stadium / 54-Highlight film / 65-Kick / 73-Ball / 76-Ball / 77-Arena / 81-Athlete / \n", "10-Animal / 31-Disc jockey / \n", "1-Vehicle / 62-Train / 64-Transport / 103-Rail transport / 119-Locomotive / 143-Railroad car / 428-Rail transport modelling / \n", "0-Games / 1-Vehicle / 2-Video game / 4-Car / 34-Action-adventure game / 41-Sports car / 118-Grand Theft Auto V / \n", "3-Concert / 5-Dance / 7-Musician / 16-Performance art / \n", "1-Vehicle / 11-Motorsport / 50-Aircraft / 139-Airport / 157-Jet aircraft / 175-Airliner / 188-Airline / \n", "3-Concert / 68-Lighting / \n", "0-Games / 2-Video game / 20-Trailer / 176-PlayStation 4 / 474-Destiny / \n", "9-Music video / \n", "0-Games / 2-Video game / 1469-X-Men / \n", "3117-Match / \n", "93-Comedy / \n", "12-Food / 1158-Wheat / \n", "5-Dance / 16-Performance art / \n", "43-Call of Duty / 134-Call of Duty: Black Ops / 138-Call of Duty: Black Ops II / \n", "0-Games / 2-Video game / 20-Trailer / 43-Call of Duty / 134-Call of Duty: Black Ops / 138-Call of Duty: Black Ops II / 3917-Eurogamer / \n", "0-Games / 1-Vehicle / 2-Video game / 4-Car / 34-Action-adventure game / 118-Grand Theft Auto V / 201-Grand Theft Auto: San Andreas / \n", "45-Cosmetics / 56-Hair / 106-Hairstyle / \n", "1-Vehicle / 4-Car / 983-Lexus / \n", "28-Fashion / 56-Hair / 106-Hairstyle / \n", "3-Concert / 7-Musician / 13-Musical ensemble / 357-DVD / \n", "28-Fashion / 45-Cosmetics / 153-Dress / 195-Mascara / 483-Costume / 757-Boot / 1586-Cowboy / \n", "3-Concert / 16-Performance art / 31-Disc jockey / \n", "10-Animal / 91-Fish / \n", "3-Concert / 7-Musician / 13-Musical ensemble / 16-Performance art / 28-Fashion / \n", "2-Video game / 171-Warcraft / \n", "1-Vehicle / 87-Boat / 817-Bulldozer / \n", "6-Animation / 9-Music video / 1934-Hunter × Hunter / \n", "0-Games / 21-Stadium / 81-Athlete / 1111-Home run / \n", "5-Dance / 49-School / 133-Nightclub / 253-Carnival / \n", "45-Cosmetics / 1038-Hijab / \n", "0-Games / 241-Fighting game / 311-Street Fighter / 486-M.U.G.E.N / 1331-Ken Masters / \n", "3-Concert / 7-Musician / 13-Musical ensemble / 46-Choir / \n", "12-Food / 26-Cooking / 32-Recipe / 52-Dish / 58-Cuisine / 184-Vegetable / 274-Meat / 454-Sauce / 500-Rice / 671-Tomato / 3021-Tomato sauce / \n", "5-Dance / \n", "9-Music video / \n", "14-Guitar / 63-Acoustic guitar / \n", "93-Comedy / \n", "3-Concert / \n", "0-Games / 2-Video game / \n", "5-Dance / 1266-Death Note / 3661-Light Yagami / \n", "0-Games / 2-Video game / 6-Animation / 2486-Scooby-Doo / \n", "5-Dance / 98-Festival / 250-Parade / 253-Carnival / \n", "1-Vehicle / 42-Fishing / 82-Airplane / 164-Beach / 223-Weather / 233-Ocean / 423-Coast / \n", "6-Animation / 1206-Cinema 4D / \n", "14-Guitar / \n", "3-Concert / \n", "9-Music video / 411-MapleStory / \n", "0-Games / 88-Machine / 535-Slot machine / \n", "0-Games / 2-Video game / 51-Strategy video game / \n", "5-Dance / \n", "0-Games / 2-Video game / 19-PC game / 241-Fighting game / 486-M.U.G.E.N / \n", "165-League of Legends / \n", "0-Games / 159-World of Warcraft / \n", "102-Building / 113-House / 416-Apartment / \n", "1918-Comet / \n", "260-Prayer / \n", "0-Games / 51-Strategy video game / 165-League of Legends / 3259-Riven / \n", "0-Games / 2-Video game / 19-PC game / 47-Personal computer / \n", "1-Vehicle / 11-Motorsport / 17-Racing / 18-Outdoor recreation / 27-Motorcycle / 105-Motorcycling / 172-Motocross / 197-Trail / 491-Enduro / 1663-Two-stroke engine / 4308-Husaberg / \n", "3-Concert / \n", "287-Human swimming / 329-Swimming pool / 2605-Medley swimming / \n", "18-Outdoor recreation / 267-Mountain / 283-Rock / 347-Climbing / \n", "1-Vehicle / 50-Aircraft / 82-Airplane / 129-Aviation / 139-Airport / 151-Landing / 157-Jet aircraft / 175-Airliner / 187-Takeoff / 188-Airline / 254-Runway / 471-Wing / 719-Boeing 747 / 1340-Airbus A380 / \n", "5-Dance / \n", "38-Orchestra / \n", "22-Nature / \n", "6-Animation / 547-Bleach / \n", "56-Hair / 106-Hairstyle / \n", "359-Harry Potter (Literary Series) / 1436-Draco Malfoy / \n", "3-Concert / 38-Orchestra / \n", "23-Mobile phone / 29-Smartphone / 72-iPhone / 131-Computer / 189-iPad / 272-iPod / 331-iPod touch / \n", "726-Clarinet / \n", "10-Animal / 48-Pet / 132-Bird / \n", "1-Vehicle / 4-Car / 12-Food / 475-Cheese / 1843-Ham / 2852-Burger King / \n", "0-Games / 2-Video game / 141-Microsoft Windows / \n", "9-Music video / 12-Food / 226-Dessert / \n", "347-Climbing / 434-Ninja / \n", "371-Firefighter / 415-Sewing / \n", "0-Games / 256-Tennis / 544-Table tennis / \n", "113-House / 707-Roof / \n", "0-Games / 8-Football / 18-Outdoor recreation / 65-Kick / 73-Ball / 76-Ball / \n", "12-Food / 26-Cooking / 2246-Snail / \n", "9-Music video / \n", "3-Concert / 21-Stadium / 68-Lighting / 365-Fireworks / \n", "1-Vehicle / 10-Animal / 18-Outdoor recreation / 42-Fishing / 87-Boat / 91-Fish / 275-Diving / 288-Underwater / 481-Kayak / \n", "78-Wedding / 263-Light / 1375-Laser lighting display / \n", "3-Concert / 98-Festival / 594-Music festival / \n", "3-Concert / 7-Musician / 13-Musical ensemble / 46-Choir / \n", "10-Animal / 537-Zoo / \n", "96-Soldier / \n", "12-Food / 640-Bottle / 1363-Candle / \n", "0-Games / 2-Video game / 20-Trailer / 51-Strategy video game / 1570-Command & Conquer / 2990-Command & Conquer: Red Alert / 3430-Command & Conquer: Red Alert 3 / \n", "6-Animation / 15-Cartoon / 1696-Raven / 2782-Starfire / \n", "5-Dance / \n", "0-Games / 2-Video game / 335-Nintendo 3DS / 398-The Legend of Zelda / 2289-Super Mario Bros. 3 / 2589-Yoshi's Island / \n", "0-Games / 2-Video game / 34-Action-adventure game / 656-Metal Gear / 3284-Metal Gear Solid 2: Sons of Liberty / \n", "3-Concert / 7-Musician / \n", "1-Vehicle / 4-Car / 336-Scooter / 1448-Vespa / \n", "506-Garry's Mod / \n", "0-Games / 1-Vehicle / 2-Video game / 4-Car / 17-Racing / 34-Action-adventure game / 111-PlayStation 3 / 118-Grand Theft Auto V / \n", "28-Fashion / \n", "1-Vehicle / 11-Motorsport / 27-Motorcycle / 40-Road / \n", "3-Concert / 5-Dance / 16-Performance art / \n", "170-Home improvement / 246-Furniture / \n", "3-Concert / 7-Musician / 13-Musical ensemble / 950-Blu-ray disc / \n", "1-Vehicle / 4-Car / 11-Motorsport / 40-Road / 70-Driving / 94-Dashcam / 473-Ford Mustang / \n", "8-Football / \n", "3-Concert / \n", "1-Vehicle / 62-Train / 64-Transport / 103-Rail transport / 119-Locomotive / 428-Rail transport modelling / \n", "243-Sonic the Hedgehog / 2236-Emerald / \n", "6-Animation / 23-Mobile phone / 29-Smartphone / 101-Telephone / 1366-LG Optimus series / 4407-LG Optimus G / \n", "66-Bollywood / \n", "246-Furniture / 276-Room / 888-Living room / 1281-Couch / \n", "14-Guitar / 465-Microphone / \n", "0-Games / 8-Football / 21-Stadium / 65-Kick / 73-Ball / 76-Ball / 81-Athlete / \n", "6-Animation / 950-Blu-ray disc / 1311-Amazon.com; Inc. / \n", "0-Games / 942-Silkroad Online / \n", "13-Musical ensemble / \n", "3-Concert / \n", "1-Vehicle / 74-Truck / 249-Heavy equipment / 433-Construction / 817-Bulldozer / \n", "205-Newscaster / 766-President of the United States / \n", "3-Concert / 219-Accordion / \n", "219-Accordion / 655-Diatonic button accordion / \n", "307-Hand / \n", "1-Vehicle / 4-Car / 11-Motorsport / 17-Racing / 41-Sports car / 231-Supercar / 569-Lamborghini / \n", "0-Games / 2-Video game / 685-Luigi / \n", "689-Harmonica / 1023-Rose / \n", "1-Vehicle / 4-Car / 11-Motorsport / 17-Racing / 74-Truck / 167-Four-wheel drive / 198-Off-road vehicle / 327-Mud / 565-Mud bogging / \n", "15-Cartoon / 4598-Beavis / 4622-Butt-head / \n", "1300-Earrings / 2814-Keychain / \n", "75-Wrestling / 777-WWE 2K / \n", "0-Games / 2-Video game / 89-Comics / 123-Naruto / 591-Naruto: Ultimate Ninja / \n", "259-Fire / \n", "0-Games / 8-Football / 60-Basketball / \n", "10-Animal / 22-Nature / 86-Plant / 91-Fish / 258-Aquarium / \n", "1-Vehicle / 4-Car / 11-Motorsport / 17-Racing / 41-Sports car / 70-Driving / 135-Drifting / 517-Audi / 982-V8 engine / \n", "46-Choir / \n", "10-Animal / 18-Outdoor recreation / 42-Fishing / 91-Fish / 288-Underwater / \n", "12-Food / \n", "0-Games / 8-Football / 21-Stadium / \n", "31-Disc jockey / \n", "0-Games / 2-Video game / \n", "13-Musical ensemble / \n", "66-Bollywood / \n", "0-Games / 2-Video game / 6-Animation / 20-Trailer / 123-Naruto / 185-Dragon Ball / 244-Goku / 285-Sasuke Uchiha / 396-One Piece / 682-PlayStation Vita / 1930-Rurouni Kenshin / 3232-Naruto Uzumaki / 3469-J-Stars Victory VS / 4078-Toriko / 4652-Weekly Shōnen Jump / \n", "10-Animal / 80-Horse / 1434-Saddle / \n", "14-Guitar / 46-Choir / 1231-Mandolin / \n", "5-Dance / \n", "9-Music video / \n", "23-Mobile phone / 29-Smartphone / 72-iPhone / \n", "411-MapleStory / \n", "1-Vehicle / 4-Car / 11-Motorsport / 17-Racing / 57-Race track / 135-Drifting / 169-Rallying / \n", "3-Concert / 31-Disc jockey / 133-Nightclub / \n", "242-Paper / 476-Textile / 1057-Embroidery / \n", "0-Games / 8-Football / \n", "9-Music video / \n", "76-Ball / 256-Tennis / 974-Racket / 1337-Serve / \n", "3-Concert / 342-Windows Media Video / \n", "402-Jewellery / 829-Bead / 859-Beadwork / \n", "1-Vehicle / 4-Car / 11-Motorsport / 40-Road / 41-Sports car / 207-Exhaust system / 234-Ford / 473-Ford Mustang / 647-Muscle car / 1361-Ford GT / 1877-Shelby Mustang / 3755-AC Cobra / \n", "458-Gold / 768-Diamond / \n", "14-Guitar / \n", "36-Piano / 53-Keyboard / \n", "0-Games / 2-Video game / 34-Action-adventure game / 243-Sonic the Hedgehog / 3366-Amnesia: The Dark Descent / \n", "9-Music video / \n", "3-Concert / \n", "9-Music video / \n", "18-Outdoor recreation / 59-Winter sport / 84-Snow / 177-Skiing / 281-Snowboarding / 305-Ski / 501-Snowboard / \n", "3-Concert / 7-Musician / 13-Musical ensemble / 38-Orchestra / 46-Choir / \n", "3-Concert / 14-Guitar / 24-String instrument / \n", "483-Costume / 2556-Halloween costume / \n", "0-Games / 8-Football / 21-Stadium / 73-Ball / 77-Arena / 90-Sports game / \n", "3-Concert / \n", "23-Mobile phone / 29-Smartphone / 37-Gadget / 155-Camera / 278-GoPro / \n", "5-Dance / \n", "5-Dance / \n", "5-Dance / 16-Performance art / \n", "5-Dance / 16-Performance art / 115-Ballet / 678-Latin dance / \n", "209-University / \n", "3-Concert / 7-Musician / 13-Musical ensemble / 30-Drummer / \n", "23-Mobile phone / 29-Smartphone / 37-Gadget / 101-Telephone / \n", "2-Video game / 37-Gadget / 55-Video game console / 292-Handheld game console / 362-PlayStation Portable / \n", "0-Games / 2-Video game / 6-Animation / 20-Trailer / 55-Video game console / 111-PlayStation 3 / 362-PlayStation Portable / \n", "110-Album / \n", "1-Vehicle / 4-Car / 11-Motorsport / 17-Racing / 41-Sports car / \n", "14-Guitar / 24-String instrument / 63-Acoustic guitar / 606-String / \n", "1-Vehicle / 4-Car / 11-Motorsport / 17-Racing / 25-Toy / 92-Radio-controlled model / 202-Radio-controlled car / \n", "45-Cosmetics / 56-Hair / 385-Face / 2108-Hair removal / \n", "0-Games / 60-Basketball / 76-Ball / \n", "1-Vehicle / 27-Motorcycle / \n", "147-River / \n", "23-Mobile phone / 29-Smartphone / 37-Gadget / 101-Telephone / 178-Samsung Galaxy / 932-Microsoft Lumia / 4611-Nokia Lumia 800 / \n", "9-Music video / \n", "938-Steel / 2221-Deck / \n", "0-Games / 2-Video game / 34-Action-adventure game / 1396-Lara Croft / 3892-Tomb Raider / \n", "122-Weight training / 204-Gym / \n", "1836-Fly tying / \n", "1680-Seafight / \n", "5-Dance / 361-Flamenco / \n", "8-Football / \n", "1-Vehicle / 11-Motorsport / 27-Motorcycle / 135-Drifting / 1887-Tricycle / 3614-Drift trikes / \n", "0-Games / 49-School / 54-Highlight film / 60-Basketball / 168-Basketball moves / \n", "2337-Passport / \n", "912-Silver / \n", "7-Musician / 14-Guitar / \n", "3-Concert / 7-Musician / 14-Guitar / 63-Acoustic guitar / \n", "28-Fashion / 68-Lighting / 121-Photography / 155-Camera / 263-Light / \n", "215-Eating / \n", "45-Cosmetics / 56-Hair / 193-Eye liner / 196-Eye / 218-Eyelash / \n", "31-Disc jockey / \n", "37-Gadget / 121-Photography / 155-Camera / 629-Camera lens / 1855-Video camera / \n", "12-Food / 45-Cosmetics / 173-Eye shadow / 195-Mascara / 215-Eating / 218-Eyelash / 235-Lipstick / \n", "121-Photography / 155-Camera / \n", "5-Dance / 1199-Kinect / \n", "8-Football / 21-Stadium / 77-Arena / \n", "3-Concert / 5-Dance / 16-Performance art / \n", "5-Dance / \n", "23-Mobile phone / 29-Smartphone / \n", "9-Music video / \n", "6-Animation / 15-Cartoon / 750-Sailor Moon / 1342-Sailor Moon / 1551-The Phantom of the Opera / \n", "6-Animation / 10-Animal / 15-Cartoon / 166-Cat / 1209-Bean / \n", "282-Skin / \n", "3-Concert / 5-Dance / 16-Performance art / 93-Comedy / \n", "3-Concert / 14-Guitar / 63-Acoustic guitar / \n", "1057-Embroidery / \n", "10-Animal / 17-Racing / 384-Horse racing / \n", "1-Vehicle / 11-Motorsport / 17-Racing / 27-Motorcycle / 105-Motorcycling / 172-Motocross / \n", "512-Printing / 525-Printer / 1751-Image scanner / 2634-Inkjet printing / 3195-Multifunction printer / \n", "3-Concert / 5-Dance / 13-Musical ensemble / 14-Guitar / 16-Performance art / 39-Drums / 44-Drums / 100-Electric guitar / 376-Flute / \n", "0-Games / 51-Strategy video game / 353-Wall / 373-Clash of Clans / \n", "121-Photography / \n", "42-Fishing / 573-Paragliding / \n", "18-Outdoor recreation / 164-Beach / 222-Surfing / \n", "5-Dance / 95-Talent show / 191-Cheerleading / \n", "6-Animation / 15-Cartoon / \n", "250-Parade / 253-Carnival / \n", "205-Newscaster / \n", "9-Music video / \n", "10-Animal / 540-Insect / 775-Bee / \n", "3-Concert / 7-Musician / 13-Musical ensemble / 46-Choir / 49-School / \n", "6-Animation / 15-Cartoon / 412-Monster / 503-Monster High / \n", "3-Concert / 8-Football / \n", "9-Music video / 78-Wedding / 190-Bride / \n", "1-Vehicle / 64-Transport / 205-Newscaster / 286-Bus / \n", "1095-Shiva / \n", "9-Music video / \n", "13-Musical ensemble / \n", "10-Animal / 588-Reptile / 1444-Lizard / 3576-Iguana / \n", "15-Cartoon / 61-Art / 2550-EarthBound / \n", "20-Trailer / \n", "60-Basketball / 1399-Streetball / 3652-3x3 / \n", "12-Food / 26-Cooking / 32-Recipe / 52-Dish / 58-Cuisine / 184-Vegetable / 542-Potato / 1355-Vegetarian food / 4700-Shashlik / \n", "1-Vehicle / 2-Video game / 4-Car / 11-Motorsport / 17-Racing / 40-Road / 41-Sports car / 57-Race track / 70-Driving / 135-Drifting / 231-Supercar / 705-Forza Motorsport / 2105-Forza Horizon / 2106-Forza Horizon 2 / 2210-McLaren / 3748-McLaren 12C / \n", "39-Drums / 724-Rock Band / 1703-Rock Band / \n", "209-University / \n", "1459-Magazine / 2493-The Hobbit / \n", "93-Comedy / \n", "3342-The Mad Hatter / \n", "0-Games / 2-Video game / 19-PC game / 4140-Quake Live / \n", "122-Weight training / 204-Gym / 214-Muscle / 1430-Pull-up / \n", "1-Vehicle / 249-Heavy equipment / 363-Soil / 1093-Loader / \n", "9-Music video / \n", "3-Concert / 16-Performance art / \n", "7-Musician / 13-Musical ensemble / 46-Choir / \n", "28-Fashion / 153-Dress / 390-Model / 495-Runway / \n", "56-Hair / 106-Hairstyle / 340-Afro-textured hair / 954-Dreadlocks / \n", "3-Concert / \n", "5-Dance / 6-Animation / 126-The Walt Disney Company / \n", "23-Mobile phone / 29-Smartphone / 380-Assassin's Creed / 553-Sony Xperia / 4440-Sony Ericsson Xperia X8 / \n", "1-Vehicle / 4-Car / 11-Motorsport / 270-Sedan / 403-Toyota / 2104-Toyota Camry / \n", "3-Concert / 16-Performance art / \n", "7-Musician / 14-Guitar / 24-String instrument / \n", "3-Concert / 16-Performance art / \n", "1-Vehicle / 4-Car / 11-Motorsport / 27-Motorcycle / 40-Road / 64-Transport / 94-Dashcam / \n", "0-Games / 535-Slot machine / \n", "0-Games / 8-Football / 79-American football / \n", "1-Vehicle / 4-Car / 74-Truck / 154-Wheel / 167-Four-wheel drive / 234-Ford / 375-Pickup truck / 515-Dodge / 1437-Ram Trucks / 1651-Dodge Ram / \n", "3-Concert / 46-Choir / \n", "45-Cosmetics / 173-Eye shadow / 238-Nail / 304-Nail polish / \n", "0-Games / 686-Diablo III / 872-Diablo / \n", "6-Animation / 15-Cartoon / 3689-Woody Woodpecker / \n", "5-Dance / 115-Ballet / \n", "1026-Foreign Exchange Market / \n", "9-Music video / \n", "0-Games / 2-Video game / 782-Super Smash Bros. Melee / 1166-Dark Souls II / \n", "458-Gold / 466-Gift / \n", "72-iPhone / 508-iPhone 5 / \n", "0-Games / 8-Football / 73-Ball / 90-Sports game / 111-PlayStation 3 / 453-Pro Evolution Soccer / \n", "14-Guitar / \n", "3-Concert / 7-Musician / \n", "6-Animation / 477-Transformers / 990-Optimus Prime / \n", "1-Vehicle / 4-Car / 40-Road / 74-Truck / 575-Traffic / \n", "102-Building / 113-House / 170-Home improvement / 276-Room / \n", "1-Vehicle / 50-Aircraft / 82-Airplane / 151-Landing / 175-Airliner / 719-Boeing 747 / 923-Boeing 777 / \n", "0-Games / 60-Basketball / 168-Basketball moves / 255-Slam dunk / \n", "0-Games / 2-Video game / 903-Mass Effect / 1655-Mass Effect 2 / \n", "1-Vehicle / 11-Motorsport / 27-Motorcycle / 105-Motorcycling / \n", "0-Games / 47-Personal computer / 131-Computer / 1293-RollerCoaster Tycoon 3 / \n", "1-Vehicle / 4-Car / 9-Music video / 531-Hatchback / 988-Ibiza / 3990-SEAT Ibiza / \n", "0-Games / 8-Football / 21-Stadium / 65-Kick / 73-Ball / \n", "83-Skateboarding / \n", "28-Fashion / \n", "0-Games / 201-Grand Theft Auto: San Andreas / 1612-Multi Theft Auto / \n", "14-Guitar / 24-String instrument / 63-Acoustic guitar / \n", "10-Animal / 16-Performance art / 61-Art / 636-Rabbit / 981-Sculpture / \n", "376-Flute / \n", "156-Agriculture / 174-Farm / 715-Farming Simulator / 3260-Barley / \n", "7-Musician / 14-Guitar / 24-String instrument / 63-Acoustic guitar / 401-Cello / \n", "1-Vehicle / 50-Aircraft / 82-Airplane / 129-Aviation / 151-Landing / 157-Jet aircraft / 175-Airliner / 187-Takeoff / 388-Fighter aircraft / 758-Air force / \n", "1-Vehicle / 4-Car / 11-Motorsport / 17-Racing / 57-Race track / \n", "12-Food / 482-Asphalt / \n", "0-Games / 35-Minecraft / \n", "0-Games / 51-Strategy video game / \n", "3-Concert / 13-Musical ensemble / 46-Choir / \n", "783-Zee TV / \n", "0-Games / 2-Video game / 37-Gadget / 55-Video game console / 292-Handheld game console / 362-PlayStation Portable / \n", "0-Games / 8-Football / 73-Ball / 90-Sports game / 180-FIFA 15 / \n", "1-Vehicle / 4-Car / 11-Motorsport / 17-Racing / 41-Sports car / 506-Garry's Mod / \n", "6-Animation / 10-Animal / 15-Cartoon / 538-Illustration / 2697-Zebra / \n", "20-Trailer / \n", "489-Board game / 799-Chess / \n", "9-Music video / \n", "45-Cosmetics / 56-Hair / 871-Hair coloring / \n", "10-Animal / 48-Pet / 132-Bird / 3121-Homing pigeon / \n", "0-Games / 33-Weapon / 664-CrossFire / \n", "1-Vehicle / 4-Car / 349-Chevrolet / 406-Chevrolet / 647-Muscle car / 761-Chevrolet Camaro / \n", "163-Running / 1013-Walking / \n", "3-Concert / 7-Musician / \n", "1550-Newspaper / \n", "1-Vehicle / 87-Boat / 949-Rowing / \n", "7-Musician / 14-Guitar / 24-String instrument / 36-Piano / 53-Keyboard / 100-Electric guitar / 1103-Gibson Les Paul / \n", "3-Concert / \n", "5-Dance / 16-Performance art / \n", "6-Animation / 15-Cartoon / 97-Drawing / 183-Manga / 1987-Reborn! / \n", "3-Concert / 5-Dance / 16-Performance art / 361-Flamenco / \n", "10-Animal / 48-Pet / 280-Wildlife / 706-Deer / \n", "10-Animal / 48-Pet / 71-Dog / 2350-Flying disc / \n", "18-Outdoor recreation / 83-Skateboarding / 108-Skateboard / 221-Skatepark / 556-Skateboarding trick / \n", "0-Games / 2-Video game / 83-Skateboarding / 2140-Tony Hawk's / \n", "12-Food / 26-Cooking / 52-Dish / 58-Cuisine / \n", "0-Games / 2-Video game / 89-Comics / 2368-Marvel Universe / \n", "96-Soldier / 1573-United States Army / \n", "0-Games / 2-Video game / 19-PC game / \n", "1-Vehicle / 4-Car / 11-Motorsport / 40-Road / 41-Sports car / 207-Exhaust system / 403-Toyota / 3416-Toyota MR2 / \n", "45-Cosmetics / 173-Eye shadow / 193-Eye liner / 195-Mascara / 196-Eye / 218-Eyelash / 291-Rouge / \n", "0-Games / 319-Robot / \n", "23-Mobile phone / \n", "0-Games / 2-Video game / 116-Pokémon / 186-Pokémon / 1387-Pokémon Battle Revolution / \n", "17-Racing / 23-Mobile phone / 29-Smartphone / 37-Gadget / 67-Cycling / 69-Bicycle / \n", "0-Games / 8-Football / 21-Stadium / 54-Highlight film / 65-Kick / 73-Ball / 76-Ball / 81-Athlete / \n", "9-Music video / \n", "116-Pokémon / 1104-Ash Ketchum / 3347-Misty / \n", "5-Dance / 49-School / 95-Talent show / 191-Cheerleading / \n", "16-Performance art / 49-School / \n", "9-Music video / 905-Flash Video / \n", "3-Concert / 7-Musician / 465-Microphone / \n", "7-Musician / 46-Choir / \n", "20-Trailer / \n", "1-Vehicle / 62-Train / 64-Transport / 103-Rail transport / 119-Locomotive / 143-Railroad car / 152-Track / 161-Train station / 224-Rapid transit / \n", "262-Ice / 312-Ice rink / 833-Ice skate / \n", "7-Musician / 14-Guitar / 24-String instrument / 63-Acoustic guitar / 842-Harp / \n", "75-Wrestling / \n", "0-Games / 8-Football / 21-Stadium / 77-Arena / \n", "5-Dance / \n", "1-Vehicle / 4-Car / 6-Animation / 25-Toy / 140-LEGO / 940-Clay animation / 1079-Lego City / \n", "9-Music video / 938-Steel / \n", "754-Gramophone record / \n", "14-Guitar / 100-Electric guitar / \n", "1-Vehicle / 4-Car / 864-Speedometer / 3229-Opel Kadett / \n", "12-Food / 1162-Dietary supplement / 3498-Whey protein / \n", "20-Trailer / \n", "222-Surfing / \n", "313-Book / 748-Iron Man / 855-Hulk / 1803-Ultron / 3109-Clint Barton / \n", "0-Games / 1-Vehicle / 2-Video game / 6-Animation / 271-Tank / 409-World of Tanks / \n", "28-Fashion / 206-Shoe / 333-Nike; Inc. / 368-Sneakers / \n", "23-Mobile phone / 29-Smartphone / 37-Gadget / 101-Telephone / 487-Samsung Electronics / \n", "3-Concert / 7-Musician / 38-Orchestra / \n", "3-Concert / \n", "3-Concert / 68-Lighting / 263-Light / \n", "1-Vehicle / 4-Car / 11-Motorsport / 17-Racing / 57-Race track / \n", "163-Running / \n", "0-Games / 2-Video game / 33-Weapon / 43-Call of Duty / 194-Call of Duty: Modern Warfare 2 / 302-Call of Duty 4: Modern Warfare / 1425-Gears of War 2 / \n", "60-Basketball / 3905-Slam Dunk / \n", "15-Cartoon / 111-PlayStation 3 / 877-Hatsune Miku: Project DIVA / \n", "0-Games / 2-Video game / \n", "1-Vehicle / 50-Aircraft / 82-Airplane / 129-Aviation / 139-Airport / 151-Landing / 157-Jet aircraft / 175-Airliner / 188-Airline / 254-Runway / \n", "12-Food / \n", "1-Vehicle / 4-Car / 74-Truck / 167-Four-wheel drive / 234-Ford / 375-Pickup truck / \n", "8-Football / 21-Stadium / \n", "3-Concert / 7-Musician / \n", "0-Games / 2-Video game / 20-Trailer / 43-Call of Duty / 138-Call of Duty: Black Ops II / 394-Map / \n", "0-Games / 2-Video game / 75-Wrestling / 85-Combat / 90-Sports game / 111-PlayStation 3 / 117-Boxing / 127-Xbox 360 / 181-Xbox / 700-Hard disk drive / 3907-UFC Undisputed 3 / \n", "5-Dance / \n", "0-Games / 2-Video game / 432-Kingdom Hearts / 665-Kingdom Hearts / 836-Sora / 941-Kingdom Hearts II / 3299-Kingdom Hearts: Chain of Memories / \n", "3-Concert / 5-Dance / 16-Performance art / \n", "5-Dance / 16-Performance art / 980-Solo dance / \n", "0-Games / 2-Video game / 89-Comics / 185-Dragon Ball / \n", "12-Food / 26-Cooking / 32-Recipe / 52-Dish / 58-Cuisine / 109-Cooking show / 179-Kitchen / 210-Cookware and bakeware / 239-Roasting / 533-Oil / 662-Pizza / 2483-Deep frying / \n", "0-Games / 2-Video game / 34-Action-adventure game / 297-Grand Theft Auto IV / \n", "5-Dance / 16-Performance art / 123-Naruto / \n", "6-Animation / 102-Building / 433-Construction / 666-Concrete / \n", "415-Sewing / 571-Bear / 1152-Plush / 1212-Stuffed toy / 2329-Felt / \n", "5-Dance / 31-Disc jockey / \n", "18-Outdoor recreation / 42-Fishing / 395-Camping / 914-Tent / \n", "444-Family / \n", "1-Vehicle / 11-Motorsport / 124-Tractor / 156-Agriculture / 252-Tractor pulling / 459-Plough / \n", "455-Unidentified flying object / \n", "1-Vehicle / 7-Musician / 13-Musical ensemble / 49-School / 54-Highlight film / 79-American football / 203-Marching band / 306-High school / \n", "59-Winter sport / 84-Snow / 177-Skiing / 305-Ski / 789-Alpine skiing / \n", "3-Concert / \n", "0-Games / 2-Video game / 19-PC game / 33-Weapon / 43-Call of Duty / 704-Call of Duty: World at War / 1248-Call of Duty: Zombies / \n", "30-Drummer / 39-Drums / 44-Drums / 128-Cymbal / 146-Snare drum / \n", "0-Games / 2-Video game / 23-Mobile phone / 29-Smartphone / 37-Gadget / \n", "59-Winter sport / 83-Skateboarding / 108-Skateboard / 112-Ice skating / 222-Surfing / \n", "673-Saw / 4161-Harbor Freight Tools / \n", "3002-Essential oil / \n", "0-Games / 8-Football / 21-Stadium / 65-Kick / 81-Athlete / 358-Rugby football / \n", "5-Dance / 49-School / \n", "0-Games / 1-Vehicle / 2-Video game / 4-Car / 11-Motorsport / 17-Racing / 40-Road / 57-Race track / 169-Rallying / 2518-Dirt 3 / \n", "455-Unidentified flying object / \n", "1-Vehicle / 25-Toy / 27-Motorcycle / 140-LEGO / 1478-Lego Ninjago / 1925-Skull / \n", "1-Vehicle / 4-Car / 6-Animation / \n", "1-Vehicle / 6-Animation / 25-Toy / 50-Aircraft / 82-Airplane / 139-Airport / 140-LEGO / 1079-Lego City / 1302-The Lego Group / \n", "20-Trailer / 229-Halo / 866-Halo 4 / \n", "1-Vehicle / 4-Car / 11-Motorsport / 17-Racing / 27-Motorcycle / 57-Race track / 105-Motorcycling / \n", "8-Football / 21-Stadium / \n", "78-Wedding / \n", "45-Cosmetics / 56-Hair / 173-Eye shadow / 193-Eye liner / 195-Mascara / 196-Eye / 218-Eyelash / 235-Lipstick / 282-Skin / 291-Rouge / 385-Face / 436-Concealer / \n", "0-Games / 2-Video game / 19-PC game / 33-Weapon / 43-Call of Duty / 96-Soldier / 216-Call of Duty: Modern Warfare 3 / \n", "0-Games / 2-Video game / 19-PC game / 192-Battlefield / 374-Battlefield 3 / \n", "0-Games / 1-Vehicle / 11-Motorsport / 17-Racing / 57-Race track / \n", "10-Animal / 48-Pet / 71-Dog / \n", "12-Food / 26-Cooking / 52-Dish / 58-Cuisine / 215-Eating / 771-Noodle / 1438-Street food / 2178-Chinese food / \n", "1-Vehicle / 4-Car / 11-Motorsport / 17-Racing / 57-Race track / 569-Lamborghini / \n", "0-Games / 4685-Shot put / \n", "2642-CNBC / \n", "10-Animal / 48-Pet / 71-Dog / 182-Puppy / 1084-Bulldog / 2438-French Bulldog / \n", "0-Games / 2-Video game / 19-PC game / 33-Weapon / 96-Soldier / 638-Gears of War / 773-Gears of War / 1425-Gears of War 2 / \n", "0-Games / 35-Minecraft / \n", "10-Animal / 91-Fish / 258-Aquarium / \n", "3-Concert / 5-Dance / 78-Wedding / \n", "273-News program / \n", "1-Vehicle / 4-Car / 345-Loudspeaker / 484-Subwoofer / \n", "45-Cosmetics / 196-Eye / \n", "3-Concert / \n", "5-Dance / \n", "0-Games / 1-Vehicle / 2-Video game / 34-Action-adventure game / 297-Grand Theft Auto IV / 658-Grand Theft Auto: The Lost and Damned / \n", "1107-United States Navy / \n", "9-Music video / 843-The Doctor / \n", "205-Newscaster / 307-Hand / \n", "9-Music video / 121-Photography / 153-Dress / \n", "242-Paper / 370-Origami / \n", "9-Music video / \n", "45-Cosmetics / 56-Hair / 106-Hairstyle / \n", "23-Mobile phone / 29-Smartphone / 37-Gadget / 101-Telephone / 2834-Huawei Ascend / \n", "110-Album / \n", "23-Mobile phone / 29-Smartphone / 37-Gadget / 72-iPhone / 101-Telephone / 2191-Text messaging / \n", "1-Vehicle / 27-Motorcycle / \n", "8-Football / 21-Stadium / \n", "205-Newscaster / 354-Egg / 500-Rice / \n", "0-Games / 60-Basketball / 168-Basketball moves / 255-Slam dunk / \n", "10-Animal / 12-Food / 26-Cooking / 32-Recipe / 91-Fish / 274-Meat / 893-Fish as food / \n", "10-Animal / 91-Fish / 258-Aquarium / 716-Coral / 1051-Reef aquarium / 1368-Gallon / \n", "79-American football / \n", "60-Basketball / 3291-Basketball Coach / \n", "22-Nature / 275-Diving / 288-Underwater / 381-Underwater diving / 407-Scuba diving / 425-City / \n", "75-Wrestling / 1042-Jumbotron / \n", "59-Winter sport / 177-Skiing / 789-Alpine skiing / 1764-Ski jumping / \n", "78-Wedding / \n", "8-Football / 350-Circus / 788-Juggling / \n", "222-Surfing / 672-Bodyboarding / \n", "46-Choir / 1501-Military parade / \n", "5-Dance / \n", "12-Food / 26-Cooking / 533-Oil / \n", "0-Games / 1-Vehicle / 4-Car / 11-Motorsport / 17-Racing / 1817-Initial D Arcade Stage / \n", "66-Bollywood / \n", "9-Music video / \n", "45-Cosmetics / 919-Hair conditioner / \n", "8-Football / \n", "10-Animal / 71-Dog / \n", "76-Ball / 3264-Golf ball / \n", "5-Dance / 16-Performance art / 1289-Caporales / \n", "5-Dance / 16-Performance art / 115-Ballet / \n", "12-Food / 26-Cooking / 32-Recipe / 52-Dish / 179-Kitchen / 210-Cookware and bakeware / 239-Roasting / 274-Meat / 1186-Seafood / \n", "5-Dance / 6-Animation / 15-Cartoon / 243-Sonic the Hedgehog / 251-Sonic the Hedgehog / \n", "1-Vehicle / 4-Car / 270-Sedan / 3134-Mazda6 / \n", "12-Food / 136-Gardening / 666-Concrete / \n", "20-Trailer / \n", "3-Concert / 7-Musician / 13-Musical ensemble / 30-Drummer / 38-Orchestra / 39-Drums / \n", "10-Animal / 14-Guitar / 48-Pet / 71-Dog / 2271-Boxer / \n", "67-Cycling / 69-Bicycle / 83-Skateboarding / 221-Skatepark / 377-BMX bike / \n", "42-Fishing / 217-Hunting / \n", "3-Concert / 68-Lighting / 258-Aquarium / \n", "9-Music video / \n", "59-Winter sport / 84-Snow / 177-Skiing / 305-Ski / 789-Alpine skiing / \n", "5-Dance / \n", "3-Concert / 7-Musician / 13-Musical ensemble / 46-Choir / 49-School / \n", "28-Fashion / \n", "5-Dance / 95-Talent show / \n", "10-Animal / 709-Bull / \n", "6-Animation / \n", "0-Games / 2086-Dominoes / \n", "0-Games / 8-Football / 21-Stadium / 65-Kick / 76-Ball / 81-Athlete / \n", "0-Games / 2-Video game / 19-PC game / 3362-WildStar / 4655-The Secret World / \n", "28-Fashion / 78-Wedding / 153-Dress / 190-Bride / 421-Gown / 590-Wedding dress / 607-Lace / \n", "5-Dance / 2341-Fine art / \n", "259-Fire / 371-Firefighter / \n", "1-Vehicle / 4-Car / 270-Sedan / 517-Audi / 1587-Audi Quattro / 2306-Audi A6 / \n", "75-Wrestling / 85-Combat / 287-Human swimming / \n", "6-Animation / 15-Cartoon / 61-Art / 89-Comics / 183-Manga / 1696-Raven / 2782-Starfire / 2945-Fan art / \n", "6-Animation / 15-Cartoon / 1866-Katara / 3734-Aang / \n", "12-Food / 301-Meal / \n", "35-Minecraft / \n", "5-Dance / \n", "3-Concert / 7-Musician / 13-Musical ensemble / 14-Guitar / 39-Drums / \n", "9-Music video / \n", "61-Art / 479-Graffiti / \n", "0-Games / 8-Football / 21-Stadium / 73-Ball / 77-Arena / 90-Sports game / 180-FIFA 15 / \n", "0-Games / 2-Video game / 19-PC game / 166-Cat / 850-Left 4 Dead / 963-Left 4 Dead 2 / \n", "1-Vehicle / 4-Car / 40-Road / 74-Truck / 167-Four-wheel drive / 198-Off-road vehicle / 277-Sport utility vehicle / 375-Pickup truck / 901-Mitsubishi / \n", "0-Games / 2-Video game / 635-Achievement / 831-Plants vs. Zombies / \n", "8-Football / 164-Beach / \n", "12-Food / 475-Cheese / \n", "0-Games / 2-Video game / 35-Minecraft / 47-Personal computer / \n", "3-Concert / 7-Musician / 16-Performance art / \n", "3-Concert / 5-Dance / 16-Performance art / \n", "8-Football / \n", "6-Animation / 15-Cartoon / 89-Comics / 801-Inuyasha / \n", "0-Games / 8-Football / 54-Highlight film / 79-American football / \n", "56-Hair / 106-Hairstyle / 340-Afro-textured hair / 405-Braid / 857-Afro / 1812-Hair twists / 2351-Moisturizer / 2356-Comb / \n", "1-Vehicle / 11-Motorsport / 67-Cycling / 69-Bicycle / 220-Mountain bike / \n", "12-Food / 26-Cooking / 32-Recipe / 226-Dessert / 232-Cake / 437-Cream / 475-Cheese / 560-Sugar / 759-Icing / 841-Butter / 1853-Whipped cream / \n", "31-Disc jockey / \n", "6-Animation / 15-Cartoon / 3734-Aang / \n", "1076-Wakeboarding / \n", "10-Animal / 751-Desert / 2410-Camel / \n", "12-Food / 52-Dish / \n", "0-Games / 2-Video game / 6-Animation / 15-Cartoon / 89-Comics / 185-Dragon Ball / 244-Goku / 869-Dragon Ball Z: Budokai Tenkaichi / 876-Freeza / \n", "0-Games / 2-Video game / 19-PC game / 1071-Hearthstone: Heroes of Warcraft / \n", "5-Dance / 191-Cheerleading / \n", "23-Mobile phone / 29-Smartphone / 37-Gadget / 47-Personal computer / 145-Tablet computer / 178-Samsung Galaxy / 1287-Samsung Galaxy Tab series / 3863-Temple Run / \n", "12-Food / \n", "125-Television / 609-Tattoo / \n", "3-Concert / 7-Musician / 16-Performance art / \n", "0-Games / 2-Video game / 2712-Just Cause / \n", "102-Building / 113-House / 339-Architecture / \n", "9-Music video / \n", "0-Games / 8-Football / 79-American football / 81-Athlete / \n", "68-Lighting / 3803-Jaguar / \n", "6-Animation / \n", "5-Dance / \n", "122-Weight training / \n", "0-Games / 1680-Seafight / \n", "1684-Hillsong Church / \n", "0-Games / 2-Video game / 320-Super Smash Bros. / 782-Super Smash Bros. Melee / 2085-Falco Lombardi / 2165-Ganon / \n", "0-Games / 6-Animation / 141-Microsoft Windows / \n", "3-Concert / 7-Musician / 3489-Rock and Roll Hall of Fame / \n", "0-Games / 1-Vehicle / 2-Video game / 11-Motorsport / 456-Go-kart / 518-Mario Kart / 1831-Mario Kart 8 / 2901-Mario Kart 64 / 3095-Mario Kart DS / 3346-Super Mario Kart / \n", "1-Vehicle / 345-Loudspeaker / \n", "1-Vehicle / 62-Train / 102-Building / \n", "359-Harry Potter (Literary Series) / 1436-Draco Malfoy / 1856-Hogwarts School of Witchcraft and Wizardry / 1912-Lord Voldemort / \n", "3-Concert / 7-Musician / 13-Musical ensemble / 36-Piano / 46-Choir / \n", "3-Concert / 7-Musician / 13-Musical ensemble / 16-Performance art / \n", "1-Vehicle / 4-Car / 245-BMW / \n", "0-Games / 577-Penguin / \n", "1-Vehicle / 4-Car / 11-Motorsport / 17-Racing / 40-Road / 41-Sports car / 57-Race track / 70-Driving / 1537-Audi R8 / 2949-V10 engine / \n", "3-Concert / 5-Dance / 16-Performance art / \n", "1-Vehicle / 11-Motorsport / 27-Motorcycle / 84-Snow / 172-Motocross / 278-GoPro / 4014-Yamaha YZ250 / 4436-Yamaha YZ450F / 4536-Kawasaki KX250F / \n", "0-Games / 8-Football / \n", "0-Games / 2-Video game / \n", "1-Vehicle / 67-Cycling / 69-Bicycle / 1036-Moped / \n", "0-Games / 2-Video game / 23-Mobile phone / \n", "3-Concert / 7-Musician / 13-Musical ensemble / 46-Choir / \n", "3-Concert / 7-Musician / 13-Musical ensemble / 14-Guitar / 30-Drummer / 128-Cymbal / \n", "93-Comedy / \n", "3-Concert / 7-Musician / 13-Musical ensemble / 24-String instrument / \n", "1-Vehicle / 87-Boat / 308-Ship / 579-Yacht / \n", "9-Music video / \n", "10-Animal / 18-Outdoor recreation / 42-Fishing / 91-Fish / 478-Fishing rod / \n", "25-Toy / 140-LEGO / \n", "12-Food / 86-Plant / 225-Lake / \n", "5-Dance / 31-Disc jockey / \n", "12-Food / 26-Cooking / 32-Recipe / 52-Dish / 58-Cuisine / 239-Roasting / 500-Rice / 572-Indian cuisine / 727-Curry / 808-Breakfast / \n", "6-Animation / \n", "9-Music video / \n", "49-School / 1640-Library / \n", "102-Building / 259-Fire / \n", "248-Hotel / \n", "66-Bollywood / \n", "925-Brain / \n", "341-Sketch comedy / \n", "0-Games / 2-Video game / 19-PC game / 192-Battlefield / 383-Battlefield 4 / \n", "75-Wrestling / 1441-Mattel / \n", "720-Wine / \n", "2-Video game / 55-Video game console / 111-PlayStation 3 / 127-Xbox 360 / 199-Wii / 317-PlayStation / \n", "3218-Dragonica / \n", "2-Video game / 9-Music video / 229-Halo / 644-Halo: Reach / 866-Halo 4 / 2240-Halo 5: Guardians / 2796-Halo: The Master Chief Collection / \n", "250-Parade / \n", "36-Piano / 53-Keyboard / 104-Pianist / 107-Musical keyboard / \n", "467-Chipmunk / \n", "5-Dance / \n", "0-Games / 2-Video game / 19-PC game / 159-World of Warcraft / 171-Warcraft / \n", "0-Games / 19-PC game / 211-RuneScape / \n", "0-Games / 2-Video game / 51-Strategy video game / \n", "144-Amusement park / 321-Roller coaster / 334-Amusement ride / \n", "0-Games / 2-Video game / 34-Action-adventure game / 118-Grand Theft Auto V / \n", "0-Games / 2-Video game / 19-PC game / 33-Weapon / 96-Soldier / 1394-Alliance of Valiant Arms / \n", "28-Fashion / 206-Shoe / 757-Boot / \n", "3-Concert / 5-Dance / 16-Performance art / \n", "33-Weapon / 693-Sword / \n", "3-Concert / 5-Dance / \n", "3-Concert / \n", "23-Mobile phone / 29-Smartphone / 802-Google Nexus / \n", "325-Vampire / \n", "1-Vehicle / 4-Car / 11-Motorsport / 17-Racing / 41-Sports car / 135-Drifting / 169-Rallying / 1041-Subaru Impreza / 1286-Subaru / \n", "9-Music video / \n", "516-Headphones / 832-Headset / \n", "2-Video game / 83-Skateboarding / 90-Sports game / 2140-Tony Hawk's / \n", "7-Musician / 13-Musical ensemble / 38-Orchestra / 219-Accordion / \n", "1-Vehicle / 88-Machine / 512-Printing / \n", "28-Fashion / 56-Hair / 106-Hairstyle / \n", "3-Concert / 5-Dance / \n", "3895-Curling / \n", "1-Vehicle / 4-Car / 70-Driving / 270-Sedan / 332-Mercedes-Benz / 2365-Mercedes-Benz S-Class / \n", "89-Comics / 269-Comic book / 387-Spider-Man / \n", "350-Circus / \n", "0-Games / 2-Video game / 127-Xbox 360 / 386-Xbox / 948-Far Cry / 1593-Digital video recorder / \n", "0-Games / 2-Video game / 6-Animation / 19-PC game / 1343-Super Robot Wars / 3128-Super Robot / \n", "1325-Gel / 2731-Aloe vera / 3770-Toothpaste / \n", "1-Vehicle / 4-Car / \n", "273-News program / \n", "68-Lighting / 133-Nightclub / \n", "9-Music video / \n", "5-Dance / 16-Performance art / 153-Dress / 762-Angel / \n", "7-Musician / 13-Musical ensemble / 46-Choir / \n", "298-Television advertisement / 337-Advertising / \n", "1-Vehicle / 2903-Umbrella / \n", "23-Mobile phone / 29-Smartphone / 37-Gadget / 101-Telephone / 178-Samsung Galaxy / 1207-Samsung Galaxy S5 / \n", "1-Vehicle / 12-Food / 2937-Banknote / \n", "9-Music video / \n", "0-Games / 379-Touhou Project / 3849-Perfect Cherry Blossom / \n", "3-Concert / 7-Musician / \n", "28-Fashion / 495-Runway / \n", "5-Dance / 28-Fashion / \n", "102-Building / 113-House / 170-Home improvement / \n", "3-Concert / \n", "1-Vehicle / 62-Train / 64-Transport / 103-Rail transport / 119-Locomotive / 143-Railroad car / 2413-Union Pacific Railroad / \n", "0-Games / 95-Talent show / 191-Cheerleading / \n", "38-Orchestra / \n", "5-Dance / 153-Dress / 430-Ballroom dance / 678-Latin dance / \n", "571-Bear / 2134-Teddy bear / \n", "12-Food / 1022-Chili pepper / \n", "1816-Fire Emblem / \n", "3-Concert / 7-Musician / 13-Musical ensemble / 36-Piano / 46-Choir / \n", "452-Watch / 627-Clock / \n", "1-Vehicle / 4-Car / 27-Motorcycle / \n", "1-Vehicle / 62-Train / 259-Fire / \n", "1-Vehicle / 11-Motorsport / 17-Racing / 27-Motorcycle / 105-Motorcycling / 172-Motocross / 197-Trail / 213-Forest / 363-Soil / \n", "0-Games / 2-Video game / 55-Video game console / 237-Final Fantasy / 335-Nintendo 3DS / \n", "3-Concert / \n", "121-Photography / 155-Camera / \n", "0-Games / 211-RuneScape / \n", "93-Comedy / 264-Star Wars / \n", "0-Games / 2-Video game / 55-Video game console / 199-Wii / 362-PlayStation Portable / \n", "6-Animation / 15-Cartoon / 307-Hand / 733-Washing / \n", "28-Fashion / 45-Cosmetics / 56-Hair / \n", "93-Comedy / 293-Slide show / \n", "0-Games / 1-Vehicle / 2-Video game / 124-Tractor / 174-Farm / 715-Farming Simulator / \n", "0-Games / 2-Video game / \n", "133-Nightclub / \n", "1-Vehicle / 4-Car / 11-Motorsport / 40-Road / 94-Dashcam / 212-Highway / 668-Volkswagen Golf / \n", "3-Concert / 7-Musician / \n", "9-Music video / \n", "3-Concert / \n", "341-Sketch comedy / 364-Sitcom / \n", "5-Dance / \n", "1-Vehicle / 18-Outdoor recreation / 67-Cycling / 69-Bicycle / 197-Trail / 220-Mountain bike / 284-Mountain biking / \n", "3-Concert / 7-Musician / 13-Musical ensemble / 46-Choir / \n", "1-Vehicle / 4-Car / 27-Motorcycle / 40-Road / 336-Scooter / 3685-Pulley / \n", "12-Food / 26-Cooking / 32-Recipe / 52-Dish / 58-Cuisine / 109-Cooking show / 210-Cookware and bakeware / 727-Curry / 1024-Culinary art / 1209-Bean / 3100-Chickpea / \n", "10-Animal / 18-Outdoor recreation / 42-Fishing / 91-Fish / 268-Recreational fishing / 601-Shark / \n", "345-Loudspeaker / 484-Subwoofer / \n", "3-Concert / 5-Dance / 16-Performance art / 115-Ballet / \n", "6-Animation / 2951-Black Rock Shooter / \n", "3-Concert / 7-Musician / 13-Musical ensemble / 46-Choir / 318-Church / \n", "22-Nature / \n", "1-Vehicle / 4-Car / 11-Motorsport / 41-Sports car / 207-Exhaust system / 234-Ford / 473-Ford Mustang / 647-Muscle car / 650-Convertible / 1361-Ford GT / \n", "0-Games / 2-Video game / 903-Mass Effect / \n", "1-Vehicle / 4-Car / 11-Motorsport / 41-Sports car / 952-Chevrolet Corvette / \n", "20-Trailer / \n", "1-Vehicle / 67-Cycling / 69-Bicycle / 490-Rain / 645-Road bicycle / 1887-Tricycle / \n", "72-iPhone / \n", "31-Disc jockey / \n", "14-Guitar / \n", "12-Food / 32-Recipe / 226-Dessert / 239-Roasting / 295-Chocolate / 770-Banana / 1085-Barbecue grill / \n", "3-Concert / 68-Lighting / \n", "2671-Digital electronics / \n", "0-Games / 1-Vehicle / 2-Video game / 19-PC game / 33-Weapon / 51-Strategy video game / 271-Tank / 409-World of Tanks / \n", "0-Games / 54-Highlight film / 60-Basketball / 81-Athlete / 168-Basketball moves / \n", "6-Animation / 15-Cartoon / 89-Comics / 123-Naruto / 272-iPod / \n", "0-Games / 2-Video game / 35-Minecraft / \n", "0-Games / 5-Dance / 191-Cheerleading / \n", "6-Animation / 15-Cartoon / 123-Naruto / 285-Sasuke Uchiha / \n", "3-Concert / 98-Festival / \n", "9-Music video / \n", "0-Games / 60-Basketball / 666-Concrete / \n", "1-Vehicle / 68-Lighting / 92-Radio-controlled model / 149-Model aircraft / 160-Radio-controlled aircraft / 263-Light / 393-Unmanned aerial vehicle / 604-Quadcopter / \n", "25-Toy / 290-Pony / 343-My Little Pony / 417-Play-Doh / \n", "3-Concert / \n", "10-Animal / 550-Dinosaur / \n", "1-Vehicle / 4-Car / 11-Motorsport / 252-Tractor pulling / 585-Lawn / 741-Mower / 805-Lawn mower / \n", "1100-Hat / 3817-Garage sale / \n", "641-Mask / 2242-Pillow / \n", "9-Music video / 163-Running / \n", "0-Games / 49-School / 60-Basketball / 168-Basketball moves / \n", "10-Animal / 18-Outdoor recreation / 42-Fishing / 91-Fish / 147-River / 510-Fishing lure / 844-Bass / \n", "10-Animal / 88-Machine / 319-Robot / 354-Egg / \n", "5-Dance / \n", "3-Concert / 5-Dance / \n", "5-Dance / \n", "47-Personal computer / 1045-Computer case / 2481-Cooler Master / \n", "1-Vehicle / 4-Car / 11-Motorsport / 17-Racing / \n", "99-Christmas / 2424-Greeting card / \n", "1-Vehicle / 4-Car / \n", "18-Outdoor recreation / 83-Skateboarding / 108-Skateboard / 221-Skatepark / \n", "3-Concert / 68-Lighting / \n", "8-Football / \n", "0-Games / 2-Video game / 55-Video game console / 386-Xbox / \n", "3-Concert / \n", "0-Games / 90-Sports game / 180-FIFA 15 / \n", "205-Newscaster / 273-News program / \n", "1-Vehicle / 4-Car / 11-Motorsport / 41-Sports car / 70-Driving / 2261-Nissan Silvia / \n", "9-Music video / \n", "28-Fashion / 206-Shoe / 333-Nike; Inc. / 368-Sneakers / \n", "1-Vehicle / 11-Motorsport / 27-Motorcycle / 105-Motorcycling / \n", "1-Vehicle / 308-Ship / \n", "0-Games / 2-Video game / 47-Personal computer / 192-Battlefield / 602-Central processing unit / 1996-Intel Core i7 / \n", "0-Games / 2-Video game / 34-Action-adventure game / 111-PlayStation 3 / 413-PlayStation 2 / 3263-Shadow of the Colossus / \n", "0-Games / 2-Video game / 34-Action-adventure game / 398-The Legend of Zelda / \n", "9-Music video / \n", "0-Games / 60-Basketball / 168-Basketball moves / 255-Slam dunk / 1101-Point guard / \n", "1-Vehicle / 4-Car / 40-Road / 70-Driving / 94-Dashcam / 1415-Kia Motors / 4060-Kia Sportage / \n", "36-Piano / 53-Keyboard / 104-Pianist / 107-Musical keyboard / \n", "10-Animal / 18-Outdoor recreation / 22-Nature / 217-Hunting / 1215-Wild boar / \n", "83-Skateboarding / 108-Skateboard / \n", "237-Final Fantasy / 730-Final Fantasy VII / 1857-Sephiroth / 2041-Crisis Core: Final Fantasy VII / \n", "0-Games / 2-Video game / 34-Action-adventure game / 297-Grand Theft Auto IV / \n", "6-Animation / 15-Cartoon / 61-Art / 97-Drawing / \n", "5-Dance / 31-Disc jockey / 78-Wedding / \n", "1-Vehicle / 4-Car / 11-Motorsport / 17-Racing / 57-Race track / 787-Monster truck / \n", "0-Games / 34-Action-adventure game / 201-Grand Theft Auto: San Andreas / \n", "1-Vehicle / 4-Car / 11-Motorsport / 17-Racing / 40-Road / 41-Sports car / 57-Race track / 299-Volkswagen Passenger Cars / 526-Volkswagen / 711-Gran Turismo / 953-Gran Turismo 5 / 1519-Gran Turismo 6 / 1761-Volkswagen Jetta / \n", "0-Games / 2-Video game / 19-PC game / 33-Weapon / 43-Call of Duty / 704-Call of Duty: World at War / \n", "0-Games / 6-Animation / 15-Cartoon / 877-Hatsune Miku: Project DIVA / \n", "298-Television advertisement / 337-Advertising / 640-Bottle / 720-Wine / \n", "1033-Drill / 1226-Metalworking / 1995-Hammer / \n", "861-Mixing console / \n", "164-Beach / 447-Sand / \n", "5-Dance / 16-Performance art / 95-Talent show / \n", "2389-Exercise ball / \n", "15-Cartoon / 61-Art / 97-Drawing / 408-Sketch / \n", "4490-Nier / \n", "1-Vehicle / 4-Car / \n", "28-Fashion / 56-Hair / 919-Hair conditioner / \n", "3-Concert / 5-Dance / 16-Performance art / 412-Monster / \n", "15-Cartoon / 89-Comics / 123-Naruto / \n", "0-Games / 2-Video game / 1000-AdventureQuest Worlds / \n", "2-Video game / 420-The Elder Scrolls V: Skyrim / 441-The Elder Scrolls / \n", "0-Games / 530-Cue sports / 568-Pool / 994-Cue stick / 2524-Carom billiards / \n", "1-Vehicle / 4-Car / 34-Action-adventure game / 1661-Volkswagen Gol / \n", "0-Games / 90-Sports game / 180-FIFA 15 / \n", "14-Guitar / 345-Loudspeaker / 689-Harmonica / 738-Guitar amplifier / \n", "3-Concert / 7-Musician / 14-Guitar / \n", "3-Concert / 14-Guitar / \n", "7-Musician / 36-Piano / 53-Keyboard / 878-Composer / \n", "66-Bollywood / \n", "0-Games / 1-Vehicle / 2-Video game / 4-Car / 34-Action-adventure game / 118-Grand Theft Auto V / 663-Street / \n", "9-Music video / \n", "2227-Multimedios Televisión / \n", "1-Vehicle / 4-Car / 40-Road / 70-Driving / 270-Sedan / \n", "25-Toy / 323-Toddler / \n", "46-Choir / \n", "46-Choir / \n", "1-Vehicle / 27-Motorcycle / 114-Engine / \n", "5-Dance / \n", "14-Guitar / 63-Acoustic guitar / 2744-Bajo sexto / \n", "311-Street Fighter / 496-Super Street Fighter IV / \n", "3-Concert / 16-Performance art / \n", "8-Football / \n", "1-Vehicle / 18-Outdoor recreation / 87-Boat / 225-Lake / 233-Ocean / \n", "1-Vehicle / 4-Car / 303-Paint / \n", "8-Football / \n", "3-Concert / 7-Musician / 13-Musical ensemble / 14-Guitar / \n", "126-The Walt Disney Company / 2904-Express train / \n", "2102-IKEA / \n", "9-Music video / \n", "75-Wrestling / \n", "14-Guitar / 24-String instrument / 639-Fiddle / 670-Banjo / \n", "13-Musical ensemble / \n", "14-Guitar / 24-String instrument / 63-Acoustic guitar / \n", "6-Animation / 25-Toy / 477-Transformers / 990-Optimus Prime / \n", "3-Concert / 16-Performance art / \n", "6-Animation / 15-Cartoon / \n", "1-Vehicle / 4-Car / 11-Motorsport / 27-Motorcycle / 864-Speedometer / 2543-2001 GSX-R1000 / \n", "13-Musical ensemble / \n", "12-Food / 26-Cooking / 32-Recipe / 58-Cuisine / 109-Cooking show / 179-Kitchen / 557-Salad / \n", "12-Food / 86-Plant / 137-Tree / 348-Fruit / 770-Banana / \n", "10-Animal / 80-Horse / 203-Marching band / 217-Hunting / \n", "415-Sewing / 476-Textile / 1479-Quilt / \n", "0-Games / 8-Football / 21-Stadium / 65-Kick / 76-Ball / 81-Athlete / \n", "0-Games / 2-Video game / 6-Animation / 15-Cartoon / 708-Mermaid / 1092-Game Boy Advance / \n", "0-Games / 35-Minecraft / \n", "0-Games / 2-Video game / 19-PC game / 51-Strategy video game / 1072-Total War / \n", "31-Disc jockey / \n", "5-Dance / 889-Swing / \n", "56-Hair / 583-Wig / \n", "6-Animation / 15-Cartoon / 61-Art / 97-Drawing / 123-Naruto / 183-Manga / 408-Sketch / \n", "75-Wrestling / 269-Comic book / 313-Book / \n", "8-Football / 623-Coach / \n", "6-Animation / 15-Cartoon / \n", "90-Sports game / 99-Christmas / 180-FIFA 15 / 466-Gift / \n", "1-Vehicle / 4-Car / 680-Subaru / 1286-Subaru / \n", "47-Personal computer / 97-Drawing / 145-Tablet computer / 189-iPad / \n", "31-Disc jockey / 3458-DMC World DJ Championships / \n", "1-Vehicle / 4-Car / 449-Honda / \n", "12-Food / 66-Bollywood / \n", "0-Games / 60-Basketball / 93-Comedy / \n", "59-Winter sport / 84-Snow / 177-Skiing / 281-Snowboarding / 305-Ski / 501-Snowboard / \n", "1-Vehicle / 4-Car / 40-Road / 64-Transport / 74-Truck / 114-Engine / 259-Fire / 371-Firefighter / 575-Traffic / 663-Street / 796-Fire engine / 1020-Emergency vehicle / \n", "1-Vehicle / 25-Toy / 62-Train / 119-Locomotive / 507-Thomas the Tank Engine / 1279-Plarail / \n", "12-Food / 539-Home appliance / 818-Washing machine / 1142-Laundry / 1456-Clothes dryer / \n", "5-Dance / 9-Music video / \n", "6-Animation / 15-Cartoon / 126-The Walt Disney Company / 412-Monster / \n", "1-Vehicle / 4-Car / 11-Motorsport / 17-Racing / 41-Sports car / 207-Exhaust system / 618-Fiat Automobiles / \n", "1-Vehicle / 11-Motorsport / 40-Road / 59-Winter sport / 84-Snow / 120-Winter / 610-Snowmobile / 1201-Ski-Doo / \n", "99-Christmas / 137-Tree / 964-Christmas tree / 2776-Pine / \n", "1-Vehicle / 50-Aircraft / 88-Machine / 1257-Turbine / 2194-Wind power / 2441-Wind chime / \n", "3-Concert / \n", "7-Musician / 9-Music video / 14-Guitar / \n", "1649-Fast food / \n", "9-Music video / \n", "12-Food / \n", "115-Ballet / \n", "0-Games / 8-Football / 21-Stadium / 65-Kick / 73-Ball / 81-Athlete / \n", "98-Festival / \n", "330-Money / \n", "1-Vehicle / 4-Car / 10-Animal / 48-Pet / 71-Dog / \n", "13-Musical ensemble / 296-Saxophone / \n", "18-Outdoor recreation / 83-Skateboarding / 108-Skateboard / 221-Skatepark / \n", "3-Concert / 7-Musician / 13-Musical ensemble / 14-Guitar / \n", "1026-Foreign Exchange Market / \n", "1-Vehicle / 4-Car / 1461-Electric car / \n", "0-Games / 8-Football / 21-Stadium / 65-Kick / 76-Ball / 81-Athlete / \n", "6-Animation / 15-Cartoon / 1674-Ouran High School Host Club / \n", "5-Dance / 16-Performance art / 678-Latin dance / \n", "1-Vehicle / 4-Car / 11-Motorsport / 40-Road / 124-Tractor / 252-Tractor pulling / \n", "30-Drummer / 39-Drums / 44-Drums / 128-Cymbal / 146-Snare drum / 1825-Cajón / \n", "12-Food / 122-Weight training / \n", "743-Edward Cullen / \n", "5-Dance / \n", "1-Vehicle / 50-Aircraft / 223-Weather / \n", "6-Animation / 350-Circus / \n", "8-Football / 21-Stadium / 77-Arena / \n", "1-Vehicle / 4-Car / 406-Chevrolet / 1502-Chrysler / \n", "162-Medicine / 2372-Dentist / \n", "903-Mass Effect / 4061-Mass Effect / \n", "8-Football / \n", "110-Album / \n", "6-Animation / 15-Cartoon / 123-Naruto / 183-Manga / 4598-Beavis / 4622-Butt-head / \n", "6-Animation / \n", "6-Animation / 15-Cartoon / \n", "6-Animation / 20-Trailer / \n", "6-Animation / 25-Toy / 140-LEGO / 264-Star Wars / 852-Lego Star Wars / 940-Clay animation / \n", "6-Animation / 15-Cartoon / 61-Art / 290-Pony / 343-My Little Pony / \n", "3-Concert / \n", "0-Games / 2-Video game / 51-Strategy video game / 1072-Total War / 2674-Rome: Total War / \n", "1-Vehicle / 4-Car / 11-Motorsport / 17-Racing / 41-Sports car / 57-Race track / 135-Drifting / \n", "0-Games / 2-Video game / 686-Diablo III / 872-Diablo / \n", "0-Games / 2-Video game / 35-Minecraft / \n", "1-Vehicle / 4-Car / 11-Motorsport / 17-Racing / \n", "31-Disc jockey / \n", "9-Music video / \n", "31-Disc jockey / \n", "1-Vehicle / 4-Car / 1510-Liquid / 2303-Motor oil / \n", "223-Weather / \n", "1-Vehicle / 50-Aircraft / 149-Model aircraft / \n", "0-Games / 2-Video game / 2572-Lupin III / \n", "676-Handbag / 829-Bead / \n", "6-Animation / 15-Cartoon / 89-Comics / 294-Batman / 630-Superman / 643-Hero / 748-Iron Man / 843-The Doctor / \n", "5-Dance / 16-Performance art / \n", "78-Wedding / 190-Bride / 421-Gown / 590-Wedding dress / \n", "5-Dance / 59-Winter sport / 112-Ice skating / 262-Ice / 316-Figure skating / 632-Ice dancing / \n", "1013-Walking / \n", "267-Mountain / 347-Climbing / 4563-Mont Blanc / \n", "12-Food / 86-Plant / 348-Fruit / 4021-Pomegranate / \n", "18-Outdoor recreation / 83-Skateboarding / 108-Skateboard / 221-Skatepark / \n", "0-Games / 240-Arcade game / 1802-Jubeat / \n", "0-Games / 8-Football / 54-Highlight film / 65-Kick / 73-Ball / 76-Ball / \n", "0-Games / 23-Mobile phone / 35-Minecraft / \n", "46-Choir / 219-Accordion / 318-Church / \n", "0-Games / 8-Football / 21-Stadium / 54-Highlight film / 73-Ball / \n", "6-Animation / 10-Animal / 15-Cartoon / 723-Pig / \n", "1-Vehicle / 4-Car / 299-Volkswagen Passenger Cars / 526-Volkswagen / 668-Volkswagen Golf / \n", "0-Games / 116-Pokémon / 186-Pokémon / 1387-Pokémon Battle Revolution / \n", "1-Vehicle / 4-Car / 11-Motorsport / 40-Road / 70-Driving / \n", "159-World of Warcraft / 171-Warcraft / 811-Mouse / \n", "9-Music video / \n", "6-Animation / 25-Toy / 200-Doll / 323-Toddler / \n", "0-Games / 8-Football / 21-Stadium / 65-Kick / 77-Arena / \n", "3-Concert / 7-Musician / 39-Drums / \n", "110-Album / \n", "10-Animal / 12-Food / 132-Bird / 540-Insect / 3017-Beetle / \n", "0-Games / 142-Hockey / \n", "8-Football / \n", "61-Art / 97-Drawing / 148-Painting / 528-Pencil / 628-Portrait / 890-Brush / \n", "3-Concert / 7-Musician / 13-Musical ensemble / 24-String instrument / 38-Orchestra / 150-Violin / 450-Viola / \n", "20-Trailer / \n", "0-Games / 1-Vehicle / 2-Video game / 34-Action-adventure game / 118-Grand Theft Auto V / \n", "9-Music video / \n", "3-Concert / 68-Lighting / 263-Light / \n", "10-Animal / 25-Toy / 48-Pet / 166-Cat / 346-Kitten / \n", "0-Games / 2-Video game / 19-PC game / 43-Call of Duty / 138-Call of Duty: Black Ops II / \n", "38-Orchestra / \n", "18-Outdoor recreation / 22-Nature / 86-Plant / 137-Tree / \n", "9-Music video / \n", "0-Games / 1232-Perfect World / \n", "1-Vehicle / 50-Aircraft / 82-Airplane / 129-Aviation / 139-Airport / 151-Landing / 188-Airline / 1547-Passenger / \n", "429-Website / \n", "0-Games / 211-RuneScape / 1873-Crossbow / \n", "0-Games / 2-Video game / 33-Weapon / 35-Minecraft / \n", "3-Concert / 7-Musician / 13-Musical ensemble / 16-Performance art / 30-Drummer / 38-Orchestra / \n", "23-Mobile phone / 29-Smartphone / 37-Gadget / 72-iPhone / 83-Skateboarding / 108-Skateboard / 272-iPod / 331-iPod touch / \n", "170-Home improvement / 1671-Drywall / \n", "12-Food / 179-Kitchen / \n", "8-Football / \n", "0-Games / 2-Video game / 21-Stadium / 60-Basketball / 79-American football / 90-Sports game / 176-PlayStation 4 / 314-Xbox One / 561-Madden NFL / \n", "0-Games / 8-Football / 21-Stadium / 65-Kick / 595-Champion / \n", "98-Festival / 250-Parade / 253-Carnival / \n", "6-Animation / \n", "110-Album / \n", "31-Disc jockey / \n", "275-Diving / 287-Human swimming / 329-Swimming pool / \n", "10-Animal / 48-Pet / 132-Bird / \n", "5-Dance / 16-Performance art / \n", "1-Vehicle / 62-Train / 64-Transport / 103-Rail transport / 119-Locomotive / 143-Railroad car / 152-Track / 161-Train station / \n", "0-Games / 8-Football / \n", "28-Fashion / 206-Shoe / 333-Nike; Inc. / 368-Sneakers / 1423-Nike Air Max / \n", "23-Mobile phone / 29-Smartphone / 37-Gadget / 72-iPhone / \n", "130-Gymnastics / \n", "3-Concert / 7-Musician / \n", "5-Dance / 49-School / 95-Talent show / 191-Cheerleading / 306-High school / \n", "476-Textile / 3822-Mold / \n", "1-Vehicle / 113-House / 707-Roof / 1168-Neighbourhood / \n", "0-Games / 2-Video game / 19-PC game / 55-Video game console / \n", "0-Games / 2-Video game / 55-Video game console / 199-Wii / \n", "0-Games / 599-Dofus / \n", "12-Food / 26-Cooking / 32-Recipe / 52-Dish / 58-Cuisine / 226-Dessert / 389-Restaurant / 684-Ice cream / \n", "14-Guitar / 24-String instrument / 63-Acoustic guitar / \n", "153-Dress / 1535-Prize / \n", "20-Trailer / 66-Bollywood / 1059-Movie theater / \n", "6-Animation / 15-Cartoon / \n", "264-Star Wars / 1393-Lightsaber / \n", "0-Games / 308-Ship / 605-Cruise ship / \n", "3-Concert / 5-Dance / 98-Festival / \n", "5-Dance / 76-Ball / 130-Gymnastics / 1451-Ribbon / \n", "0-Games / 781-Pachinko / \n", "1-Vehicle / 4-Car / 11-Motorsport / 17-Racing / 41-Sports car / 57-Race track / 3881-Porsche 930 / \n", "0-Games / 2-Video game / 159-World of Warcraft / 171-Warcraft / 217-Hunting / \n", "163-Running / \n", "6-Animation / 123-Naruto / \n", "10-Animal / 48-Pet / 132-Bird / \n", "567-Compact disc / \n", "49-School / 293-Slide show / \n", "9-Music video / 79-American football / 209-University / \n", "36-Piano / 53-Keyboard / 3953-Kawai Musical Instruments / \n", "23-Mobile phone / 29-Smartphone / 37-Gadget / 553-Sony Xperia / 3867-Sony Ericsson Xperia arc / \n", "79-American football / \n", "12-Food / 26-Cooking / 32-Recipe / 236-Baking / 442-Bread / 485-Dough / 600-Flour / 4022-Bun / \n", "1-Vehicle / 4-Car / 234-Ford / 1634-Ford Fiesta / \n", "0-Games / 2-Video game / 55-Video game console / 292-Handheld game console / 320-Super Smash Bros. / \n", "6-Animation / 15-Cartoon / \n", "0-Games / 8-Football / 21-Stadium / \n", "61-Art / 97-Drawing / 148-Painting / 303-Paint / 2124-Spray painting / \n", "75-Wrestling / \n", "0-Games / 2-Video game / 19-PC game / 229-Halo / 1161-Halo 2 / \n", "10-Animal / 48-Pet / 71-Dog / 182-Puppy / \n", "0-Games / 2-Video game / 43-Call of Duty / 138-Call of Duty: Black Ops II / \n", "1-Vehicle / \n", "0-Games / 2-Video game / 19-PC game / 33-Weapon / 43-Call of Duty / 96-Soldier / 431-Call of Duty: Advanced Warfare / \n", "14-Guitar / 24-String instrument / 100-Electric guitar / \n", "6-Animation / 25-Toy / \n", "2-Video game / 55-Video game console / 292-Handheld game console / 317-PlayStation / 362-PlayStation Portable / 596-Monster Hunter / 682-PlayStation Vita / 1339-Monster Hunter Freedom Unite / \n", "219-Accordion / 655-Diatonic button accordion / \n", "93-Comedy / \n", "31-Disc jockey / \n", "0-Games / 33-Weapon / 257-Counter-Strike / 1109-Counter-Strike Online / \n", "3-Concert / 7-Musician / 13-Musical ensemble / 16-Performance art / \n", "1-Vehicle / 4-Car / 344-Coupé / 2517-Chevrolet Opala / \n", "5-Dance / \n", "12-Food / 500-Rice / 2000-Mango / \n", "55-Video game console / 127-Xbox 360 / 386-Xbox / \n", "1-Vehicle / 62-Train / 64-Transport / 103-Rail transport / 119-Locomotive / 143-Railroad car / 152-Track / 854-Diesel engine / 2824-Diesel locomotive / \n", "0-Games / 21-Stadium / 54-Highlight film / 562-Pitcher / 946-Baseball park / \n", "110-Album / \n", "0-Games / 2-Video game / 730-Final Fantasy VII / 1312-Dynasty Warriors / \n", "5-Dance / 892-Aerobics / \n", "3-Concert / 7-Musician / 13-Musical ensemble / \n", "99-Christmas / 558-Santa Claus / \n", "0-Games / 2267-Geometry Dash / 2347-Marble / 2870-Marble / \n", "6-Animation / 15-Cartoon / 398-The Legend of Zelda / 877-Hatsune Miku: Project DIVA / \n", "12-Food / 26-Cooking / 32-Recipe / 301-Meal / \n", "2037-Straw / \n", "5-Dance / \n", "3-Concert / 7-Musician / 13-Musical ensemble / \n", "31-Disc jockey / \n", "45-Cosmetics / \n", "56-Hair / 340-Afro-textured hair / \n", "5-Dance / 16-Performance art / 978-Love song / \n", "1909-Analog synthesizer / \n", "230-Brass instrument / 296-Saxophone / 1006-Tenor saxophone / \n", "46-Choir / \n", "0-Games / 51-Strategy video game / 165-League of Legends / \n", "20-Trailer / 1469-X-Men / 3507-Superhero movie / \n", "1-Vehicle / 50-Aircraft / 82-Airplane / 84-Snow / 129-Aviation / 139-Airport / 157-Jet aircraft / 175-Airliner / 187-Takeoff / 188-Airline / 254-Runway / 472-Boeing 737 / 1145-Winter storm / 1529-Jet engine / \n", "9-Music video / 45-Cosmetics / \n", "0-Games / 33-Weapon / 664-CrossFire / \n", "3-Concert / 16-Performance art / \n", "1-Vehicle / 4-Car / 167-Four-wheel drive / 197-Trail / 198-Off-road vehicle / 277-Sport utility vehicle / 327-Mud / \n", "5-Dance / 9-Music video / \n", "22-Nature / \n", "3-Concert / 68-Lighting / 98-Festival / \n", "18-Outdoor recreation / 83-Skateboarding / 108-Skateboard / 555-Longboard / 667-Longboarding / 784-Honey / \n", "9-Music video / \n", "0-Games / 2-Video game / 35-Minecraft / \n", "6-Animation / 15-Cartoon / 1160-Giant panda / \n", "14-Guitar / 1231-Mandolin / \n", "1-Vehicle / 74-Truck / 249-Heavy equipment / 976-Demolition / \n", "768-Diamond / \n", "689-Harmonica / \n", "1-Vehicle / 22-Nature / 62-Train / 94-Dashcam / 103-Rail transport / \n", "3-Concert / 7-Musician / 13-Musical ensemble / 14-Guitar / 24-String instrument / \n", "1-Vehicle / 4-Car / 517-Audi / \n", "0-Games / 47-Personal computer / 141-Microsoft Windows / 4072-Outlast / \n", "1-Vehicle / 11-Motorsport / 18-Outdoor recreation / 27-Motorcycle / 105-Motorcycling / 197-Trail / 198-Off-road vehicle / \n", "1-Vehicle / 11-Motorsport / 17-Racing / 27-Motorcycle / 172-Motocross / \n", "1874-Mattress / \n", "3-Concert / 5-Dance / 16-Performance art / \n", "315-Knife / \n", "3-Concert / \n", "36-Piano / 53-Keyboard / 104-Pianist / \n", "10-Animal / 18-Outdoor recreation / 42-Fishing / 91-Fish / 601-Shark / 3292-Rock fishing / \n", "6-Animation / 15-Cartoon / 887-King Koopa / \n", "10-Animal / 1716-Rat / \n", "1-Vehicle / 4-Car / 270-Sedan / 299-Volkswagen Passenger Cars / 526-Volkswagen / \n", "1-Vehicle / 4-Car / 88-Machine / 114-Engine / \n", "0-Games / 49-School / 60-Basketball / \n", "0-Games / 2-Video game / 19-PC game / 43-Call of Duty / 134-Call of Duty: Black Ops / 138-Call of Duty: Black Ops II / 176-PlayStation 4 / 216-Call of Duty: Modern Warfare 3 / \n", "59-Winter sport / 112-Ice skating / 262-Ice / 316-Figure skating / 833-Ice skate / \n", "0-Games / 60-Basketball / \n", "3-Concert / \n", "10-Animal / \n", "47-Personal computer / 145-Tablet computer / 189-iPad / 1150-iPad 2 / \n", "122-Weight training / 214-Muscle / \n", "6-Animation / 1809-Tokyo Mew Mew / \n", "8-Football / \n", "1-Vehicle / 4-Car / 64-Transport / 286-Bus / \n", "1-Vehicle / 4-Car / 11-Motorsport / 70-Driving / 349-Chevrolet / 406-Chevrolet / 509-Classic car / 647-Muscle car / 2061-Chevrolet Chevelle / \n", "23-Mobile phone / 101-Telephone / 986-Suit / 2191-Text messaging / \n", "0-Games / 2-Video game / 33-Weapon / \n", "59-Winter sport / 84-Snow / 120-Winter / 177-Skiing / 278-GoPro / 281-Snowboarding / 305-Ski / 501-Snowboard / \n", "5-Dance / 95-Talent show / \n", "1-Vehicle / 4-Car / 659-Van / \n", "0-Games / 2-Video game / 35-Minecraft / 308-Ship / 2767-Captain Hook / \n", "9-Music video / \n", "20-Trailer / \n", "1-Vehicle / 4-Car / 11-Motorsport / 27-Motorcycle / 105-Motorcycling / \n", "47-Personal computer / \n", "1-Vehicle / 50-Aircraft / 92-Radio-controlled model / 149-Model aircraft / 160-Radio-controlled aircraft / 809-Logo / \n", "9-Music video / \n", "6-Animation / 15-Cartoon / \n", "47-Personal computer / 145-Tablet computer / 189-iPad / 1464-iPad Mini / \n", "9-Music video / 31-Disc jockey / \n", "122-Weight training / \n", "20-Trailer / \n", "978-Love song / \n", "1-Vehicle / 4-Car / 531-Hatchback / \n", "10-Animal / 48-Pet / 166-Cat / 2859-Black cat / \n", "380-Assassin's Creed / 1585-Assassin's Creed Unity / \n", "10-Animal / 48-Pet / 166-Cat / 346-Kitten / \n", "30-Drummer / 39-Drums / 44-Drums / 128-Cymbal / 146-Snare drum / 911-Studio / \n", "28-Fashion / \n", "3-Concert / \n", "6-Animation / 4714-Air Gear / \n", "0-Games / 141-Microsoft Windows / \n", "3-Concert / \n", "3-Concert / \n", "162-Medicine / \n", "196-Eye / \n", "0-Games / 1029-Aion: The Tower of Eternity / \n", "1-Vehicle / 4-Car / 11-Motorsport / 74-Truck / 154-Wheel / 158-Tire / 234-Ford / 375-Pickup truck / 1493-Supercharger / 2427-Ford F-Series / \n", "45-Cosmetics / 173-Eye shadow / 193-Eye liner / 195-Mascara / \n", "1-Vehicle / 11-Motorsport / 17-Racing / 67-Cycling / 69-Bicycle / 220-Mountain bike / \n", "98-Festival / 1453-Balloon / 1879-Hot air balloon / 1972-Toy balloon / \n", "0-Games / 2-Video game / 23-Mobile phone / 29-Smartphone / 37-Gadget / 101-Telephone / 4424-Nexus One / \n", "12-Food / 26-Cooking / 32-Recipe / 208-Wood / 274-Meat / 589-Grilling / 620-Barbecue / 1256-Smoking / \n", "45-Cosmetics / \n", "1-Vehicle / 4-Car / 11-Motorsport / 17-Racing / 57-Race track / 3190-Chevrolet Cruze / \n", "28-Fashion / \n", "61-Art / 457-Knitting / 546-Stitch / 653-Thread / 1795-Sock / 1804-Sewing needle / \n", "820-Bagpipes / \n", "10-Animal / 80-Horse / 247-Stallion / 289-Dressage / 404-Livestock / \n", "3-Concert / 68-Lighting / 133-Nightclub / \n", "88-Machine / 208-Wood / 1700-Lumber / \n", "10-Animal / 571-Bear / \n", "1-Vehicle / 4-Car / 11-Motorsport / 17-Racing / 27-Motorcycle / 57-Race track / 105-Motorcycling / \n", "22-Nature / 213-Forest / \n", "14-Guitar / 458-Gold / \n", "523-Cattle / 534-Milk / \n", "0-Games / 2-Video game / 51-Strategy video game / \n", "0-Games / 2-Video game / 47-Personal computer / 127-Xbox 360 / 4134-Ni no Kuni: Wrath of the White Witch / \n", "31-Disc jockey / \n", "3996-Fandango / \n", "2-Video game / 276-Room / \n", "3-Concert / \n", "7-Musician / 13-Musical ensemble / \n", "10-Animal / 18-Outdoor recreation / 22-Nature / 42-Fishing / 91-Fish / 147-River / 510-Fishing lure / \n", "68-Lighting / \n", "3850-Weighing scale / \n", "0-Games / 2-Video game / 35-Minecraft / \n", "2-Video game / \n", "0-Games / 2-Video game / \n", "214-Muscle / \n", "3-Concert / 68-Lighting / \n", "10-Animal / 80-Horse / 384-Horse racing / \n", "1-Vehicle / 4-Car / 40-Road / 94-Dashcam / 1960-Traffic light / \n", "1-Vehicle / 28-Fashion / 663-Street / 1168-Neighbourhood / \n", "3-Concert / 5-Dance / \n", "13-Musical ensemble / 376-Flute / 1326-Recorder / \n", "1-Vehicle / 250-Parade / \n", "1-Vehicle / 11-Motorsport / 27-Motorcycle / 114-Engine / 135-Drifting / 1887-Tricycle / 3614-Drift trikes / \n", "296-Saxophone / 1139-Belle / 1265-Kyle Kingson / \n", "0-Games / 2-Video game / 55-Video game console / 199-Wii / 456-Go-kart / 518-Mario Kart / \n", "624-Number / 1818-Calculator / \n", "12-Food / 26-Cooking / 32-Recipe / 52-Dish / 58-Cuisine / 109-Cooking show / 239-Roasting / 1381-Garlic / \n", "66-Bollywood / \n", "5-Dance / \n", "5-Dance / \n", "66-Bollywood / \n", "3-Concert / 7-Musician / 13-Musical ensemble / 38-Orchestra / 150-Violin / 264-Star Wars / \n", "121-Photography / 628-Portrait / \n", "93-Comedy / 125-Television / \n", "9-Music video / \n", "14-Guitar / \n", "0-Games / 2-Video game / 19-PC game / 1196-Dragon Age / 1735-Dragon Age: Inquisition / \n", "23-Mobile phone / 29-Smartphone / 37-Gadget / 47-Personal computer / 145-Tablet computer / 178-Samsung Galaxy / 487-Samsung Electronics / 1287-Samsung Galaxy Tab series / \n", "0-Games / 8-Football / \n", "913-President / \n", "0-Games / 2-Video game / 51-Strategy video game / 165-League of Legends / 1494-Jungle / \n", "1-Vehicle / 1036-Moped / \n", "1513-Hitman / 2470-Hitman: Absolution / \n", "1453-Balloon / \n", "0-Games / 1-Vehicle / 2-Video game / 4-Car / 19-PC game / 34-Action-adventure game / 118-Grand Theft Auto V / 127-Xbox 360 / \n", "0-Games / 6-Animation / 1946-Total Drama Island / \n", "5-Dance / 16-Performance art / 115-Ballet / \n", "5-Dance / \n", "3653-Masih ad-Dajjal / \n", "303-Paint / 323-Toddler / 324-Finger / \n", "6-Animation / 15-Cartoon / 538-Illustration / \n", "3-Concert / 7-Musician / 13-Musical ensemble / 38-Orchestra / \n", "7-Musician / 30-Drummer / 39-Drums / 44-Drums / 46-Choir / 128-Cymbal / 146-Snare drum / \n", "6-Animation / 15-Cartoon / \n", "3-Concert / 5-Dance / 16-Performance art / \n", "0-Games / 2-Video game / 43-Call of Duty / 134-Call of Duty: Black Ops / 138-Call of Duty: Black Ops II / 422-Call of Duty: Ghosts / \n", "5-Dance / 16-Performance art / \n", "1-Vehicle / 11-Motorsport / 27-Motorcycle / 207-Exhaust system / \n", "25-Toy / 106-Hairstyle / 200-Doll / 439-Barbie / \n", "46-Choir / \n", "67-Cycling / 69-Bicycle / 377-BMX bike / \n", "99-Christmas / 466-Gift / \n", "0-Games / 2-Video game / 19-PC game / 35-Minecraft / \n", "3-Concert / 2738-Fender Jazz Bass / \n", "6-Animation / 15-Cartoon / 61-Art / 243-Sonic the Hedgehog / 251-Sonic the Hedgehog / \n", "3-Concert / 7-Musician / 14-Guitar / 24-String instrument / 100-Electric guitar / \n", "7-Musician / 13-Musical ensemble / 46-Choir / \n", "14-Guitar / 63-Acoustic guitar / 100-Electric guitar / \n", "1-Vehicle / 50-Aircraft / 778-Gliding / \n", "219-Accordion / \n", "102-Building / 113-House / 353-Wall / 433-Construction / 666-Concrete / \n", "10-Animal / 48-Pet / 71-Dog / 166-Cat / 182-Puppy / \n", "0-Games / 2-Video game / 14-Guitar / 445-Guitar Hero / \n", "1572-Husband / \n", "505-Mega Man / 2602-Mega Man X / \n", "7-Musician / 14-Guitar / 24-String instrument / 63-Acoustic guitar / 1562-Viola caipira / \n", "8-Football / \n", "1-Vehicle / 4-Car / 11-Motorsport / 17-Racing / 92-Radio-controlled model / 149-Model aircraft / 160-Radio-controlled aircraft / 393-Unmanned aerial vehicle / \n", "9-Music video / 31-Disc jockey / \n", "9-Music video / \n", "25-Toy / 354-Egg / 410-Teacher / 417-Play-Doh / 611-Kinder Surprise / \n", "1043-Router / \n", "45-Cosmetics / 238-Nail / 300-Nail art / 304-Nail polish / 338-Manicure / \n", "5-Dance / \n", "25-Toy / 455-Unidentified flying object / 1152-Plush / 1212-Stuffed toy / 1668-Claw crane / \n", "10-Animal / 540-Insect / 775-Bee / \n", "22-Nature / 141-Microsoft Windows / \n", "110-Album / \n", "13-Musical ensemble / 38-Orchestra / 203-Marching band / 634-Trombone / 886-James Bond / \n", "10-Animal / 132-Bird / 718-Duck / 1947-Goose / \n", "0-Games / 8-Football / \n", "10-Animal / 12-Food / 26-Cooking / 91-Fish / 893-Fish as food / 2125-Fillet / \n", "3-Concert / 7-Musician / 13-Musical ensemble / 14-Guitar / \n", "0-Games / 6-Animation / 23-Mobile phone / 29-Smartphone / 553-Sony Xperia / 4440-Sony Ericsson Xperia X8 / \n", "66-Bollywood / \n", "9-Music video / \n", "0-Games / 2-Video game / 6-Animation / 432-Kingdom Hearts / 665-Kingdom Hearts / 836-Sora / 941-Kingdom Hearts II / 2779-Universe of Kingdom Hearts / \n", "5-Dance / 16-Performance art / 59-Winter sport / 95-Talent show / 112-Ice skating / 115-Ballet / 316-Figure skating / \n", "75-Wrestling / 117-Boxing / \n", "0-Games / 2043-Mabinogi / \n", "0-Games / 2-Video game / 19-PC game / 2334-Phantasy Star Online 2 / \n", "12-Food / 1162-Dietary supplement / \n", "23-Mobile phone / 29-Smartphone / 37-Gadget / 101-Telephone / 932-Microsoft Lumia / \n", "0-Games / 960-Badminton / \n", "0-Games / 2-Video game / 39-Drums / 445-Guitar Hero / 724-Rock Band / 1762-Rock Band 3 / \n", "0-Games / 5-Dance / 49-School / 191-Cheerleading / \n", "83-Skateboarding / 108-Skateboard / 221-Skatepark / 703-Roller skating / 1937-Inline skates / \n", "22-Nature / 279-Garden / \n", "0-Games / 739-Guild Wars / \n", "3-Concert / 7-Musician / 13-Musical ensemble / 14-Guitar / 16-Performance art / 39-Drums / 44-Drums / \n", "12-Food / 323-Toddler / 500-Rice / 2928-Cereal / \n", "315-Knife / 676-Handbag / \n", "0-Games / 489-Board game / 799-Chess / 922-Tabletop game / \n", "61-Art / 97-Drawing / 148-Painting / 159-World of Warcraft / 171-Warcraft / \n", "7-Musician / 13-Musical ensemble / 24-String instrument / 150-Violin / 401-Cello / 450-Viola / \n", "9-Music video / \n", "6-Animation / 15-Cartoon / 4504-Xenosaga / \n", "3-Concert / 7-Musician / 13-Musical ensemble / 38-Orchestra / \n", "36-Piano / 46-Choir / 53-Keyboard / 104-Pianist / 107-Musical keyboard / \n", "0-Games / 2-Video game / 35-Minecraft / \n", "18-Outdoor recreation / 20-Trailer / 22-Nature / 86-Plant / \n", "9-Music video / \n", "3-Concert / 7-Musician / 14-Guitar / 24-String instrument / \n", "122-Weight training / \n", "66-Bollywood / \n", "3-Concert / 16-Performance art / 3599-The Wonderful Wizard of Oz / \n", "38-Orchestra / \n", "0-Games / 2-Video game / 33-Weapon / 43-Call of Duty / 96-Soldier / 134-Call of Duty: Black Ops / \n", "3-Concert / 7-Musician / 36-Piano / 104-Pianist / \n", "5-Dance / \n", "3-Concert / 21-Stadium / 77-Arena / \n", "148-Painting / 2538-Painter / \n", "1-Vehicle / 4-Car / 114-Engine / 1491-Steam engine / \n", "1-Vehicle / 4-Car / 11-Motorsport / 27-Motorcycle / 105-Motorcycling / 207-Exhaust system / \n", "16-Performance art / 93-Comedy / 355-Comedian / \n", "9-Music video / \n", "0-Games / 10-Animal / 48-Pet / \n", "18-Outdoor recreation / 83-Skateboarding / 108-Skateboard / 221-Skatepark / \n", "3-Concert / 7-Musician / 24-String instrument / 44-Drums / 1442-Tabla / \n", "1731-Haunted house / \n", "10-Animal / 550-Dinosaur / \n", "113-House / \n", "319-Robot / 4438-Humanoid robot / \n", "0-Games / 2-Video game / 2880-Worms / \n", "0-Games / 8-Football / 21-Stadium / 65-Kick / 1247-Penalty kick / \n", "0-Games / 2-Video game / 6-Animation / 685-Luigi / 3506-Luigi's Mansion / \n", "214-Muscle / 661-Stretching / \n", "264-Star Wars / \n", "43-Call of Duty / 134-Call of Duty: Black Ops / \n", "0-Games / 1-Vehicle / 2-Video game / 4-Car / 11-Motorsport / 17-Racing / 1333-Wangan Midnight / \n", "762-Angel / \n", "5-Dance / 16-Performance art / \n", "6-Animation / 15-Cartoon / 264-Star Wars / 2576-Clone trooper / \n", "22-Nature / 499-Cloud / \n", "12-Food / 26-Cooking / 32-Recipe / 236-Baking / 560-Sugar / 578-Cookie / 867-Tea / 1081-Lemon / \n", "23-Mobile phone / 29-Smartphone / 37-Gadget / 553-Sony Xperia / 1097-Touchscreen / 3619-Xperia Play / \n", "439-Barbie / \n", "0-Games / 2-Video game / 83-Skateboarding / 90-Sports game / 1708-Skate 3 / \n", "227-Water / \n", "15-Cartoon / \n", "12-Food / 389-Restaurant / 462-Table / \n", "10-Animal / 22-Nature / 86-Plant / 91-Fish / 258-Aquarium / \n", "83-Skateboarding / 221-Skatepark / \n", "3-Concert / \n", "0-Games / 35-Minecraft / \n", "7-Musician / 14-Guitar / 24-String instrument / \n", "1-Vehicle / 4-Car / 11-Motorsport / 41-Sports car / 352-Need for Speed / 473-Ford Mustang / 2739-Midnight Club: Los Angeles / \n", "2641-Radha / \n", "3009-LED lamp / \n", "23-Mobile phone / 29-Smartphone / 3623-BlackBerry Z10 / \n", "1-Vehicle / 4-Car / 207-Exhaust system / \n", "1219-Ariel / \n", "10-Animal / 22-Nature / 588-Reptile / 1354-Frog / \n", "1002-String quartet / \n", "86-Plant / 136-Gardening / 468-Biology / \n", "1-Vehicle / 4-Car / 11-Motorsport / 17-Racing / 41-Sports car / 57-Race track / 70-Driving / 277-Sport utility vehicle / 332-Mercedes-Benz / \n", "0-Games / 8-Football / 595-Champion / \n", "0-Games / 19-PC game / 192-Battlefield / 383-Battlefield 4 / \n", "10-Animal / 1354-Frog / \n", "0-Games / 2-Video game / 33-Weapon / 43-Call of Duty / 96-Soldier / 134-Call of Duty: Black Ops / 138-Call of Duty: Black Ops II / 194-Call of Duty: Modern Warfare 2 / 216-Call of Duty: Modern Warfare 3 / 422-Call of Duty: Ghosts / 431-Call of Duty: Advanced Warfare / \n", "0-Games / 60-Basketball / \n", "10-Animal / 71-Dog / 217-Hunting / 1037-Grass / \n", "147-River / \n", "0-Games / 2-Video game / 43-Call of Duty / 96-Soldier / 302-Call of Duty 4: Modern Warfare / 635-Achievement / \n", "3-Concert / 7-Musician / 14-Guitar / 30-Drummer / 39-Drums / \n", "342-Windows Media Video / \n", "0-Games / 8-Football / 21-Stadium / 65-Kick / 73-Ball / \n", "1-Vehicle / 4-Car / 344-Coupé / 683-Renault / 1388-V6 engine / \n", "12-Food / 26-Cooking / 32-Recipe / 52-Dish / 210-Cookware and bakeware / \n", "9-Music video / 310-The Sims / 543-The Sims 2 / \n", "1-Vehicle / 4-Car / 40-Road / 64-Transport / 74-Truck / 328-Simulation video game / 1099-Euro Truck Simulator 2 / 3886-Mercedes-Benz Actros / \n", "2-Video game / \n", "2685-Giant / \n", "1927-Atari / \n", "0-Games / 51-Strategy video game / 373-Clash of Clans / \n", "6-Animation / 264-Star Wars / 717-Darth Vader / \n", "45-Cosmetics / 238-Nail / 300-Nail art / 304-Nail polish / 307-Hand / 324-Finger / 338-Manicure / 823-Artificial nails / \n", "97-Drawing / 920-Chair / \n", "0-Games / 2-Video game / 111-PlayStation 3 / 127-Xbox 360 / 176-PlayStation 4 / 181-Xbox / 314-Xbox One / \n", "0-Games / 1-Vehicle / 2-Video game / 4-Car / 11-Motorsport / 17-Racing / 40-Road / 352-Need for Speed / 1770-Need for Speed: Hot Pursuit / 3146-Need for Speed: Undercover / \n", "14-Guitar / 24-String instrument / 63-Acoustic guitar / 4680-A&E / \n", "3-Concert / 5-Dance / 16-Performance art / \n", "8-Football / 623-Coach / \n", "7-Musician / 36-Piano / 53-Keyboard / 104-Pianist / \n", "36-Piano / 376-Flute / 446-Sheet music / 878-Composer / 1326-Recorder / 1942-Oboe / \n", "12-Food / 358-Rugby football / \n", "6-Animation / 20-Trailer / 1266-Death Note / \n", "3-Concert / 13-Musical ensemble / 38-Orchestra / \n", "0-Games / 1-Vehicle / 2-Video game / 4-Car / 34-Action-adventure game / 84-Snow / 118-Grand Theft Auto V / 198-Off-road vehicle / \n", "10-Animal / 48-Pet / 132-Bird / \n", "3362-WildStar / \n", "1-Vehicle / 4-Car / 27-Motorcycle / 88-Machine / 114-Engine / 154-Wheel / 158-Tire / \n", "3480-Sound system / \n", "0-Games / 2-Video game / 626-Puzzle / 4392-The Evil Within / \n", "20-Trailer / 2368-Marvel Universe / 3276-Punisher / \n", "0-Games / 1-Vehicle / 2-Video game / 4-Car / 297-Grand Theft Auto IV / 1611-Garbage truck / \n", "25-Toy / 140-LEGO / 735-Kamen Rider Series / \n", "85-Combat / \n", "1-Vehicle / 50-Aircraft / 82-Airplane / 92-Radio-controlled model / 149-Model aircraft / 160-Radio-controlled aircraft / \n", "6-Animation / 22-Nature / 877-Hatsune Miku: Project DIVA / \n", "28-Fashion / 1525-Wallet / \n", "233-Ocean / 672-Bodyboarding / \n", "0-Games / 2-Video game / 6-Animation / 126-The Walt Disney Company / \n", "1-Vehicle / 1772-Forklift / 2297-Elf / \n", "6-Animation / 15-Cartoon / \n", "61-Art / 97-Drawing / 148-Painting / 479-Graffiti / 1362-Canvas / \n", "10-Animal / 48-Pet / 71-Dog / 182-Puppy / \n", "1-Vehicle / 27-Motorcycle / 40-Road / \n", "3-Concert / 30-Drummer / 39-Drums / 44-Drums / 2656-Guitar Center / \n", "18-Outdoor recreation / 96-Soldier / 217-Hunting / \n", "1-Vehicle / 455-Unidentified flying object / \n", "9-Music video / 14-Guitar / \n", "9-Music video / \n", "8-Football / 54-Highlight film / 79-American football / \n", "1-Vehicle / 4-Car / 6-Animation / 11-Motorsport / 27-Motorcycle / 105-Motorcycling / 154-Wheel / \n", "1-Vehicle / 50-Aircraft / 778-Gliding / 1200-Ultralight aviation / \n", "222-Surfing / \n", "0-Games / 8-Football / 21-Stadium / 54-Highlight film / 65-Kick / 76-Ball / \n", "0-Games / 2-Video game / 33-Weapon / 96-Soldier / 229-Halo / 638-Gears of War / 773-Gears of War / \n", "53-Keyboard / 397-Organ / 646-Pipe organ / \n", "162-Medicine / \n", "46-Choir / \n", "512-Printing / \n", "12-Food / 26-Cooking / 32-Recipe / 52-Dish / 58-Cuisine / 109-Cooking show / 475-Cheese / 2500-Pear / 3382-Risotto / \n", "10-Animal / 132-Bird / 4362-Turmeric / \n", "1-Vehicle / 11-Motorsport / 17-Racing / 27-Motorcycle / 105-Motorcycling / 172-Motocross / 197-Trail / \n", "1-Vehicle / 27-Motorcycle / 207-Exhaust system / 336-Scooter / \n", "1-Vehicle / \n", "3-Concert / 5-Dance / 191-Cheerleading / \n", "7-Musician / 14-Guitar / 24-String instrument / 63-Acoustic guitar / \n", "0-Games / 8-Football / 21-Stadium / 73-Ball / \n", "3-Concert / 5-Dance / \n", "14-Guitar / 63-Acoustic guitar / \n", "121-Photography / 155-Camera / 566-Digital camera / \n", "398-The Legend of Zelda / 1228-The Legend of Zelda: Ocarina of Time / \n", "122-Weight training / 204-Gym / \n", "25-Toy / 140-LEGO / \n", "670-Banjo / \n", "1-Vehicle / 4-Car / 449-Honda / 1614-Honda Accord / \n", "0-Games / 116-Pokémon / 186-Pokémon / \n", "10-Animal / 45-Cosmetics / 48-Pet / 166-Cat / \n", "1-Vehicle / 4-Car / 70-Driving / 94-Dashcam / 552-Honda Civic / \n", "3-Concert / \n", "1-Vehicle / 1573-United States Army / \n", "6-Animation / \n", "162-Medicine / 342-Windows Media Video / \n", "3-Concert / 7-Musician / 13-Musical ensemble / 14-Guitar / 24-String instrument / 38-Orchestra / 150-Violin / 639-Fiddle / \n", "85-Combat / 1832-Hand-to-hand combat / \n", "61-Art / 97-Drawing / 512-Printing / 1063-Ink / \n", "0-Games / 8-Football / 54-Highlight film / 65-Kick / 73-Ball / 90-Sports game / \n", "36-Piano / 53-Keyboard / 104-Pianist / \n", "18-Outdoor recreation / 83-Skateboarding / 108-Skateboard / 221-Skatepark / 915-Town / \n", "355-Comedian / \n", "12-Food / 86-Plant / 136-Gardening / 227-Water / 279-Garden / 363-Soil / 1720-Grape / \n", "3-Concert / 7-Musician / 13-Musical ensemble / \n", "12-Food / 26-Cooking / 32-Recipe / 52-Dish / 58-Cuisine / 109-Cooking show / 179-Kitchen / 184-Vegetable / 454-Sauce / 752-Pasta / 771-Noodle / 1539-Ramen / 2178-Chinese food / \n", "0-Games / 59-Winter sport / 112-Ice skating / 142-Hockey / 312-Ice rink / \n", "12-Food / 32-Recipe / 367-Drink / 720-Wine / 786-Juice / 904-Cocktail / 1065-Bartender / 1669-Vodka / 2406-Peach / \n", "0-Games / 2-Video game / 33-Weapon / 664-CrossFire / \n", "3-Concert / 7-Musician / 13-Musical ensemble / 46-Choir / \n", "9-Music video / \n", "1-Vehicle / 62-Train / 68-Lighting / 152-Track / 161-Train station / 428-Rail transport modelling / 1450-Model building / \n", "6-Animation / 9-Music video / 15-Cartoon / \n", "0-Games / 90-Sports game / 180-FIFA 15 / \n", "38-Orchestra / 376-Flute / \n", "45-Cosmetics / 56-Hair / \n", "0-Games / 2-Video game / 43-Call of Duty / 111-PlayStation 3 / 194-Call of Duty: Modern Warfare 2 / \n", "4343-Canon EOS 700D / \n", "23-Mobile phone / 29-Smartphone / 37-Gadget / 72-iPhone / 508-iPhone 5 / \n", "6-Animation / 15-Cartoon / \n", "0-Games / 2-Video game / 47-Personal computer / 2869-Bomberman / \n", "8-Football / \n", "122-Weight training / 204-Gym / 214-Muscle / 532-Barbell / 1250-Biceps / 1264-Head / 1389-Dumbbell / 2045-Biceps curl / \n", "85-Combat / 117-Boxing / \n", "5-Dance / 115-Ballet / 1253-Ballet Dancer / \n", "0-Games / 1-Vehicle / 2-Video game / 34-Action-adventure game / 201-Grand Theft Auto: San Andreas / \n", "0-Games / 2-Video game / 33-Weapon / 43-Call of Duty / 96-Soldier / 194-Call of Duty: Modern Warfare 2 / 302-Call of Duty 4: Modern Warfare / \n", "0-Games / 2-Video game / 886-James Bond / 4485-Warhawk / \n", "45-Cosmetics / 238-Nail / 300-Nail art / 304-Nail polish / 338-Manicure / \n", "0-Games / 792-DarkOrbit / \n", "10-Animal / 71-Dog / 2049-Police dog / \n", "1697-Astrological sign / \n", "85-Combat / 1327-Minnie Mouse / \n", "18-Outdoor recreation / 59-Winter sport / 84-Snow / 120-Winter / 177-Skiing / 281-Snowboarding / 305-Ski / \n", "47-Personal computer / 131-Computer / \n", "12-Food / \n", "15-Cartoon / \n", "0-Games / 2-Video game / 515-Dodge / 2527-Loadout / \n", "12-Food / 1510-Liquid / \n", "1-Vehicle / 4-Car / \n", "14-Guitar / 63-Acoustic guitar / \n", "3-Concert / 7-Musician / 13-Musical ensemble / 38-Orchestra / \n", "75-Wrestling / \n", "1-Vehicle / 4-Car / 864-Speedometer / 1286-Subaru / 3040-Subaru Legacy / \n", "185-Dragon Ball / \n", "1-Vehicle / 33-Weapon / 64-Transport / 271-Tank / \n", "10-Animal / 48-Pet / 71-Dog / 458-Gold / 3173-Great Dane / \n", "35-Minecraft / \n", "0-Games / 3227-I Wanna Be the Guy / \n", "1-Vehicle / 4-Car / 15-Cartoon / \n", "4583-TV9 / \n", "9-Music video / \n", "78-Wedding / \n", "13-Musical ensemble / 38-Orchestra / \n", "8-Football / \n", "1-Vehicle / 4-Car / 11-Motorsport / 17-Racing / 57-Race track / 728-Dirt track racing / \n", "12-Food / 26-Cooking / 32-Recipe / 226-Dessert / 232-Cake / 236-Baking / 437-Cream / 759-Icing / \n", "47-Personal computer / 228-Laptop / 3000-ThinkPad / \n", "0-Games / 2-Video game / 264-Star Wars / 714-Jedi / 2117-Star Wars Battlefront / 2340-Star Wars: Battlefront II / \n", "9-Music video / \n", "0-Games / 35-Minecraft / \n", "807-Seed / \n", "25-Toy / 2525-Gashapon / 3396-Ressha Sentai ToQger / \n", "6-Animation / 10-Animal / 15-Cartoon / 61-Art / 89-Comics / 290-Pony / 343-My Little Pony / \n", "6-Animation / 7-Musician / 9-Music video / \n", "1992-Space Shuttle / \n", "14-Guitar / 100-Electric guitar / \n", "25-Toy / 89-Comics / 435-Action figure / 2121-Marvel Legends / 3276-Punisher / \n", "3-Concert / \n", "0-Games / 8-Football / 21-Stadium / 54-Highlight film / 65-Kick / 73-Ball / \n", "0-Games / 8-Football / \n", "0-Games / 8-Football / 21-Stadium / 65-Kick / 73-Ball / 76-Ball / 81-Athlete / \n", "122-Weight training / 390-Model / \n", "10-Animal / 970-Monkey / \n", "23-Mobile phone / 29-Smartphone / 37-Gadget / 72-iPhone / 101-Telephone / 3441-HTC HD2 / \n", "0-Games / 2-Video game / 102-Building / 113-House / 174-Farm / 310-The Sims / 328-Simulation video game / 339-Architecture / 564-The Sims 3 / \n", "8-Football / \n", "1-Vehicle / 4-Car / 11-Motorsport / 17-Racing / 40-Road / 57-Race track / 94-Dashcam / 135-Drifting / 2288-Ford Escort / \n", "46-Choir / \n", "9-Music video / \n", "38-Orchestra / \n", "37-Gadget / 141-Microsoft Windows / \n", "0-Games / 2-Video game / 6-Animation / 34-Action-adventure game / \n", "415-Sewing / \n", "1-Vehicle / 4-Car / 11-Motorsport / 17-Racing / 27-Motorcycle / 927-Supermoto / \n", "0-Games / 413-PlayStation 2 / 849-Tales / \n", "36-Piano / \n", "1-Vehicle / 4-Car / 154-Wheel / 201-Grand Theft Auto: San Andreas / 299-Volkswagen Passenger Cars / 493-Rim / \n", "227-Water / \n", "3-Concert / 7-Musician / \n", "6-Animation / 10-Animal / 15-Cartoon / 1354-Frog / \n", "477-Transformers / 990-Optimus Prime / 1900-Autobot / \n", "110-Album / 322-Mixtape / \n", "0-Games / 2-Video game / 8-Football / 21-Stadium / 65-Kick / 76-Ball / 90-Sports game / 180-FIFA 15 / \n", "9-Music video / \n", "33-Weapon / 261-Tool / 315-Knife / 673-Saw / \n", "1-Vehicle / 4-Car / 114-Engine / 245-BMW / 827-Pump / 1233-BMW 3 Series / 1622-BMW M5 / 2305-BMW 5 Series / 2736-BMW 5 Series (E39) / 3097-Thermostat / \n", "1-Vehicle / 4-Car / 11-Motorsport / 70-Driving / \n", "1-Vehicle / 4-Car / 11-Motorsport / 17-Racing / 169-Rallying / 1321-Škoda Auto / \n", "256-Tennis / 974-Racket / \n", "36-Piano / 53-Keyboard / 107-Musical keyboard / 265-Electronic keyboard / 592-Synthesia / \n", "28-Fashion / 206-Shoe / 333-Nike; Inc. / 368-Sneakers / \n", "28-Fashion / 45-Cosmetics / 335-Nintendo 3DS / \n", "3-Concert / 357-DVD / \n", "5-Dance / 13-Musical ensemble / 78-Wedding / \n", "61-Art / 200-Doll / \n", "315-Knife / \n", "189-iPad / \n", "3-Concert / \n", "6-Animation / 15-Cartoon / 126-The Walt Disney Company / 616-The Bible / \n", "5-Dance / \n", "137-Tree / 208-Wood / 279-Garden / 2887-Swing / \n", "259-Fire / 371-Firefighter / \n", "22-Nature / 137-Tree / \n", "3-Concert / \n", "1-Vehicle / 4-Car / 11-Motorsport / 329-Swimming pool / 1081-Lemon / \n", "6-Animation / \n", "885-Bell / \n", "5-Dance / 6-Animation / 440-Lion / \n", "3-Concert / 7-Musician / 13-Musical ensemble / 46-Choir / \n", "262-Ice / 981-Sculpture / 1073-Chainsaw / 2339-Swan / \n", "31-Disc jockey / \n", "0-Games / 2-Video game / 734-DayZ / \n", "22-Nature / 1606-Butterfly / \n", "10-Animal / 18-Outdoor recreation / 80-Horse / 247-Stallion / 289-Dressage / 290-Pony / 351-Jumping / 404-Livestock / \n", "9-Music video / \n", "18-Outdoor recreation / 67-Cycling / 69-Bicycle / 377-BMX bike / \n", "14-Guitar / 63-Acoustic guitar / \n", "9-Music video / 78-Wedding / 121-Photography / \n", "9-Music video / \n", "1-Vehicle / \n", "3-Concert / \n", "1-Vehicle / 4-Car / 70-Driving / 344-Coupé / 945-MINI Cooper / 1221-MINI / \n", "7-Musician / 14-Guitar / 24-String instrument / 100-Electric guitar / 738-Guitar amplifier / \n", "3-Concert / 7-Musician / 13-Musical ensemble / 30-Drummer / 39-Drums / 44-Drums / \n", "1-Vehicle / 11-Motorsport / 447-Sand / 975-Dune / \n", "56-Hair / 106-Hairstyle / 405-Braid / 909-Updo / 1064-French braid / 1292-Ponytail / \n", "2882-Gladiator / \n", "1-Vehicle / 4-Car / 245-BMW / \n", "973-Nintendo 64 / \n", "1-Vehicle / 4-Car / 449-Honda / 552-Honda Civic / \n", "2-Video game / \n", "1-Vehicle / 10-Animal / 18-Outdoor recreation / 42-Fishing / 87-Boat / 91-Fish / 147-River / 225-Lake / 268-Recreational fishing / 844-Bass / 896-Largemouth bass / 2836-Walleye / \n", "36-Piano / 53-Keyboard / 104-Pianist / 107-Musical keyboard / \n", "0-Games / 2-Video game / 111-PlayStation 3 / 127-Xbox 360 / 386-Xbox / \n", "23-Mobile phone / 47-Personal computer / 72-iPhone / 131-Computer / 819-World Wide Web / \n", "7-Musician / 14-Guitar / 24-String instrument / 63-Acoustic guitar / \n", "0-Games / 2-Video game / 43-Call of Duty / 194-Call of Duty: Modern Warfare 2 / \n", "99-Christmas / 137-Tree / 365-Fireworks / 964-Christmas tree / \n", "3-Concert / 1431-Madison Square Garden / \n", "176-PlayStation 4 / 386-Xbox / \n", "79-American football / \n", "9-Music video / \n", "5-Dance / \n", "0-Games / 2-Video game / 47-Personal computer / 241-Fighting game / 486-M.U.G.E.N / 4413-Ultimate Mortal Kombat 3 / \n", "0-Games / 159-World of Warcraft / 171-Warcraft / 1213-World of Warcraft: Wrath of the Lich King / \n", "208-Wood / \n", "0-Games / 2267-Geometry Dash / \n", "5-Dance / \n", "0-Games / 703-Roller skating / 3013-Roller derby / \n", "223-Weather / 267-Mountain / 1058-Volcano / \n", "5-Dance / \n", "0-Games / 2-Video game / 237-Final Fantasy / 2506-Final Fantasy XIII-2 / \n", "10-Animal / 80-Horse / \n", "9-Music video / \n", "61-Art / 479-Graffiti / 1385-Street art / \n", "59-Winter sport / 177-Skiing / \n", "1181-Mars / \n", "56-Hair / 871-Hair coloring / \n", "10-Animal / 48-Pet / 71-Dog / 824-Terrier / 2388-Bull Terrier / 3052-Staffordshire Bull Terrier / \n", "5-Dance / 16-Performance art / \n", "7-Musician / \n", "0-Games / 8-Football / 54-Highlight film / 79-American football / \n", "1-Vehicle / 4-Car / 154-Wheel / 650-Convertible / 2016-Volvo Cars / \n", "31-Disc jockey / \n", "10-Animal / 22-Nature / 91-Fish / 275-Diving / 288-Underwater / 381-Underwater diving / 407-Scuba diving / 716-Coral / \n", "0-Games / 20-Trailer / 191-Cheerleading / 199-Wii / \n", "9-Music video / \n", "12-Food / 26-Cooking / 32-Recipe / 52-Dish / 58-Cuisine / 109-Cooking show / 274-Meat / 893-Fish as food / 1115-Steak / \n", "0-Games / 2-Video game / 19-PC game / 33-Weapon / 96-Soldier / 192-Battlefield / 736-ARMA / \n", "36-Piano / 339-Architecture / \n", "9-Music video / 1514-Crab / \n", "3-Concert / 6-Animation / \n", "455-Unidentified flying object / 502-Sky / \n", "1-Vehicle / \n", "1-Vehicle / 519-Recreational vehicle / 1376-Safari / \n", "1-Vehicle / 4-Car / 64-Transport / 286-Bus / \n", "5-Dance / \n", "12-Food / 812-Plastic / 1362-Canvas / \n", "5-Dance / \n", "8-Football / 79-American football / \n", "1-Vehicle / 62-Train / 64-Transport / 103-Rail transport / 161-Train station / 224-Rapid transit / \n", "31-Disc jockey / \n", "23-Mobile phone / 25-Toy / 29-Smartphone / 88-Machine / 154-Wheel / 319-Robot / \n", "15-Cartoon / 379-Touhou Project / \n", "0-Games / 8-Football / 21-Stadium / 65-Kick / 73-Ball / 76-Ball / 77-Arena / \n", "23-Mobile phone / 29-Smartphone / 387-Spider-Man / 3561-The Amazing Spider-Man / \n", "0-Games / 2-Video game / 19-PC game / 33-Weapon / 271-Tank / 409-World of Tanks / \n", "3-Concert / 7-Musician / 30-Drummer / 39-Drums / 44-Drums / 128-Cymbal / \n", "1-Vehicle / 4-Car / 11-Motorsport / 17-Racing / \n", "5-Dance / 78-Wedding / 190-Bride / 421-Gown / \n", "0-Games / 2-Video game / 19-PC game / 33-Weapon / 96-Soldier / 257-Counter-Strike / \n", "1-Vehicle / 4-Car / 158-Tire / \n", "574-Elevator / 1746-Otis Elevator Company / \n", "14-Guitar / 24-String instrument / 63-Acoustic guitar / 100-Electric guitar / \n", "102-Building / 174-Farm / 534-Milk / \n", "0-Games / 2-Video game / 19-PC game / 51-Strategy video game / 382-Dota 2 / 549-Defense of the Ancients / \n", "25-Toy / 434-Ninja / 951-Teenage Mutant Ninja Turtles / \n", "3-Concert / \n", "1-Vehicle / 87-Boat / \n", "0-Games / 2-Video game / 34-Action-adventure game / 118-Grand Theft Auto V / \n", "2-Video game / \n", "1-Vehicle / 50-Aircraft / 102-Building / 2326-Royal Air Force / \n", "512-Printing / 525-Printer / 1063-Ink / 2634-Inkjet printing / \n", "516-Headphones / \n", "0-Games / 17-Racing / 19-PC game / 837-Roblox / \n", "819-World Wide Web / \n", "0-Games / 2-Video game / 19-PC game / 34-Action-adventure game / 118-Grand Theft Auto V / \n", "227-Water / 1119-Sensor / 2472-Tap / \n", "61-Art / 242-Paper / 370-Origami / 2604-Envelope / \n", "23-Mobile phone / 29-Smartphone / 178-Samsung Galaxy / \n", "907-Necklace / 1082-Knot / \n", "0-Games / 88-Machine / 535-Slot machine / \n", "126-The Walt Disney Company / 144-Amusement park / 576-Walt Disney World / \n", "3-Concert / 7-Musician / 36-Piano / 53-Keyboard / 104-Pianist / \n", "66-Bollywood / \n", "5-Dance / 115-Ballet / \n", "1-Vehicle / 124-Tractor / 156-Agriculture / 363-Soil / 906-Sowing / \n", "0-Games / 2-Video game / 67-Cycling / 69-Bicycle / \n", "28-Fashion / 45-Cosmetics / 195-Mascara / 196-Eye / \n", "703-Roller skating / \n", "9-Music video / 31-Disc jockey / 293-Slide show / \n", "827-Pump / \n", "0-Games / 2-Video game / 1690-Warhammer Fantasy Battle / \n", "0-Games / 21-Stadium / 60-Basketball / 77-Arena / \n", "0-Games / 8-Football / 21-Stadium / \n", "0-Games / 2-Video game / 6-Animation / 15-Cartoon / 159-World of Warcraft / 171-Warcraft / \n", "1-Vehicle / 62-Train / 64-Transport / 103-Rail transport / 119-Locomotive / 152-Track / 154-Wheel / 428-Rail transport modelling / \n", "5-Dance / 16-Performance art / \n", "0-Games / 2-Video game / 34-Action-adventure game / 85-Combat / 380-Assassin's Creed / 1048-Assassin's Creed / 1140-Assassin's Creed III / 1435-Ezio Auditore da Firenze / \n", "0-Games / 2-Video game / 19-PC game / 257-Counter-Strike / \n", "0-Games / 2-Video game / \n", "0-Games / 8-Football / 90-Sports game / 176-PlayStation 4 / 180-FIFA 15 / \n", "6-Animation / 379-Touhou Project / 3528-Imperishable Night / \n", "1528-Euro / \n", "9-Music video / \n", "10-Animal / 3710-Kangaroo / 4253-Koala / \n", "1-Vehicle / 4-Car / 40-Road / 64-Transport / 74-Truck / 250-Parade / 796-Fire engine / 827-Pump / 1020-Emergency vehicle / \n", "5-Dance / 16-Performance art / \n", "3-Concert / 5-Dance / \n", "0-Games / 2-Video game / 20-Trailer / 903-Mass Effect / \n", "7-Musician / 14-Guitar / 24-String instrument / \n", "0-Games / 102-Building / 1938-Rust / 3418-The Elder Scrolls III: Morrowind / \n", "1-Vehicle / 22-Nature / 137-Tree / 433-Construction / 794-Bridge / \n", "31-Disc jockey / 133-Nightclub / 988-Ibiza / \n", "7-Musician / 14-Guitar / 24-String instrument / 63-Acoustic guitar / \n", "10-Animal / 48-Pet / 71-Dog / 182-Puppy / 824-Terrier / \n", "3-Concert / 16-Performance art / \n", "5-Dance / 49-School / 95-Talent show / 191-Cheerleading / \n", "1-Vehicle / 92-Radio-controlled model / 202-Radio-controlled car / \n", "93-Comedy / 341-Sketch comedy / 364-Sitcom / 2764-Mad / \n", "165-League of Legends / 1100-Hat / \n", "1-Vehicle / 11-Motorsport / 17-Racing / 27-Motorcycle / 105-Motorcycling / 172-Motocross / \n", "10-Animal / 132-Bird / 3378-Pigeon racing / \n", "0-Games / 1-Vehicle / 2-Video game / 4-Car / 34-Action-adventure game / 41-Sports car / 297-Grand Theft Auto IV / 658-Grand Theft Auto: The Lost and Damned / \n", "1841-Path of Exile / \n", "0-Games / 2-Video game / 19-PC game / 241-Fighting game / 486-M.U.G.E.N / 764-The King of Fighters / \n", "110-Album / \n", "6-Animation / 15-Cartoon / \n", "8-Football / 21-Stadium / 77-Arena / \n", "12-Food / 26-Cooking / 32-Recipe / 52-Dish / 58-Cuisine / 109-Cooking show / 179-Kitchen / 184-Vegetable / 210-Cookware and bakeware / 301-Meal / 557-Salad / 752-Pasta / 1024-Culinary art / \n", "6-Animation / 15-Cartoon / 183-Manga / 3239-Hakuōki / \n", "3-Concert / 5-Dance / 16-Performance art / \n", "36-Piano / \n", "648-Clown / \n", "1-Vehicle / 10-Animal / 18-Outdoor recreation / 42-Fishing / 87-Boat / 91-Fish / 268-Recreational fishing / 481-Kayak / \n", "1-Vehicle / 4-Car / 11-Motorsport / 17-Racing / 40-Road / 41-Sports car / 57-Race track / 711-Gran Turismo / 2718-Gran Turismo 4 / \n", "0-Games / 2-Video game / 34-Action-adventure game / 413-PlayStation 2 / 1125-God of War / 1432-Kratos / 1743-God of War / 2754-God of War II / \n", "3-Concert / 7-Musician / \n", "1-Vehicle / 87-Boat / 753-Sailboat / \n", "7-Musician / 13-Musical ensemble / 203-Marching band / \n", "23-Mobile phone / 29-Smartphone / 37-Gadget / 101-Telephone / 178-Samsung Galaxy / \n", "10-Animal / 2867-Wasp / \n", "0-Games / 2-Video game / 19-PC game / 43-Call of Duty / 96-Soldier / 194-Call of Duty: Modern Warfare 2 / \n", "0-Games / 8-Football / 21-Stadium / \n", "6-Animation / \n", "413-PlayStation 2 / \n", "34-Action-adventure game / \n", "1-Vehicle / 4-Car / 11-Motorsport / 209-University / 365-Fireworks / \n", "8-Football / \n", "0-Games / 35-Minecraft / 102-Building / \n", "0-Games / 8-Football / 54-Highlight film / 76-Ball / \n", "159-World of Warcraft / 171-Warcraft / 1213-World of Warcraft: Wrath of the Lich King / \n", "8-Football / \n", "273-News program / \n", "102-Building / 208-Wood / 279-Garden / 462-Table / \n", "8-Football / \n", "3-Concert / \n", "86-Plant / 136-Gardening / 279-Garden / 906-Sowing / 1022-Chili pepper / \n", "5-Dance / \n", "28-Fashion / 45-Cosmetics / 49-School / 206-Shoe / 880-Trousers / \n", "0-Games / 6-Animation / 3863-Temple Run / \n", "47-Personal computer / 145-Tablet computer / 900-MacBook / 1086-MacBook Pro / \n", "12-Food / \n", "415-Sewing / 830-Bed / \n", "6-Animation / 15-Cartoon / 20-Trailer / 1164-Cartoon Network / \n", "1095-Shiva / 1524-Zee Bangla / \n", "1-Vehicle / 27-Motorcycle / 245-BMW / \n", "93-Comedy / \n", "1-Vehicle / 4-Car / 11-Motorsport / 41-Sports car / 207-Exhaust system / 2272-Fifth Generation Nissan Z-car (Z33) / \n", "205-Newscaster / \n", "13-Musical ensemble / \n", "7-Musician / 36-Piano / 53-Keyboard / 104-Pianist / \n", "12-Food / 52-Dish / 215-Eating / 3385-Burrito / \n", "1-Vehicle / \n", "22-Nature / 225-Lake / \n", "1-Vehicle / 50-Aircraft / 139-Airport / 151-Landing / 254-Runway / \n", "0-Games / 756-Tibia / \n", "0-Games / 2-Video game / 19-PC game / 686-Diablo III / 872-Diablo / \n", "6-Animation / 1456-Clothes dryer / \n", "0-Games / 2-Video game / 33-Weapon / 47-Personal computer / 80-Horse / 96-Soldier / \n", "3-Concert / \n", "222-Surfing / 225-Lake / 1667-Standup paddleboarding / \n", "99-Christmas / \n", "293-Slide show / \n", "3-Concert / 5-Dance / \n", "0-Games / 2-Video game / 90-Sports game / 241-Fighting game / 2301-Dead or Alive / \n", "320-Super Smash Bros. / \n", "341-Sketch comedy / \n", "0-Games / 240-Arcade game / 687-Taiko no Tatsujin / \n", "121-Photography / 155-Camera / 629-Camera lens / \n", "0-Games / 2-Video game / 21-Stadium / 60-Basketball / 77-Arena / 81-Athlete / 90-Sports game / 168-Basketball moves / 255-Slam dunk / \n", "3-Concert / 7-Musician / 39-Drums / \n", "0-Games / 35-Minecraft / \n", "9-Music video / \n", "110-Album / \n", "9-Music video / 310-The Sims / 543-The Sims 2 / 564-The Sims 3 / \n", "0-Games / 15-Cartoon / 687-Taiko no Tatsujin / \n", "1-Vehicle / 4-Car / 25-Toy / 92-Radio-controlled model / 167-Four-wheel drive / 202-Radio-controlled car / 2742-Mini 4WD / \n", "3286-Stunt Performer / \n", "1-Vehicle / 27-Motorcycle / 522-Helmet / \n", "664-CrossFire / \n", "1-Vehicle / 4-Car / 25-Toy / 2742-Mini 4WD / \n", "86-Plant / 136-Gardening / 279-Garden / \n", "95-Talent show / \n", "6-Animation / 3363-Typography / \n", "209-University / \n", "78-Wedding / 1460-Herb / \n", "205-Newscaster / 2642-CNBC / \n", "9-Music video / 18-Outdoor recreation / \n", "122-Weight training / 204-Gym / 529-Squat / 532-Barbell / \n", "12-Food / 329-Swimming pool / 2258-Patio / \n", "9-Music video / \n", "33-Weapon / 261-Tool / 315-Knife / 2387-Military tactics / \n", "573-Paragliding / \n", "222-Surfing / 762-Angel / \n", "3-Concert / 5-Dance / 16-Performance art / \n", "1-Vehicle / 4-Car / 11-Motorsport / 41-Sports car / 231-Supercar / \n", "3-Concert / 7-Musician / 30-Drummer / \n", "199-Wii / 737-Joker / 1199-Kinect / \n", "28-Fashion / 45-Cosmetics / 56-Hair / 173-Eye shadow / 193-Eye liner / 195-Mascara / 196-Eye / 218-Eyelash / 235-Lipstick / 282-Skin / 291-Rouge / 436-Concealer / 3655-Makeup brush / \n", "28-Fashion / 45-Cosmetics / 282-Skin / \n", "0-Games / 2-Video game / 229-Halo / 470-Halo 3 / \n", "7-Musician / 13-Musical ensemble / 46-Choir / \n", "1-Vehicle / \n", "1-Vehicle / 11-Motorsport / 27-Motorcycle / 154-Wheel / 158-Tire / 493-Rim / \n", "25-Toy / 229-Halo / 435-Action figure / 749-Figurine / 1161-Halo 2 / 2121-Marvel Legends / \n", "264-Star Wars / 468-Biology / 1130-Armour / 2576-Clone trooper / \n", "1-Vehicle / 4-Car / 40-Road / 212-Highway / 519-Recreational vehicle / \n", "61-Art / 628-Portrait / 873-Acrylic paint / 1362-Canvas / \n", "0-Games / 2-Video game / 51-Strategy video game / 537-Zoo / 4255-Zoo Tycoon 2 / 4535-Zoo Tycoon / \n", "1-Vehicle / 4-Car / 40-Road / 64-Transport / 70-Driving / 74-Truck / 94-Dashcam / 212-Highway / 575-Traffic / \n", "0-Games / 2-Video game / 19-PC game / 33-Weapon / 3168-Resistance / \n", "0-Games / 2-Video game / 359-Harry Potter (Literary Series) / 4502-Harry Potter and the Order of the Phoenix / \n", "28-Fashion / 206-Shoe / 681-Leather / 757-Boot / 919-Hair conditioner / \n", "0-Games / 2-Video game / 1087-Donkey Kong / 2036-Donkey Kong Country / 2274-Donkey Kong / 2970-Donkey Kong / \n", "3-Concert / 7-Musician / 14-Guitar / 24-String instrument / 63-Acoustic guitar / \n", "159-World of Warcraft / 171-Warcraft / \n", "45-Cosmetics / 218-Eyelash / 3106-Eyelash extensions / \n", "23-Mobile phone / 101-Telephone / \n", "5-Dance / 20-Trailer / \n", "0-Games / 2-Video game / 240-Arcade game / 311-Street Fighter / 461-Street Fighter IV / 496-Super Street Fighter IV / \n", "18-Outdoor recreation / 42-Fishing / 222-Surfing / \n", "1-Vehicle / 62-Train / 64-Transport / 94-Dashcam / \n", "3-Concert / 16-Performance art / 30-Drummer / \n", "0-Games / 2-Video game / 237-Final Fantasy / 432-Kingdom Hearts / 665-Kingdom Hearts / 1857-Sephiroth / 2127-Final Fantasy / \n", "1-Vehicle / 4-Car / 11-Motorsport / 17-Racing / 27-Motorcycle / 105-Motorcycling / \n", "0-Games / 942-Silkroad Online / \n", "0-Games / 2-Video game / 55-Video game console / 199-Wii / 320-Super Smash Bros. / 335-Nintendo 3DS / 443-Wii U / 697-Super Smash Bros. for Nintendo 3DS and Wii U / 782-Super Smash Bros. Melee / \n", "6-Animation / 15-Cartoon / 571-Bear / \n", "6-Animation / 20-Trailer / 126-The Walt Disney Company / \n", "93-Comedy / 355-Comedian / \n", "93-Comedy / \n", "376-Flute / 4605-Bansuri / \n", "165-League of Legends / \n", "1-Vehicle / 27-Motorcycle / 207-Exhaust system / \n", "18-Outdoor recreation / 22-Nature / 275-Diving / 288-Underwater / 381-Underwater diving / 407-Scuba diving / 468-Biology / \n", "9-Music video / 78-Wedding / \n", "10-Animal / 71-Dog / \n", "609-Tattoo / \n", "0-Games / 35-Minecraft / 488-Door / \n", "3-Concert / 7-Musician / \n", "67-Cycling / 69-Bicycle / 83-Skateboarding / 108-Skateboard / 555-Longboard / \n", "1-Vehicle / 4-Car / 11-Motorsport / 17-Racing / 169-Rallying / \n", "3-Concert / 7-Musician / 13-Musical ensemble / 16-Performance art / 38-Orchestra / \n", "1-Vehicle / 11-Motorsport / 17-Racing / 27-Motorcycle / 105-Motorcycling / 172-Motocross / 491-Enduro / 669-Extreme sport / \n", "10-Animal / 12-Food / 48-Pet / 52-Dish / 184-Vegetable / 215-Eating / 636-Rabbit / \n", "1017-Yo-yo / \n", "0-Games / 2-Video game / 34-Action-adventure game / 921-Grand Theft Auto: Vice City / \n", "8-Football / 21-Stadium / 77-Arena / \n", "4559-Federal Senate / \n", "0-Games / 34-Action-adventure game / 201-Grand Theft Auto: San Andreas / \n", "0-Games / 2-Video game / 33-Weapon / 43-Call of Duty / 216-Call of Duty: Modern Warfare 3 / \n", "0-Games / 2-Video game / 10-Animal / \n", "162-Medicine / 784-Honey / \n", "56-Hair / 798-Shampoo / \n", "25-Toy / 88-Machine / 140-LEGO / 319-Robot / 357-DVD / 3753-Plotter / \n", "93-Comedy / \n", "1-Vehicle / 164-Beach / 233-Ocean / 423-Coast / 573-Paragliding / 1561-Paramotor / \n", "93-Comedy / \n", "3-Concert / 7-Musician / 16-Performance art / \n", "0-Games / 2-Video game / 19-PC game / 33-Weapon / 664-CrossFire / \n", "0-Games / 2-Video game / 34-Action-adventure game / 55-Video game console / 201-Grand Theft Auto: San Andreas / \n", "0-Games / 8-Football / 21-Stadium / 65-Kick / \n", "0-Games / 10-Animal / 35-Minecraft / 80-Horse / \n", "8-Football / 1193-Orange / \n", "3-Concert / 110-Album / \n", "205-Newscaster / 463-Cricket / \n", "3-Concert / 7-Musician / 13-Musical ensemble / 30-Drummer / 39-Drums / \n", "347-Climbing / 828-Rock climbing / \n", "0-Games / 8-Football / 21-Stadium / \n", "45-Cosmetics / 173-Eye shadow / 193-Eye liner / 196-Eye / 218-Eyelash / 235-Lipstick / \n", "1-Vehicle / 4-Car / 11-Motorsport / 17-Racing / 41-Sports car / 57-Race track / 135-Drifting / \n", "0-Games / 2-Video game / 19-PC game / 33-Weapon / 464-First-person Shooter / \n", "150-Violin / \n", "0-Games / 2-Video game / 34-Action-adventure game / 774-Galaxy / 1397-Super Mario Galaxy / 1759-Super Mario Galaxy 2 / \n", "61-Art / 148-Painting / 303-Paint / 873-Acrylic paint / 936-Watercolor paint / 1362-Canvas / 3177-Tulip / \n", "261-Tool / \n", "1-Vehicle / 25-Toy / 62-Train / 507-Thomas the Tank Engine / 1279-Plarail / 1685-Toy train / \n", "66-Bollywood / \n", "9-Music video / \n", "133-Nightclub / 559-Night / \n", "3-Concert / 9-Music video / \n", "1-Vehicle / 4-Car / 11-Motorsport / 41-Sports car / 70-Driving / 231-Supercar / 344-Coupé / 517-Audi / 1537-Audi R8 / 1587-Audi Quattro / 2949-V10 engine / \n", "5-Dance / 16-Performance art / 361-Flamenco / \n", "25-Toy / 89-Comics / 435-Action figure / 803-DC Comics / \n", "5-Dance / 16-Performance art / \n", "3857-Muskmelon / \n", "12-Food / 747-Retail / 1328-Grocery store / 1782-Supermarket / \n", "1183-Karaoke box / \n", "748-Iron Man / \n", "1-Vehicle / 4-Car / 27-Motorcycle / 40-Road / 274-Meat / \n", "0-Games / 1-Vehicle / 2-Video game / 4-Car / 6-Animation / 17-Racing / \n", "0-Games / 2-Video game / 51-Strategy video game / 72-iPhone / 1120-Tower defense / \n", "2478-Red hair / \n", "55-Video game console / 176-PlayStation 4 / 317-PlayStation / 677-Game controller / 1262-NBA 2K14 / 1482-PlayStation Network / 2074-Killzone / \n", "3-Concert / 46-Choir / 209-University / \n", "45-Cosmetics / 173-Eye shadow / 193-Eye liner / 195-Mascara / 218-Eyelash / 291-Rouge / 436-Concealer / \n", "83-Skateboarding / 108-Skateboard / \n", "170-Home improvement / 276-Room / 353-Wall / 2743-Wallpaper / \n", "0-Games / 51-Strategy video game / 165-League of Legends / \n", "28-Fashion / 61-Art / \n", "0-Games / 60-Basketball / 168-Basketball moves / \n", "14-Guitar / \n", "0-Games / 2-Video game / 35-Minecraft / \n", "3-Concert / 7-Musician / \n", "14-Guitar / 24-String instrument / 490-Rain / \n", "1-Vehicle / 4-Car / 40-Road / 64-Transport / 70-Driving / 74-Truck / 94-Dashcam / \n", "88-Machine / 261-Tool / 972-Lathe / \n", "10-Animal / 1373-Torero / \n", "0-Games / 2-Video game / 34-Action-adventure game / \n", "10-Animal / 22-Nature / 280-Wildlife / \n", "0-Games / 2-Video game / 9-Music video / 43-Call of Duty / 138-Call of Duty: Black Ops II / \n", "23-Mobile phone / 29-Smartphone / 72-iPhone / \n", "0-Games / 2-Video game / 928-Borderlands 2 / 2613-Borderlands: The Pre-Sequel! / \n", "1-Vehicle / 67-Cycling / 69-Bicycle / \n", "49-School / 113-House / 131-Computer / 209-University / \n", "9-Music video / \n", "22-Nature / 86-Plant / 689-Harmonica / \n", "8-Football / 65-Kick / \n", "36-Piano / 53-Keyboard / \n", "0-Games / 9-Music video / \n", "13-Musical ensemble / \n", "28-Fashion / \n", "23-Mobile phone / 29-Smartphone / 121-Photography / 155-Camera / 566-Digital camera / \n", "3151-Kermit the Frog / \n", "0-Games / 2-Video game / 1789-Slender: The Eight Pages / \n", "7-Musician / 93-Comedy / 355-Comedian / \n", "25-Toy / 116-Pokémon / 1152-Plush / \n", "1-Vehicle / 4-Car / 11-Motorsport / 17-Racing / 41-Sports car / 4395-Sim racing / \n", "3-Concert / 7-Musician / 13-Musical ensemble / 38-Orchestra / \n", "66-Bollywood / \n", "0-Games / 8-Football / 54-Highlight film / 81-Athlete / \n", "0-Games / 1-Vehicle / 4-Car / 11-Motorsport / 17-Racing / 41-Sports car / 57-Race track / 2010-Chart / \n", "0-Games / 35-Minecraft / 637-Five Nights at Freddy's / \n", "6-Animation / 123-Naruto / \n", "3-Concert / 46-Choir / \n", "89-Comics / \n", "7-Musician / 13-Musical ensemble / 230-Brass instrument / \n", "2-Video game / 18-Outdoor recreation / 51-Strategy video game / 144-Amusement park / 321-Roller coaster / 334-Amusement ride / 1293-RollerCoaster Tycoon 3 / \n", "3-Concert / 7-Musician / 13-Musical ensemble / 46-Choir / 49-School / \n", "25-Toy / 33-Weapon / 140-LEGO / \n", "3-Concert / 361-Flamenco / \n", "6-Animation / 25-Toy / 200-Doll / 439-Barbie / 1272-Cabinetry / \n", "6-Animation / 9-Music video / 15-Cartoon / \n", "1-Vehicle / 4-Car / 403-Toyota / 1840-Toyota Hilux / \n", "49-School / \n", "6-Animation / 477-Transformers / \n", "242-Paper / 370-Origami / \n", "5-Dance / 98-Festival / 130-Gymnastics / 250-Parade / 1835-Majorette / \n", "7-Musician / 1621-Ork / \n", "0-Games / 159-World of Warcraft / 171-Warcraft / 643-Hero / \n", "246-Furniture / 276-Room / 416-Apartment / \n", "1030-Trampoline / 1998-Canon EOS 600D / \n", "0-Games / 1-Vehicle / 2-Video game / 4-Car / 17-Racing / 41-Sports car / 231-Supercar / 352-Need for Speed / 406-Chevrolet / 952-Chevrolet Corvette / 1600-Need for Speed: World / 4135-Chevrolet Corvette C5 Z06 / \n", "5-Dance / 9-Music video / \n", "5-Dance / 16-Performance art / \n", "10-Animal / 17-Racing / 80-Horse / 384-Horse racing / \n", "1-Vehicle / 4-Car / 154-Wheel / 158-Tire / \n", "1-Vehicle / 4-Car / 74-Truck / 167-Four-wheel drive / 327-Mud / 375-Pickup truck / \n", "3-Concert / 5-Dance / 16-Performance art / 98-Festival / 209-University / \n", "1-Vehicle / 11-Motorsport / 17-Racing / 27-Motorcycle / 57-Race track / 105-Motorcycling / 172-Motocross / \n", "1225-Antenna / \n", "5-Dance / 115-Ballet / \n", "0-Games / 2-Video game / 19-PC game / 35-Minecraft / \n", "0-Games / 2-Video game / 4465-ourWorld / \n", "78-Wedding / 121-Photography / 633-Photograph / 2269-Wedding photography / \n", "3-Concert / 68-Lighting / 77-Arena / \n", "1-Vehicle / 4-Car / 11-Motorsport / 17-Racing / 27-Motorcycle / 40-Road / 92-Radio-controlled model / 202-Radio-controlled car / 3839-Volkswagen Golf Mk5 / \n", "0-Games / 8-Football / 54-Highlight film / 79-American football / \n", "1-Vehicle / 4-Car / 11-Motorsport / 17-Racing / 40-Road / 57-Race track / 378-Drag racing / 406-Chevrolet / \n", "1-Vehicle / 87-Boat / \n", "3-Concert / \n", "971-Alphabet / \n", "42-Fishing / 1082-Knot / 1369-Necktie / \n", "3-Concert / 31-Disc jockey / 68-Lighting / 133-Nightclub / \n", "14-Guitar / 24-String instrument / 63-Acoustic guitar / \n", "15-Cartoon / 877-Hatsune Miku: Project DIVA / 1648-Hatsune Miku: Project DIVA F / 2682-Hatsune Miku: Project DIVA F 2nd / \n", "3-Concert / 5-Dance / 16-Performance art / \n", "1-Vehicle / 11-Motorsport / 124-Tractor / \n", "1-Vehicle / 144-Amusement park / \n", "465-Microphone / \n", "12-Food / 442-Bread / 2928-Cereal / \n", "5-Dance / 16-Performance art / 95-Talent show / \n", "0-Games / 2-Video game / 33-Weapon / 43-Call of Duty / 176-PlayStation 4 / 181-Xbox / 314-Xbox One / 422-Call of Duty: Ghosts / \n", "45-Cosmetics / \n", "419-Earth / \n", "12-Food / 26-Cooking / 32-Recipe / 274-Meat / \n", "0-Games / 88-Machine / 240-Arcade game / 387-Spider-Man / 535-Slot machine / \n", "23-Mobile phone / 29-Smartphone / 37-Gadget / 72-iPhone / 508-iPhone 5 / \n", "0-Games / 35-Minecraft / \n", "3-Concert / 7-Musician / 13-Musical ensemble / 38-Orchestra / \n", "3-Concert / \n", "3-Concert / 31-Disc jockey / 68-Lighting / 133-Nightclub / 227-Water / \n", "56-Hair / 106-Hairstyle / \n", "3-Concert / 7-Musician / \n", "3-Concert / 7-Musician / 13-Musical ensemble / 38-Orchestra / 46-Choir / \n", "3-Concert / 7-Musician / 13-Musical ensemble / 38-Orchestra / \n", "10-Animal / 48-Pet / 132-Bird / \n", "113-House / \n", "452-Watch / 627-Clock / \n", "84-Snow / 120-Winter / 365-Fireworks / \n", "1-Vehicle / 62-Train / 64-Transport / 103-Rail transport / 119-Locomotive / 143-Railroad car / 152-Track / 161-Train station / 224-Rapid transit / \n", "93-Comedy / \n", "1-Vehicle / 11-Motorsport / 50-Aircraft / 82-Airplane / 102-Building / 139-Airport / 151-Landing / 157-Jet aircraft / 175-Airliner / 187-Takeoff / 188-Airline / 254-Runway / \n", "126-The Walt Disney Company / \n", "8-Football / \n", "0-Games / 1-Vehicle / 2-Video game / 11-Motorsport / 456-Go-kart / 518-Mario Kart / 2909-Mario Kart 7 / \n", "34-Action-adventure game / 47-Personal computer / 131-Computer / 201-Grand Theft Auto: San Andreas / \n", "273-News program / \n", "14-Guitar / 24-String instrument / 100-Electric guitar / \n", "14-Guitar / 24-String instrument / 63-Acoustic guitar / 2744-Bajo sexto / \n", "0-Games / 2-Video game / \n", "1-Vehicle / 4-Car / 92-Radio-controlled model / 202-Radio-controlled car / \n", "1088-Infomercial / \n", "5-Dance / 78-Wedding / 190-Bride / 430-Ballroom dance / \n", "12-Food / 26-Cooking / 519-Recreational vehicle / \n", "3-Concert / 16-Performance art / \n", "36-Piano / \n", "287-Human swimming / 293-Slide show / 710-Handball / 2273-Polo / 2360-Water polo / \n", "6-Animation / 113-House / 170-Home improvement / 179-Kitchen / 742-Countertop / \n", "1-Vehicle / 4-Car / 11-Motorsport / 114-Engine / 650-Convertible / 1521-Jaguar Cars / \n", "3-Concert / 5-Dance / 16-Performance art / \n", "10-Animal / 18-Outdoor recreation / 42-Fishing / 233-Ocean / 1471-Tuna / \n", "12-Food / 88-Machine / 262-Ice / \n", "12-Food / 534-Milk / 1617-Fungus / \n", "45-Cosmetics / 193-Eye liner / 195-Mascara / 218-Eyelash / 235-Lipstick / 291-Rouge / 641-Mask / \n", "0-Games / 1-Vehicle / 2-Video game / 4-Car / 11-Motorsport / 17-Racing / 41-Sports car / 57-Race track / 231-Supercar / 352-Need for Speed / 1637-GameCube / 1770-Need for Speed: Hot Pursuit / \n", "9-Music video / \n", "1183-Karaoke box / \n", "0-Games / 2-Video game / 19-PC game / \n", "0-Games / 1-Vehicle / 2-Video game / 4-Car / 34-Action-adventure game / 118-Grand Theft Auto V / \n", "18-Outdoor recreation / 83-Skateboarding / 108-Skateboard / \n", "0-Games / 734-DayZ / 736-ARMA / 1034-ARMA 2 / 1216-DayZ / \n", "2-Video game / \n", "0-Games / 6-Animation / 411-MapleStory / \n", "242-Paper / 370-Origami / 1449-Rubber band / \n", "1868-Preacher / \n", "322-Mixtape / \n", "10-Animal / 80-Horse / \n", "96-Soldier / \n", "1-Vehicle / 64-Transport / 74-Truck / 1070-Semi-trailer truck / \n", "1043-Router / \n", "0-Games / 366-Dragon / \n", "31-Disc jockey / \n", "0-Games / 2-Video game / 34-Action-adventure game / 921-Grand Theft Auto: Vice City / \n", "18-Outdoor recreation / 83-Skateboarding / 108-Skateboard / 1031-Fisheye lens / \n", "99-Christmas / 729-Christmas decoration / 964-Christmas tree / \n", "23-Mobile phone / 29-Smartphone / 37-Gadget / 47-Personal computer / 72-iPhone / 145-Tablet computer / 189-iPad / \n", "0-Games / 2-Video game / 19-PC game / 34-Action-adventure game / 201-Grand Theft Auto: San Andreas / 297-Grand Theft Auto IV / \n", "12-Food / 26-Cooking / \n", "0-Games / 60-Basketball / 191-Cheerleading / \n", "1-Vehicle / 50-Aircraft / 82-Airplane / 2391-American Airlines / \n", "130-Gymnastics / 350-Circus / 788-Juggling / \n", "0-Games / 2-Video game / 2658-Mirror's Edge / \n", "66-Bollywood / \n", "102-Building / 113-House / 170-Home improvement / 279-Garden / \n", "3-Concert / 133-Nightclub / 925-Brain / \n", "103-Rail transport / 428-Rail transport modelling / \n", "5-Dance / 85-Combat / 430-Ballroom dance / 4235-American Kenpo / \n", "1-Vehicle / 25-Toy / 264-Star Wars / \n", "0-Games / 2-Video game / 426-Resident Evil / 506-Garry's Mod / 1276-Resident Evil / \n", "3-Concert / 16-Performance art / \n", "0-Games / 59-Winter sport / 112-Ice skating / 142-Hockey / 3612-NHL 15 / \n", "1-Vehicle / 4-Car / \n", "0-Games / 59-Winter sport / 90-Sports game / 112-Ice skating / 142-Hockey / 2546-NHL 14 / \n", "1-Vehicle / 4-Car / \n", "1-Vehicle / 62-Train / 64-Transport / 103-Rail transport / 119-Locomotive / 143-Railroad car / 161-Train station / 224-Rapid transit / \n", "376-Flute / \n", "13-Musical ensemble / 250-Parade / 820-Bagpipes / \n", "6-Animation / 717-Darth Vader / \n", "3-Concert / 5-Dance / 16-Performance art / 115-Ballet / \n", "3622-Def Jam Recordings / \n", "102-Building / \n", "20-Trailer / 541-Black-and-white / \n", "242-Paper / 370-Origami / 2454-Notebook / \n", "3-Concert / \n", "113-House / 170-Home improvement / 303-Paint / 353-Wall / 1350-Plaster / 4139-Stucco / \n", "5-Dance / 16-Performance art / 980-Solo dance / \n", "117-Boxing / \n", "1-Vehicle / 4-Car / 27-Motorcycle / 4436-Yamaha YZ450F / \n", "359-Harry Potter (Literary Series) / \n", "9-Music video / \n", "3-Concert / 5-Dance / 31-Disc jockey / 133-Nightclub / \n", "66-Bollywood / \n", "162-Medicine / \n", "28-Fashion / 206-Shoe / 333-Nike; Inc. / 368-Sneakers / \n", "0-Games / 2-Video game / 19-PC game / 33-Weapon / 420-The Elder Scrolls V: Skyrim / 441-The Elder Scrolls / \n", "23-Mobile phone / 29-Smartphone / 37-Gadget / 72-iPhone / \n", "6-Animation / 15-Cartoon / \n", "59-Winter sport / 112-Ice skating / 142-Hockey / \n", "102-Building / 318-Church / 397-Organ / 646-Pipe organ / 870-Chapel / \n", "322-Mixtape / \n", "1061-Marriage proposal / \n", "0-Games / 2-Video game / 55-Video game console / 116-Pokémon / 199-Wii / 1298-Pikachu / \n", "3-Concert / 7-Musician / 13-Musical ensemble / 30-Drummer / 39-Drums / \n", "5-Dance / 4004-Tug of war / \n", "1-Vehicle / 573-Paragliding / \n", "0-Games / 2-Video game / 19-PC game / 51-Strategy video game / 382-Dota 2 / 549-Defense of the Ancients / 1378-Multiplayer online battle arena / \n", "12-Food / 26-Cooking / 32-Recipe / 52-Dish / 58-Cuisine / 109-Cooking show / 179-Kitchen / 1024-Culinary art / 1843-Ham / 3356-Cauliflower / 4153-Gratin / \n", "9-Music video / \n", "125-Television / \n", "42-Fishing / \n", "341-Sketch comedy / \n", "9-Music video / \n", "28-Fashion / 45-Cosmetics / 193-Eye liner / 195-Mascara / 218-Eyelash / \n", "1-Vehicle / 50-Aircraft / \n", "93-Comedy / \n", "5-Dance / 78-Wedding / \n", "1-Vehicle / 4-Car / 11-Motorsport / 154-Wheel / 349-Chevrolet / 406-Chevrolet / 509-Classic car / 647-Muscle car / 1148-Hot rod / \n", "0-Games / 8-Football / 21-Stadium / \n", "0-Games / 2-Video game / 241-Fighting game / 320-Super Smash Bros. / 782-Super Smash Bros. Melee / 973-Nintendo 64 / \n", "5-Dance / \n", "12-Food / 26-Cooking / 32-Recipe / 52-Dish / 58-Cuisine / 109-Cooking show / 210-Cookware and bakeware / \n", "9-Music video / \n", "0-Games / 1-Vehicle / 2-Video game / 4-Car / 34-Action-adventure game / 118-Grand Theft Auto V / \n", "1-Vehicle / 4-Car / 2737-GAZ / \n", "85-Combat / \n", "3-Concert / 68-Lighting / \n", "28-Fashion / 153-Dress / \n", "0-Games / 8-Football / \n", "14-Guitar / 24-String instrument / 63-Acoustic guitar / \n", "6-Animation / 1139-Belle / 3975-Sleeping Beauty / 4264-Princess Aurora / \n", "462-Table / 544-Table tennis / \n", "6-Animation / \n", "31-Disc jockey / \n", "12-Food / 1596-Cinnamon / 2475-Spoon / \n", "323-Toddler / 567-Compact disc / 695-Balloon / \n", "5-Dance / 16-Performance art / \n", "113-House / 170-Home improvement / \n", "3-Concert / 7-Musician / 13-Musical ensemble / 696-T-shirt / \n", "99-Christmas / 120-Winter / 150-Violin / \n", "125-Television / \n", "0-Games / 141-Microsoft Windows / \n", "7-Musician / 14-Guitar / 24-String instrument / \n", "10-Animal / 91-Fish / 1258-Goldfish / \n", "164-Beach / 222-Surfing / 233-Ocean / \n", "66-Bollywood / \n", "0-Games / 2-Video game / 310-The Sims / 328-Simulation video game / 1092-Game Boy Advance / \n", "1-Vehicle / 50-Aircraft / 92-Radio-controlled model / 149-Model aircraft / 160-Radio-controlled aircraft / 266-Helicopter / 392-Radio-controlled helicopter / \n", "22-Nature / 155-Camera / 275-Diving / 278-GoPro / 288-Underwater / 407-Scuba diving / 716-Coral / \n", "85-Combat / 117-Boxing / 399-Kickboxing / \n", "6-Animation / 183-Manga / \n", "9-Music video / \n", "66-Bollywood / \n", "3714-Chalk / \n", "0-Games / 2-Video game / 240-Arcade game / \n", "0-Games / 2-Video game / 4674-Ape Escape / \n", "5-Dance / 16-Performance art / \n", "1-Vehicle / 11-Motorsport / 17-Racing / 27-Motorcycle / 105-Motorcycling / 172-Motocross / 608-Wheelie / 1520-Minibike / \n", "117-Boxing / \n", "355-Comedian / \n", "107-Musical keyboard / 219-Accordion / \n", "770-Banana / \n", "0-Games / 2-Video game / 33-Weapon / 736-ARMA / \n", "25-Toy / 75-Wrestling / 435-Action figure / 1441-Mattel / \n", "3-Concert / 7-Musician / 13-Musical ensemble / 30-Drummer / 98-Festival / \n", "5-Dance / 253-Carnival / \n", "1-Vehicle / 4-Car / 234-Ford / \n", "5-Dance / 49-School / 191-Cheerleading / 306-High school / \n", "0-Games / 1644-ArcheAge / \n", "12-Food / 215-Eating / 323-Toddler / \n", "15-Cartoon / \n", "3-Concert / 3208-Lowlands / \n", "45-Cosmetics / 218-Eyelash / \n", "8-Football / 79-American football / \n", "3-Concert / \n", "14-Guitar / 24-String instrument / 100-Electric guitar / \n", "0-Games / 271-Tank / 1167-Tanki Online / \n", "0-Games / 8-Football / \n", "0-Games / 2-Video game / 34-Action-adventure game / 734-DayZ / 1612-Multi Theft Auto / \n", "0-Games / 2-Video game / 90-Sports game / 453-Pro Evolution Soccer / 3526-Pro Evolution Soccer 2009 / \n", "1-Vehicle / 62-Train / 64-Transport / 103-Rail transport / 119-Locomotive / 143-Railroad car / 161-Train station / \n", "10-Animal / 91-Fish / 814-Pond / 1258-Goldfish / 1568-Koi / \n", "1-Vehicle / 4-Car / 11-Motorsport / 17-Racing / 41-Sports car / 231-Supercar / 901-Mitsubishi / \n", "0-Games / 75-Wrestling / \n", "0-Games / 8-Football / \n", "28-Fashion / 56-Hair / 106-Hairstyle / \n", "15-Cartoon / 123-Naruto / 285-Sasuke Uchiha / 1180-Itachi Uchiha / \n", "0-Games / 1-Vehicle / 2-Video game / 4-Car / 17-Racing / 19-PC game / 34-Action-adventure game / 41-Sports car / 111-PlayStation 3 / 118-Grand Theft Auto V / \n", "38-Orchestra / \n", "12-Food / 2706-Starbucks / \n", "1-Vehicle / 4-Car / 40-Road / 94-Dashcam / \n", "1-Vehicle / 4-Car / 11-Motorsport / 207-Exhaust system / 515-Dodge / \n", "1-Vehicle / 4-Car / 705-Forza Motorsport / 1530-Forza Motorsport 4 / \n", "0-Games / 8-Football / 21-Stadium / 54-Highlight film / 65-Kick / 73-Ball / 76-Ball / 77-Arena / 81-Athlete / \n", "0-Games / 19-PC game / 792-DarkOrbit / \n", "79-American football / \n", "28-Fashion / 469-Bag / 676-Handbag / 681-Leather / \n", "0-Games / \n", "46-Choir / \n", "9-Music video / \n", "372-Playing card / 746-Card manipulation / \n", "10-Animal / 48-Pet / 166-Cat / 346-Kitten / 2154-Gorilla / 2924-Persian cat / \n", "0-Games / 2-Video game / 34-Action-adventure game / 118-Grand Theft Auto V / \n", "0-Games / 2-Video game / 237-Final Fantasy / 432-Kingdom Hearts / 665-Kingdom Hearts / 941-Kingdom Hearts II / 1658-Dissidia Final Fantasy / 1908-Kingdom Hearts Birth by Sleep / 2127-Final Fantasy / 3299-Kingdom Hearts: Chain of Memories / \n", "0-Games / 60-Basketball / 77-Arena / 81-Athlete / 168-Basketball moves / 255-Slam dunk / \n", "117-Boxing / \n", "2478-Red hair / \n", "153-Dress / \n", "10-Animal / 132-Bird / 242-Paper / 370-Origami / \n", "9-Music video / 78-Wedding / \n", "359-Harry Potter (Literary Series) / \n", "6-Animation / 396-One Piece / \n", "0-Games / 351-Jumping / 2401-Wolfenstein / \n", "1-Vehicle / 50-Aircraft / 92-Radio-controlled model / 149-Model aircraft / 160-Radio-controlled aircraft / 266-Helicopter / 392-Radio-controlled helicopter / \n", "3-Concert / 7-Musician / 13-Musical ensemble / 46-Choir / 466-Gift / \n", "10-Animal / 18-Outdoor recreation / 22-Nature / 42-Fishing / 91-Fish / 225-Lake / 268-Recreational fishing / 1306-Carp fishing / \n", "0-Games / 2-Video game / 19-PC game / 33-Weapon / 43-Call of Duty / 96-Soldier / 194-Call of Duty: Modern Warfare 2 / 216-Call of Duty: Modern Warfare 3 / \n", "12-Food / 226-Dessert / 232-Cake / \n", "51-Strategy video game / 373-Clash of Clans / \n", "3-Concert / 7-Musician / 13-Musical ensemble / 14-Guitar / 24-String instrument / 30-Drummer / 39-Drums / \n", "60-Basketball / \n", "0-Games / 2-Video game / 19-PC game / 33-Weapon / \n", "0-Games / 2-Video game / 55-Video game console / 292-Handheld game console / 584-Super Mario Bros. / \n", "885-Bell / \n", "30-Drummer / 39-Drums / 44-Drums / 128-Cymbal / \n", "725-Human / \n", "2488-Fitness Center / \n", "3-Concert / \n", "121-Photography / \n", "3-Concert / \n", "0-Games / 2-Video game / 116-Pokémon / 4461-Pokémon Stadium / \n", "0-Games / 2-Video game / 685-Luigi / \n", "170-Home improvement / 208-Wood / 1700-Lumber / \n", "1-Vehicle / 11-Motorsport / 27-Motorcycle / 105-Motorcycling / 172-Motocross / \n", "1-Vehicle / 4-Car / 84-Snow / 198-Off-road vehicle / 360-Jeep / 3828-Amphibious ATV / \n", "0-Games / 1-Vehicle / 2-Video game / 4-Car / 34-Action-adventure game / 41-Sports car / 111-PlayStation 3 / 297-Grand Theft Auto IV / 635-Achievement / \n", "5-Dance / 10-Animal / 80-Horse / 356-Star / \n", "144-Amusement park / 321-Roller coaster / \n", "12-Food / 86-Plant / 137-Tree / 348-Fruit / 1319-Apple / \n", "259-Fire / 995-Attack on Titan / \n", "0-Games / 8-Football / 21-Stadium / 54-Highlight film / \n", "12-Food / 2032-Pepsi / \n", "6-Animation / 15-Cartoon / 97-Drawing / \n", "0-Games / 35-Minecraft / \n", "14-Guitar / 100-Electric guitar / 1721-Tetris / \n", "5-Dance / \n", "0-Games / 256-Tennis / 544-Table tennis / \n", "6-Animation / 10-Animal / 15-Cartoon / 538-Illustration / 624-Number / 3105-Numerical digit / \n", "5-Dance / 13-Musical ensemble / 78-Wedding / \n", "0-Games / 2-Video game / 43-Call of Duty / \n", "45-Cosmetics / 56-Hair / 798-Shampoo / 919-Hair conditioner / \n", "121-Photography / 2107-Grey / \n", "23-Mobile phone / 29-Smartphone / 72-iPhone / 272-iPod / 331-iPod touch / \n", "67-Cycling / 69-Bicycle / \n", "10-Animal / 22-Nature / 366-Dragon / 588-Reptile / 763-Snake / 1444-Lizard / \n", "18-Outdoor recreation / 283-Rock / 347-Climbing / \n", "3-Concert / 5-Dance / 68-Lighting / \n", "66-Bollywood / \n", "426-Resident Evil / 1068-Resident Evil 4 / \n", "10-Animal / 18-Outdoor recreation / 80-Horse / 247-Stallion / 289-Dressage / 351-Jumping / 404-Livestock / 438-Mare / \n", "20-Trailer / \n", "0-Games / 2-Video game / 34-Action-adventure game / 84-Snow / 118-Grand Theft Auto V / \n", "5-Dance / 9-Music video / 28-Fashion / 153-Dress / \n", "6-Animation / 15-Cartoon / 880-Trousers / 4615-Bakura / \n", "8-Football / 73-Ball / 76-Ball / 333-Nike; Inc. / \n", "3-Concert / \n", "3-Concert / \n", "1416-Cliff / \n", "9-Music video / \n", "47-Personal computer / 131-Computer / 228-Laptop / \n", "3-Concert / 13-Musical ensemble / 30-Drummer / \n", "51-Strategy video game / \n", "14-Guitar / 63-Acoustic guitar / \n", "0-Games / 2-Video game / 19-PC game / 33-Weapon / 35-Minecraft / 165-League of Legends / 506-Garry's Mod / \n", "0-Games / 88-Machine / 240-Arcade game / 535-Slot machine / 4247-Bullion / \n", "14-Guitar / 100-Electric guitar / \n", "9-Music video / \n", "455-Unidentified flying object / \n", "1-Vehicle / 11-Motorsport / 17-Racing / 27-Motorcycle / 40-Road / 172-Motocross / 669-Extreme sport / 1815-Pit bike / \n", "12-Food / 26-Cooking / 32-Recipe / 236-Baking / 485-Dough / 578-Cookie / 600-Flour / 841-Butter / 1081-Lemon / \n", "20-Trailer / \n", "5-Dance / \n", "10-Animal / 18-Outdoor recreation / 42-Fishing / 91-Fish / 510-Fishing lure / 3224-European perch / \n", "1-Vehicle / 4-Car / 406-Chevrolet / 509-Classic car / 1418-Model car / \n", "1-Vehicle / 4-Car / 207-Exhaust system / 680-Subaru / 947-Muffler / \n", "3-Concert / \n", "23-Mobile phone / 29-Smartphone / 37-Gadget / 1097-Touchscreen / 4128-Nokia N97 / \n", "12-Food / 26-Cooking / 32-Recipe / 236-Baking / 442-Bread / \n", "86-Plant / 136-Gardening / \n", "641-Mask / 723-Pig / \n", "227-Water / \n", "1-Vehicle / 27-Motorcycle / 751-Desert / \n", "56-Hair / 106-Hairstyle / 340-Afro-textured hair / 405-Braid / 857-Afro / 909-Updo / 954-Dreadlocks / 1812-Hair twists / \n", "9-Music video / \n", "6-Animation / 9-Music video / \n", "0-Games / 1-Vehicle / 2-Video game / 4-Car / 11-Motorsport / 17-Racing / 41-Sports car / 57-Race track / 135-Drifting / 2518-Dirt 3 / \n", "3-Concert / 7-Musician / 14-Guitar / 24-String instrument / 743-Edward Cullen / \n", "0-Games / 2-Video game / 19-PC game / 192-Battlefield / 383-Battlefield 4 / \n", "14-Guitar / 24-String instrument / 63-Acoustic guitar / 1313-Strum / \n", "20-Trailer / \n", "12-Food / \n", "9-Music video / \n", "0-Games / 2-Video game / 506-Garry's Mod / \n", "6-Animation / 15-Cartoon / 851-Digimon / \n", "3-Concert / \n", "18-Outdoor recreation / 42-Fishing / 84-Snow / 120-Winter / 262-Ice / \n", "33-Weapon / 217-Hunting / \n", "898-Pilates / \n", "1-Vehicle / 4-Car / 9-Music video / \n", "1-Vehicle / 42-Fishing / 87-Boat / 92-Radio-controlled model / 202-Radio-controlled car / 308-Ship / 1490-Radio-controlled boat / 3163-Tugboat / \n", "616-The Bible / \n", "1-Vehicle / 4-Car / 70-Driving / 94-Dashcam / 277-Sport utility vehicle / 983-Lexus / \n", "31-Disc jockey / 322-Mixtape / \n", "102-Building / 113-House / 339-Architecture / \n", "10-Animal / 48-Pet / 71-Dog / 166-Cat / 182-Puppy / \n", "3-Concert / 7-Musician / 30-Drummer / 39-Drums / \n", "47-Personal computer / \n", "18-Outdoor recreation / 83-Skateboarding / 108-Skateboard / 1031-Fisheye lens / \n", "5-Dance / \n", "541-Black-and-white / \n", "783-Zee TV / \n", "31-Disc jockey / \n", "23-Mobile phone / 29-Smartphone / 72-iPhone / \n", "20-Trailer / \n", "3-Concert / 14-Guitar / \n", "1054-Vacuum cleaner / \n", "30-Drummer / 39-Drums / 44-Drums / 1825-Cajón / \n", "85-Combat / 117-Boxing / \n", "0-Games / 2-Video game / 577-Penguin / \n", "467-Chipmunk / \n", "28-Fashion / 67-Cycling / 69-Bicycle / \n", "0-Games / 2-Video game / 43-Call of Duty / 111-PlayStation 3 / 194-Call of Duty: Modern Warfare 2 / 216-Call of Duty: Modern Warfare 3 / \n", "222-Surfing / 2415-Big wave surfing / \n", "2-Video game / 1188-Uncharted / \n", "0-Games / 311-Street Fighter / 461-Street Fighter IV / 496-Super Street Fighter IV / 1331-Ken Masters / 3936-Gouken / \n", "3-Concert / 13-Musical ensemble / \n", "0-Games / 2-Video game / 51-Strategy video game / \n", "425-City / \n", "31-Disc jockey / 68-Lighting / 263-Light / 988-Ibiza / \n", "3-Concert / 7-Musician / 13-Musical ensemble / 24-String instrument / 38-Orchestra / 150-Violin / 450-Viola / 2348-Chrono Trigger / \n", "28-Fashion / 56-Hair / 106-Hairstyle / \n", "36-Piano / 53-Keyboard / 104-Pianist / 107-Musical keyboard / \n", "0-Games / 2-Video game / 19-PC game / 33-Weapon / 43-Call of Duty / 96-Soldier / 194-Call of Duty: Modern Warfare 2 / \n", "31-Disc jockey / 754-Gramophone record / 861-Mixing console / 3458-DMC World DJ Championships / \n", "12-Food / \n", "6-Animation / 15-Cartoon / 4183-Tamagotchi / \n", "31-Disc jockey / \n", "5-Dance / 115-Ballet / 430-Ballroom dance / 678-Latin dance / \n", "22-Nature / 36-Piano / 376-Flute / \n", "12-Food / 770-Banana / \n", "1-Vehicle / 4-Car / 531-Hatchback / 3197-Citroën C4 / \n", "8-Football / \n", "22-Nature / 84-Snow / 99-Christmas / 120-Winter / 729-Christmas decoration / \n", "0-Games / 25-Toy / 140-LEGO / 319-Robot / \n", "30-Drummer / 39-Drums / 44-Drums / 128-Cymbal / 146-Snare drum / \n", "18-Outdoor recreation / 42-Fishing / 1236-Fishing reel / \n", "3-Concert / 7-Musician / \n", "1500-Cherry blossom / 2640-Petal / \n", "1-Vehicle / 4-Car / 11-Motorsport / 17-Racing / 40-Road / 41-Sports car / 1207-Samsung Galaxy S5 / \n", "25-Toy / 140-LEGO / \n", "0-Games / \n", "1314-Garage / 2196-Boiler / \n", "1-Vehicle / 286-Bus / 2226-Bus driver / 2850-Chauffeur / \n", "1465-FOX / \n", "0-Games / 25-Toy / 323-Toddler / 3303-Table football / \n", "6-Animation / 183-Manga / 3588-Berserk / \n", "0-Games / 8-Football / \n", "0-Games / 51-Strategy video game / \n", "0-Games / 2-Video game / 10-Animal / 23-Mobile phone / 178-Samsung Galaxy / 899-Samsung Galaxy S III / \n", "764-The King of Fighters / 961-Sega Genesis / \n", "6-Animation / 183-Manga / \n", "5-Dance / \n", "0-Games / 2-Video game / 837-Roblox / \n", "0-Games / 2-Video game / 111-PlayStation 3 / 192-Battlefield / 374-Battlefield 3 / \n", "1-Vehicle / 62-Train / 103-Rail transport / 119-Locomotive / 152-Track / \n", "6-Animation / 15-Cartoon / 61-Art / 97-Drawing / \n", "3-Concert / 7-Musician / 16-Performance art / \n", "739-Guild Wars / \n", "0-Games / 8-Football / 21-Stadium / 65-Kick / 73-Ball / \n", "3-Concert / 7-Musician / 13-Musical ensemble / 16-Performance art / 38-Orchestra / 46-Choir / \n", "303-Paint / 696-T-shirt / 793-Shirt / \n", "75-Wrestling / \n", "45-Cosmetics / 173-Eye shadow / 193-Eye liner / 195-Mascara / 196-Eye / 218-Eyelash / 235-Lipstick / 291-Rouge / \n", "3-Concert / \n", "402-Jewellery / 859-Beadwork / 907-Necklace / 1300-Earrings / 3428-Costume jewelry / \n", "10-Animal / 80-Horse / 4694-Cruelty to animals / \n", "1-Vehicle / 50-Aircraft / 129-Aviation / \n", "0-Games / 6-Animation / 15-Cartoon / 290-Pony / 343-My Little Pony / \n", "97-Drawing / 479-Graffiti / \n", "110-Album / \n", "3-Concert / 16-Performance art / \n", "30-Drummer / 39-Drums / 44-Drums / 128-Cymbal / 146-Snare drum / 2073-Sabian / \n", "1-Vehicle / 4-Car / 509-Classic car / 3350-Rolls-Royce Limited / 4501-The Adventures of Tintin / \n", "1-Vehicle / 62-Train / 64-Transport / 103-Rail transport / 161-Train station / 224-Rapid transit / \n", "0-Games / 1-Vehicle / 2-Video game / 4-Car / 34-Action-adventure game / 118-Grand Theft Auto V / \n", "1157-Rope / \n", "0-Games / 8-Football / 21-Stadium / \n", "9-Music video / 10-Animal / 22-Nature / 280-Wildlife / 1603-Fox / \n", "3027-Olive oil / \n", "9-Music video / \n", "9-Music video / \n", "6-Animation / \n", "20-Trailer / \n", "359-Harry Potter (Literary Series) / 1436-Draco Malfoy / \n", "614-Hiking / 1203-Backpacking / \n", "0-Games / 60-Basketball / 77-Arena / 168-Basketball moves / 255-Slam dunk / \n", "6-Animation / 10-Animal / 25-Toy / 140-LEGO / 550-Dinosaur / 940-Clay animation / 1126-Tyrannosaurus / \n", "3-Concert / \n", "1-Vehicle / 4-Car / 11-Motorsport / 27-Motorcycle / 70-Driving / 349-Chevrolet / 406-Chevrolet / \n", "293-Slide show / \n", "14-Guitar / \n", "36-Piano / 53-Keyboard / 104-Pianist / 318-Church / 397-Organ / 646-Pipe organ / \n", "1-Vehicle / 4-Car / 449-Honda / \n", "1-Vehicle / 11-Motorsport / 17-Racing / 59-Winter sport / 1829-Sledding / 4709-Luge / \n", "0-Games / 199-Wii / 256-Tennis / \n", "66-Bollywood / \n", "0-Games / 463-Cricket / \n", "121-Photography / \n", "25-Toy / 126-The Walt Disney Company / 749-Figurine / \n", "7-Musician / 13-Musical ensemble / 24-String instrument / 38-Orchestra / 1747-Erhu / \n", "0-Games / 2-Video game / 35-Minecraft / 174-Farm / 1108-Mushroom / \n", "36-Piano / \n", "916-Spa / 1271-Beauty salon / \n", "3-Concert / 7-Musician / 13-Musical ensemble / 16-Performance art / 30-Drummer / 39-Drums / 44-Drums / \n", "0-Games / 8-Football / 21-Stadium / 65-Kick / 73-Ball / 76-Ball / 81-Athlete / \n", "3-Concert / 21-Stadium / 366-Dragon / \n", "0-Games / 2-Video game / 6-Animation / 20-Trailer / 43-Call of Duty / \n", "23-Mobile phone / 29-Smartphone / 37-Gadget / 178-Samsung Galaxy / 3184-Moto X / \n", "6-Animation / 25-Toy / 35-Minecraft / 140-LEGO / 1302-The Lego Group / 3899-Lego Minecraft / \n", "1-Vehicle / 4-Car / 40-Road / 64-Transport / 74-Truck / 332-Mercedes-Benz / 3886-Mercedes-Benz Actros / \n", "6-Animation / 4708-Visual novel / \n", "56-Hair / 106-Hairstyle / 583-Wig / 607-Lace / \n", "8-Football / \n", "28-Fashion / 45-Cosmetics / 238-Nail / 300-Nail art / 304-Nail polish / 307-Hand / 324-Finger / 338-Manicure / \n", "1-Vehicle / 4-Car / 11-Motorsport / 41-Sports car / 70-Driving / \n", "1721-Tetris / \n", "0-Games / 2-Video game / 35-Minecraft / \n", "3-Concert / \n", "1-Vehicle / 4-Car / 2616-Chevrolet S-10 / \n", "372-Playing card / 746-Card manipulation / \n", "28-Fashion / 956-Glasses / \n", "0-Games / 8-Football / 90-Sports game / 180-FIFA 15 / 480-FIFA 13 / \n", "1-Vehicle / 4-Car / 11-Motorsport / 17-Racing / 40-Road / 57-Race track / 169-Rallying / \n", "9-Music video / 322-Mixtape / \n", "613-Bar / \n", "59-Winter sport / 112-Ice skating / 142-Hockey / 162-Medicine / 179-Kitchen / \n", "1143-Marimba / 1734-Xylophone / \n", "0-Games / 34-Action-adventure game / 330-Money / \n", "5-Dance / 31-Disc jockey / 133-Nightclub / \n", "191-Cheerleading / \n", "975-Dune / \n", "9-Music video / \n", "9-Music video / \n", "234-Ford / 4005-Mercury / \n", "6-Animation / 15-Cartoon / 243-Sonic the Hedgehog / 251-Sonic the Hedgehog / 1475-Knuckles the Echidna / \n", "12-Food / 261-Tool / 673-Saw / 1073-Chainsaw / 1346-Wood carving / \n", "1-Vehicle / 4-Car / 11-Motorsport / 74-Truck / 375-Pickup truck / \n", "9-Music video / \n", "1-Vehicle / 4-Car / 11-Motorsport / 234-Ford / 473-Ford Mustang / 1877-Shelby Mustang / 3755-AC Cobra / \n", "0-Games / 2-Video game / 19-PC game / 211-RuneScape / 330-Money / \n", "6-Animation / 9-Music video / \n", "1-Vehicle / 50-Aircraft / 92-Radio-controlled model / 129-Aviation / 149-Model aircraft / 160-Radio-controlled aircraft / 266-Helicopter / 393-Unmanned aerial vehicle / \n", "110-Album / \n", "1-Vehicle / 10-Animal / 18-Outdoor recreation / 42-Fishing / 87-Boat / 91-Fish / 233-Ocean / \n", "7-Musician / 14-Guitar / \n", "227-Water / 259-Fire / \n", "1-Vehicle / 4-Car / 41-Sports car / 332-Mercedes-Benz / \n", "0-Games / 2-Video game / 35-Minecraft / 55-Video game console / 111-PlayStation 3 / 394-Map / \n", "5-Dance / 16-Performance art / 95-Talent show / \n", "85-Combat / 117-Boxing / 204-Gym / 399-Kickboxing / 1239-K-1 / \n", "6-Animation / 25-Toy / 701-Beyblade / \n", "10-Animal / 48-Pet / 71-Dog / 182-Puppy / 1845-Pug / \n", "85-Combat / 117-Boxing / \n", "0-Games / 2-Video game / 229-Halo / 470-Halo 3 / 644-Halo: Reach / \n", "3-Concert / 7-Musician / 13-Musical ensemble / \n", "12-Food / 170-Home improvement / 179-Kitchen / 276-Room / 742-Countertop / \n", "46-Choir / 318-Church / \n", "25-Toy / 4652-Weekly Shōnen Jump / \n", "10-Animal / 48-Pet / 71-Dog / \n", "3-Concert / 7-Musician / 14-Guitar / 24-String instrument / 63-Acoustic guitar / \n", "0-Games / 2-Video game / 19-PC game / \n", "28-Fashion / 121-Photography / 155-Camera / 566-Digital camera / 633-Photograph / 721-Digital SLR / \n", "273-News program / \n", "1459-Magazine / \n", "9-Music video / \n", "0-Games / 8-Football / 21-Stadium / 65-Kick / \n", "36-Piano / 53-Keyboard / \n", "33-Weapon / 261-Tool / 315-Knife / \n", "137-Tree / 640-Bottle / \n", "22-Nature / \n", "0-Games / 1-Vehicle / 4-Car / 17-Racing / 41-Sports car / 70-Driving / 245-BMW / \n", "1-Vehicle / 4-Car / 11-Motorsport / 17-Racing / 447-Sand / 751-Desert / 975-Dune / \n", "0-Games / 1732-Oni / 4197-Ao Oni / \n", "0-Games / 19-PC game / 35-Minecraft / \n", "8-Football / 21-Stadium / \n", "5-Dance / 95-Talent show / \n", "3-Concert / 7-Musician / 13-Musical ensemble / 38-Orchestra / 46-Choir / 870-Chapel / \n", "3-Concert / 16-Performance art / \n", "5-Dance / \n", "3-Concert / \n", "31-Disc jockey / \n", "5-Dance / \n", "162-Medicine / \n", "0-Games / 2-Video game / 518-Mario Kart / 3964-Diddy Kong / \n", "0-Games / 35-Minecraft / \n", "60-Basketball / 76-Ball / 168-Basketball moves / 1399-Streetball / \n", "1-Vehicle / 50-Aircraft / 82-Airplane / 129-Aviation / 139-Airport / 151-Landing / 157-Jet aircraft / 175-Airliner / 187-Takeoff / 188-Airline / 254-Runway / 472-Boeing 737 / \n", "0-Games / 2-Video game / 847-Angry Birds / 4064-Angry Birds Space / \n", "0-Games / 2-Video game / 335-Nintendo 3DS / \n", "0-Games / 2-Video game / 19-PC game / 35-Minecraft / \n", "1-Vehicle / 4-Car / 11-Motorsport / 40-Road / 559-Night / 1761-Volkswagen Jetta / \n", "144-Amusement park / 321-Roller coaster / 334-Amusement ride / 366-Dragon / 3794-Legoland / \n", "1-Vehicle / 18-Outdoor recreation / 67-Cycling / 69-Bicycle / 197-Trail / 220-Mountain bike / 284-Mountain biking / \n", "213-Forest / \n", "0-Games / 2-Video game / 19-PC game / 33-Weapon / 96-Soldier / 111-PlayStation 3 / 2074-Killzone / \n", "14-Guitar / 100-Electric guitar / \n", "3-Concert / 7-Musician / 16-Performance art / \n", "117-Boxing / \n", "10-Animal / 91-Fish / 215-Eating / 258-Aquarium / 716-Coral / 1051-Reef aquarium / \n", "1026-Foreign Exchange Market / \n", "3-Concert / 7-Musician / 16-Performance art / 350-Circus / 2358-Contortion / \n", "205-Newscaster / 273-News program / \n", "3-Concert / \n", "14-Guitar / 36-Piano / 53-Keyboard / 107-Musical keyboard / 265-Electronic keyboard / \n", "0-Games / 159-World of Warcraft / 1232-Perfect World / \n", "1-Vehicle / 68-Lighting / 263-Light / \n", "164-Beach / 1481-Sunglasses / \n", "0-Games / 2-Video game / 8-Football / 21-Stadium / 90-Sports game / 180-FIFA 15 / 453-Pro Evolution Soccer / 1719-Pro Evolution Soccer 2015 / \n", "567-Compact disc / 2447-Guzheng / \n", "20-Trailer / \n", "28-Fashion / 45-Cosmetics / 676-Handbag / \n", "85-Combat / \n", "0-Games / 2-Video game / 34-Action-adventure game / 359-Harry Potter (Literary Series) / \n", "6-Animation / 15-Cartoon / 123-Naruto / \n", "205-Newscaster / 752-Pasta / \n", "9-Music video / \n", "93-Comedy / \n", "5-Dance / 694-Tai chi / \n", "0-Games / 211-RuneScape / \n", "5-Dance / 8-Football / \n", "3-Concert / \n", "1-Vehicle / 4-Car / \n", "3-Concert / 7-Musician / 14-Guitar / 24-String instrument / \n", "7-Musician / 13-Musical ensemble / 21-Stadium / 49-School / 54-Highlight film / 203-Marching band / 209-University / \n", "14-Guitar / 63-Acoustic guitar / \n", "20-Trailer / 78-Wedding / 196-Eye / \n", "3-Concert / 68-Lighting / 98-Festival / \n", "93-Comedy / 125-Television / 341-Sketch comedy / \n", "0-Games / 6-Animation / 624-Number / 971-Alphabet / \n", "483-Costume / 3153-Freddy Krueger / \n", "9-Music video / \n", "78-Wedding / 232-Cake / \n", "6-Animation / \n", "8-Football / \n", "0-Games / 2-Video game / 43-Call of Duty / 422-Call of Duty: Ghosts / \n", "66-Bollywood / \n", "3-Concert / 16-Performance art / \n", "7-Musician / 9-Music video / 14-Guitar / \n", "444-Family / \n", "0-Games / 2-Video game / 6-Animation / 816-Gundam / 1564-Mobile Suit Gundam: Extreme Vs. / \n", "6-Animation / 15-Cartoon / 61-Art / 97-Drawing / 123-Naruto / \n", "1-Vehicle / 18-Outdoor recreation / 42-Fishing / 87-Boat / 208-Wood / 481-Kayak / 753-Sailboat / 917-Sail / 1309-Sailing ship / \n", "1-Vehicle / 50-Aircraft / 82-Airplane / 92-Radio-controlled model / 149-Model aircraft / 160-Radio-controlled aircraft / 1450-Model building / 1497-Wool / 2552-Dyeing / \n", "5-Dance / 115-Ballet / \n", "12-Food / 26-Cooking / 32-Recipe / 52-Dish / 58-Cuisine / 109-Cooking show / 184-Vegetable / \n", "28-Fashion / 56-Hair / 106-Hairstyle / \n", "325-Vampire / \n", "0-Games / 2-Video game / 34-Action-adventure game / 362-PlayStation Portable / 1125-God of War / 1432-Kratos / 1743-God of War / \n", "0-Games / 1-Vehicle / 2-Video game / 11-Motorsport / 55-Video game console / 199-Wii / 456-Go-kart / 518-Mario Kart / \n", "12-Food / 354-Egg / 1597-Sausage / \n", "3-Concert / \n", "5-Dance / 16-Performance art / \n", "9-Music video / 68-Lighting / 263-Light / 1016-Glove / 1375-Laser lighting display / \n", "1-Vehicle / 11-Motorsport / 59-Winter sport / 177-Skiing / 278-GoPro / 305-Ski / \n", "10-Animal / 48-Pet / 71-Dog / 182-Puppy / 1377-Golden Retriever / \n", "3-Concert / 1019-Christ / \n", "25-Toy / 140-LEGO / \n", "12-Food / 26-Cooking / 32-Recipe / 109-Cooking show / 179-Kitchen / 2588-Mayonnaise / \n", "1-Vehicle / 4-Car / 11-Motorsport / 332-Mercedes-Benz / 650-Convertible / 891-Mercedes-AMG / 3995-Mercedes-Benz W124 / \n", "0-Games / 54-Highlight film / 60-Basketball / 168-Basketball moves / 255-Slam dunk / \n", "3-Concert / \n", "489-Board game / 799-Chess / \n", "5-Dance / 6-Animation / 1527-Pretty Cure / \n", "337-Advertising / \n", "1-Vehicle / 4-Car / 11-Motorsport / 41-Sports car / 70-Driving / 155-Camera / 650-Convertible / 1521-Jaguar Cars / 4386-Jaguar F-Type / \n", "116-Pokémon / 186-Pokémon / \n", "0-Games / 8-Football / 54-Highlight film / 79-American football / 813-Running back / \n", "1-Vehicle / 74-Truck / 1070-Semi-trailer truck / 2920-Food grain / \n", "1-Vehicle / 4-Car / 41-Sports car / 231-Supercar / 332-Mercedes-Benz / \n", "9-Music video / 375-Pickup truck / \n", "4476-Shipyard / \n", "88-Machine / 283-Rock / 666-Concrete / 2679-Pavement / \n", "3-Concert / 7-Musician / 13-Musical ensemble / 14-Guitar / 24-String instrument / 38-Orchestra / 1231-Mandolin / \n", "1-Vehicle / 25-Toy / 62-Train / 1279-Plarail / \n", "5-Dance / 49-School / \n", "1-Vehicle / 18-Outdoor recreation / 22-Nature / 87-Boat / 147-River / 614-Hiking / \n", "1-Vehicle / 4-Car / 114-Engine / \n", "3-Concert / 30-Drummer / \n", "122-Weight training / 214-Muscle / \n", "1-Vehicle / 2-Video game / 4-Car / 11-Motorsport / 17-Racing / 41-Sports car / 57-Race track / 352-Need for Speed / 2772-Shift 2: Unleashed / \n", "512-Printing / 525-Printer / 1424-Coupon / \n", "1-Vehicle / 50-Aircraft / 82-Airplane / 129-Aviation / 157-Jet aircraft / 175-Airliner / 388-Fighter aircraft / 758-Air force / 2326-Royal Air Force / \n", "6-Animation / \n", "0-Games / 2-Video game / 55-Video game console / 2431-Metal Slug / \n", "3-Concert / \n", "88-Machine / 400-Woodturning / \n", "3-Concert / 68-Lighting / 133-Nightclub / 988-Ibiza / \n", "0-Games / 19-PC game / 51-Strategy video game / 165-League of Legends / \n", "9-Music video / \n", "28-Fashion / 45-Cosmetics / 56-Hair / 78-Wedding / 106-Hairstyle / 405-Braid / 909-Updo / \n", "10-Animal / 18-Outdoor recreation / 80-Horse / 247-Stallion / 289-Dressage / 351-Jumping / 438-Mare / \n", "0-Games / 2-Video game / 1981-Commodore 64 / 3918-ZX Spectrum / \n", "10-Animal / 91-Fish / 275-Diving / 288-Underwater / 381-Underwater diving / 407-Scuba diving / 468-Biology / \n", "0-Games / 2-Video game / 20-Trailer / 872-Diablo / 2054-Diablo II / \n", "1-Vehicle / 4-Car / 84-Snow / 120-Winter / 344-Coupé / \n", "0-Games / 12-Food / 1223-Pastry / \n", "14-Guitar / \n", "25-Toy / 626-Puzzle / 631-Cube / 740-Rubik's Cube / \n", "5-Dance / 6-Animation / 679-Mickey Mouse / 1327-Minnie Mouse / 2793-Disguise / \n", "1-Vehicle / 4-Car / 41-Sports car / 231-Supercar / \n", "25-Toy / 140-LEGO / \n", "0-Games / 8-Football / 21-Stadium / 65-Kick / 73-Ball / 76-Ball / 77-Arena / 81-Athlete / \n", "0-Games / 159-World of Warcraft / 171-Warcraft / 1356-Paladin / \n", "9-Music video / \n", "3-Concert / 5-Dance / \n", "6-Animation / 1887-Tricycle / \n", "9-Music video / \n", "0-Games / 211-RuneScape / 1178-Quest / \n", "0-Games / 2-Video game / 19-PC game / 734-DayZ / 736-ARMA / 1034-ARMA 2 / 1216-DayZ / \n", "1183-Karaoke box / \n", "1-Vehicle / 62-Train / 64-Transport / 103-Rail transport / 119-Locomotive / 161-Train station / \n", "61-Art / 232-Cake / 242-Paper / 370-Origami / 795-Cupcake / \n", "1-Vehicle / 4-Car / 56-Hair / 70-Driving / 106-Hairstyle / \n", "3-Concert / 99-Christmas / \n", "6-Animation / 10-Animal / 15-Cartoon / \n", "3-Concert / 7-Musician / 14-Guitar / 24-String instrument / \n", "5-Dance / 16-Performance art / \n", "3-Concert / 7-Musician / 13-Musical ensemble / \n", "16-Performance art / 95-Talent show / \n", "0-Games / 2-Video game / 55-Video game console / \n", "25-Toy / 113-House / 140-LEGO / 1079-Lego City / 2055-Mansion / \n", "83-Skateboarding / 108-Skateboard / 555-Longboard / 667-Longboarding / \n", "7-Musician / 14-Guitar / 24-String instrument / 63-Acoustic guitar / 465-Microphone / 1013-Walking / \n", "6-Animation / 61-Art / 97-Drawing / \n", "3-Concert / 7-Musician / \n", "28-Fashion / 402-Jewellery / 907-Necklace / 3428-Costume jewelry / \n", "45-Cosmetics / 235-Lipstick / \n", "12-Food / 25-Toy / 389-Restaurant / 417-Play-Doh / 439-Barbie / 485-Dough / 578-Cookie / 944-Hamburger / 962-McDonald's / 1185-Plasticine / 1701-French fries / 3102-Cookie Monster / 3822-Mold / \n", "12-Food / 367-Drink / 2992-Energy drink / \n", "0-Games / 2-Video game / 420-The Elder Scrolls V: Skyrim / 441-The Elder Scrolls / \n", "0-Games / 2-Video game / 43-Call of Duty / 127-Xbox 360 / 134-Call of Duty: Black Ops / 192-Battlefield / 194-Call of Duty: Modern Warfare 2 / 216-Call of Duty: Modern Warfare 3 / 229-Halo / 464-First-person Shooter / 466-Gift / 2081-Infinity Ward / \n", "130-Gymnastics / \n", "1-Vehicle / 4-Car / 11-Motorsport / 515-Dodge / 2078-Dodge Challenger / 2530-Chrysler Hemi engine / \n", "0-Games / 19-PC game / 311-Street Fighter / 496-Super Street Fighter IV / \n", "7-Musician / 38-Orchestra / \n", "30-Drummer / 39-Drums / 44-Drums / 128-Cymbal / 146-Snare drum / \n", "3-Concert / 5-Dance / 49-School / 95-Talent show / \n", "14-Guitar / 100-Electric guitar / \n", "3-Concert / 5-Dance / 46-Choir / \n", "45-Cosmetics / 238-Nail / 300-Nail art / 304-Nail polish / 307-Hand / 324-Finger / \n", "5-Dance / 16-Performance art / \n", "0-Games / 8-Football / 73-Ball / \n", "0-Games / 2-Video game / 9-Music video / 477-Transformers / 990-Optimus Prime / 2186-Megatron / \n", "0-Games / 2-Video game / 1637-GameCube / 4347-Rocky Balboa / \n", "9-Music video / \n", "1-Vehicle / 11-Motorsport / 17-Racing / 27-Motorcycle / 172-Motocross / 927-Supermoto / \n", "3-Concert / \n", "18-Outdoor recreation / 42-Fishing / 222-Surfing / 1008-Kite / \n", "142-Hockey / 1610-Field hockey / \n", "233-Ocean / \n", "296-Saxophone / 1006-Tenor saxophone / \n", "66-Bollywood / 1281-Couch / \n", "3-Concert / 7-Musician / 16-Performance art / 39-Drums / 4412-Alcatraz Island / \n", "22-Nature / 458-Gold / \n", "12-Food / 26-Cooking / 32-Recipe / 52-Dish / 58-Cuisine / 109-Cooking show / 597-Soup / \n", "23-Mobile phone / 29-Smartphone / 125-Television / \n", "0-Games / 8-Football / \n", "0-Games / 5-Dance / 257-Counter-Strike / \n", "3-Concert / 5-Dance / 7-Musician / 13-Musical ensemble / 46-Choir / 78-Wedding / 99-Christmas / \n", "10-Animal / 537-Zoo / 970-Monkey / 1948-Ape / \n", "633-Photograph / \n", "25-Toy / 200-Doll / 412-Monster / 503-Monster High / 2476-Maze / \n", "6-Animation / 20-Trailer / \n", "0-Games / 1-Vehicle / 19-PC game / 33-Weapon / 271-Tank / 409-World of Tanks / \n", "0-Games / 2-Video game / 25-Toy / 120-Winter / 140-LEGO / 359-Harry Potter (Literary Series) / \n", "1-Vehicle / 4-Car / 114-Engine / 618-Fiat Automobiles / \n", "0-Games / 8-Football / \n", "81-Athlete / 222-Surfing / 815-Surfboard / \n", "1731-Haunted house / \n", "130-Gymnastics / 1558-Hula hoop / \n", "0-Games / 2-Video game / 6-Animation / \n", "1-Vehicle / 4-Car / 11-Motorsport / 17-Racing / 135-Drifting / \n", "5-Dance / \n", "227-Water / \n", "12-Food / 26-Cooking / 32-Recipe / 52-Dish / 58-Cuisine / 109-Cooking show / 301-Meal / 597-Soup / 2040-Korean food / 3539-Clam / \n", "36-Piano / 53-Keyboard / \n", "2-Video game / 320-Super Smash Bros. / 782-Super Smash Bros. Melee / 2289-Super Mario Bros. 3 / \n", "0-Games / 2-Video game / 6-Animation / 111-PlayStation 3 / 2628-Sly Cooper / \n", "3-Concert / 7-Musician / 9-Music video / \n", "23-Mobile phone / 29-Smartphone / 37-Gadget / 101-Telephone / 1032-BlackBerry / 3732-BlackBerry Bold / \n", "3-Concert / 7-Musician / \n", "3-Concert / 68-Lighting / \n", "12-Food / 26-Cooking / 32-Recipe / 49-School / 52-Dish / 808-Breakfast / 1408-Snack / \n", "467-Chipmunk / \n", "0-Games / 60-Basketball / \n", "14-Guitar / 24-String instrument / 63-Acoustic guitar / \n", "1-Vehicle / 18-Outdoor recreation / 67-Cycling / 69-Bicycle / 220-Mountain bike / 284-Mountain biking / \n", "6-Animation / 10-Animal / 15-Cartoon / \n", "10-Animal / 91-Fish / 258-Aquarium / \n", "1-Vehicle / 11-Motorsport / 17-Racing / 27-Motorcycle / 57-Race track / 172-Motocross / 278-GoPro / 669-Extreme sport / 3548-Yamaha YZ125 / \n", "6-Animation / 15-Cartoon / 61-Art / 97-Drawing / \n", "7-Musician / 13-Musical ensemble / 38-Orchestra / \n", "6-Animation / \n", "0-Games / 8-Football / \n", "12-Food / 1208-Coconut / \n", "1-Vehicle / 4-Car / 18-Outdoor recreation / 25-Toy / 74-Truck / 92-Radio-controlled model / 167-Four-wheel drive / 198-Off-road vehicle / 202-Radio-controlled car / 787-Monster truck / \n", "10-Animal / 636-Rabbit / \n", "9-Music video / 323-Toddler / \n", "10-Animal / 91-Fish / 258-Aquarium / \n", "6-Animation / 61-Art / 97-Drawing / 148-Painting / 1716-Rat / \n", "83-Skateboarding / \n", "8-Football / 76-Ball / \n", "5-Dance / 16-Performance art / \n", "14-Guitar / \n", "28-Fashion / 206-Shoe / \n", "0-Games / 116-Pokémon / 186-Pokémon / \n", "7-Musician / 13-Musical ensemble / 38-Orchestra / \n", "0-Games / 2-Video game / 34-Action-adventure game / 774-Galaxy / 1397-Super Mario Galaxy / \n", "0-Games / 8-Football / 21-Stadium / 54-Highlight film / 65-Kick / 73-Ball / \n", "3-Concert / 7-Musician / \n", "3-Concert / 13-Musical ensemble / 30-Drummer / 295-Chocolate / \n", "122-Weight training / 204-Gym / 214-Muscle / \n", "22-Nature / 86-Plant / 136-Gardening / 279-Garden / \n", "30-Drummer / 39-Drums / 44-Drums / 128-Cymbal / 146-Snare drum / \n", "0-Games / 256-Tennis / \n", "6-Animation / \n", "28-Fashion / 56-Hair / 106-Hairstyle / 405-Braid / 909-Updo / 1292-Ponytail / \n", "162-Medicine / 468-Biology / \n", "12-Food / 26-Cooking / 32-Recipe / 442-Bread / 1567-Cook / \n", "276-Room / 888-Living room / 1739-Fireplace / \n", "121-Photography / 155-Camera / \n", "14-Guitar / \n", "5-Dance / 10-Animal / 80-Horse / 247-Stallion / \n", "3-Concert / 726-Clarinet / \n", "46-Choir / \n", "0-Games / 2-Video game / 34-Action-adventure game / 1588-Prince of Persia / 1653-Prince / 3814-Prince of Persia: Warrior Within / \n", "93-Comedy / \n", "56-Hair / 106-Hairstyle / \n", "0-Games / 2-Video game / 43-Call of Duty / 194-Call of Duty: Modern Warfare 2 / \n", "0-Games / 8-Football / 21-Stadium / 54-Highlight film / 65-Kick / 73-Ball / 76-Ball / 81-Athlete / \n", "9-Music video / 233-Ocean / \n", "8-Football / \n", "5-Dance / \n", "1-Vehicle / 88-Machine / 92-Radio-controlled model / 202-Radio-controlled car / 271-Tank / \n", "3-Concert / 7-Musician / 14-Guitar / 100-Electric guitar / \n", "20-Trailer / 66-Bollywood / \n", "1-Vehicle / 4-Car / 40-Road / 94-Dashcam / 212-Highway / 223-Weather / 490-Rain / 958-Thunderstorm / \n", "3-Concert / 7-Musician / 13-Musical ensemble / 44-Drums / 46-Choir / 1019-Christ / \n", "0-Games / 1-Vehicle / 2-Video game / 286-Bus / \n", "1-Vehicle / 4-Car / 245-BMW / 1285-Parking / 3475-BMW X6 / \n", "0-Games / 54-Highlight film / 60-Basketball / 79-American football / \n", "5-Dance / 6-Animation / 1527-Pretty Cure / \n", "0-Games / 1-Vehicle / 2-Video game / 11-Motorsport / 55-Video game console / 199-Wii / 456-Go-kart / 518-Mario Kart / \n", "0-Games / 5-Dance / 739-Guild Wars / 839-Guild Wars 2 / \n", "359-Harry Potter (Literary Series) / 1436-Draco Malfoy / \n", "12-Food / 26-Cooking / 32-Recipe / 52-Dish / 58-Cuisine / 109-Cooking show / 179-Kitchen / 210-Cookware and bakeware / 274-Meat / \n", "2-Video game / 309-Computer hardware / \n", "9-Music video / \n", "3-Concert / 7-Musician / 14-Guitar / \n", "512-Printing / 525-Printer / 619-Manufacturing / \n", "6-Animation / 25-Toy / 524-Littlest Pet Shop / \n", "5-Dance / 130-Gymnastics / \n", "45-Cosmetics / \n", "28-Fashion / 153-Dress / \n", "12-Food / \n", "23-Mobile phone / 29-Smartphone / 70-Driving / 223-Weather / \n", "0-Games / 2-Video game / 55-Video game console / 251-Sonic the Hedgehog / 320-Super Smash Bros. / \n", "3-Concert / 7-Musician / 13-Musical ensemble / \n", "49-School / \n", "9-Music video / 14-Guitar / \n", "12-Food / 2138-Compost / \n", "0-Games / 2-Video game / 19-PC game / 33-Weapon / \n", "2-Video game / 561-Madden NFL / 2916-Madden NFL 12 / \n", "0-Games / 33-Weapon / 664-CrossFire / \n", "10-Animal / 166-Cat / 397-Organ / 502-Sky / 646-Pipe organ / \n", "0-Games / 6-Animation / \n", "1-Vehicle / 4-Car / 40-Road / 64-Transport / 70-Driving / 74-Truck / 212-Highway / \n", "36-Piano / 446-Sheet music / \n", "0-Games / 8-Football / 21-Stadium / 54-Highlight film / 65-Kick / \n", "3-Concert / 7-Musician / \n", "3-Concert / 13-Musical ensemble / 46-Choir / \n", "3-Concert / 7-Musician / 13-Musical ensemble / \n", "2716-The Hunger Games / \n", "6-Animation / 15-Cartoon / 4671-Little Busters! / \n", "3-Concert / 7-Musician / 30-Drummer / 39-Drums / \n", "0-Games / 2-Video game / 34-Action-adventure game / 127-Xbox 360 / 181-Xbox / 297-Grand Theft Auto IV / \n", "66-Bollywood / \n", "93-Comedy / \n", "0-Games / 8-Football / 54-Highlight film / 79-American football / \n", "1-Vehicle / 11-Motorsport / 17-Racing / 27-Motorcycle / 57-Race track / 105-Motorcycling / \n", "47-Personal computer / 1043-Router / \n", "0-Games / 2-Video game / 55-Video game console / 116-Pokémon / 292-Handheld game console / 320-Super Smash Bros. / 335-Nintendo 3DS / 443-Wii U / 697-Super Smash Bros. for Nintendo 3DS and Wii U / \n", "12-Food / 26-Cooking / 32-Recipe / 109-Cooking show / 239-Roasting / 542-Potato / 1701-French fries / 2266-Potato chip / \n", "1-Vehicle / 11-Motorsport / 17-Racing / 27-Motorcycle / 608-Wheelie / 927-Supermoto / 4308-Husaberg / \n", "3-Concert / 7-Musician / 13-Musical ensemble / 38-Orchestra / 203-Marching band / \n", "1-Vehicle / 4-Car / 40-Road / 64-Transport / 575-Traffic / 663-Street / 1020-Emergency vehicle / 3245-Mercedes-Benz Sprinter / \n", "6-Animation / 15-Cartoon / 123-Naruto / 185-Dragon Ball / 244-Goku / \n", "45-Cosmetics / 238-Nail / 300-Nail art / 304-Nail polish / 307-Hand / 324-Finger / 338-Manicure / 402-Jewellery / 823-Artificial nails / \n", "0-Games / 2-Video game / 19-PC game / 1861-Smite / \n", "0-Games / 6-Animation / 1676-Aikatsu! / \n", "0-Games / 2-Video game / 34-Action-adventure game / 1028-Devil May Cry / 1893-Devil May Cry 4 / \n", "3-Concert / 5-Dance / 16-Performance art / 66-Bollywood / \n", "1-Vehicle / 18-Outdoor recreation / 124-Tractor / 156-Agriculture / \n", "6-Animation / 25-Toy / 84-Snow / 99-Christmas / 120-Winter / 126-The Walt Disney Company / 354-Egg / 417-Play-Doh / 507-Thomas the Tank Engine / 611-Kinder Surprise / 2145-Olaf / \n", "5-Dance / 16-Performance art / \n", "7-Musician / 38-Orchestra / \n", "429-Website / \n", "246-Furniture / 369-Resort / \n", "1-Vehicle / 4-Car / 11-Motorsport / 17-Racing / 27-Motorcycle / 105-Motorcycling / 158-Tire / \n", "5-Dance / \n", "3-Concert / 7-Musician / 13-Musical ensemble / 38-Orchestra / 46-Choir / \n", "325-Vampire / \n", "3-Concert / 7-Musician / 14-Guitar / \n", "9-Music video / \n", "12-Food / 389-Restaurant / \n", "8-Football / \n", "1-Vehicle / 286-Bus / \n", "1-Vehicle / 62-Train / 64-Transport / 103-Rail transport / 152-Track / 868-Tram / \n", "1726-Harlem Shake (meme) / \n", "7-Musician / 14-Guitar / 24-String instrument / 100-Electric guitar / 1259-Fender Telecaster / \n", "5-Dance / \n", "0-Games / 89-Comics / 1676-Aikatsu! / \n", "18-Outdoor recreation / 1367-Golf club / 3140-Putter / 3264-Golf ball / \n", "3-Concert / \n", "14-Guitar / 24-String instrument / 96-Soldier / \n", "14-Guitar / 100-Electric guitar / 1754-Rocksmith / 2050-Rocksmith 2014 / \n", "1-Vehicle / 4-Car / 277-Sport utility vehicle / 403-Toyota / 1461-Electric car / 3758-Toyota RAV4 / \n", "3-Concert / 5-Dance / 31-Disc jockey / 133-Nightclub / \n", "66-Bollywood / \n", "23-Mobile phone / 29-Smartphone / 37-Gadget / 101-Telephone / 932-Microsoft Lumia / \n", "1970-Axe / \n", "83-Skateboarding / 555-Longboard / 667-Longboarding / \n", "10-Animal / 48-Pet / 71-Dog / 1603-Fox / \n", "3-Concert / 5-Dance / \n", "6-Animation / 1303-Fullmetal Alchemist / \n", "928-Borderlands 2 / \n", "1-Vehicle / 11-Motorsport / 27-Motorcycle / 197-Trail / 608-Wheelie / \n", "8-Football / \n", "12-Food / 26-Cooking / 32-Recipe / 58-Cuisine / 226-Dessert / 232-Cake / 236-Baking / 295-Chocolate / 485-Dough / 560-Sugar / 600-Flour / 1728-Chocolate cake / 2777-Mousse / \n", "5-Dance / \n", "1791-Stargate / \n", "9-Music video / \n", "95-Talent show / 130-Gymnastics / 1474-Acrobatic gymnastics / \n", "7-Musician / 230-Brass instrument / 296-Saxophone / \n", "0-Games / 2-Video game / 192-Battlefield / 374-Battlefield 3 / \n", "3-Concert / \n", "12-Food / 26-Cooking / 32-Recipe / 52-Dish / 210-Cookware and bakeware / 274-Meat / 572-Indian cuisine / 1748-Carrot / 3205-Beetroot / \n", "1-Vehicle / 11-Motorsport / 198-Off-road vehicle / 213-Forest / 327-Mud / 565-Mud bogging / \n", "23-Mobile phone / 29-Smartphone / 37-Gadget / 72-iPhone / 508-iPhone 5 / 675-iPhone 4S / 760-iPhone 5s / 1009-iPhone 3G / 1774-iPhone 5c / \n", "0-Games / 2-Video game / 35-Minecraft / 2398-Personal armor / \n", "28-Fashion / 785-Jeans / 1534-H&M / 2188-Coat / \n", "1-Vehicle / 4-Car / 27-Motorcycle / 207-Exhaust system / 336-Scooter / 1036-Moped / \n", "23-Mobile phone / 29-Smartphone / 72-iPhone / 131-Computer / 141-Microsoft Windows / 272-iPod / 331-iPod touch / \n", "84-Snow / 120-Winter / 3493-Tubing / \n", "452-Watch / 1260-Smartwatch / \n", "9-Music video / \n", "1-Vehicle / 4-Car / 11-Motorsport / 17-Racing / 41-Sports car / 57-Race track / 651-Stock car racing / \n", "7-Musician / 14-Guitar / 24-String instrument / 63-Acoustic guitar / \n", "0-Games / 8-Football / 65-Kick / 76-Ball / \n", "10-Animal / 48-Pet / 71-Dog / 1377-Golden Retriever / \n", "1-Vehicle / 40-Road / 62-Train / 64-Transport / 103-Rail transport / 119-Locomotive / 161-Train station / 224-Rapid transit / 663-Street / 1960-Traffic light / \n", "2-Video game / 43-Call of Duty / 111-PlayStation 3 / 127-Xbox 360 / 302-Call of Duty 4: Modern Warfare / \n", "110-Album / \n", "1-Vehicle / 4-Car / 20-Trailer / 74-Truck / 796-Fire engine / \n", "3-Concert / \n", "5-Dance / 16-Performance art / 115-Ballet / 980-Solo dance / \n", "0-Games / 6-Animation / 185-Dragon Ball / 241-Fighting game / 244-Goku / 418-Gohan / 1462-Dragon Ball XenoVerse / \n", "3-Concert / 7-Musician / 13-Musical ensemble / 38-Orchestra / \n", "66-Bollywood / \n", "3-Concert / 7-Musician / 13-Musical ensemble / 30-Drummer / 39-Drums / 3640-Akira Cine-Manga NeoTokyo 2019 / \n", "36-Piano / 53-Keyboard / \n", "3-Concert / 7-Musician / 13-Musical ensemble / 14-Guitar / \n", "23-Mobile phone / 29-Smartphone / 37-Gadget / 101-Telephone / \n", "121-Photography / 155-Camera / 566-Digital camera / 3796-Sony NEX-5 / \n", "1-Vehicle / 4-Car / 344-Coupé / 3229-Opel Kadett / \n", "7-Musician / 14-Guitar / 24-String instrument / 63-Acoustic guitar / 361-Flamenco / 2380-Flamenco guitar / \n", "0-Games / 1-Vehicle / 2-Video game / 4-Car / 11-Motorsport / 17-Racing / 40-Road / 41-Sports car / 231-Supercar / 1400-rFactor / 4395-Sim racing / \n", "16-Performance art / 66-Bollywood / \n", "12-Food / 246-Furniture / 389-Restaurant / 1230-Italian food / \n", "0-Games / 2-Video game / 55-Video game console / 586-Nintendo Entertainment System / \n", "0-Games / 59-Winter sport / 112-Ice skating / 142-Hockey / 262-Ice / 312-Ice rink / \n", "28-Fashion / 153-Dress / 361-Flamenco / 495-Runway / \n", "25-Toy / 477-Transformers / \n", "31-Disc jockey / \n", "5-Dance / 68-Lighting / 115-Ballet / \n", "61-Art / 276-Room / 353-Wall / \n", "6-Animation / 15-Cartoon / 183-Manga / \n", "0-Games / 8-Football / 21-Stadium / 54-Highlight film / 65-Kick / 76-Ball / \n", "45-Cosmetics / 436-Concealer / \n", "3-Concert / \n", "2449-Chuck E. Cheese's / \n", "2-Video game / 134-Call of Duty: Black Ops / 138-Call of Duty: Black Ops II / \n", "0-Games / 2-Video game / 55-Video game console / 292-Handheld game console / 320-Super Smash Bros. / 335-Nintendo 3DS / 443-Wii U / 697-Super Smash Bros. for Nintendo 3DS and Wii U / \n", "0-Games / 6-Animation / \n", "5-Dance / 22-Nature / 98-Festival / \n", "36-Piano / \n", "264-Star Wars / 2117-Star Wars Battlefront / \n", "8-Football / \n", "3-Concert / \n", "6-Animation / 126-The Walt Disney Company / 576-Walt Disney World / 3545-Pluto / \n", "1-Vehicle / 4-Car / 299-Volkswagen Passenger Cars / 531-Hatchback / \n", "3-Concert / \n", "0-Games / 582-Candy / 1112-Candy Crush Saga / \n", "60-Basketball / \n", "0-Games / 2-Video game / 35-Minecraft / 366-Dragon / 420-The Elder Scrolls V: Skyrim / \n", "0-Games / 1698-Darts / \n", "5-Dance / 9-Music video / \n", "163-Running / \n", "0-Games / 2-Video game / 23-Mobile phone / 29-Smartphone / 37-Gadget / 2285-LG G3 / \n", "1-Vehicle / 6-Animation / 136-Gardening / 585-Lawn / 741-Mower / 805-Lawn mower / \n", "23-Mobile phone / 29-Smartphone / 37-Gadget / \n", "23-Mobile phone / 29-Smartphone / 37-Gadget / 101-Telephone / 1820-Motorola Droid / 3186-Motorola Razr / \n", "22-Nature / 113-House / 136-Gardening / \n", "20-Trailer / 159-World of Warcraft / 171-Warcraft / \n", "3-Concert / \n", "3-Concert / 5-Dance / 16-Performance art / \n", "10-Animal / 48-Pet / 71-Dog / 182-Puppy / 2319-Zara / \n", "78-Wedding / \n", "56-Hair / 106-Hairstyle / 583-Wig / \n", "0-Games / 2-Video game / 43-Call of Duty / 134-Call of Duty: Black Ops / 138-Call of Duty: Black Ops II / 768-Diamond / \n", "5-Dance / 16-Performance art / \n", "0-Games / 25-Toy / 417-Play-Doh / 723-Pig / \n", "28-Fashion / 415-Sewing / 880-Trousers / 1627-Shorts / \n", "6-Animation / 15-Cartoon / 20-Trailer / 1724-Homer Simpson / \n", "1-Vehicle / 4-Car / 11-Motorsport / 167-Four-wheel drive / 198-Off-road vehicle / 327-Mud / 935-Off-road racing / \n", "10-Animal / 412-Monster / 550-Dinosaur / \n", "0-Games / 8-Football / 21-Stadium / 77-Arena / \n", "3-Concert / 7-Musician / 14-Guitar / 24-String instrument / 63-Acoustic guitar / 256-Tennis / \n", "125-Television / \n", "0-Games / 2-Video game / 55-Video game console / 292-Handheld game console / 317-PlayStation / 362-PlayStation Portable / 3980-Sega CD / \n", "405-Braid / 457-Knitting / 546-Stitch / 653-Thread / \n", "3-Concert / 5-Dance / 7-Musician / 13-Musical ensemble / \n", "376-Flute / 1326-Recorder / 3344-Tin whistle / \n", "3-Concert / 7-Musician / 13-Musical ensemble / 46-Choir / \n", "0-Games / 2-Video game / 34-Action-adventure game / 3649-Metroid Prime / \n", "253-Carnival / \n", "0-Games / 489-Board game / 799-Chess / \n", "1-Vehicle / 4-Car / 11-Motorsport / 17-Racing / 40-Road / 41-Sports car / 169-Rallying / \n", "920-Chair / \n", "842-Harp / \n", "6-Animation / 9-Music video / 15-Cartoon / \n", "0-Games / 8-Football / 21-Stadium / 77-Arena / \n", "56-Hair / 78-Wedding / 106-Hairstyle / 190-Bride / 405-Braid / 909-Updo / \n", "66-Bollywood / 1524-Zee Bangla / \n", "9-Music video / 15-Cartoon / \n", "0-Games / 2-Video game / 90-Sports game / 3764-EA Sports UFC / 4132-Fight Night Champion / \n", "1-Vehicle / 62-Train / 64-Transport / 94-Dashcam / 103-Rail transport / 152-Track / 161-Train station / \n", "85-Combat / 3604-Chokehold / \n", "3-Concert / \n", "163-Running / \n", "0-Games / 8-Football / 54-Highlight film / 79-American football / 813-Running back / \n", "31-Disc jockey / \n", "804-Shopping mall / \n", "7-Musician / 13-Musical ensemble / 46-Choir / \n", "1-Vehicle / 18-Outdoor recreation / 27-Motorcycle / 83-Skateboarding / 108-Skateboard / 608-Wheelie / \n", "1721-Tetris / \n", "3-Concert / 7-Musician / 13-Musical ensemble / 30-Drummer / 44-Drums / \n", "1-Vehicle / 11-Motorsport / 27-Motorcycle / 105-Motorcycling / 955-Police officer / \n", "110-Album / \n", "159-World of Warcraft / 171-Warcraft / \n", "5-Dance / 16-Performance art / \n", "1-Vehicle / 4-Car / 6-Animation / 20-Trailer / 126-The Walt Disney Company / 1077-Lightning McQueen / \n", "10-Animal / 467-Chipmunk / 846-Squirrel / \n", "7-Musician / 14-Guitar / 24-String instrument / \n", "7-Musician / 14-Guitar / 24-String instrument / 63-Acoustic guitar / \n", "6-Animation / 10-Animal / 636-Rabbit / \n", "315-Knife / \n", "7-Musician / 14-Guitar / 24-String instrument / 39-Drums / 63-Acoustic guitar / \n", "3-Concert / 7-Musician / 39-Drums / 44-Drums / 110-Album / \n", "33-Weapon / 85-Combat / 96-Soldier / \n", "5-Dance / 12-Food / 30-Drummer / 1465-FOX / \n", "0-Games / 2-Video game / 10-Animal / 51-Strategy video game / \n", "1-Vehicle / 10-Animal / 11-Motorsport / \n", "5-Dance / 16-Performance art / 66-Bollywood / \n", "3-Concert / 620-Barbecue / \n", "1-Vehicle / 27-Motorcycle / \n", "66-Bollywood / \n", "5-Dance / \n", "113-House / \n", "12-Food / 26-Cooking / 32-Recipe / 52-Dish / 58-Cuisine / 109-Cooking show / 179-Kitchen / 232-Cake / 239-Roasting / 742-Countertop / 1018-Salmon / \n", "1-Vehicle / 67-Cycling / 69-Bicycle / 197-Trail / 220-Mountain bike / 284-Mountain biking / \n", "0-Games / 8-Football / 21-Stadium / 54-Highlight film / 65-Kick / 73-Ball / \n", "3-Concert / 7-Musician / 30-Drummer / 39-Drums / 44-Drums / \n", "5-Dance / 227-Water / 490-Rain / \n", "411-MapleStory / \n", "0-Games / 8-Football / 73-Ball / 81-Athlete / 358-Rugby football / \n", "2-Video game / 441-The Elder Scrolls / 1410-The Elder Scrolls IV: Oblivion / \n", "14-Guitar / \n", "10-Animal / 18-Outdoor recreation / 42-Fishing / 91-Fish / \n", "23-Mobile phone / 29-Smartphone / 37-Gadget / 72-iPhone / 101-Telephone / 760-iPhone 5s / 1774-iPhone 5c / \n", "147-River / 283-Rock / 347-Climbing / \n", "163-Running / 206-Shoe / \n", "3-Concert / \n", "0-Games / 2-Video game / 34-Action-adventure game / 118-Grand Theft Auto V / \n", "25-Toy / 55-Video game console / 4183-Tamagotchi / \n", "1-Vehicle / 50-Aircraft / 82-Airplane / 129-Aviation / 139-Airport / 151-Landing / 157-Jet aircraft / 175-Airliner / 187-Takeoff / 188-Airline / 254-Runway / 451-Cockpit / 719-Boeing 747 / \n", "9-Music video / \n", "25-Toy / 626-Puzzle / 631-Cube / 740-Rubik's Cube / \n", "1-Vehicle / 62-Train / 119-Locomotive / 507-Thomas the Tank Engine / 2332-Percy / \n", "116-Pokémon / 186-Pokémon / 690-Pokémon X and Y / 1005-Pokémon Omega Ruby and Alpha Sapphire / \n", "9-Music video / 125-Television / 322-Mixtape / \n", "10-Animal / 48-Pet / 71-Dog / 182-Puppy / 1252-Dog agility / 1839-Rottweiler / \n", "28-Fashion / \n", "9-Music video / 31-Disc jockey / \n", "49-School / \n", "1-Vehicle / 4-Car / 40-Road / 70-Driving / 94-Dashcam / \n", "6-Animation / 15-Cartoon / \n", "3-Concert / 16-Performance art / \n", "113-House / 339-Architecture / \n", "0-Games / 211-RuneScape / \n", "0-Games / 12-Food / 389-Restaurant / \n", "5-Dance / 16-Performance art / \n", "6-Animation / 15-Cartoon / 49-School / \n", "0-Games / 88-Machine / 240-Arcade game / 489-Board game / 535-Slot machine / 3235-Monopoly / 4541-Multiplayer game / \n", "25-Toy / 140-LEGO / 462-Table / 920-Chair / \n", "12-Food / 26-Cooking / 32-Recipe / 58-Cuisine / 109-Cooking show / 533-Oil / 1765-Muffin / \n", "36-Piano / 53-Keyboard / 107-Musical keyboard / 397-Organ / \n", "9-Music video / \n", "0-Games / 3274-DragonFable / \n", "3-Concert / 16-Performance art / 4284-Arte / \n", "59-Winter sport / 112-Ice skating / 142-Hockey / 522-Helmet / \n", "12-Food / 26-Cooking / 52-Dish / 58-Cuisine / 215-Eating / 301-Meal / 727-Curry / \n", "2-Video game / \n", "0-Games / 2-Video game / 19-PC game / 43-Call of Duty / 431-Call of Duty: Advanced Warfare / 1843-Ham / \n", "497-Macintosh / 3233-Screencast / \n", "1-Vehicle / 4-Car / \n", "122-Weight training / 204-Gym / 214-Muscle / 1055-Back / \n", "9-Music video / \n", "1-Vehicle / 11-Motorsport / 27-Motorcycle / 1124-Yamaha YZF-R1 / \n", "47-Personal computer / 145-Tablet computer / 189-iPad / 2262-E-reader / \n", "3-Concert / 68-Lighting / \n", "527-Battery / 1642-Cigarette / \n", "1-Vehicle / 42-Fishing / 87-Boat / 498-Motorboat / 579-Yacht / 753-Sailboat / \n", "3-Concert / \n", "3-Concert / 38-Orchestra / \n", "4214-Spike / \n", "0-Games / 1-Vehicle / 2-Video game / 2101-Ace Combat / \n", "113-House / 179-Kitchen / 246-Furniture / \n", "1-Vehicle / 62-Train / 64-Transport / 103-Rail transport / 152-Track / 428-Rail transport modelling / \n", "93-Comedy / 364-Sitcom / \n", "3-Concert / 7-Musician / 13-Musical ensemble / 38-Orchestra / 150-Violin / 639-Fiddle / \n", "3-Concert / 16-Performance art / \n", "0-Games / 326-Metin2 / \n", "7-Musician / 46-Choir / 126-The Walt Disney Company / 153-Dress / 576-Walt Disney World / \n", "0-Games / 1-Vehicle / 2-Video game / 4-Car / 11-Motorsport / 17-Racing / 111-PlayStation 3 / 127-Xbox 360 / 181-Xbox / 2621-GameTrailers / \n", "0-Games / 1-Vehicle / 2-Video game / 4-Car / 11-Motorsport / 17-Racing / 41-Sports car / 57-Race track / 111-PlayStation 3 / 352-Need for Speed / 1419-GameSpot / 2219-Need for Speed: Shift / \n", "31-Disc jockey / \n", "9-Music video / 322-Mixtape / \n", "293-Slide show / \n", "20-Trailer / 359-Harry Potter (Literary Series) / \n", "12-Food / 26-Cooking / 32-Recipe / 226-Dessert / 534-Milk / 770-Banana / 1283-Pancake / \n", "8-Football / \n", "37-Gadget / 47-Personal computer / 131-Computer / 141-Microsoft Windows / 145-Tablet computer / \n", "36-Piano / 88-Machine / 400-Woodturning / 673-Saw / 1700-Lumber / 2874-Sawmill / 3789-Circular saw / \n", "13-Musical ensemble / \n", "3-Concert / 7-Musician / \n", "144-Amusement park / 329-Swimming pool / 1053-Water park / \n", "47-Personal computer / 228-Laptop / 513-Computer keyboard / \n", "6-Animation / 15-Cartoon / \n", "3-Concert / \n", "25-Toy / 417-Play-Doh / 3571-Lollipop / \n", "2-Video game / 43-Call of Duty / 111-PlayStation 3 / 127-Xbox 360 / 134-Call of Duty: Black Ops / \n", "0-Games / 1605-Second Life / \n", "66-Bollywood / \n", "3-Concert / 7-Musician / 14-Guitar / 16-Performance art / \n", "5-Dance / \n", "514-Tekken / \n", "0-Games / 159-World of Warcraft / 171-Warcraft / 3615-Monk / \n", "1-Vehicle / 4-Car / 11-Motorsport / 154-Wheel / 158-Tire / 270-Sedan / 1455-Cadillac / 3681-Cadillac CTS / \n", "3-Concert / \n", "0-Games / 1-Vehicle / 2-Video game / 4-Car / 34-Action-adventure game / 201-Grand Theft Auto: San Andreas / 297-Grand Theft Auto IV / 332-Mercedes-Benz / \n", "1-Vehicle / 4-Car / \n", "122-Weight training / \n", "9-Music video / \n", "72-iPhone / \n", "0-Games / 51-Strategy video game / 165-League of Legends / \n", "6-Animation / 15-Cartoon / \n", "5-Dance / \n", "1-Vehicle / 4-Car / 403-Toyota / 3758-Toyota RAV4 / \n", "3-Concert / 7-Musician / \n", "455-Unidentified flying object / 822-Crane / 1668-Claw crane / \n", "9-Music video / \n", "0-Games / 2-Video game / 192-Battlefield / 374-Battlefield 3 / \n", "253-Carnival / \n", "23-Mobile phone / 29-Smartphone / 37-Gadget / 101-Telephone / 553-Sony Xperia / \n", "170-Home improvement / 276-Room / 652-Bathroom / 1011-Tile / 1113-Toilet / 1272-Cabinetry / 2153-Sink / \n", "122-Weight training / 204-Gym / \n", "3-Concert / 16-Performance art / 594-Music festival / \n", "0-Games / 8-Football / 21-Stadium / 77-Arena / \n", "1-Vehicle / 67-Cycling / 69-Bicycle / 108-Skateboard / 305-Ski / 336-Scooter / 667-Longboarding / 703-Roller skating / \n", "0-Games / 2-Video game / 51-Strategy video game / 1072-Total War / 1882-Warrior / 2810-Total War: Rome II / \n", "6-Animation / 25-Toy / 140-LEGO / \n", "0-Games / 20-Trailer / 4304-Civilization V / \n", "2811-Llanero / \n", "1-Vehicle / 4-Car / 170-Home improvement / 519-Recreational vehicle / \n", "3-Concert / 13-Musical ensemble / \n", "0-Games / 2-Video game / 43-Call of Duty / 134-Call of Duty: Black Ops / 138-Call of Duty: Black Ops II / 1248-Call of Duty: Zombies / \n", "75-Wrestling / 85-Combat / \n", "1-Vehicle / 62-Train / 64-Transport / 103-Rail transport / 119-Locomotive / 143-Railroad car / 152-Track / 161-Train station / 224-Rapid transit / \n", "45-Cosmetics / 173-Eye shadow / 195-Mascara / 196-Eye / 218-Eyelash / 3106-Eyelash extensions / \n", "3-Concert / 7-Musician / 13-Musical ensemble / 46-Choir / 2610-Rosary / \n", "9-Music video / \n", "6-Animation / 15-Cartoon / 643-Hero / \n", "14-Guitar / 96-Soldier / \n", "0-Games / 2-Video game / 43-Call of Duty / 138-Call of Duty: Black Ops II / \n", "122-Weight training / \n", "527-Battery / 2484-Electronic circuit / \n", "0-Games / 2-Video game / 34-Action-adventure game / 201-Grand Theft Auto: San Andreas / 858-Carl Johnson / \n", "5-Dance / 95-Talent show / 130-Gymnastics / 191-Cheerleading / 892-Aerobics / 2228-Sport aerobics / \n", "33-Weapon / 217-Hunting / \n", "0-Games / 2-Video game / 257-Counter-Strike / 315-Knife / \n", "47-Personal computer / 345-Loudspeaker / 414-Computer monitor / \n", "5-Dance / 49-School / 77-Arena / 95-Talent show / 130-Gymnastics / 191-Cheerleading / 306-High school / \n", "5-Dance / 13-Musical ensemble / 78-Wedding / \n", "23-Mobile phone / 29-Smartphone / 37-Gadget / 47-Personal computer / 101-Telephone / 145-Tablet computer / 621-Asus / \n", "0-Games / 2-Video game / 19-PC game / 43-Call of Duty / 96-Soldier / 704-Call of Duty: World at War / \n", "0-Games / 75-Wrestling / \n", "10-Animal / 25-Toy / 166-Cat / 524-Littlest Pet Shop / 749-Figurine / \n", "6-Animation / \n", "5-Dance / 16-Performance art / 115-Ballet / 430-Ballroom dance / \n", "3-Concert / 68-Lighting / \n", "429-Website / 1601-Social media / \n", "1-Vehicle / 4-Car / 11-Motorsport / 234-Ford / 473-Ford Mustang / 647-Muscle car / 1877-Shelby Mustang / \n", "99-Christmas / \n", "696-T-shirt / \n", "351-Jumping / 529-Squat / \n", "0-Games / 2-Video game / 35-Minecraft / 102-Building / 1067-Lock / \n", "0-Games / 60-Basketball / 81-Athlete / 168-Basketball moves / 255-Slam dunk / \n", "1-Vehicle / 4-Car / \n", "696-T-shirt / 793-Shirt / 2693-Puma SE / \n", "1-Vehicle / 27-Motorcycle / 154-Wheel / \n", "33-Weapon / 1873-Crossbow / \n", "121-Photography / \n", "1-Vehicle / 18-Outdoor recreation / 67-Cycling / 69-Bicycle / 197-Trail / 220-Mountain bike / 284-Mountain biking / \n", "238-Nail / 300-Nail art / 304-Nail polish / 307-Hand / 324-Finger / 338-Manicure / \n", "93-Comedy / 355-Comedian / \n", "1-Vehicle / 4-Car / 11-Motorsport / 17-Racing / 70-Driving / 2286-Grand Tourer Injection / 2826-Suzuki Swift / \n", "1-Vehicle / 4-Car / 519-Recreational vehicle / 707-Roof / \n", "9-Music video / 78-Wedding / 190-Bride / \n", "3-Concert / 68-Lighting / \n", "242-Paper / 370-Origami / \n", "1-Vehicle / 27-Motorcycle / \n", "0-Games / 1-Vehicle / 2-Video game / 4-Car / 70-Driving / 2249-Driver / 3365-Driver: San Francisco / \n", "25-Toy / 140-LEGO / 249-Heavy equipment / \n", "12-Food / 26-Cooking / 32-Recipe / 52-Dish / 58-Cuisine / 109-Cooking show / 4236-Biryani / \n", "1-Vehicle / 4-Car / 11-Motorsport / 17-Racing / 40-Road / 57-Race track / 378-Drag racing / \n", "18-Outdoor recreation / 42-Fishing / 164-Beach / 222-Surfing / 233-Ocean / 423-Coast / \n", "22-Nature / 1235-Rainbow / \n", "5-Dance / \n", "0-Games / 637-Five Nights at Freddy's / \n", "28-Fashion / 206-Shoe / 333-Nike; Inc. / 368-Sneakers / \n", "31-Disc jockey / \n", "22-Nature / 490-Rain / 998-Landscape / \n", "10-Animal / 18-Outdoor recreation / 22-Nature / 42-Fishing / 147-River / 478-Fishing rod / 825-Stream / 895-Trout / \n", "1-Vehicle / 4-Car / 11-Motorsport / 531-Hatchback / 618-Fiat Automobiles / 3127-Fiat Uno / \n", "2449-Chuck E. Cheese's / \n", "0-Games / 2-Video game / 229-Halo / 470-Halo 3 / 2035-Forge / \n", "1-Vehicle / 18-Outdoor recreation / 67-Cycling / 69-Bicycle / 197-Trail / 213-Forest / 220-Mountain bike / 284-Mountain biking / \n", "30-Drummer / 39-Drums / 44-Drums / 128-Cymbal / \n", "9-Music video / \n", "0-Games / 2-Video game / 192-Battlefield / 374-Battlefield 3 / 383-Battlefield 4 / \n", "10-Animal / 20-Trailer / 80-Horse / \n", "18-Outdoor recreation / 22-Nature / 222-Surfing / 233-Ocean / 672-Bodyboarding / \n", "75-Wrestling / \n", "0-Games / 2-Video game / 1959-Phantasy Star / 2142-Sega Saturn / \n", "0-Games / 2-Video game / 55-Video game console / 413-PlayStation 2 / \n", "5-Dance / 115-Ballet / \n", "110-Album / \n", "130-Gymnastics / \n", "12-Food / 26-Cooking / 52-Dish / 58-Cuisine / 2959-Wok / \n", "0-Games / 2-Video game / 19-PC game / 20-Trailer / \n", "7-Musician / 13-Musical ensemble / \n", "10-Animal / \n", "3-Concert / 77-Arena / \n", "5-Dance / \n", "0-Games / 2-Video game / 34-Action-adventure game / 294-Batman / 838-Batman: Arkham / 1660-Batman: Arkham Asylum / 1881-Arkham Asylum / \n", "989-Warface / \n", "6-Animation / 15-Cartoon / \n", "130-Gymnastics / 1255-Ribbon / 1451-Ribbon / \n", "36-Piano / 53-Keyboard / 107-Musical keyboard / 379-Touhou Project / 592-Synthesia / \n", "22-Nature / 86-Plant / 136-Gardening / 137-Tree / \n", "102-Building / 2347-Marble / \n", "12-Food / 18-Outdoor recreation / 26-Cooking / 32-Recipe / 136-Gardening / 395-Camping / 968-Pie / \n", "9-Music video / \n", "242-Paper / 315-Knife / 512-Printing / 525-Printer / \n", "5-Dance / 16-Performance art / \n", "0-Games / 116-Pokémon / 186-Pokémon / \n", "860-Christian Church / 3783-Blazer / \n", "59-Winter sport / 112-Ice skating / 142-Hockey / 262-Ice / 312-Ice rink / \n", "1-Vehicle / 4-Car / \n", "30-Drummer / 39-Drums / 44-Drums / 128-Cymbal / 146-Snare drum / \n", "0-Games / 19-PC game / 51-Strategy video game / 165-League of Legends / \n", "36-Piano / 53-Keyboard / 104-Pianist / 107-Musical keyboard / \n", "12-Food / 25-Toy / 26-Cooking / 1567-Cook / \n", "7-Musician / 14-Guitar / 24-String instrument / 63-Acoustic guitar / \n", "95-Talent show / 130-Gymnastics / 1417-Balance beam / \n", "2-Video game / 989-Warface / \n", "10-Animal / 48-Pet / 162-Medicine / 166-Cat / \n", "93-Comedy / \n", "3-Concert / 7-Musician / 30-Drummer / 39-Drums / 44-Drums / \n", "14-Guitar / 24-String instrument / 100-Electric guitar / 2738-Fender Jazz Bass / \n", "3-Concert / 5-Dance / 99-Christmas / \n", "0-Games / 8-Football / 81-Athlete / 358-Rugby football / \n", "3-Concert / 68-Lighting / \n", "3-Concert / 7-Musician / 13-Musical ensemble / 16-Performance art / \n", "5-Dance / \n", "6-Animation / 15-Cartoon / 185-Dragon Ball / 244-Goku / 418-Gohan / 1244-Trunks / 1304-Cell / \n", "242-Paper / 313-Book / 415-Sewing / 2454-Notebook / \n", "20-Trailer / \n", "261-Tool / 673-Saw / 1033-Drill / \n", "3-Concert / 7-Musician / 219-Accordion / 545-Quartet (ensemble) / \n", "624-Number / \n", "1-Vehicle / 4-Car / 27-Motorcycle / \n", "31-Disc jockey / \n", "0-Games / 2-Video game / 1007-Portal / 1069-Easter egg / 1274-Portal 2 / 3555-Valve Corporation / \n", "10-Animal / 71-Dog / \n", "1249-Rainbow Loom / 1449-Rubber band / \n", "3-Concert / \n", "3-Concert / 7-Musician / \n", "358-Rugby football / \n", "1-Vehicle / 27-Motorcycle / 263-Light / \n", "1-Vehicle / 11-Motorsport / 64-Transport / 74-Truck / 822-Crane / \n", "6-Animation / 15-Cartoon / 123-Naruto / \n", "3-Concert / 5-Dance / 16-Performance art / 1572-Husband / \n", "259-Fire / 371-Firefighter / \n", "0-Games / 35-Minecraft / \n", "1338-Devil / 2793-Disguise / \n", "10-Animal / 48-Pet / 1147-Hamster / \n", "1783-Roller skates / \n", "6-Animation / 15-Cartoon / 801-Inuyasha / \n", "1-Vehicle / 4-Car / 912-Silver / \n", "5-Dance / 16-Performance art / 980-Solo dance / \n", "1-Vehicle / 12-Food / 1054-Vacuum cleaner / \n", "2128-Pollution / \n", "28-Fashion / 56-Hair / 106-Hairstyle / \n", "0-Games / 8-Football / 21-Stadium / \n", "1-Vehicle / 4-Car / 11-Motorsport / 74-Truck / 349-Chevrolet / 375-Pickup truck / 406-Chevrolet / \n", "1-Vehicle / 18-Outdoor recreation / 67-Cycling / 69-Bicycle / 197-Trail / 220-Mountain bike / 284-Mountain biking / \n", "0-Games / 8-Football / 54-Highlight film / \n", "9-Music video / \n", "0-Games / 2-Video game / 266-Helicopter / 383-Battlefield 4 / \n", "780-Fat / \n", "5-Dance / 20-Trailer / 78-Wedding / \n", "6-Animation / 15-Cartoon / 1946-Total Drama Island / \n", "31-Disc jockey / \n", "1-Vehicle / 50-Aircraft / 1409-Missile / 4341-Anti-aircraft warfare / \n", "9-Music video / \n", "31-Disc jockey / 322-Mixtape / \n", "1883-Gothic / 2515-Gothic / \n", "45-Cosmetics / 195-Mascara / \n", "0-Games / 60-Basketball / \n", "34-Action-adventure game / 111-PlayStation 3 / 1125-God of War / 3927-God of War: Ascension / \n", "0-Games / 2-Video game / 116-Pokémon / 186-Pokémon / \n", "18-Outdoor recreation / 83-Skateboarding / 108-Skateboard / 221-Skatepark / \n", "9-Music video / 78-Wedding / 153-Dress / 190-Bride / \n", "5-Dance / \n", "79-American football / \n", "3-Concert / 13-Musical ensemble / 46-Choir / 49-School / \n", "5-Dance / \n", "1442-Tabla / \n", "47-Personal computer / 957-Desktop computer / \n", "12-Food / 26-Cooking / 32-Recipe / 52-Dish / 58-Cuisine / 109-Cooking show / 301-Meal / 597-Soup / \n", "25-Toy / 75-Wrestling / 435-Action figure / 1844-WWE action figures / \n", "223-Weather / \n", "341-Sketch comedy / \n", "0-Games / \n", "2-Video game / 55-Video game console / 413-PlayStation 2 / \n", "20-Trailer / \n", "0-Games / 1-Vehicle / 2-Video game / 4-Car / 11-Motorsport / 17-Racing / 413-PlayStation 2 / 886-James Bond / \n", "0-Games / 8-Football / 21-Stadium / 54-Highlight film / 65-Kick / 76-Ball / 81-Athlete / \n", "99-Christmas / 729-Christmas decoration / 964-Christmas tree / \n", "17-Racing / 67-Cycling / 69-Bicycle / \n", "1-Vehicle / 11-Motorsport / 17-Racing / 27-Motorcycle / 105-Motorcycling / \n", "12-Food / 1596-Cinnamon / \n", "3-Concert / 7-Musician / 13-Musical ensemble / 36-Piano / 38-Orchestra / \n", "1-Vehicle / 4-Car / \n", "3-Concert / 7-Musician / 13-Musical ensemble / 14-Guitar / 30-Drummer / 39-Drums / 44-Drums / \n", "341-Sketch comedy / 364-Sitcom / 463-Cricket / \n", "322-Mixtape / \n", "3-Concert / 7-Musician / 13-Musical ensemble / 14-Guitar / 30-Drummer / \n", "75-Wrestling / \n", "1-Vehicle / 4-Car / 11-Motorsport / 17-Racing / 169-Rallying / 1122-Mitsubishi Lancer Evolution / \n", "1-Vehicle / \n", "298-Television advertisement / 4145-Tape recorder / \n", "1-Vehicle / 4-Car / 40-Road / 64-Transport / 70-Driving / 94-Dashcam / 212-Highway / \n", "0-Games / 6-Animation / 15-Cartoon / 243-Sonic the Hedgehog / 251-Sonic the Hedgehog / \n", "420-The Elder Scrolls V: Skyrim / \n", "102-Building / 113-House / 433-Construction / \n", "0-Games / 2-Video game / 34-Action-adventure game / \n", "0-Games / 8-Football / 81-Athlete / \n", "1-Vehicle / 50-Aircraft / 139-Airport / 151-Landing / 175-Airliner / 188-Airline / 472-Boeing 737 / \n", "1-Vehicle / 11-Motorsport / 27-Motorcycle / \n", "10-Animal / 350-Circus / \n", "5-Dance / \n", "1-Vehicle / 4-Car / 40-Road / 70-Driving / 94-Dashcam / 212-Highway / \n", "14-Guitar / 24-String instrument / 100-Electric guitar / \n", "0-Games / 8-Football / 21-Stadium / 54-Highlight film / 73-Ball / \n", "56-Hair / 583-Wig / 607-Lace / 871-Hair coloring / \n", "260-Prayer / 616-The Bible / \n", "3-Concert / 7-Musician / 13-Musical ensemble / 30-Drummer / \n", "84-Snow / 120-Winter / \n", "0-Games / 9-Music video / 326-Metin2 / \n", "117-Boxing / \n", "13-Musical ensemble / 110-Album / \n", "6-Animation / 35-Minecraft / \n", "0-Games / 2-Video game / 1745-Assassin's Creed: Revelations / \n", "1-Vehicle / 4-Car / 1067-Lock / \n", "89-Comics / 269-Comic book / 313-Book / \n", "1546-Line / \n", "42-Fishing / 510-Fishing lure / 902-Fishing bait / 1836-Fly tying / \n", "0-Games / 59-Winter sport / 112-Ice skating / 142-Hockey / 312-Ice rink / \n", "5-Dance / 9-Music video / 1443-Dance studio / \n", "1-Vehicle / 10-Animal / 18-Outdoor recreation / 42-Fishing / 87-Boat / 91-Fish / 268-Recreational fishing / 698-Fisherman / 1471-Tuna / \n", "1-Vehicle / 259-Fire / 371-Firefighter / \n", "3-Concert / 1684-Hillsong Church / \n", "0-Games / 2-Video game / 19-PC game / 51-Strategy video game / 171-Warcraft / 1691-Warcraft III: Reign of Chaos / 2549-Warcraft III: The Frozen Throne / \n", "725-Human / 1264-Head / \n", "36-Piano / \n", "102-Building / 627-Clock / \n", "0-Games / 2-Video game / 89-Comics / 185-Dragon Ball / 244-Goku / 876-Freeza / 1462-Dragon Ball XenoVerse / \n", "2507-Macramé / 3063-Friendship bracelet / \n", "31-Disc jockey / \n", "25-Toy / 387-Spider-Man / 435-Action figure / 4288-Carnage / \n", "25-Toy / 89-Comics / 185-Dragon Ball / \n", "45-Cosmetics / 173-Eye shadow / 193-Eye liner / 195-Mascara / 196-Eye / 218-Eyelash / 235-Lipstick / \n", "3-Concert / 7-Musician / 13-Musical ensemble / 2381-Menu / \n", "1-Vehicle / 18-Outdoor recreation / 67-Cycling / 69-Bicycle / 197-Trail / 208-Wood / 220-Mountain bike / 284-Mountain biking / \n", "9-Music video / \n", "10-Animal / 48-Pet / 71-Dog / 182-Puppy / \n", "10-Animal / 56-Hair / \n", "1430-Pull-up / \n", "950-Blu-ray disc / \n", "1-Vehicle / 271-Tank / 273-News program / \n", "1-Vehicle / 4-Car / 40-Road / 64-Transport / 74-Truck / \n", "10-Animal / 48-Pet / 71-Dog / 2211-Fur / 2735-Raccoon / \n", "0-Games / 756-Tibia / \n", "238-Nail / 304-Nail polish / 324-Finger / \n", "0-Games / 2-Video game / 35-Minecraft / \n", "6-Animation / 15-Cartoon / 2710-Shikamaru Nara / 3702-Temari / \n", "110-Album / \n", "6-Animation / 200-Doll / 439-Barbie / \n", "1-Vehicle / 4-Car / 744-Window / 918-Brake / 4327-Window film / \n", "98-Festival / \n", "13-Musical ensemble / \n", "4357-Nezha / \n", "5-Dance / 49-School / 95-Talent show / \n", "8-Football / 14-Guitar / \n", "25-Toy / 1154-Anpanman / \n", "12-Food / 26-Cooking / 32-Recipe / 770-Banana / \n", "12-Food / 26-Cooking / 32-Recipe / 236-Baking / 485-Dough / 578-Cookie / \n", "15-Cartoon / 116-Pokémon / 1104-Ash Ketchum / \n", "1-Vehicle / 4-Car / 11-Motorsport / 167-Four-wheel drive / 198-Off-road vehicle / 213-Forest / 327-Mud / 565-Mud bogging / \n", "217-Hunting / 718-Duck / 1914-Waterfowl hunting / \n", "14-Guitar / 100-Electric guitar / \n", "7-Musician / 14-Guitar / 24-String instrument / 100-Electric guitar / \n", "130-Gymnastics / \n", "6-Animation / 20-Trailer / \n", "5-Dance / 16-Performance art / \n", "22-Nature / 84-Snow / 120-Winter / 918-Brake / 2058-Snowman / \n", "12-Food / 232-Cake / 795-Cupcake / 1707-Microwave oven / \n", "211-RuneScape / \n", "1-Vehicle / 4-Car / 27-Motorcycle / 105-Motorcycling / \n", "99-Christmas / \n", "23-Mobile phone / 29-Smartphone / 72-iPhone / 121-Photography / \n", "276-Room / 476-Textile / \n", "1386-ABS-CBN / 1849-ABS-CBN News and Current Affairs / \n", "5-Dance / 115-Ballet / \n", "3-Concert / 7-Musician / 13-Musical ensemble / 46-Choir / 318-Church / 397-Organ / 870-Chapel / \n", "462-Table / 544-Table tennis / 1615-Pong / \n", "113-House / 200-Doll / 1842-Dollhouse / 2743-Wallpaper / \n", "0-Games / 2-Video game / 34-Action-adventure game / 685-Luigi / 1638-Soulcalibur / 2033-Princess Peach / 2877-Soulcalibur IV / 4628-Soulcalibur III / \n", "0-Games / 2-Video game / 33-Weapon / 43-Call of Duty / 111-PlayStation 3 / 138-Call of Duty: Black Ops II / \n", "8-Football / \n", "0-Games / 35-Minecraft / \n", "1-Vehicle / 4-Car / 40-Road / 64-Transport / 70-Driving / 94-Dashcam / \n", "0-Games / 19-PC game / 211-RuneScape / \n", "3-Concert / 8-Football / 21-Stadium / 77-Arena / \n", "0-Games / 2-Video game / 55-Video game console / 317-PlayStation / 682-PlayStation Vita / \n", "219-Accordion / \n", "3-Concert / 5-Dance / 16-Performance art / \n", "66-Bollywood / \n", "126-The Walt Disney Company / 369-Resort / 574-Elevator / 576-Walt Disney World / 1746-Otis Elevator Company / \n", "0-Games / 2-Video game / 420-The Elder Scrolls V: Skyrim / 441-The Elder Scrolls / \n", "10-Animal / 80-Horse / 247-Stallion / 404-Livestock / 438-Mare / 1871-Foal / \n", "286-Bus / 3694-Burton Snowboards / \n", "122-Weight training / 954-Dreadlocks / \n", "2-Video game / 34-Action-adventure game / 362-PlayStation Portable / 1125-God of War / \n", "1-Vehicle / 4-Car / 11-Motorsport / 41-Sports car / 231-Supercar / 1056-Porsche 911 / \n", "7-Musician / 14-Guitar / 24-String instrument / 100-Electric guitar / \n", "45-Cosmetics / 173-Eye shadow / 193-Eye liner / 195-Mascara / 196-Eye / 218-Eyelash / 291-Rouge / \n", "1-Vehicle / 74-Truck / \n", "10-Animal / 80-Horse / 261-Tool / 1995-Hammer / 2035-Forge / \n", "5-Dance / 16-Performance art / \n", "1-Vehicle / 27-Motorcycle / \n", "0-Games / 2-Video game / 19-PC game / 33-Weapon / 1631-PlanetSide 2 / 2026-PlanetSide / \n", "10-Animal / 22-Nature / 86-Plant / 258-Aquarium / 588-Reptile / \n", "66-Bollywood / \n", "14-Guitar / 63-Acoustic guitar / 1231-Mandolin / \n", "14-Guitar / 24-String instrument / \n", "40-Road / 492-Floor / \n", "983-Lexus / \n", "36-Piano / 53-Keyboard / 107-Musical keyboard / 265-Electronic keyboard / 592-Synthesia / \n", "10-Animal / 86-Plant / 136-Gardening / 174-Farm / 279-Garden / 363-Soil / 2138-Compost / \n", "1-Vehicle / 67-Cycling / 69-Bicycle / 113-House / 351-Jumping / 585-Lawn / \n", "0-Games / 51-Strategy video game / \n", "1-Vehicle / 18-Outdoor recreation / 87-Boat / 147-River / 225-Lake / 949-Rowing / 1090-Canoe / 1305-Canoeing / \n", "1076-Wakeboarding / \n", "0-Games / 2-Video game / 19-PC game / 33-Weapon / 43-Call of Duty / 96-Soldier / 111-PlayStation 3 / 127-Xbox 360 / 216-Call of Duty: Modern Warfare 3 / \n", "12-Food / 88-Machine / \n", "5-Dance / 16-Performance art / \n", "5-Dance / 31-Disc jockey / 98-Festival / \n", "28-Fashion / \n", "205-Newscaster / 2762-CBS News / \n", "130-Gymnastics / \n", "5-Dance / \n", "1035-Bank / \n", "0-Games / 2-Video game / 6-Animation / 89-Comics / 123-Naruto / 591-Naruto: Ultimate Ninja / \n", "36-Piano / 53-Keyboard / 104-Pianist / 107-Musical keyboard / 265-Electronic keyboard / \n", "28-Fashion / 206-Shoe / 333-Nike; Inc. / 368-Sneakers / \n", "9-Music video / \n", "46-Choir / \n", "1-Vehicle / 4-Car / 11-Motorsport / 22-Nature / 86-Plant / 154-Wheel / 158-Tire / 167-Four-wheel drive / 198-Off-road vehicle / 327-Mud / 360-Jeep / 565-Mud bogging / 1406-Jeep Cherokee / 2925-Jeep Cherokee (XJ) / \n", "1-Vehicle / 4-Car / 227-Water / \n", "0-Games / 1-Vehicle / 2-Video game / 34-Action-adventure game / 118-Grand Theft Auto V / \n", "18-Outdoor recreation / 22-Nature / 347-Climbing / 828-Rock climbing / \n", "10-Animal / 18-Outdoor recreation / 80-Horse / 289-Dressage / 1434-Saddle / \n", "36-Piano / \n", "9-Music video / \n", "3-Concert / 7-Musician / 13-Musical ensemble / \n", "31-Disc jockey / \n", "3-Concert / 7-Musician / 13-Musical ensemble / 14-Guitar / 670-Banjo / \n", "0-Games / 2-Video game / \n", "766-President of the United States / \n", "14-Guitar / 24-String instrument / 100-Electric guitar / \n", "1-Vehicle / 18-Outdoor recreation / 67-Cycling / 69-Bicycle / 84-Snow / 220-Mountain bike / 284-Mountain biking / 669-Extreme sport / \n", "0-Games / 6-Animation / 15-Cartoon / 35-Minecraft / 2312-SpongeBob SquarePants / \n", "6-Animation / 10-Animal / 15-Cartoon / 440-Lion / 1156-Simba / \n", "6-Animation / \n", "93-Comedy / \n", "0-Games / 562-Pitcher / \n", "504-Web page / 819-World Wide Web / \n", "76-Ball / 772-Bowling / 930-Ten-pin bowling / 1217-Bowling ball / \n", "23-Mobile phone / 29-Smartphone / \n", "232-Cake / \n", "0-Games / 2-Video game / 34-Action-adventure game / 118-Grand Theft Auto V / \n", "0-Games / 10-Animal / 48-Pet / \n", "6-Animation / 15-Cartoon / 1617-Fungus / \n", "273-News program / \n", "12-Food / 26-Cooking / 122-Weight training / 215-Eating / 301-Meal / \n", "1-Vehicle / 11-Motorsport / 27-Motorcycle / 40-Road / 94-Dashcam / 105-Motorcycling / \n", "3-Concert / 5-Dance / 13-Musical ensemble / \n", "0-Games / 8-Football / \n", "0-Games / 326-Metin2 / \n", "12-Food / 26-Cooking / 32-Recipe / 52-Dish / 58-Cuisine / 184-Vegetable / 1295-Stir frying / 2397-Cabbage / 2614-Lettuce / \n", "3-Concert / 7-Musician / \n", "31-Disc jockey / \n", "6-Animation / 25-Toy / 298-Television advertisement / 507-Thomas the Tank Engine / 2332-Percy / \n", "2-Video game / 55-Video game console / 292-Handheld game console / 320-Super Smash Bros. / 782-Super Smash Bros. Melee / \n", "0-Games / 2-Video game / 33-Weapon / 734-DayZ / 1216-DayZ / \n", "66-Bollywood / \n", "1-Vehicle / 50-Aircraft / 102-Building / 451-Cockpit / 472-Boeing 737 / \n", "1-Vehicle / 11-Motorsport / \n", "1-Vehicle / 50-Aircraft / 129-Aviation / 388-Fighter aircraft / 840-Museum / \n", "5-Dance / \n", "1-Vehicle / 158-Tire / \n", "3-Concert / 7-Musician / 13-Musical ensemble / 14-Guitar / 38-Orchestra / \n", "0-Games / 241-Fighting game / 764-The King of Fighters / 2354-The King of Fighters 2002 / \n", "23-Mobile phone / 29-Smartphone / 37-Gadget / 2265-Subway Surfers / \n", "18-Outdoor recreation / 33-Weapon / 137-Tree / 213-Forest / \n", "1-Vehicle / 4-Car / 11-Motorsport / 17-Racing / 74-Truck / 252-Tractor pulling / 375-Pickup truck / 854-Diesel engine / \n", "5-Dance / 16-Performance art / \n", "1-Vehicle / 50-Aircraft / 82-Airplane / 92-Radio-controlled model / 129-Aviation / 149-Model aircraft / 157-Jet aircraft / 160-Radio-controlled aircraft / 502-Sky / \n", "28-Fashion / 78-Wedding / 153-Dress / 190-Bride / 421-Gown / 495-Runway / 590-Wedding dress / \n", "3-Concert / \n", "1151-Temple / \n", "18-Outdoor recreation / 144-Amusement park / 321-Roller coaster / 334-Amusement ride / \n", "93-Comedy / \n", "0-Games / 530-Cue sports / 568-Pool / 994-Cue stick / 1364-Nine-ball / \n", "3-Concert / 5-Dance / \n", "79-American football / \n", "1-Vehicle / 18-Outdoor recreation / 67-Cycling / 69-Bicycle / 520-Bicycle frame / 570-Road bicycle racing / 645-Road bicycle / \n", "1-Vehicle / 4-Car / 84-Snow / 120-Winter / 245-BMW / 1145-Winter storm / \n", "1-Vehicle / 4-Car / 74-Truck / 124-Tractor / 158-Tire / 167-Four-wheel drive / 249-Heavy equipment / 327-Mud / \n", "20-Trailer / \n", "0-Games / 19-PC game / 35-Minecraft / 48-Pet / 366-Dragon / \n", "0-Games / 8-Football / 21-Stadium / 65-Kick / 76-Ball / \n", "1-Vehicle / 4-Car / 11-Motorsport / 17-Racing / 41-Sports car / 403-Toyota / 3416-Toyota MR2 / \n", "47-Personal computer / 145-Tablet computer / 189-iPad / \n", "66-Bollywood / \n", "0-Games / 2-Video game / 47-Personal computer / 131-Computer / 957-Desktop computer / \n", "309-Computer hardware / 908-Video card / \n", "3-Concert / \n", "3-Concert / 16-Performance art / 95-Talent show / \n", "0-Games / 2-Video game / 19-PC game / 1392-Injustice: Gods Among Us / \n", "28-Fashion / \n", "6-Animation / 849-Tales / \n", "1-Vehicle / 62-Train / 67-Cycling / 69-Bicycle / \n", "0-Games / 35-Minecraft / 464-First-person Shooter / \n", "1-Vehicle / 11-Motorsport / \n", "9-Music video / \n", "0-Games / 60-Basketball / 142-Hockey / \n", "14-Guitar / 100-Electric guitar / 1103-Gibson Les Paul / \n", "1-Vehicle / 87-Boat / 579-Yacht / \n", "1-Vehicle / 4-Car / 11-Motorsport / 17-Racing / \n", "12-Food / 1151-Temple / 3039-Wat / \n", "15-Cartoon / \n", "144-Amusement park / 321-Roller coaster / 334-Amusement ride / 1293-RollerCoaster Tycoon 3 / \n", "79-American football / \n", "8-Football / \n", "5-Dance / 490-Rain / \n", "5-Dance / 16-Performance art / 115-Ballet / 191-Cheerleading / \n", "1-Vehicle / 154-Wheel / 249-Heavy equipment / 1093-Loader / \n", "237-Final Fantasy / 1706-Final Fantasy X-2 / \n", "1-Vehicle / 4-Car / 11-Motorsport / 17-Racing / 25-Toy / 92-Radio-controlled model / 202-Radio-controlled car / 926-Traxxas / \n", "0-Games / 2-Video game / 19-PC game / 229-Halo / 470-Halo 3 / \n", "8-Football / 79-American football / \n", "0-Games / 2-Video game / 34-Action-adventure game / 685-Luigi / 3506-Luigi's Mansion / \n", "59-Winter sport / 84-Snow / 177-Skiing / 305-Ski / 789-Alpine skiing / 1830-Downhill / \n", "2-Video game / 43-Call of Duty / 111-PlayStation 3 / 127-Xbox 360 / 134-Call of Duty: Black Ops / 138-Call of Duty: Black Ops II / 181-Xbox / 194-Call of Duty: Modern Warfare 2 / 431-Call of Duty: Advanced Warfare / \n", "10-Animal / \n", "0-Games / 2-Video game / 34-Action-adventure game / 320-Super Smash Bros. / 782-Super Smash Bros. Melee / 2165-Ganon / \n", "22-Nature / 259-Fire / 419-Earth / 1058-Volcano / \n", "18-Outdoor recreation / 283-Rock / 810-Village / \n", "1-Vehicle / 4-Car / 11-Motorsport / 41-Sports car / 952-Chevrolet Corvette / \n", "3-Concert / 7-Musician / 16-Performance art / \n", "84-Snow / 120-Winter / 1145-Winter storm / \n", "56-Hair / 106-Hairstyle / 607-Lace / 1886-Bob cut / \n", "6-Animation / 61-Art / 97-Drawing / 148-Painting / 528-Pencil / 628-Portrait / \n", "9-Music video / \n", "3-Concert / 7-Musician / 16-Performance art / \n", "743-Edward Cullen / \n", "0-Games / 613-Bar / 3538-S.T.A.L.K.E.R.: Call of Pripyat / \n", "30-Drummer / 39-Drums / 44-Drums / 128-Cymbal / 146-Snare drum / 1127-Drum stick / \n", "10-Animal / 48-Pet / 71-Dog / 182-Puppy / 2097-Dachshund / \n", "5-Dance / 9-Music video / \n", "12-Food / 1596-Cinnamon / \n", "99-Christmas / 729-Christmas decoration / 964-Christmas tree / \n", "1-Vehicle / 50-Aircraft / 388-Fighter aircraft / 1483-United States Air Force / \n", "0-Games / 33-Weapon / 257-Counter-Strike / \n", "3-Concert / 5-Dance / 3804-Music of Eritrea / \n", "46-Choir / \n", "0-Games / 8-Football / 21-Stadium / 65-Kick / \n", "3-Concert / 7-Musician / 13-Musical ensemble / \n", "31-Disc jockey / \n", "8-Football / 79-American football / \n", "1-Vehicle / 4-Car / \n", "1-Vehicle / 4-Car / 11-Motorsport / 17-Racing / 40-Road / 41-Sports car / 94-Dashcam / 378-Drag racing / \n", "219-Accordion / \n", "3-Concert / 16-Performance art / \n", "0-Games / 1876-Beatmania IIDX / \n", "14-Guitar / 25-Toy / 1154-Anpanman / \n", "9-Music video / \n", "0-Games / 8-Football / 90-Sports game / 180-FIFA 15 / \n", "1-Vehicle / 4-Car / 11-Motorsport / 17-Racing / 41-Sports car / 57-Race track / 378-Drag racing / \n", "6-Animation / 15-Cartoon / \n", "3-Concert / \n", "3-Concert / 7-Musician / 30-Drummer / 39-Drums / \n", "3-Concert / 5-Dance / \n", "8-Football / \n", "1-Vehicle / 83-Skateboarding / 108-Skateboard / 555-Longboard / \n", "1-Vehicle / 11-Motorsport / 20-Trailer / 27-Motorcycle / \n", "0-Games / 2-Video game / 637-Five Nights at Freddy's / 1027-Terraria / \n", "8-Football / \n", "1142-Laundry / \n", "5-Dance / \n", "0-Games / 2-Video game / 43-Call of Duty / 134-Call of Duty: Black Ops / \n", "0-Games / 60-Basketball / \n", "1-Vehicle / 4-Car / 11-Motorsport / 17-Racing / 41-Sports car / 70-Driving / \n", "1-Vehicle / 18-Outdoor recreation / 42-Fishing / \n", "8-Football / \n", "61-Art / 479-Graffiti / 2853-Aerosol paint / \n", "7-Musician / 219-Accordion / 655-Diatonic button accordion / \n", "7-Musician / 9-Music video / 14-Guitar / \n", "1-Vehicle / 11-Motorsport / 27-Motorcycle / 197-Trail / 327-Mud / 491-Enduro / \n", "18-Outdoor recreation / 22-Nature / 86-Plant / 137-Tree / 722-Waterfall / \n", "12-Food / 26-Cooking / 32-Recipe / 58-Cuisine / 239-Roasting / 274-Meat / 391-Chicken meat / 589-Grilling / 620-Barbecue / 1085-Barbecue grill / \n", "10-Animal / 22-Nature / 137-Tree / 213-Forest / 1108-Mushroom / 1617-Fungus / \n", "12-Food / \n", "0-Games / 6-Animation / 15-Cartoon / 877-Hatsune Miku: Project DIVA / \n", "2649-Nativity scene / \n", "3-Concert / \n", "36-Piano / 53-Keyboard / 104-Pianist / 107-Musical keyboard / \n", "3-Concert / \n", "1-Vehicle / 11-Motorsport / 92-Radio-controlled model / 149-Model aircraft / 160-Radio-controlled aircraft / 278-GoPro / 393-Unmanned aerial vehicle / \n", "18-Outdoor recreation / 94-Dashcam / 144-Amusement park / 278-GoPro / 321-Roller coaster / 334-Amusement ride / \n", "2273-Polo / \n", "0-Games / 2-Video game / 20-Trailer / 34-Action-adventure game / 111-PlayStation 3 / 127-Xbox 360 / 380-Assassin's Creed / 1048-Assassin's Creed / 1140-Assassin's Creed III / 1435-Ezio Auditore da Firenze / 1745-Assassin's Creed: Revelations / 2554-F.E.A.R. / \n", "12-Food / 968-Pie / \n", "1-Vehicle / 8-Football / \n", "1-Vehicle / 50-Aircraft / 92-Radio-controlled model / 149-Model aircraft / 160-Radio-controlled aircraft / 266-Helicopter / 392-Radio-controlled helicopter / 1257-Turbine / \n", "5-Dance / \n", "1-Vehicle / 4-Car / 41-Sports car / 154-Wheel / 245-BMW / 493-Rim / \n", "5-Dance / 16-Performance art / \n", "1-Vehicle / 4-Car / 25-Toy / 200-Doll / 217-Hunting / 269-Comic book / 290-Pony / 343-My Little Pony / \n", "0-Games / 2-Video game / 51-Strategy video game / 603-StarCraft II: Wings of Liberty / 4353-Sarah Kerrigan / \n", "1-Vehicle / 4-Car / 270-Sedan / 864-Speedometer / 983-Lexus / \n", "12-Food / 26-Cooking / 32-Recipe / 52-Dish / 391-Chicken meat / 424-Chicken / 500-Rice / \n", "1-Vehicle / 50-Aircraft / 82-Airplane / 129-Aviation / 139-Airport / 151-Landing / 157-Jet aircraft / 175-Airliner / 187-Takeoff / 188-Airline / 254-Runway / 471-Wing / \n", "0-Games / 8-Football / 21-Stadium / 77-Arena / \n", "2424-Greeting card / \n", "18-Outdoor recreation / 117-Boxing / \n", "78-Wedding / \n", "35-Minecraft / 1007-Portal / \n", "30-Drummer / 39-Drums / 44-Drums / \n", "0-Games / 8-Football / 54-Highlight film / 79-American football / \n", "6-Animation / 15-Cartoon / 1534-H&M / \n", "8-Football / 358-Rugby football / \n", "0-Games / 43-Call of Duty / 138-Call of Duty: Black Ops II / \n", "121-Photography / \n", "3-Concert / 7-Musician / \n", "2-Video game / 55-Video game console / 111-PlayStation 3 / \n", "810-Village / \n", "293-Slide show / \n", "3-Concert / 7-Musician / 14-Guitar / \n", "1-Vehicle / 4-Car / 94-Dashcam / 915-Town / \n", "0-Games / 1000-AdventureQuest Worlds / \n", "9-Music video / \n", "12-Food / 533-Oil / 640-Bottle / 1769-Olive / 3027-Olive oil / \n", "30-Drummer / 39-Drums / 44-Drums / \n", "60-Basketball / \n", "14-Guitar / \n", "45-Cosmetics / 173-Eye shadow / 193-Eye liner / 195-Mascara / 196-Eye / 218-Eyelash / 291-Rouge / \n", "9-Music video / \n", "28-Fashion / 4200-London Fashion Week / \n", "7-Musician / 14-Guitar / 24-String instrument / 219-Accordion / 2744-Bajo sexto / \n", "0-Games / 2-Video game / 19-PC game / 464-First-person Shooter / 2460-Unreal / 3624-Quake III Arena / 3906-Unreal Tournament / \n", "1-Vehicle / 4-Car / 124-Tractor / 156-Agriculture / 174-Farm / 523-Cattle / \n", "1-Vehicle / 18-Outdoor recreation / 59-Winter sport / 84-Snow / 120-Winter / 610-Snowmobile / 1201-Ski-Doo / \n", "5-Dance / 16-Performance art / 98-Festival / \n", "0-Games / 23-Mobile phone / 29-Smartphone / 35-Minecraft / \n", "319-Robot / \n", "1-Vehicle / 4-Car / 1391-Kia / 1415-Kia Motors / \n", "12-Food / \n", "23-Mobile phone / 29-Smartphone / 37-Gadget / 101-Telephone / 178-Samsung Galaxy / 487-Samsung Electronics / 899-Samsung Galaxy S III / \n", "23-Mobile phone / 29-Smartphone / 101-Telephone / \n", "5-Dance / \n", "31-Disc jockey / 559-Night / \n", "2-Video game / 380-Assassin's Creed / 1048-Assassin's Creed / 1522-Assassin's Creed II / \n", "20-Trailer / \n", "14-Guitar / 63-Acoustic guitar / \n", "22-Nature / \n", "10-Animal / 91-Fish / 258-Aquarium / \n", "0-Games / 2-Video game / 37-Gadget / 55-Video game console / 292-Handheld game console / 847-Angry Birds / \n", "2668-Duke Nukem / 3554-Duke Nukem Forever / \n", "28-Fashion / \n", "310-The Sims / 543-The Sims 2 / \n", "3-Concert / 7-Musician / \n", "122-Weight training / 204-Gym / 214-Muscle / 1389-Dumbbell / \n", "1-Vehicle / 4-Car / 27-Motorcycle / 40-Road / 105-Motorcycling / \n", "1-Vehicle / 22-Nature / 74-Truck / \n", "0-Games / 2-Video game / 505-Mega Man / 1664-Zero / 3035-Mega Man Zero / \n", "110-Album / \n", "36-Piano / \n", "1-Vehicle / 4-Car / \n", "1-Vehicle / 11-Motorsport / 17-Racing / 27-Motorcycle / 172-Motocross / \n", "1-Vehicle / 4-Car / 473-Ford Mustang / \n", "122-Weight training / 204-Gym / 532-Barbell / \n", "60-Basketball / \n", "3-Concert / 7-Musician / 13-Musical ensemble / \n", "0-Games / 326-Metin2 / \n", "0-Games / 2-Video game / 35-Minecraft / \n", "8-Football / \n", "0-Games / 15-Cartoon / 1650-The Idolmaster / 1755-The Idolmaster / \n", "3-Concert / 5-Dance / 16-Performance art / 350-Circus / \n", "1-Vehicle / 50-Aircraft / 82-Airplane / 129-Aviation / 139-Airport / 151-Landing / 157-Jet aircraft / 175-Airliner / 187-Takeoff / 188-Airline / 254-Runway / 472-Boeing 737 / 3433-Southwest Airlines / \n", "3-Concert / \n", "23-Mobile phone / 29-Smartphone / 37-Gadget / 72-iPhone / 101-Telephone / 554-iPhone 4 / \n", "10-Animal / 174-Farm / 523-Cattle / 534-Milk / 4055-Milking / \n", "205-Newscaster / \n", "0-Games / 2-Video game / 43-Call of Duty / 302-Call of Duty 4: Modern Warfare / \n", "12-Food / 26-Cooking / 32-Recipe / \n", "1-Vehicle / 4-Car / 74-Truck / \n", "31-Disc jockey / 39-Drums / \n", "3-Concert / \n", "14-Guitar / 24-String instrument / 63-Acoustic guitar / \n", "0-Games / 8-Football / 21-Stadium / 54-Highlight film / 65-Kick / 73-Ball / 76-Ball / 180-FIFA 15 / \n", "5-Dance / \n", "0-Games / 2-Video game / 19-PC game / 33-Weapon / 96-Soldier / 1543-Crysis / 2854-Stealth game / \n", "12-Food / 26-Cooking / 32-Recipe / 52-Dish / 348-Fruit / 968-Pie / 1319-Apple / 3613-Apple pie / \n", "3-Concert / 16-Performance art / \n", "1-Vehicle / 4-Car / \n", "3-Concert / 5-Dance / \n", "5-Dance / 16-Performance art / \n", "0-Games / 2-Video game / 19-PC game / 33-Weapon / 43-Call of Duty / 194-Call of Duty: Modern Warfare 2 / 216-Call of Duty: Modern Warfare 3 / \n", "18-Outdoor recreation / 83-Skateboarding / 221-Skatepark / 703-Roller skating / 1477-Aggressive inline skating / 2747-Inline skating / \n", "2-Video game / 43-Call of Duty / 704-Call of Duty: World at War / \n", "99-Christmas / \n", "130-Gymnastics / \n", "20-Trailer / 843-The Doctor / \n", "0-Games / 2-Video game / 19-PC game / 55-Video game console / 320-Super Smash Bros. / 2165-Ganon / \n", "1-Vehicle / 87-Boat / 444-Family / 753-Sailboat / \n", "12-Food / \n", "9-Music video / \n", "0-Games / 8-Football / \n", "75-Wrestling / 85-Combat / 117-Boxing / \n", "3-Concert / 1854-Centella / \n", "0-Games / 60-Basketball / 76-Ball / 168-Basketball moves / 623-Coach / 3291-Basketball Coach / \n", "3-Concert / 21-Stadium / 46-Choir / 260-Prayer / \n", "1-Vehicle / 4-Car / 11-Motorsport / 17-Racing / 41-Sports car / 403-Toyota / \n", "0-Games / 8-Football / 21-Stadium / \n", "9-Music video / 28-Fashion / \n", "0-Games / 2-Video game / 1638-Soulcalibur / 2877-Soulcalibur IV / \n", "14-Guitar / 22-Nature / 156-Agriculture / 267-Mountain / \n", "0-Games / 43-Call of Duty / 138-Call of Duty: Black Ops II / \n", "10-Animal / 80-Horse / 384-Horse racing / 438-Mare / \n", "10-Animal / 48-Pet / 71-Dog / \n", "3-Concert / \n", "9-Music video / \n", "0-Games / 21-Stadium / 54-Highlight film / 81-Athlete / 562-Pitcher / \n", "49-School / \n", "9-Music video / \n", "1-Vehicle / 11-Motorsport / 124-Tractor / 156-Agriculture / 249-Heavy equipment / 252-Tractor pulling / 363-Soil / 459-Plough / 618-Fiat Automobiles / \n", "1-Vehicle / 4-Car / 25-Toy / \n", "0-Games / 23-Mobile phone / 29-Smartphone / 1113-Toilet / \n", "61-Art / 148-Painting / \n", "1-Vehicle / 27-Motorcycle / 608-Wheelie / \n", "93-Comedy / \n", "0-Games / 2-Video game / 6-Animation / 241-Fighting game / 514-Tekken / 797-Tekken / 1470-Tekken Tag Tournament 2 / 4292-Heihachi Mishima / \n", "1-Vehicle / 4-Car / 70-Driving / 118-Grand Theft Auto V / 297-Grand Theft Auto IV / 3074-Porsche Cayenne / \n", "4501-The Adventures of Tintin / \n", "9-Music video / \n", "10-Animal / 80-Horse / 247-Stallion / \n", "36-Piano / 53-Keyboard / 104-Pianist / 107-Musical keyboard / 1303-Fullmetal Alchemist / \n", "0-Games / 2-Video game / 8-Football / 90-Sports game / 180-FIFA 15 / 480-FIFA 13 / \n", "1-Vehicle / 4-Car / 337-Advertising / 663-Street / \n", "863-Body piercing / \n", "0-Games / 2-Video game / 43-Call of Duty / 111-PlayStation 3 / 127-Xbox 360 / 194-Call of Duty: Modern Warfare 2 / 216-Call of Duty: Modern Warfare 3 / \n", "6-Animation / \n", "0-Games / 2-Video game / 72-iPhone / 189-iPad / 992-Rocket / \n", "12-Food / 26-Cooking / 367-Drink / 904-Cocktail / 4295-Mojito / \n", "9-Music video / \n", "0-Games / 2-Video game / 25-Toy / 140-LEGO / 243-Sonic the Hedgehog / 251-Sonic the Hedgehog / \n", "14-Guitar / 36-Piano / \n", "61-Art / 148-Painting / \n", "10-Animal / 22-Nature / \n", "0-Games / 1-Vehicle / 2-Video game / 19-PC game / 34-Action-adventure game / 201-Grand Theft Auto: San Andreas / 858-Carl Johnson / \n", "31-Disc jockey / \n", "1-Vehicle / 11-Motorsport / 17-Racing / 27-Motorcycle / 172-Motocross / 669-Extreme sport / \n", "1-Vehicle / 4-Car / 11-Motorsport / 40-Road / 64-Transport / 286-Bus / \n", "96-Soldier / \n", "1-Vehicle / 50-Aircraft / 82-Airplane / 175-Airliner / \n", "20-Trailer / \n", "3-Concert / 804-Shopping mall / \n", "2-Video game / 257-Counter-Strike / \n", "0-Games / 78-Wedding / 271-Tank / 1167-Tanki Online / \n", "0-Games / 6-Animation / 15-Cartoon / 3492-Gnome / \n", "10-Animal / 80-Horse / \n", "3-Concert / \n", "28-Fashion / 3063-Friendship bracelet / \n", "8-Football / \n", "5-Dance / \n", "238-Nail / 300-Nail art / 338-Manicure / \n", "6-Animation / 2489-Sheriff Woody / \n", "56-Hair / \n", "3-Concert / 7-Musician / 13-Musical ensemble / 30-Drummer / \n", "337-Advertising / 776-Kindergarten / \n", "1-Vehicle / 4-Car / 11-Motorsport / 64-Transport / 124-Tractor / 252-Tractor pulling / \n", "1-Vehicle / 4-Car / 349-Chevrolet / 406-Chevrolet / 761-Chevrolet Camaro / 864-Speedometer / 982-V8 engine / \n", "1-Vehicle / 4-Car / 88-Machine / 114-Engine / 154-Wheel / \n", "122-Weight training / 214-Muscle / \n", "12-Food / \n", "3-Concert / 7-Musician / 14-Guitar / 24-String instrument / \n", "5-Dance / 2119-Splatter film / \n", "7-Musician / 14-Guitar / 63-Acoustic guitar / 488-Door / \n", "22-Nature / 28-Fashion / 206-Shoe / 333-Nike; Inc. / 368-Sneakers / 985-Strawberry / 1423-Nike Air Max / \n", "3-Concert / \n", "25-Toy / 42-Fishing / 140-LEGO / \n", "5-Dance / 13-Musical ensemble / 98-Festival / 203-Marching band / 250-Parade / \n", "0-Games / 1-Vehicle / 4-Car / 6-Animation / 15-Cartoon / 25-Toy / 74-Truck / \n", "31-Disc jockey / \n", "3-Concert / \n", "1-Vehicle / 4-Car / 11-Motorsport / 17-Racing / 41-Sports car / 1056-Porsche 911 / \n", "67-Cycling / 69-Bicycle / 154-Wheel / 519-Recreational vehicle / \n", "1-Vehicle / 4-Car / 11-Motorsport / 17-Racing / 57-Race track / \n", "1-Vehicle / 4-Car / 74-Truck / \n", "8-Football / \n", "0-Games / 8-Football / 21-Stadium / 81-Athlete / \n", "6-Animation / 15-Cartoon / 111-PlayStation 3 / 4286-Atelier / \n", "9-Music video / \n", "0-Games / \n", "3-Concert / 7-Musician / 14-Guitar / \n", "133-Nightclub / \n", "0-Games / 1852-Final Fantasy XI / \n", "1-Vehicle / 4-Car / 23-Mobile phone / 29-Smartphone / \n", "3-Concert / \n", "12-Food / 26-Cooking / 32-Recipe / 52-Dish / 58-Cuisine / 210-Cookware and bakeware / 597-Soup / 1381-Garlic / \n", "1-Vehicle / 11-Motorsport / 17-Racing / \n", "1-Vehicle / 20-Trailer / 50-Aircraft / 766-President of the United States / \n", "9-Music video / 15-Cartoon / 524-Littlest Pet Shop / \n", "1484-Final Fantasy VIII / \n", "260-Prayer / \n", "342-Windows Media Video / 366-Dragon / \n", "1-Vehicle / 1699-Cart / 3573-Golf cart / \n", "12-Food / 26-Cooking / 52-Dish / 58-Cuisine / 215-Eating / 301-Meal / 389-Restaurant / \n", "6-Animation / 15-Cartoon / 1724-Homer Simpson / \n", "22-Nature / 46-Choir / \n", "3-Concert / 7-Musician / 14-Guitar / \n", "0-Games / 19-PC game / 51-Strategy video game / 373-Clash of Clans / \n", "12-Food / 3250-Chubby Bunny / \n", "1669-Vodka / \n", "0-Games / 8-Football / 21-Stadium / 54-Highlight film / 65-Kick / 73-Ball / 76-Ball / 81-Athlete / \n", "10-Animal / 18-Outdoor recreation / 80-Horse / 247-Stallion / 289-Dressage / 438-Mare / \n", "23-Mobile phone / 29-Smartphone / 37-Gadget / 72-iPhone / 331-iPod touch / \n", "36-Piano / 53-Keyboard / 107-Musical keyboard / 265-Electronic keyboard / 592-Synthesia / \n", "1-Vehicle / 4-Car / 1243-Mirror / \n", "125-Television / \n", "22-Nature / 59-Winter sport / 177-Skiing / 267-Mountain / 347-Climbing / \n", "110-Album / \n", "2155-Unicorn / \n", "1137-Crash Bandicoot / 4629-Crash Twinsanity / \n", "6-Animation / 15-Cartoon / 877-Hatsune Miku: Project DIVA / 1648-Hatsune Miku: Project DIVA F / \n", "5-Dance / 78-Wedding / 190-Bride / 889-Swing / \n", "0-Games / 2-Video game / 34-Action-adventure game / 118-Grand Theft Auto V / \n", "2-Video game / 23-Mobile phone / 29-Smartphone / 37-Gadget / 101-Telephone / \n", "0-Games / 3512-Just Dance 2015 / \n", "9-Music video / 31-Disc jockey / \n", "3-Concert / 7-Musician / 16-Performance art / 30-Drummer / \n", "3-Concert / 7-Musician / \n", "9-Music video / \n", "9-Music video / \n", "23-Mobile phone / 29-Smartphone / 37-Gadget / 101-Telephone / \n", "0-Games / 2-Video game / 19-PC game / 734-DayZ / \n", "61-Art / 148-Painting / 303-Paint / 873-Acrylic paint / 936-Watercolor paint / \n", "93-Comedy / 355-Comedian / \n", "273-News program / 913-President / \n", "9-Music video / \n", "1-Vehicle / 4-Car / 11-Motorsport / 327-Mud / 853-Dune buggy / \n", "28-Fashion / 327-Mud / 757-Boot / 785-Jeans / 3505-Wellington boot / \n", "39-Drums / 724-Rock Band / \n", "1-Vehicle / 87-Boat / 1643-Submarine / \n", "1-Vehicle / 4-Car / 27-Motorcycle / 40-Road / \n", "1-Vehicle / 4-Car / 1511-Toyota Land Cruiser / \n", "12-Food / \n", "6-Animation / 15-Cartoon / 2176-Fist of the North Star / 3349-Kenshiro / \n", "3-Concert / 7-Musician / 14-Guitar / 24-String instrument / 63-Acoustic guitar / \n", "12-Food / 26-Cooking / 389-Restaurant / \n", "6-Animation / 701-Beyblade / \n", "10-Animal / 48-Pet / 71-Dog / 488-Door / \n", "1229-Magnet / \n", "14-Guitar / 24-String instrument / 63-Acoustic guitar / \n", "237-Final Fantasy / \n", "3-Concert / 16-Performance art / \n", "0-Games / 2-Video game / 33-Weapon / 43-Call of Duty / 422-Call of Duty: Ghosts / \n", "10-Animal / 48-Pet / 71-Dog / \n", "130-Gymnastics / \n", "1-Vehicle / 11-Motorsport / 17-Racing / 21-Stadium / 57-Race track / 77-Arena / \n", "68-Lighting / 263-Light / 2703-Light fixture / \n", "6-Animation / 23-Mobile phone / 29-Smartphone / \n", "1-Vehicle / 4-Car / 67-Cycling / 69-Bicycle / 154-Wheel / 520-Bicycle frame / \n", "2224-Sherlock Holmes / \n", "49-School / \n", "10-Animal / 71-Dog / \n", "5-Dance / 16-Performance art / 95-Talent show / \n", "8-Football / 205-Newscaster / \n", "5-Dance / 115-Ballet / \n", "248-Hotel / \n", "1975-Amiga / \n", "3-Concert / 16-Performance art / \n", "15-Cartoon / 99-Christmas / 2047-Popeye / \n", "1-Vehicle / 50-Aircraft / 121-Photography / 1955-Letter / \n", "6-Animation / 15-Cartoon / 183-Manga / 396-One Piece / \n", "93-Comedy / \n", "1-Vehicle / 4-Car / 201-Grand Theft Auto: San Andreas / \n", "42-Fishing / 293-Slide show / \n", "5-Dance / 16-Performance art / \n", "20-Trailer / 126-The Walt Disney Company / \n", "3-Concert / 7-Musician / 16-Performance art / 30-Drummer / \n", "130-Gymnastics / \n", "0-Games / 2-Video game / 320-Super Smash Bros. / 584-Super Mario Bros. / 685-Luigi / 887-King Koopa / 3072-Super Smash Bros. / \n", "0-Games / 54-Highlight film / 256-Tennis / \n", "3-Concert / \n", "45-Cosmetics / 2184-Lush / \n", "10-Animal / 22-Nature / 132-Bird / \n", "9-Music video / \n", "0-Games / 2-Video game / 19-PC game / 51-Strategy video game / 1072-Total War / 2674-Rome: Total War / \n", "12-Food / 28-Fashion / 45-Cosmetics / 282-Skin / 385-Face / 784-Honey / 3404-Oatmeal / \n", "6-Animation / 15-Cartoon / 4714-Air Gear / \n", "1-Vehicle / 33-Weapon / 271-Tank / \n", "14-Guitar / 36-Piano / 53-Keyboard / 107-Musical keyboard / 150-Violin / \n", "0-Games / 2-Video game / 19-PC game / 43-Call of Duty / 134-Call of Duty: Black Ops / 138-Call of Duty: Black Ops II / \n", "2-Video game / 320-Super Smash Bros. / \n", "0-Games / 1-Vehicle / 2-Video game / 35-Minecraft / 87-Boat / 308-Ship / \n", "0-Games / 88-Machine / 240-Arcade game / 3591-Pump It Up Fiesta 2 / \n", "9-Music video / \n", "0-Games / 2-Video game / 43-Call of Duty / 138-Call of Duty: Black Ops II / \n", "1-Vehicle / 42-Fishing / 87-Boat / 2755-Lithium polymer battery / \n", "3-Concert / 68-Lighting / \n", "1-Vehicle / 4-Car / 449-Honda / \n", "3-Concert / 7-Musician / 13-Musical ensemble / 46-Choir / \n", "3-Concert / \n", "6-Animation / 15-Cartoon / 61-Art / 1164-Cartoon Network / \n", "0-Games / 10-Animal / 17-Racing / 80-Horse / 384-Horse racing / \n", "0-Games / 2-Video game / 6-Animation / 89-Comics / 396-One Piece / 1312-Dynasty Warriors / 4203-One Piece: Pirate Warriors / \n", "1-Vehicle / 4-Car / 11-Motorsport / 406-Chevrolet / 952-Chevrolet Corvette / \n", "6-Animation / 97-Drawing / 183-Manga / 408-Sketch / 1674-Ouran High School Host Club / \n", "28-Fashion / 45-Cosmetics / 56-Hair / 106-Hairstyle / 193-Eye liner / \n", "47-Personal computer / 145-Tablet computer / 189-iPad / 525-Printer / 2095-Cash / 2913-Drawer / \n", "93-Comedy / 355-Comedian / \n", "1-Vehicle / 4-Car / 245-BMW / 270-Sedan / 1233-BMW 3 Series / 2130-Car dealership / \n", "45-Cosmetics / 56-Hair / 291-Rouge / \n", "1507-Calligraphy / 4460-Chinese calligraphy / \n", "12-Food / 26-Cooking / 32-Recipe / 52-Dish / 58-Cuisine / 236-Baking / 442-Bread / 485-Dough / 600-Flour / 662-Pizza / 2728-Yeast / \n", "1-Vehicle / 18-Outdoor recreation / 42-Fishing / 548-Parachuting / 669-Extreme sport / 765-Parachute / \n", "0-Games / 2-Video game / 35-Minecraft / 84-Snow / \n", "5-Dance / 16-Performance art / \n", "3-Concert / 7-Musician / 13-Musical ensemble / 46-Choir / 318-Church / 397-Organ / \n", "694-Tai chi / \n", "9-Music video / \n", "2019-Dental braces / \n", "3-Concert / 16-Performance art / 68-Lighting / \n", "0-Games / 2-Video game / 90-Sports game / 180-FIFA 15 / \n", "1-Vehicle / 4-Car / 70-Driving / 3439-Compact car / \n", "28-Fashion / 206-Shoe / 333-Nike; Inc. / 368-Sneakers / \n", "9-Music video / \n", "5-Dance / 16-Performance art / \n", "639-Fiddle / \n", "60-Basketball / \n", "25-Toy / 435-Action figure / \n", "7-Musician / 230-Brass instrument / 1574-French horn / \n", "0-Games / 489-Board game / 1966-The Lord of the Rings / \n", "0-Games / 756-Tibia / \n", "201-Grand Theft Auto: San Andreas / \n", "30-Drummer / 39-Drums / 44-Drums / 128-Cymbal / 146-Snare drum / \n", "10-Animal / 48-Pet / 71-Dog / \n", "1-Vehicle / 18-Outdoor recreation / 87-Boat / 114-Engine / 147-River / 225-Lake / 498-Motorboat / \n", "0-Games / 60-Basketball / 77-Arena / 168-Basketball moves / 255-Slam dunk / \n", "9-Music video / \n", "1-Vehicle / 4-Car / 11-Motorsport / 17-Racing / 41-Sports car / 135-Drifting / \n", "1-Vehicle / 4-Car / \n", "1-Vehicle / 4-Car / 13-Musical ensemble / 27-Motorcycle / 1924-Symbol / \n", "12-Food / 587-Clay / 695-Balloon / \n", "3-Concert / 7-Musician / 16-Performance art / 30-Drummer / \n", "155-Camera / 1348-Panasonic Corporation / \n", "6-Animation / 15-Cartoon / 123-Naruto / 310-The Sims / 543-The Sims 2 / 622-Sakura Haruno / \n", "3-Concert / 5-Dance / 7-Musician / \n", "1-Vehicle / 4-Car / 1245-Alarm device / \n", "0-Games / 2-Video game / 34-Action-adventure game / 434-Ninja / \n", "36-Piano / 53-Keyboard / 104-Pianist / 107-Musical keyboard / \n", "47-Personal computer / 145-Tablet computer / \n", "61-Art / 2905-Warhammer Online: Age of Reckoning / \n", "0-Games / 256-Tennis / \n", "0-Games / 59-Winter sport / 81-Athlete / 112-Ice skating / 142-Hockey / 312-Ice rink / \n", "0-Games / 6-Animation / \n", "337-Advertising / \n", "7-Musician / 13-Musical ensemble / 21-Stadium / 49-School / 79-American football / 203-Marching band / 306-High school / \n", "3-Concert / 7-Musician / 13-Musical ensemble / 39-Drums / 44-Drums / \n", "1-Vehicle / 62-Train / 64-Transport / 103-Rail transport / 119-Locomotive / 161-Train station / \n", "2830-Super Mario Sunshine / \n", "1-Vehicle / 4-Car / 234-Ford / 344-Coupé / 406-Chevrolet / 509-Classic car / 1148-Hot rod / \n", "12-Food / 215-Eating / 1701-French fries / \n", "6-Animation / 10-Animal / 15-Cartoon / 132-Bird / 718-Duck / 1445-Donald Duck / \n", "1-Vehicle / 50-Aircraft / 499-Cloud / \n", "0-Games / 2-Video game / 19-PC game / 1630-Warframe / \n", "136-Gardening / 2928-Cereal / \n", "1-Vehicle / 4-Car / 11-Motorsport / 40-Road / 94-Dashcam / 169-Rallying / \n", "22-Nature / 86-Plant / 136-Gardening / \n", "5-Dance / 98-Festival / 250-Parade / 253-Carnival / \n", "12-Food / 88-Machine / 539-Home appliance / 818-Washing machine / 827-Pump / 1456-Clothes dryer / \n", "0-Games / 2-Video game / 211-RuneScape / \n", "8-Football / 54-Highlight film / 65-Kick / \n", "3-Concert / 16-Performance art / 38-Orchestra / \n", "93-Comedy / \n", "1-Vehicle / 4-Car / 1547-Passenger / \n", "36-Piano / 53-Keyboard / 104-Pianist / 107-Musical keyboard / \n", "61-Art / 97-Drawing / 242-Paper / 2162-Marker pen / \n", "122-Weight training / 204-Gym / 481-Kayak / \n", "0-Games / 5-Dance / \n", "122-Weight training / 214-Muscle / 1250-Biceps / \n", "22-Nature / 86-Plant / 136-Gardening / 137-Tree / 1460-Herb / 1969-Shrub / 4644-Vine / \n", "1-Vehicle / 162-Medicine / \n", "3-Concert / 7-Musician / 24-String instrument / \n", "1-Vehicle / 4-Car / 40-Road / 64-Transport / 74-Truck / 482-Asphalt / \n", "3-Concert / 5-Dance / 13-Musical ensemble / 98-Festival / 1289-Caporales / \n", "0-Games / 2-Video game / 19-PC game / 33-Weapon / 989-Warface / \n", "0-Games / 2-Video game / \n", "0-Games / 8-Football / 21-Stadium / 65-Kick / \n", "1-Vehicle / 120-Winter / 262-Ice / 308-Ship / \n", "14-Guitar / \n", "567-Compact disc / \n", "9-Music video / \n", "624-Number / 1507-Calligraphy / \n", "16-Performance art / 78-Wedding / \n", "0-Games / 6-Animation / 319-Robot / \n", "9-Music video / 110-Album / \n", "15-Cartoon / 61-Art / 97-Drawing / 237-Final Fantasy / 408-Sketch / 538-Illustration / \n", "120-Winter / \n", "1-Vehicle / 4-Car / 11-Motorsport / 17-Racing / 40-Road / 41-Sports car / 57-Race track / 169-Rallying / \n", "1-Vehicle / 27-Motorcycle / \n", "0-Games / 35-Minecraft / 246-Furniture / \n", "0-Games / 2-Video game / 6-Animation / 15-Cartoon / 185-Dragon Ball / 244-Goku / 418-Gohan / 869-Dragon Ball Z: Budokai Tenkaichi / \n", "14-Guitar / 162-Medicine / \n", "0-Games / 774-Galaxy / 1397-Super Mario Galaxy / 1759-Super Mario Galaxy 2 / \n", "1-Vehicle / 4-Car / 25-Toy / \n", "1-Vehicle / 11-Motorsport / 27-Motorcycle / 40-Road / 105-Motorcycling / 207-Exhaust system / \n", "6-Animation / 15-Cartoon / \n", "1-Vehicle / 11-Motorsport / 27-Motorcycle / 197-Trail / \n", "9-Music video / \n", "0-Games / 2-Video game / 33-Weapon / 43-Call of Duty / 134-Call of Duty: Black Ops / \n", "3-Concert / 7-Musician / 13-Musical ensemble / 38-Orchestra / 46-Choir / \n", "8-Football / \n", "7-Musician / 14-Guitar / 63-Acoustic guitar / 2039-Soybean / \n", "3-Concert / 7-Musician / 30-Drummer / 39-Drums / 912-Silver / \n", "1-Vehicle / 4-Car / 11-Motorsport / 17-Racing / 40-Road / 57-Race track / 169-Rallying / \n", "3-Concert / 7-Musician / 13-Musical ensemble / 46-Choir / \n", "14-Guitar / \n", "242-Paper / 745-Writing / 882-Pen / 2454-Notebook / \n", "3-Concert / 133-Nightclub / \n", "0-Games / 2-Video game / 19-PC game / \n", "0-Games / 8-Football / 21-Stadium / 54-Highlight film / 65-Kick / \n", "4676-Elphaba / \n", "1-Vehicle / 4-Car / 11-Motorsport / 40-Road / 70-Driving / 94-Dashcam / 169-Rallying / \n", "3-Concert / \n", "38-Orchestra / \n", "0-Games / 2-Video game / 55-Video game console / 3567-SOCOM / \n", "31-Disc jockey / \n", "3-Concert / 133-Nightclub / \n", "12-Food / 26-Cooking / 32-Recipe / 52-Dish / 58-Cuisine / 86-Plant / 109-Cooking show / 184-Vegetable / 210-Cookware and bakeware / 597-Soup / \n", "28-Fashion / \n", "6-Animation / 15-Cartoon / \n", "18-Outdoor recreation / 83-Skateboarding / 108-Skateboard / 556-Skateboarding trick / 1315-Kickflip / \n", "5-Dance / 7-Musician / 16-Performance art / 38-Orchestra / \n", "3-Concert / 7-Musician / 16-Performance art / \n", "467-Chipmunk / \n", "12-Food / 26-Cooking / 32-Recipe / 232-Cake / 236-Baking / 295-Chocolate / 442-Bread / 485-Dough / 560-Sugar / 578-Cookie / 600-Flour / 759-Icing / 2699-Chocolate chip / 3211-Chocolate chip cookie / \n", "5-Dance / 98-Festival / 988-Ibiza / \n", "8-Football / \n", "59-Winter sport / 84-Snow / 281-Snowboarding / 501-Snowboard / \n", "1-Vehicle / 27-Motorcycle / \n", "2-Video game / 229-Halo / 470-Halo 3 / \n", "1-Vehicle / 4-Car / 28-Fashion / \n", "12-Food / 26-Cooking / 32-Recipe / 52-Dish / 58-Cuisine / 274-Meat / 454-Sauce / 752-Pasta / 1230-Italian food / 1517-Spaghetti / 2562-Meatball / \n", "6-Animation / 522-Helmet / \n", "1-Vehicle / 11-Motorsport / 172-Motocross / \n", "0-Games / 6-Animation / 25-Toy / \n", "1-Vehicle / 2-Video game / 6-Animation / \n", "0-Games / 19-PC game / 185-Dragon Ball / 244-Goku / 837-Roblox / \n", "23-Mobile phone / 29-Smartphone / 37-Gadget / 72-iPhone / 101-Telephone / 178-Samsung Galaxy / 802-Google Nexus / 2929-Galaxy Nexus / \n", "3-Concert / 7-Musician / 13-Musical ensemble / 30-Drummer / 39-Drums / \n", "10-Animal / 48-Pet / 132-Bird / 1799-Domestic canary / \n", "3-Concert / \n", "0-Games / 2-Video game / 6-Animation / \n", "12-Food / 26-Cooking / 32-Recipe / 475-Cheese / 534-Milk / 786-Juice / 1081-Lemon / 1193-Orange / \n", "3-Concert / \n", "0-Games / 2-Video game / 118-Grand Theft Auto V / 1393-Lightsaber / \n", "46-Choir / \n", "1-Vehicle / 4-Car / 70-Driving / 299-Volkswagen Passenger Cars / 526-Volkswagen / 1761-Volkswagen Jetta / \n", "9-Music video / 25-Toy / 524-Littlest Pet Shop / \n", "1-Vehicle / 11-Motorsport / 3311-Hovercraft / \n", "14-Guitar / 1346-Wood carving / \n", "0-Games / 2-Video game / 1981-Commodore 64 / \n", "6-Animation / 15-Cartoon / 89-Comics / 123-Naruto / 183-Manga / 1556-Dōjinshi / \n", "3-Concert / 7-Musician / \n", "2-Video game / 23-Mobile phone / 29-Smartphone / 37-Gadget / 47-Personal computer / 72-iPhone / 101-Telephone / 145-Tablet computer / 178-Samsung Galaxy / 1629-Samsung Galaxy Note / 1776-Samsung Galaxy Note II / \n", "0-Games / 2-Video game / 19-PC game / 159-World of Warcraft / 171-Warcraft / 3615-Monk / \n", "2-Video game / 37-Gadget / 55-Video game console / 292-Handheld game console / 335-Nintendo 3DS / \n", "2-Video game / 82-Airplane / 194-Call of Duty: Modern Warfare 2 / 216-Call of Duty: Modern Warfare 3 / 1673-Airport terminal / \n", "7-Musician / 14-Guitar / 63-Acoustic guitar / \n", "0-Games / 2-Video game / 2628-Sly Cooper / \n", "1-Vehicle / 4-Car / 11-Motorsport / 17-Racing / 41-Sports car / 135-Drifting / \n", "31-Disc jockey / \n", "7-Musician / 46-Choir / 1338-Devil / \n", "1-Vehicle / 4-Car / 70-Driving / 1285-Parking / 1314-Garage / \n", "9-Music video / 31-Disc jockey / 133-Nightclub / \n", "1088-Infomercial / 3301-Mop / \n", "287-Human swimming / \n", "2-Video game / 387-Spider-Man / \n", "2-Video game / 243-Sonic the Hedgehog / 251-Sonic the Hedgehog / 1299-Shadow the Hedgehog / 3321-Shadow the Hedgehog / \n", "350-Circus / 788-Juggling / \n", "5-Dance / 16-Performance art / 22-Nature / \n", "356-Star / \n", "12-Food / 26-Cooking / 32-Recipe / 52-Dish / 58-Cuisine / 109-Cooking show / 617-Maize / \n", "0-Games / 2-Video game / 910-Dragon Quest / 4411-Dragon Quest V: Hand of the Heavenly Bride / \n", "1-Vehicle / 27-Motorcycle / 342-Windows Media Video / 608-Wheelie / \n", "14-Guitar / 63-Acoustic guitar / \n", "110-Album / 242-Paper / \n", "6-Animation / 25-Toy / 200-Doll / 412-Monster / 503-Monster High / \n", "1-Vehicle / 4-Car / 74-Truck / 1070-Semi-trailer truck / \n", "8-Football / 76-Ball / 2611-Wiffle ball / \n", "9-Music video / \n", "466-Gift / 2507-Macramé / \n", "22-Nature / 84-Snow / 225-Lake / 267-Mountain / \n", "28-Fashion / \n", "59-Winter sport / 112-Ice skating / 115-Ballet / 316-Figure skating / \n", "0-Games / 165-League of Legends / 296-Saxophone / \n", "0-Games / 8-Football / 73-Ball / \n", "42-Fishing / 222-Surfing / \n", "278-GoPro / \n", "476-Textile / 1403-Flower bouquet / 2329-Felt / \n", "61-Art / 479-Graffiti / 1846-Mural / \n", "452-Watch / 2168-G-Shock / \n", "10-Animal / 80-Horse / 247-Stallion / \n", "0-Games / 2-Video game / 1777-Ratchet / 1824-Clank / 1859-Ratchet & Clank / 4423-Ratchet & Clank / \n", "0-Games / 8-Football / 20-Trailer / 79-American football / 90-Sports game / 561-Madden NFL / 1813-Madden NFL 13 / 2916-Madden NFL 12 / \n", "1-Vehicle / 4-Car / 6-Animation / 15-Cartoon / 74-Truck / 88-Machine / 124-Tractor / 433-Construction / \n", "163-Running / \n", "0-Games / 21-Stadium / 54-Highlight film / 163-Running / \n", "18-Outdoor recreation / 217-Hunting / 706-Deer / \n", "5-Dance / 430-Ballroom dance / 678-Latin dance / \n", "0-Games / 6-Animation / 15-Cartoon / \n", "10-Animal / 48-Pet / 71-Dog / 824-Terrier / 1252-Dog agility / \n", "0-Games / 2-Video game / 55-Video game console / 292-Handheld game console / 398-The Legend of Zelda / \n", "3-Concert / 7-Musician / 14-Guitar / \n", "113-House / 170-Home improvement / 879-Ceiling / \n", "3-Concert / 31-Disc jockey / \n", "60-Basketball / 255-Slam dunk / \n", "5-Dance / 9-Music video / \n", "6-Animation / 15-Cartoon / 1771-The Pink Panther / \n", "66-Bollywood / \n", "122-Weight training / 204-Gym / 1584-Kettlebell / \n", "1-Vehicle / 10-Animal / 18-Outdoor recreation / 42-Fishing / 87-Boat / 91-Fish / 268-Recreational fishing / 698-Fisherman / \n", "925-Brain / \n", "3-Concert / 7-Musician / 13-Musical ensemble / 16-Performance art / 30-Drummer / \n", "222-Surfing / 672-Bodyboarding / 815-Surfboard / \n", "102-Building / 242-Paper / 339-Architecture / 342-Windows Media Video / 1214-Skyscraper / \n", "468-Biology / \n", "1-Vehicle / 810-Village / 4324-Amish / \n", "3-Concert / \n", "9-Music video / \n", "3-Concert / 7-Musician / 13-Musical ensemble / 30-Drummer / \n", "17-Racing / 67-Cycling / 69-Bicycle / 570-Road bicycle racing / \n", "3-Concert / 16-Performance art / \n", "8-Football / \n", "3-Concert / 7-Musician / 13-Musical ensemble / 14-Guitar / 39-Drums / \n", "110-Album / \n", "12-Food / 215-Eating / 1481-Sunglasses / \n", "0-Games / 241-Fighting game / 764-The King of Fighters / 2354-The King of Fighters 2002 / \n", "114-Engine / 527-Battery / 1089-Wire / \n", "23-Mobile phone / 37-Gadget / 47-Personal computer / 145-Tablet computer / 189-iPad / \n", "10-Animal / 48-Pet / 71-Dog / \n", "1-Vehicle / 4-Car / 11-Motorsport / 17-Racing / 40-Road / 74-Truck / \n", "1-Vehicle / 50-Aircraft / 712-Outer space / 1992-Space Shuttle / \n", "1-Vehicle / 4-Car / 11-Motorsport / 17-Racing / 1134-Taxicab / \n", "61-Art / 148-Painting / \n", "0-Games / 2-Video game / 445-Guitar Hero / 965-Guitar Hero III: Legends of Rock / \n", "242-Paper / \n", "1-Vehicle / 4-Car / 11-Motorsport / 25-Toy / 92-Radio-controlled model / 202-Radio-controlled car / 482-Asphalt / \n", "0-Games / 489-Board game / 799-Chess / 922-Tabletop game / \n", "0-Games / 1000-AdventureQuest Worlds / \n", "9-Music video / \n", "5-Dance / 16-Performance art / \n", "12-Food / 26-Cooking / 32-Recipe / 226-Dessert / 232-Cake / 236-Baking / 759-Icing / 770-Banana / 795-Cupcake / \n", "3-Concert / 7-Musician / 342-Windows Media Video / 376-Flute / 726-Clarinet / \n", "45-Cosmetics / 235-Lipstick / 238-Nail / 300-Nail art / 324-Finger / 338-Manicure / 2235-Avon Products / \n", "5-Dance / \n", "36-Piano / 137-Tree / \n", "3-Concert / 7-Musician / 30-Drummer / \n", "359-Harry Potter (Literary Series) / \n", "2-Video game / 1199-Kinect / \n", "124-Tractor / \n", "209-University / \n", "0-Games / 2-Video game / 6-Animation / 257-Counter-Strike / \n", "10-Animal / 18-Outdoor recreation / 22-Nature / 42-Fishing / 225-Lake / 275-Diving / 288-Underwater / 381-Underwater diving / \n", "10-Animal / 86-Plant / 91-Fish / 258-Aquarium / 1659-Walmart / \n", "89-Comics / \n", "1-Vehicle / 4-Car / 70-Driving / 600-Flour / 630-Superman / \n", "3-Concert / 16-Performance art / \n", "0-Games / 8-Football / 54-Highlight film / 79-American football / 3987-Tight end / \n", "164-Beach / 222-Surfing / 278-GoPro / \n", "3-Concert / 7-Musician / 13-Musical ensemble / 46-Choir / \n", "5-Dance / 16-Performance art / 95-Talent show / 980-Solo dance / \n", "3-Concert / 7-Musician / 13-Musical ensemble / 38-Orchestra / \n", "1-Vehicle / 10-Animal / 18-Outdoor recreation / 42-Fishing / 87-Boat / 91-Fish / 698-Fisherman / \n", "3-Concert / 5-Dance / 46-Choir / 49-School / \n", "6-Animation / \n", "3-Concert / \n", "67-Cycling / 69-Bicycle / \n", "1-Vehicle / 4-Car / 11-Motorsport / 17-Racing / 41-Sports car / 57-Race track / 231-Supercar / 711-Gran Turismo / 953-Gran Turismo 5 / 1763-Lotus Cars / \n", "0-Games / 2-Video game / 34-Action-adventure game / \n", "0-Games / 35-Minecraft / \n", "423-Coast / 778-Gliding / \n", "45-Cosmetics / 235-Lipstick / \n", "6-Animation / 15-Cartoon / \n", "0-Games / 19-PC game / 51-Strategy video game / 165-League of Legends / \n", "3-Concert / 7-Musician / \n", "0-Games / 8-Football / 21-Stadium / 65-Kick / 73-Ball / 76-Ball / 81-Athlete / \n", "18-Outdoor recreation / 28-Fashion / 894-Jacket / \n", "2791-Transformer / \n", "13-Musical ensemble / \n", "81-Athlete / 562-Pitcher / \n", "3-Concert / 7-Musician / 13-Musical ensemble / 14-Guitar / 24-String instrument / 38-Orchestra / \n", "38-Orchestra / 541-Black-and-white / \n", "5-Dance / \n", "0-Games / 25-Toy / 126-The Walt Disney Company / 417-Play-Doh / 679-Mickey Mouse / 1234-Mouse / 1327-Minnie Mouse / 2161-Playground / \n", "96-Soldier / 257-Counter-Strike / \n", "9-Music video / 1058-Volcano / \n", "12-Food / 26-Cooking / 32-Recipe / 58-Cuisine / 109-Cooking show / 485-Dough / 662-Pizza / \n", "6-Animation / 243-Sonic the Hedgehog / 251-Sonic the Hedgehog / \n", "1-Vehicle / 2-Video game / 4-Car / 11-Motorsport / 17-Racing / 57-Race track / 111-PlayStation 3 / 1763-Lotus Cars / 3893-F1 2013 / \n", "78-Wedding / \n", "9-Music video / \n", "147-River / 481-Kayak / 1305-Canoeing / 3858-Canoe Slalom / \n", "9-Music video / \n", "12-Food / 26-Cooking / 58-Cuisine / \n", "0-Games / 756-Tibia / \n", "59-Winter sport / 84-Snow / 177-Skiing / 305-Ski / 789-Alpine skiing / \n", "0-Games / 8-Football / 21-Stadium / 65-Kick / 73-Ball / \n", "6-Animation / 89-Comics / \n", "7-Musician / 13-Musical ensemble / 30-Drummer / 39-Drums / 44-Drums / 49-School / 146-Snare drum / 203-Marching band / \n", "64-Transport / \n", "0-Games / 2-Video game / 2633-Soulcalibur V / \n", "0-Games / 2-Video game / 961-Sega Genesis / \n", "6-Animation / 15-Cartoon / 200-Doll / 412-Monster / 503-Monster High / \n", "2-Video game / 6-Animation / 477-Transformers / 990-Optimus Prime / 1900-Autobot / \n", "0-Games / 1-Vehicle / 2-Video game / 4-Car / 11-Motorsport / 17-Racing / 41-Sports car / 352-Need for Speed / 2407-Need for Speed: The Run / \n", "78-Wedding / 190-Bride / \n", "0-Games / 19-PC game / 35-Minecraft / 229-Halo / \n", "116-Pokémon / 372-Playing card / 1021-Booster pack / \n", "18-Outdoor recreation / 83-Skateboarding / 108-Skateboard / 221-Skatepark / \n", "155-Camera / 222-Surfing / 278-GoPro / \n", "5-Dance / 13-Musical ensemble / 49-School / 79-American football / 203-Marching band / \n", "0-Games / 8-Football / 54-Highlight film / 76-Ball / \n", "0-Games / 8-Football / \n", "322-Mixtape / \n", "259-Fire / 371-Firefighter / \n", "12-Food / 26-Cooking / 32-Recipe / 58-Cuisine / 109-Cooking show / 226-Dessert / 557-Salad / 671-Tomato / 1769-Olive / \n", "5-Dance / 16-Performance art / \n", "0-Games / 1938-Rust / \n", "31-Disc jockey / \n", "0-Games / 2-Video game / 35-Minecraft / \n", "0-Games / 8-Football / 65-Kick / 73-Ball / 76-Ball / \n", "28-Fashion / 45-Cosmetics / 56-Hair / 106-Hairstyle / \n", "3-Concert / 7-Musician / \n", "0-Games / 2-Video game / 1091-Saints Row / 2495-Saints Row 2 / \n", "10-Animal / 48-Pet / 71-Dog / \n", "3-Concert / 5-Dance / 31-Disc jockey / 133-Nightclub / \n", "9-Music video / \n", "0-Games / 2-Video game / 420-The Elder Scrolls V: Skyrim / 441-The Elder Scrolls / \n", "12-Food / 179-Kitchen / 539-Home appliance / 1439-Refrigerator / \n", "1-Vehicle / 4-Car / 11-Motorsport / 74-Truck / 158-Tire / 167-Four-wheel drive / 198-Off-road vehicle / 327-Mud / 375-Pickup truck / 515-Dodge / 565-Mud bogging / 854-Diesel engine / 1437-Ram Trucks / 1651-Dodge Ram / \n", "136-Gardening / \n", "278-GoPro / 283-Rock / 1416-Cliff / \n", "10-Animal / 80-Horse / 247-Stallion / \n", "780-Fat / \n", "1-Vehicle / 42-Fishing / 87-Boat / 517-Audi / 579-Yacht / 1309-Sailing ship / \n", "3-Concert / 56-Hair / 106-Hairstyle / 323-Toddler / 1113-Toilet / \n", "5-Dance / 4667-Gold's Gym / \n", "1-Vehicle / 4-Car / 11-Motorsport / 17-Racing / 40-Road / 41-Sports car / 135-Drifting / \n", "36-Piano / 53-Keyboard / 104-Pianist / 107-Musical keyboard / \n", "148-Painting / 170-Home improvement / 303-Paint / 1350-Plaster / \n", "56-Hair / 106-Hairstyle / \n", "99-Christmas / \n", "68-Lighting / 263-Light / 1375-Laser lighting display / \n", "1-Vehicle / 4-Car / 124-Tractor / 249-Heavy equipment / \n", "12-Food / 26-Cooking / 32-Recipe / 600-Flour / \n", "85-Combat / \n", "0-Games / 2-Video game / 426-Resident Evil / 2903-Umbrella / \n", "271-Tank / 409-World of Tanks / \n", "0-Games / 59-Winter sport / 81-Athlete / 112-Ice skating / 142-Hockey / 312-Ice rink / 1973-Hockey puck / \n", "3-Concert / 7-Musician / 30-Drummer / 39-Drums / 133-Nightclub / \n", "6-Animation / 15-Cartoon / 61-Art / 97-Drawing / \n", "162-Medicine / 890-Brush / 3879-Tooth brushing / \n", "3-Concert / 7-Musician / 13-Musical ensemble / 46-Choir / \n", "24-String instrument / 465-Microphone / \n", "0-Games / 2-Video game / 596-Monster Hunter / 1339-Monster Hunter Freedom Unite / \n", "2-Video game / 6-Animation / 1150-iPad 2 / \n", "1721-Tetris / \n", "3296-Transistor / \n", "516-Headphones / \n", "253-Carnival / \n", "1638-Soulcalibur / 4628-Soulcalibur III / \n", "3-Concert / 7-Musician / 14-Guitar / 63-Acoustic guitar / 356-Star / \n", "45-Cosmetics / 173-Eye shadow / 193-Eye liner / 195-Mascara / 218-Eyelash / 235-Lipstick / 291-Rouge / \n", "3-Concert / 5-Dance / 16-Performance art / 78-Wedding / \n", "330-Money / 2390-Dollar / 2722-United States Dollar / 2937-Banknote / \n", "8-Football / \n", "3-Concert / 7-Musician / 16-Performance art / \n", "23-Mobile phone / 29-Smartphone / 37-Gadget / 101-Telephone / 178-Samsung Galaxy / 1727-Samsung Galaxy Note 3 / \n", "23-Mobile phone / 29-Smartphone / 37-Gadget / 101-Telephone / 2318-Screenshot / \n", "2-Video game / 20-Trailer / 1196-Dragon Age / 2700-Dragon Age II / \n", "0-Games / 75-Wrestling / \n", "12-Food / \n", "1-Vehicle / 4-Car / 345-Loudspeaker / 484-Subwoofer / \n", "0-Games / \n", "3-Concert / 9-Music video / \n", "3-Concert / 7-Musician / \n", "0-Games / 8-Football / 54-Highlight film / 79-American football / 81-Athlete / \n", "18-Outdoor recreation / 49-School / 83-Skateboarding / 108-Skateboard / 482-Asphalt / \n", "5-Dance / \n", "5-Dance / 16-Performance art / \n", "1-Vehicle / 4-Car / 20-Trailer / \n", "10-Animal / 42-Fishing / 258-Aquarium / 288-Underwater / \n", "3-Concert / \n", "12-Food / \n", "0-Games / 1-Vehicle / 2-Video game / 4-Car / 34-Action-adventure game / 201-Grand Theft Auto: San Andreas / \n", "3-Concert / 7-Musician / \n", "5-Dance / 13-Musical ensemble / \n", "315-Knife / \n", "10-Animal / 642-Leaf / \n", "10-Animal / 22-Nature / 42-Fishing / 91-Fish / 275-Diving / 601-Shark / \n", "10-Animal / 18-Outdoor recreation / 42-Fishing / 91-Fish / 268-Recreational fishing / \n", "0-Games / 2-Video game / 6-Animation / 1777-Ratchet / 1824-Clank / 1859-Ratchet & Clank / 4423-Ratchet & Clank / \n", "783-Zee TV / \n", "5-Dance / 16-Performance art / 115-Ballet / 980-Solo dance / \n", "1-Vehicle / 4-Car / 11-Motorsport / 17-Racing / 135-Drifting / \n", "6-Animation / 15-Cartoon / \n", "259-Fire / 371-Firefighter / \n", "8-Football / \n", "0-Games / 42-Fishing / 1945-Medal game / \n", "5-Dance / \n", "1-Vehicle / 4-Car / 207-Exhaust system / 618-Fiat Automobiles / \n", "21-Stadium / 54-Highlight film / 81-Athlete / 163-Running / 1405-Primary school / \n", "3-Concert / 5-Dance / \n", "125-Television / \n", "3-Concert / 68-Lighting / \n", "45-Cosmetics / 173-Eye shadow / 325-Vampire / 2556-Halloween costume / \n", "0-Games / 2-Video game / 34-Action-adventure game / 380-Assassin's Creed / 1140-Assassin's Creed III / \n", "148-Painting / 744-Window / 1794-Curtain / \n", "6-Animation / \n", "93-Comedy / \n", "10-Animal / 48-Pet / 71-Dog / \n", "1-Vehicle / 4-Car / 27-Motorcycle / \n", "36-Piano / 53-Keyboard / 107-Musical keyboard / \n", "1-Vehicle / 4-Car / 70-Driving / 167-Four-wheel drive / 198-Off-road vehicle / 283-Rock / 360-Jeep / \n", "31-Disc jockey / \n", "0-Games / 1681-Mu Online / \n", "0-Games / 2-Video game / 34-Action-adventure game / 398-The Legend of Zelda / 692-Link / 3846-The Legend of Zelda: A Link to the Past / \n", "5-Dance / \n", "1-Vehicle / \n", "89-Comics / 269-Comic book / \n", "31-Disc jockey / \n", "78-Wedding / \n", "208-Wood / 981-Sculpture / 1073-Chainsaw / 1346-Wood carving / \n", "3-Concert / 68-Lighting / 77-Arena / \n", "209-University / \n", "9-Music video / \n", "0-Games / 116-Pokémon / 186-Pokémon / \n", "28-Fashion / 45-Cosmetics / 785-Jeans / 894-Jacket / 1822-Denim / \n", "22-Nature / 147-River / 825-Stream / \n", "0-Games / 2-Video game / 19-PC game / 33-Weapon / \n", "0-Games / 8-Football / 54-Highlight film / 76-Ball / \n", "3467-Gingerbread / \n", "1-Vehicle / \n", "89-Comics / 269-Comic book / 294-Batman / 855-Hulk / 897-Avengers / 2452-Thor / 3412-Daredevil / \n", "9-Music video / \n", "0-Games / 8-Football / 21-Stadium / 76-Ball / 77-Arena / 90-Sports game / 180-FIFA 15 / 453-Pro Evolution Soccer / 1719-Pro Evolution Soccer 2015 / \n", "47-Personal computer / 131-Computer / 228-Laptop / \n", "574-Elevator / 1285-Parking / 1746-Otis Elevator Company / \n", "7-Musician / 14-Guitar / 24-String instrument / 63-Acoustic guitar / \n", "0-Games / 2-Video game / 19-PC game / 35-Minecraft / \n", "3-Concert / 5-Dance / 16-Performance art / 49-School / 95-Talent show / \n", "3-Concert / 5-Dance / \n", "7-Musician / 14-Guitar / 24-String instrument / 100-Electric guitar / \n", "1-Vehicle / 4-Car / 11-Motorsport / 277-Sport utility vehicle / 517-Audi / 4026-Audi Q7 / \n", "97-Drawing / 148-Painting / 769-Red Bull / 2992-Energy drink / 3139-Beverage can / \n", "1-Vehicle / 2-Video game / 2401-Wolfenstein / 4432-Wolfenstein: The New Order / \n", "9-Music video / \n", "85-Combat / 1146-Wing Chun / \n", "1-Vehicle / 11-Motorsport / 124-Tractor / 156-Agriculture / 174-Farm / 252-Tractor pulling / 931-Combine Harvester / \n", "113-House / 170-Home improvement / 276-Room / 416-Apartment / 580-Bedroom / \n", "0-Games / 8-Football / 21-Stadium / 65-Kick / 73-Ball / 81-Athlete / \n", "2062-Kidney / \n", "0-Games / 2-Video game / 43-Call of Duty / 216-Call of Duty: Modern Warfare 3 / \n", "85-Combat / 117-Boxing / 681-Leather / 1016-Glove / \n", "573-Paragliding / 765-Parachute / \n", "1357-Ceramic / \n", "1-Vehicle / 4-Car / 11-Motorsport / 17-Racing / 40-Road / 41-Sports car / 57-Race track / 378-Drag racing / 449-Honda / 482-Asphalt / 552-Honda Civic / \n", "25-Toy / 2689-Furby / \n", "0-Games / 1-Vehicle / 19-PC game / 33-Weapon / 271-Tank / 409-World of Tanks / \n", "5-Dance / \n", "12-Food / 367-Drink / 560-Sugar / 1044-Coca-Cola / 1224-Cola / 1589-Soft drink / 2032-Pepsi / \n", "1-Vehicle / 27-Motorcycle / \n", "0-Games / 2-Video game / 4054-The Forest / \n", "5-Dance / 678-Latin dance / \n", "1-Vehicle / 50-Aircraft / 82-Airplane / 129-Aviation / 139-Airport / 151-Landing / 157-Jet aircraft / 175-Airliner / 187-Takeoff / 188-Airline / 254-Runway / 1340-Airbus A380 / 3671-Instrument landing system / \n", "3809-Marriott International / \n", "134-Call of Duty: Black Ops / \n", "3-Concert / 21-Stadium / \n", "1-Vehicle / \n", "18-Outdoor recreation / 163-Running / 267-Mountain / 283-Rock / 347-Climbing / \n", "1-Vehicle / 4-Car / 27-Motorcycle / 105-Motorcycling / 434-Ninja / 1380-Kawasaki motorcycles / 2345-Kawasaki Ninja 250R / 2404-Kawasaki Ninja / \n", "22-Nature / \n", "6-Animation / \n", "6-Animation / 15-Cartoon / 1266-Death Note / \n", "85-Combat / \n", "12-Food / 684-Ice cream / \n", "88-Machine / 415-Sewing / 997-Sewing machine / \n", "60-Basketball / \n", "8-Football / 21-Stadium / 77-Arena / \n", "1-Vehicle / 249-Heavy equipment / 271-Tank / 817-Bulldozer / \n", "7-Musician / 13-Musical ensemble / 38-Orchestra / \n", "287-Human swimming / 329-Swimming pool / \n", "0-Games / 8-Football / 76-Ball / 356-Star / \n", "3-Concert / 30-Drummer / \n", "3-Concert / 7-Musician / 16-Performance art / 30-Drummer / 39-Drums / \n", "14-Guitar / 4518-Shakugan no Shana / \n", "7-Musician / 14-Guitar / 44-Drums / 100-Electric guitar / \n", "20-Trailer / \n", "13-Musical ensemble / 21-Stadium / 49-School / 77-Arena / 79-American football / 203-Marching band / \n", "20-Trailer / \n", "5-Dance / 776-Kindergarten / \n", "5-Dance / 2131-Flair bartending / \n", "7-Musician / 13-Musical ensemble / 203-Marching band / 250-Parade / 376-Flute / \n", "762-Angel / \n", "75-Wrestling / 85-Combat / 117-Boxing / \n", "88-Machine / 415-Sewing / 997-Sewing machine / \n", "6-Animation / 816-Gundam / \n", "23-Mobile phone / 29-Smartphone / 37-Gadget / 47-Personal computer / 145-Tablet computer / \n", "101-Telephone / \n", "3-Concert / 5-Dance / 16-Performance art / \n", "0-Games / 2-Video game / 19-PC game / 51-Strategy video game / 165-League of Legends / \n", "102-Building / 170-Home improvement / 248-Hotel / 276-Room / 605-Cruise ship / 1402-Suite / \n", "2-Video game / 43-Call of Duty / 194-Call of Duty: Modern Warfare 2 / 216-Call of Duty: Modern Warfare 3 / 315-Knife / \n", "164-Beach / 219-Accordion / 233-Ocean / \n", "3-Concert / 5-Dance / \n", "1-Vehicle / 11-Motorsport / 50-Aircraft / 82-Airplane / 129-Aviation / 139-Airport / 151-Landing / 157-Jet aircraft / 175-Airliner / 187-Takeoff / 188-Airline / 254-Runway / 719-Boeing 747 / \n", "25-Toy / 51-Strategy video game / 264-Star Wars / \n", "93-Comedy / \n", "1-Vehicle / 11-Motorsport / 42-Fishing / 87-Boat / 1025-Personal water craft / \n", "1-Vehicle / 4-Car / 11-Motorsport / 41-Sports car / 114-Engine / 1763-Lotus Cars / \n", "6-Animation / 290-Pony / 343-My Little Pony / \n", "3-Concert / \n", "0-Games / 8-Football / 76-Ball / 81-Athlete / 1182-Softball / \n", "12-Food / 238-Nail / 827-Pump / 1510-Liquid / \n", "253-Carnival / \n", "12-Food / \n", "7-Musician / 14-Guitar / \n", "12-Food / 26-Cooking / 32-Recipe / 52-Dish / 58-Cuisine / 184-Vegetable / 239-Roasting / 301-Meal / 500-Rice / 759-Icing / 2775-Broccoli / \n", "5-Dance / 9-Music video / 115-Ballet / 130-Gymnastics / \n", "12-Food / 26-Cooking / 32-Recipe / 52-Dish / 58-Cuisine / 109-Cooking show / 179-Kitchen / 226-Dessert / 534-Milk / 1725-Custard / \n", "1975-Amiga / \n", "2-Video game / 47-Personal computer / 228-Laptop / 414-Computer monitor / \n", "25-Toy / 735-Kamen Rider Series / \n", "8-Football / \n", "3-Concert / 7-Musician / 13-Musical ensemble / 38-Orchestra / 816-Gundam / \n", "13-Musical ensemble / 203-Marching band / 253-Carnival / \n", "0-Games / 1-Vehicle / 2-Video game / 4-Car / 35-Minecraft / 245-BMW / \n", "12-Food / 720-Wine / 3722-Wine tasting / \n", "8-Football / 769-Red Bull / \n", "1-Vehicle / 4-Car / 40-Road / 70-Driving / 94-Dashcam / 164-Beach / 447-Sand / 901-Mitsubishi / 975-Dune / \n", "5-Dance / 16-Performance art / \n", "93-Comedy / 355-Comedian / \n", "9-Music video / \n", "18-Outdoor recreation / 267-Mountain / 347-Climbing / 614-Hiking / 2898-Canon EF lens mount / \n", "31-Disc jockey / 133-Nightclub / \n", "56-Hair / 106-Hairstyle / \n", "6-Animation / 15-Cartoon / 243-Sonic the Hedgehog / 251-Sonic the Hedgehog / 1197-Hedgehog / \n", "8-Football / \n", "10-Animal / 12-Food / 943-Stove / \n", "0-Games / 1991-Rappelz / \n", "6-Animation / 89-Comics / 294-Batman / 630-Superman / 803-DC Comics / 3473-Nightwing / \n", "3-Concert / 842-Harp / \n", "0-Games / 310-The Sims / 328-Simulation video game / 1553-The Sims 4 / 1713-The Sims / \n", "2-Video game / 1777-Ratchet / 1824-Clank / 1859-Ratchet & Clank / \n", "12-Food / 348-Fruit / 718-Duck / 1319-Apple / \n", "1-Vehicle / 50-Aircraft / \n", "1-Vehicle / 4-Car / 70-Driving / 94-Dashcam / 245-BMW / 344-Coupé / 1233-BMW 3 Series / 2325-BMW 3 Series (E90) / \n", "3-Concert / 5-Dance / \n", "3-Concert / 7-Musician / 13-Musical ensemble / 38-Orchestra / 46-Choir / \n", "6-Animation / 15-Cartoon / 89-Comics / \n", "3-Concert / 559-Night / 1222-Fountain / \n", "28-Fashion / 122-Weight training / 163-Running / 206-Shoe / 368-Sneakers / \n", "9-Music video / 322-Mixtape / \n", "18-Outdoor recreation / 83-Skateboarding / 108-Skateboard / 1031-Fisheye lens / \n", "0-Games / 8-Football / 21-Stadium / 65-Kick / 73-Ball / \n", "0-Games / 116-Pokémon / 186-Pokémon / \n", "117-Boxing / \n", "7-Musician / 14-Guitar / 24-String instrument / 63-Acoustic guitar / \n", "12-Food / 26-Cooking / 32-Recipe / 367-Drink / 454-Sauce / 612-Beer / \n", "10-Animal / 18-Outdoor recreation / 22-Nature / 42-Fishing / 233-Ocean / 423-Coast / 1135-Dolphin / \n", "0-Games / 2-Video game / 34-Action-adventure game / 72-iPhone / 118-Grand Theft Auto V / \n", "0-Games / 35-Minecraft / \n", "1-Vehicle / 5-Dance / 27-Motorcycle / \n", "3-Concert / 5-Dance / 16-Performance art / \n", "313-Book / \n", "47-Personal computer / 228-Laptop / 414-Computer monitor / 835-Backpack / \n", "1-Vehicle / 4-Car / 11-Motorsport / 17-Racing / 40-Road / 41-Sports car / 57-Race track / 515-Dodge / 2366-Dodge Viper / \n", "3-Concert / 7-Musician / \n", "0-Games / 75-Wrestling / \n", "127-Xbox 360 / 724-Rock Band / 1703-Rock Band / \n", "259-Fire / 350-Circus / 788-Juggling / \n", "79-American football / \n", "121-Photography / 155-Camera / 566-Digital camera / 629-Camera lens / 1117-Camcorder / 1855-Video camera / \n", "1-Vehicle / 18-Outdoor recreation / 67-Cycling / 69-Bicycle / 220-Mountain bike / 284-Mountain biking / \n", "6-Animation / 1187-Ragnarok Online / \n", "1-Vehicle / 27-Motorcycle / \n", "1-Vehicle / 139-Airport / \n", "1-Vehicle / 50-Aircraft / 82-Airplane / 139-Airport / 151-Landing / 157-Jet aircraft / 175-Airliner / 187-Takeoff / 188-Airline / 254-Runway / 472-Boeing 737 / \n", "6-Animation / 10-Animal / 15-Cartoon / \n", "415-Sewing / \n", "0-Games / 2-Video game / 10-Animal / 15-Cartoon / 132-Bird / 847-Angry Birds / \n", "9-Music video / \n", "18-Outdoor recreation / 83-Skateboarding / 108-Skateboard / \n", "0-Games / 1-Vehicle / 2-Video game / 19-PC game / 33-Weapon / 51-Strategy video game / 271-Tank / 409-World of Tanks / \n", "341-Sketch comedy / \n", "0-Games / 2-Video game / 19-PC game / 51-Strategy video game / \n", "5-Dance / \n", "0-Games / 60-Basketball / 76-Ball / 168-Basketball moves / 255-Slam dunk / \n", "9-Music video / \n", "0-Games / 2-Video game / 55-Video game console / 199-Wii / 319-Robot / 443-Wii U / 1092-Game Boy Advance / \n", "3-Concert / 5-Dance / \n", "0-Games / 8-Football / 21-Stadium / 65-Kick / 76-Ball / 77-Arena / 81-Athlete / \n", "1-Vehicle / \n", "1063-Ink / \n", "0-Games / 2-Video game / 34-Action-adventure game / 380-Assassin's Creed / 1048-Assassin's Creed / 1522-Assassin's Creed II / \n", "1-Vehicle / 4-Car / \n", "1-Vehicle / 4-Car / 114-Engine / \n", "1-Vehicle / 4-Car / 87-Boat / \n", "0-Games / 2-Video game / 426-Resident Evil / 1238-Resident Evil 5 / 2278-Albert Wesker / \n", "0-Games / 211-RuneScape / \n", "10-Animal / 48-Pet / 71-Dog / 166-Cat / 346-Kitten / 3444-Claw / \n", "1-Vehicle / 4-Car / 87-Boat / 208-Wood / \n", "0-Games / 2-Video game / 90-Sports game / 180-FIFA 15 / \n", "12-Food / 32-Recipe / 295-Chocolate / 1193-Orange / \n", "28-Fashion / 56-Hair / 106-Hairstyle / 798-Shampoo / 919-Hair conditioner / 1320-Bangs / 1359-Hair iron / \n", "12-Food / 26-Cooking / 32-Recipe / 52-Dish / 58-Cuisine / 109-Cooking show / 210-Cookware and bakeware / 239-Roasting / 771-Noodle / 1284-Frying / \n", "60-Basketball / 168-Basketball moves / 255-Slam dunk / \n", "3-Concert / 7-Musician / 13-Musical ensemble / 219-Accordion / 616-The Bible / 1019-Christ / \n", "110-Album / \n", "45-Cosmetics / 238-Nail / 300-Nail art / 304-Nail polish / 307-Hand / 324-Finger / 338-Manicure / \n", "10-Animal / 166-Cat / 3944-Beanie / \n", "209-University / \n", "7-Musician / 14-Guitar / 24-String instrument / \n", "5-Dance / \n", "12-Food / 215-Eating / 1627-Shorts / \n", "93-Comedy / \n", "1146-Wing Chun / \n", "1-Vehicle / 4-Car / 114-Engine / 299-Volkswagen Passenger Cars / 969-Volkswagen Beetle / \n", "10-Animal / 166-Cat / \n", "6-Animation / 25-Toy / 477-Transformers / 990-Optimus Prime / 4445-Transformers: Beast Wars / \n", "20-Trailer / 66-Bollywood / \n", "9-Music video / \n", "10-Animal / 48-Pet / 132-Bird / 280-Wildlife / \n", "359-Harry Potter (Literary Series) / \n", "47-Personal computer / 228-Laptop / 525-Printer / 1063-Ink / 2634-Inkjet printing / \n", "0-Games / 33-Weapon / 257-Counter-Strike / 702-Counter-Strike: Source / 826-Source Engine / 1513-Hitman / \n", "113-House / 248-Hotel / 276-Room / 416-Apartment / 652-Bathroom / \n", "18-Outdoor recreation / 347-Climbing / 828-Rock climbing / \n", "46-Choir / \n", "3-Concert / 9-Music video / \n", "1083-Role-playing game / 3524-Tabletop role-playing game / \n", "6-Animation / 15-Cartoon / \n", "0-Games / 2-Video game / 19-PC game / 89-Comics / 511-Mortal Kombat / 1392-Injustice: Gods Among Us / \n", "9-Music video / \n", "0-Games / 1-Vehicle / 2-Video game / 4-Car / 34-Action-adventure game / 118-Grand Theft Auto V / \n", "1-Vehicle / 4-Car / 88-Machine / 114-Engine / 157-Jet aircraft / 1257-Turbine / 1529-Jet engine / \n", "3-Concert / 7-Musician / \n", "293-Slide show / \n", "144-Amusement park / 321-Roller coaster / 334-Amusement ride / \n", "0-Games / 2-Video game / 34-Action-adventure game / 201-Grand Theft Auto: San Andreas / 413-PlayStation 2 / \n", "1-Vehicle / 27-Motorcycle / \n", "0-Games / 326-Metin2 / \n", "131-Computer / 141-Microsoft Windows / \n", "9-Music video / \n", "0-Games / 51-Strategy video game / 1120-Tower defense / 2988-Bloons Tower Defense / 3354-Bloons / \n", "3-Concert / 31-Disc jockey / \n", "25-Toy / 140-LEGO / 884-Lego minifigure / \n", "0-Games / 2-Video game / 6-Animation / 25-Toy / 96-Soldier / 940-Clay animation / \n", "615-Poker / \n", "6-Animation / 183-Manga / 750-Sailor Moon / 1342-Sailor Moon / \n", "12-Food / \n", "23-Mobile phone / 29-Smartphone / 37-Gadget / 47-Personal computer / 145-Tablet computer / 178-Samsung Galaxy / 487-Samsung Electronics / 1015-Samsung Galaxy Note series / 1727-Samsung Galaxy Note 3 / \n", "15-Cartoon / 185-Dragon Ball / 244-Goku / \n", "93-Comedy / \n", "10-Animal / 12-Food / 22-Nature / 84-Snow / 120-Winter / 262-Ice / 1145-Winter storm / \n", "6-Animation / 15-Cartoon / 183-Manga / \n", "1-Vehicle / 4-Car / 1089-Wire / 3187-Spark plug / \n", "335-Nintendo 3DS / 3901-Star Fox 64 / \n", "3-Concert / 13-Musical ensemble / \n", "38-Orchestra / \n", "121-Photography / 1605-Second Life / \n", "1-Vehicle / 2-Video game / 4-Car / 40-Road / 201-Grand Theft Auto: San Andreas / 297-Grand Theft Auto IV / 1425-Gears of War 2 / 3135-Call of Duty 3 / 3718-Pentium / \n", "1345-Dean Winchester / 1801-Sam Winchester / \n", "0-Games / 47-Personal computer / \n", "1-Vehicle / 4-Car / 519-Recreational vehicle / \n", "0-Games / 2-Video game / 6-Animation / 19-PC game / 123-Naruto / 591-Naruto: Ultimate Ninja / \n", "10-Animal / 48-Pet / 106-Hairstyle / 132-Bird / 1341-Parakeet / \n", "3-Concert / 7-Musician / 16-Performance art / \n", "1-Vehicle / 4-Car / 11-Motorsport / 17-Racing / 41-Sports car / 70-Driving / 378-Drag racing / \n", "429-Website / \n", "1-Vehicle / 50-Aircraft / 82-Airplane / 92-Radio-controlled model / 139-Airport / 149-Model aircraft / 157-Jet aircraft / 175-Airliner / 188-Airline / 205-Newscaster / 2206-Boeing 787 Dreamliner / \n", "3-Concert / 16-Performance art / \n", "22-Nature / 223-Weather / 296-Saxophone / 499-Cloud / 502-Sky / 1006-Tenor saxophone / \n", "2233-Derek Shepherd / \n", "31-Disc jockey / \n", "3-Concert / \n", "8-Football / \n", "93-Comedy / \n", "3-Concert / 7-Musician / 13-Musical ensemble / 38-Orchestra / 150-Violin / \n", "1-Vehicle / 4-Car / 11-Motorsport / 17-Racing / 27-Motorcycle / 57-Race track / 378-Drag racing / 2133-Road racing / \n", "85-Combat / \n", "163-Running / \n", "22-Nature / 283-Rock / \n", "0-Games / 2-Video game / 34-Action-adventure game / 118-Grand Theft Auto V / \n", "1-Vehicle / 50-Aircraft / 114-Engine / 627-Clock / \n", "23-Mobile phone / 29-Smartphone / 37-Gadget / \n", "1-Vehicle / 4-Car / 27-Motorcycle / 105-Motorcycling / 1036-Moped / \n", "0-Games / 2-Video game / 51-Strategy video game / 2503-Age of Empires / 4125-Age of Empires II: The Age of Kings / \n", "23-Mobile phone / 1820-Motorola Droid / \n", "3-Concert / \n", "113-House / 246-Furniture / 369-Resort / 416-Apartment / 580-Bedroom / \n", "95-Talent show / \n", "5-Dance / 7-Musician / 13-Musical ensemble / 203-Marching band / 1532-Military band / \n", "5-Dance / 16-Performance art / \n", "61-Art / 97-Drawing / 242-Paper / 415-Sewing / 457-Knitting / 546-Stitch / 2424-Greeting card / 3425-Christmas card / \n", "36-Piano / 53-Keyboard / 104-Pianist / 107-Musical keyboard / 265-Electronic keyboard / \n", "67-Cycling / 69-Bicycle / 377-BMX bike / \n", "2-Video game / 47-Personal computer / 111-PlayStation 3 / 194-Call of Duty: Modern Warfare 2 / 414-Computer monitor / \n", "194-Call of Duty: Modern Warfare 2 / 2550-EarthBound / \n", "0-Games / 8-Football / 21-Stadium / 79-American football / 81-Athlete / \n", "2-Video game / 55-Video game console / 2142-Sega Saturn / \n", "5-Dance / 16-Performance art / 458-Gold / \n", "5-Dance / 7-Musician / 13-Musical ensemble / 16-Performance art / \n", "1-Vehicle / 6-Animation / 87-Boat / 126-The Walt Disney Company / 308-Ship / 605-Cruise ship / \n", "9-Music video / \n", "12-Food / 534-Milk / \n", "10-Animal / 48-Pet / 71-Dog / \n", "85-Combat / \n", "18-Outdoor recreation / 573-Paragliding / \n", "12-Food / 26-Cooking / 32-Recipe / 52-Dish / 58-Cuisine / 109-Cooking show / 684-Ice cream / \n", "18-Outdoor recreation / 347-Climbing / 3318-Abseiling / \n", "6-Animation / 25-Toy / 140-LEGO / 319-Robot / \n", "3-Concert / \n", "31-Disc jockey / 861-Mixing console / \n", "3-Concert / 7-Musician / \n", "0-Games / 51-Strategy video game / 171-Warcraft / 1691-Warcraft III: Reign of Chaos / \n", "0-Games / 2-Video game / 637-Five Nights at Freddy's / 1069-Easter egg / \n", "23-Mobile phone / 29-Smartphone / 72-iPhone / 272-iPod / 331-iPod touch / \n", "18-Outdoor recreation / 22-Nature / 42-Fishing / 147-River / 268-Recreational fishing / \n", "1-Vehicle / 4-Car / 40-Road / 94-Dashcam / 3496-Chevrolet Malibu / \n", "30-Drummer / 39-Drums / 44-Drums / 128-Cymbal / 146-Snare drum / \n", "0-Games / 8-Football / \n", "1443-Dance studio / \n", "2-Video game / 55-Video game console / 127-Xbox 360 / 181-Xbox / 314-Xbox One / 386-Xbox / 677-Game controller / 1872-Xbox 360 Controller / \n", "1-Vehicle / 4-Car / 11-Motorsport / 17-Racing / \n", "1-Vehicle / 8-Football / \n", "0-Games / 2-Video game / 34-Action-adventure game / 1125-God of War / 1432-Kratos / 1743-God of War / 2103-God of War III / \n", "3-Concert / 98-Festival / \n", "1-Vehicle / 4-Car / 11-Motorsport / 17-Racing / 74-Truck / 378-Drag racing / 761-Chevrolet Camaro / 1040-Street racing / 2317-Twin-turbo / \n", "0-Games / 2-Video game / 240-Arcade game / \n", "1-Vehicle / 67-Cycling / 69-Bicycle / \n", "0-Games / 2-Video game / \n", "1-Vehicle / 11-Motorsport / 17-Racing / 27-Motorcycle / \n", "20-Trailer / 365-Fireworks / 992-Rocket / \n", "0-Games / 2-Video game / 34-Action-adventure game / 159-World of Warcraft / 171-Warcraft / \n", "1-Vehicle / 4-Car / 11-Motorsport / 74-Truck / 92-Radio-controlled model / 167-Four-wheel drive / 198-Off-road vehicle / 202-Radio-controlled car / 234-Ford / 327-Mud / 360-Jeep / 565-Mud bogging / 935-Off-road racing / 3194-Ford F-350 / \n", "978-Love song / \n", "14-Guitar / 4518-Shakugan no Shana / \n", "5-Dance / 49-School / \n", "47-Personal computer / 131-Computer / 145-Tablet computer / 228-Laptop / 309-Computer hardware / 513-Computer keyboard / \n", "208-Wood / \n", "0-Games / 1-Vehicle / 2-Video game / 4-Car / 19-PC game / 34-Action-adventure game / 387-Spider-Man / 1982-Venom / \n", "7-Musician / 14-Guitar / 24-String instrument / 100-Electric guitar / 1103-Gibson Les Paul / 3014-Gibson SG / \n", "110-Album / \n", "1-Vehicle / 4-Car / 62-Train / 428-Rail transport modelling / \n", "18-Outdoor recreation / 22-Nature / 347-Climbing / 722-Waterfall / 828-Rock climbing / \n", "1-Vehicle / 4-Car / 40-Road / 94-Dashcam / 223-Weather / 490-Rain / \n", "88-Machine / 619-Manufacturing / 625-Factory / 812-Plastic / \n", "12-Food / 215-Eating / 3385-Burrito / \n", "56-Hair / 106-Hairstyle / 340-Afro-textured hair / \n", "10-Animal / \n", "88-Machine / 512-Printing / \n", "209-University / \n", "3-Concert / 7-Musician / 13-Musical ensemble / 38-Orchestra / 98-Festival / 150-Violin / \n", "10-Animal / 12-Food / 48-Pet / 52-Dish / 71-Dog / \n", "78-Wedding / \n", "1-Vehicle / 4-Car / 11-Motorsport / 17-Racing / 41-Sports car / 135-Drifting / 1519-Gran Turismo 6 / \n", "2-Video game / 28-Fashion / 60-Basketball / 206-Shoe / 333-Nike; Inc. / 368-Sneakers / \n", "0-Games / 8-Football / 21-Stadium / 358-Rugby football / \n", "6-Animation / 15-Cartoon / 463-Cricket / \n", "0-Games / 1000-AdventureQuest Worlds / \n", "6-Animation / 15-Cartoon / 1555-Sword Art Online / \n", "1-Vehicle / 4-Car / 11-Motorsport / 17-Racing / 40-Road / 41-Sports car / 57-Race track / 212-Highway / 352-Need for Speed / 875-Need for Speed: Most Wanted / 1600-Need for Speed: World / \n", "1-Vehicle / 50-Aircraft / 82-Airplane / 102-Building / 129-Aviation / 139-Airport / 151-Landing / 157-Jet aircraft / 175-Airliner / 188-Airline / 254-Runway / \n", "60-Basketball / \n", "10-Animal / 18-Outdoor recreation / 22-Nature / 42-Fishing / 91-Fish / 275-Diving / 288-Underwater / 381-Underwater diving / 407-Scuba diving / \n", "10-Animal / 48-Pet / 166-Cat / 346-Kitten / \n", "182-Puppy / \n", "337-Advertising / 2032-Pepsi / \n", "3-Concert / 68-Lighting / 594-Music festival / \n", "2-Video game / 1087-Donkey Kong / 2970-Donkey Kong / 3950-Donkey Kong 64 / \n", "3-Concert / 9-Music video / \n", "20-Trailer / 359-Harry Potter (Literary Series) / 2001-Harry Potter and the Deathly Hallows / \n", "0-Games / 2-Video game / 34-Action-adventure game / 85-Combat / 380-Assassin's Creed / 1435-Ezio Auditore da Firenze / 1745-Assassin's Creed: Revelations / \n", "0-Games / 2-Video game / 43-Call of Duty / 55-Video game console / 176-PlayStation 4 / 292-Handheld game console / 317-PlayStation / 422-Call of Duty: Ghosts / 682-PlayStation Vita / \n", "1066-Aikido / \n", "6-Animation / 15-Cartoon / \n", "3-Concert / 68-Lighting / \n", "31-Disc jockey / 133-Nightclub / \n", "3-Concert / 7-Musician / 13-Musical ensemble / 38-Orchestra / 230-Brass instrument / 296-Saxophone / \n", "9-Music video / \n", "116-Pokémon / 186-Pokémon / 1387-Pokémon Battle Revolution / \n", "36-Piano / 53-Keyboard / 104-Pianist / \n", "45-Cosmetics / 56-Hair / 173-Eye shadow / 193-Eye liner / 195-Mascara / 235-Lipstick / 282-Skin / 291-Rouge / 385-Face / 436-Concealer / \n", "1345-Dean Winchester / \n", "141-Microsoft Windows / 513-Computer keyboard / \n", "23-Mobile phone / 29-Smartphone / 37-Gadget / 101-Telephone / 178-Samsung Galaxy / \n", "20-Trailer / \n", "0-Games / 2-Video game / 756-Tibia / 4097-Nollie / \n", "10-Animal / 132-Bird / 718-Duck / 1947-Goose / \n", "126-The Walt Disney Company / 1106-Character / \n", "16-Performance art / 93-Comedy / 355-Comedian / 2463-Smile / \n", "8-Football / 21-Stadium / 77-Arena / \n", "6-Animation / 15-Cartoon / \n", "0-Games / 2-Video game / 19-PC game / 1049-Collectible card game / \n", "10-Animal / 48-Pet / 182-Puppy / \n", "5-Dance / 16-Performance art / 200-Doll / \n", "5-Dance / 16-Performance art / \n", "1-Vehicle / 4-Car / 40-Road / 70-Driving / 94-Dashcam / \n", "0-Games / 2-Video game / 55-Video game console / 176-PlayStation 4 / 317-PlayStation / 1482-PlayStation Network / \n", "47-Personal computer / 131-Computer / 497-Macintosh / \n", "12-Food / 26-Cooking / 4138-Chorizo / \n", "0-Games / 2-Video game / 25-Toy / 140-LEGO / 294-Batman / 2440-Lego Batman: The Videogame / \n", "102-Building / 113-House / 353-Wall / 433-Construction / 488-Door / 1314-Garage / 3457-Garage door / \n", "6-Animation / 15-Cartoon / 1866-Katara / 3734-Aang / \n", "31-Disc jockey / \n", "315-Knife / 442-Bread / 2017-Sharpening / 3880-Knife sharpening / \n", "13-Musical ensemble / \n", "1-Vehicle / 4-Car / 11-Motorsport / 17-Racing / 169-Rallying / \n", "12-Food / 26-Cooking / 32-Recipe / 391-Chicken meat / 1707-Microwave oven / \n", "9-Music video / \n", "3-Concert / 7-Musician / 16-Performance art / 30-Drummer / \n", "1-Vehicle / 4-Car / 27-Motorcycle / 668-Volkswagen Golf / 2517-Chevrolet Opala / \n", "364-Sitcom / \n", "36-Piano / 53-Keyboard / 446-Sheet music / 592-Synthesia / \n", "3-Concert / 68-Lighting / \n", "9-Music video / \n", "14-Guitar / 100-Electric guitar / \n", "5-Dance / 7-Musician / 78-Wedding / \n", "61-Art / 148-Painting / 628-Portrait / \n", "0-Games / 2-Video game / 47-Personal computer / 908-Video card / 1580-Crysis 2 / \n", "1-Vehicle / 4-Car / 6-Animation / 15-Cartoon / 25-Toy / 74-Truck / 88-Machine / 1307-Police car / 1418-Model car / 2701-Tow truck / \n", "6-Animation / 15-Cartoon / 567-Compact disc / 1492-Persona / \n", "23-Mobile phone / 29-Smartphone / \n", "66-Bollywood / \n", "22-Nature / 164-Beach / 233-Ocean / 423-Coast / \n", "1-Vehicle / 4-Car / 531-Hatchback / \n", "0-Games / 8-Football / 21-Stadium / \n", "0-Games / 8-Football / 21-Stadium / 65-Kick / 73-Ball / 76-Ball / 81-Athlete / \n", "0-Games / 2-Video game / 34-Action-adventure game / 1144-Castlevania / 3768-Castlevania: Lords of Shadow / \n", "0-Games / 2-Video game / 19-PC game / 33-Weapon / 43-Call of Duty / 96-Soldier / 134-Call of Duty: Black Ops / \n", "0-Games / 2-Video game / 35-Minecraft / \n", "98-Festival / \n", "707-Roof / \n", "61-Art / \n", "2716-The Hunger Games / 4616-Catching Fire / \n", "28-Fashion / 793-Shirt / \n", "1-Vehicle / \n", "3-Concert / \n", "9-Music video / \n", "2840-Engraving / 2967-Printmaking / \n", "6-Animation / 25-Toy / 477-Transformers / 990-Optimus Prime / \n", "3-Concert / 68-Lighting / \n", "1243-Mirror / \n", "379-Touhou Project / \n", "242-Paper / 370-Origami / \n", "23-Mobile phone / 29-Smartphone / 37-Gadget / 72-iPhone / \n", "66-Bollywood / \n", "66-Bollywood / \n", "12-Food / 4609-Pepper spray / \n", "1-Vehicle / 4-Car / 74-Truck / 375-Pickup truck / 3309-Toyota Tundra / \n", "4017-Television film / \n", "23-Mobile phone / 29-Smartphone / 72-iPhone / 554-iPhone 4 / 1155-iPhone 3GS / \n", "0-Games / 2-Video game / 199-Wii / 2502-GoldenEye 007 / \n", "0-Games / 1-Vehicle / 2-Video game / 4-Car / 34-Action-adventure game / 297-Grand Theft Auto IV / 658-Grand Theft Auto: The Lost and Damned / 1923-Niko Bellic / \n", "0-Games / 2-Video game / 34-Action-adventure game / 380-Assassin's Creed / 1140-Assassin's Creed III / \n", "576-Walt Disney World / \n", "0-Games / 2-Video game / 6-Animation / 15-Cartoon / 243-Sonic the Hedgehog / 251-Sonic the Hedgehog / \n", "44-Drums / 3215-Cornet / \n", "10-Animal / 71-Dog / \n", "5-Dance / 16-Performance art / \n", "1-Vehicle / 4-Car / 11-Motorsport / 27-Motorcycle / 336-Scooter / \n", "2342-Jack Sparrow / \n", "21-Stadium / \n", "8-Football / \n", "6-Animation / 183-Manga / 269-Comic book / \n", "6-Animation / 3163-Tugboat / \n", "46-Choir / \n", "1-Vehicle / 4-Car / 234-Ford / \n", "56-Hair / 106-Hairstyle / \n", "1-Vehicle / 49-School / 286-Bus / 2199-School bus / \n", "0-Games / 1-Vehicle / 50-Aircraft / 139-Airport / 151-Landing / 188-Airline / 923-Boeing 777 / \n", "10-Animal / \n", "9-Music video / \n", "0-Games / 2-Video game / 23-Mobile phone / 29-Smartphone / 72-iPhone / 243-Sonic the Hedgehog / 251-Sonic the Hedgehog / 331-iPod touch / \n", "5-Dance / \n", "1-Vehicle / 4-Car / 11-Motorsport / 17-Racing / 935-Off-road racing / 2734-Monster Energy / 4003-Baja 1000 / \n", "3-Concert / 7-Musician / 16-Performance art / \n", "1-Vehicle / 4-Car / 1624-Minivan / 2446-Used car / \n", "102-Building / 113-House / 310-The Sims / 328-Simulation video game / 564-The Sims 3 / \n", "0-Games / 19-PC game / 35-Minecraft / \n", "1-Vehicle / 11-Motorsport / 27-Motorcycle / 447-Sand / 751-Desert / 975-Dune / \n", "0-Games / 8-Football / 21-Stadium / 77-Arena / \n", "536-Glass / 1226-Metalworking / 2855-Vase / \n", "2848-12-inch single / \n", "30-Drummer / 39-Drums / 44-Drums / 128-Cymbal / 146-Snare drum / 3454-Paiste / \n", "1-Vehicle / 62-Train / 64-Transport / 103-Rail transport / 119-Locomotive / 143-Railroad car / 152-Track / 428-Rail transport modelling / \n", "12-Food / 26-Cooking / 32-Recipe / 52-Dish / 58-Cuisine / 109-Cooking show / 1024-Culinary art / \n", "0-Games / 2-Video game / 2592-DualShock / 3862-Toph Beifong / \n", "3-Concert / 5-Dance / 98-Festival / \n", "3-Concert / 7-Musician / 13-Musical ensemble / 30-Drummer / 39-Drums / 44-Drums / \n", "0-Games / 1-Vehicle / 4-Car / 11-Motorsport / 17-Racing / 40-Road / 41-Sports car / 70-Driving / \n", "110-Album / \n", "0-Games / 21-Stadium / 60-Basketball / 77-Arena / \n", "0-Games / 2-Video game / 310-The Sims / 328-Simulation video game / 1553-The Sims 4 / \n", "7-Musician / 13-Musical ensemble / 46-Choir / 203-Marching band / 1675-Drum and bugle corps / \n", "18-Outdoor recreation / 33-Weapon / 217-Hunting / \n", "0-Games / 2-Video game / 1789-Slender: The Eight Pages / \n", "3-Concert / 7-Musician / 13-Musical ensemble / 38-Orchestra / 44-Drums / 78-Wedding / \n", "260-Prayer / \n", "3-Concert / \n", "5-Dance / \n", "31-Disc jockey / \n", "14-Guitar / \n", "31-Disc jockey / \n", "12-Food / 1227-Sushi / \n", "1-Vehicle / 4-Car / 11-Motorsport / 17-Racing / 40-Road / 41-Sports car / 57-Race track / 135-Drifting / \n", "205-Newscaster / \n", "7-Musician / 13-Musical ensemble / 14-Guitar / 24-String instrument / 63-Acoustic guitar / \n", "31-Disc jockey / 342-Windows Media Video / \n", "6-Animation / 185-Dragon Ball / 244-Goku / 876-Freeza / \n", "10-Animal / 12-Food / 26-Cooking / 32-Recipe / 91-Fish / 215-Eating / 239-Roasting / 893-Fish as food / 3825-Tilapia / \n", "0-Games / 1-Vehicle / 2-Video game / 4-Car / 19-PC game / 33-Weapon / 734-DayZ / 3202-Gesture / \n", "250-Parade / \n", "0-Games / 2-Video game / 25-Toy / 140-LEGO / 264-Star Wars / 852-Lego Star Wars / \n", "0-Games / 2-Video game / 35-Minecraft / 127-Xbox 360 / 181-Xbox / \n", "8-Football / \n", "6-Animation / 15-Cartoon / \n", "1-Vehicle / 4-Car / 11-Motorsport / 345-Loudspeaker / 349-Chevrolet / 761-Chevrolet Camaro / \n", "3-Concert / 5-Dance / 16-Performance art / 49-School / 95-Talent show / \n", "6-Animation / 4271-Inflatable castle / \n", "3-Concert / \n", "83-Skateboarding / 108-Skateboard / 555-Longboard / 667-Longboarding / \n", "0-Games / 1083-Role-playing game / \n", "6-Animation / 15-Cartoon / 396-One Piece / \n", "14-Guitar / \n", "6-Animation / 15-Cartoon / 123-Naruto / 622-Sakura Haruno / \n", "1-Vehicle / 67-Cycling / 69-Bicycle / 570-Road bicycle racing / \n", "38-Orchestra / \n", "36-Piano / 53-Keyboard / 104-Pianist / \n", "12-Food / 86-Plant / 1023-Rose / \n", "3-Concert / 7-Musician / \n", "9-Music video / \n", "5-Dance / 49-School / 95-Talent show / 191-Cheerleading / \n", "7-Musician / 30-Drummer / 39-Drums / 44-Drums / \n", "0-Games / 240-Arcade game / 1440-Pinball / \n", "10-Animal / 80-Horse / \n", "10-Animal / 48-Pet / 132-Bird / 598-Parrot / 1341-Parakeet / 2024-Cockatoo / \n", "75-Wrestling / \n", "0-Games / 8-Football / 65-Kick / 76-Ball / \n", "130-Gymnastics / 1558-Hula hoop / \n", "3-Concert / \n", "0-Games / 2-Video game / 33-Weapon / \n", "3-Concert / 38-Orchestra / \n", "6-Animation / 123-Naruto / 622-Sakura Haruno / \n", "6-Animation / 25-Toy / 35-Minecraft / 140-LEGO / 3899-Lego Minecraft / \n", "3-Concert / 1442-Tabla / \n", "0-Games / 8-Football / 21-Stadium / \n", "2-Video game / 20-Trailer / \n", "5-Dance / 115-Ballet / \n", "159-World of Warcraft / 171-Warcraft / \n", "1-Vehicle / 4-Car / 484-Subwoofer / \n", "12-Food / 26-Cooking / 32-Recipe / 226-Dessert / 232-Cake / 236-Baking / 795-Cupcake / 1192-Hello Kitty / \n", "455-Unidentified flying object / \n", "72-iPhone / \n", "0-Games / 2-Video game / 19-PC game / 33-Weapon / 96-Soldier / 217-Hunting / 601-Shark / 1543-Crysis / \n", "1-Vehicle / 92-Radio-controlled model / 149-Model aircraft / 160-Radio-controlled aircraft / 393-Unmanned aerial vehicle / 604-Quadcopter / 2025-Multirotor / \n", "88-Machine / 208-Wood / 400-Woodturning / 3061-Cylinder / \n", "9-Music video / \n", "3-Concert / 5-Dance / 16-Performance art / \n", "238-Nail / 300-Nail art / 304-Nail polish / 307-Hand / 324-Finger / 338-Manicure / \n", "45-Cosmetics / 173-Eye shadow / \n", "56-Hair / 106-Hairstyle / 164-Beach / 329-Swimming pool / 405-Braid / \n", "110-Album / \n", "546-Stitch / \n", "1-Vehicle / 4-Car / 84-Snow / 94-Dashcam / 120-Winter / 262-Ice / 680-Subaru / \n", "99-Christmas / 3425-Christmas card / \n", "5-Dance / 49-School / 95-Talent show / 130-Gymnastics / 191-Cheerleading / 306-High school / \n", "0-Games / 8-Football / 21-Stadium / 65-Kick / \n", "2454-Notebook / \n", "0-Games / 2-Video game / 6-Animation / 111-PlayStation 3 / 2368-Marvel Universe / \n", "3-Concert / 7-Musician / 13-Musical ensemble / 14-Guitar / 39-Drums / \n", "0-Games / 19-PC game / 51-Strategy video game / 382-Dota 2 / \n", "3-Concert / 5-Dance / 888-Living room / 1281-Couch / \n", "8-Football / \n", "3-Concert / 7-Musician / \n", "3-Concert / 9-Music video / \n", "66-Bollywood / \n", "0-Games / 791-Magic: The Gathering / \n", "83-Skateboarding / 108-Skateboard / 2256-Bearing / \n", "3-Concert / 68-Lighting / \n", "1-Vehicle / 27-Motorcycle / 1282-Honda CBR series / \n", "9-Music video / \n", "5-Dance / \n", "1-Vehicle / 11-Motorsport / 17-Racing / 27-Motorcycle / 927-Supermoto / 1036-Moped / \n", "385-Face / 780-Fat / \n", "8-Football / \n", "1-Vehicle / 11-Motorsport / 17-Racing / 57-Race track / 74-Truck / 854-Diesel engine / \n", "1791-Stargate / \n", "0-Games / 12-Food / 578-Cookie / \n", "12-Food / 367-Drink / 1347-Coffeehouse / \n", "0-Games / 256-Tennis / 1337-Serve / 2565-Soft tennis / \n", "125-Television / \n", "9-Music video / \n", "0-Games / 8-Football / 54-Highlight film / 79-American football / \n", "0-Games / 2-Video game / 20-Trailer / 55-Video game console / 176-PlayStation 4 / 292-Handheld game console / 317-PlayStation / 682-PlayStation Vita / \n", "681-Leather / 1525-Wallet / 2669-Leather crafting / \n", "0-Games / 2-Video game / 33-Weapon / 850-Left 4 Dead / 963-Left 4 Dead 2 / \n", "36-Piano / 53-Keyboard / 104-Pianist / \n", "122-Weight training / 204-Gym / 214-Muscle / 1430-Pull-up / \n", "10-Animal / 48-Pet / 71-Dog / 3173-Great Dane / \n", "56-Hair / 106-Hairstyle / 200-Doll / \n", "6-Animation / 15-Cartoon / 116-Pokémon / 1104-Ash Ketchum / 3347-Misty / \n", "3-Concert / 7-Musician / 13-Musical ensemble / 30-Drummer / 39-Drums / 1240-Mambo / \n", "1-Vehicle / 4-Car / 11-Motorsport / 41-Sports car / \n", "0-Games / 2-Video game / 35-Minecraft / 127-Xbox 360 / 181-Xbox / 321-Roller coaster / \n", "66-Bollywood / \n", "1-Vehicle / 4-Car / 11-Motorsport / 17-Racing / 40-Road / 57-Race track / 169-Rallying / \n", "3-Concert / \n", "9-Music video / \n", "6-Animation / 379-Touhou Project / \n", "14-Guitar / 24-String instrument / 63-Acoustic guitar / \n", "2-Video game / 20-Trailer / \n", "5-Dance / \n", "23-Mobile phone / 29-Smartphone / 178-Samsung Galaxy / 1287-Samsung Galaxy Tab series / 1390-Samsung Galaxy S II / \n", "1-Vehicle / 50-Aircraft / 82-Airplane / 129-Aviation / 139-Airport / 151-Landing / 157-Jet aircraft / 175-Airliner / 187-Takeoff / 188-Airline / 254-Runway / 719-Boeing 747 / 2790-British Airways / \n", "5-Dance / 256-Tennis / \n", "0-Games / 2-Video game / 51-Strategy video game / 127-Xbox 360 / 181-Xbox / \n", "0-Games / 8-Football / 1016-Glove / \n", "0-Games / 19-PC game / 35-Minecraft / \n", "1-Vehicle / 4-Car / 27-Motorcycle / 40-Road / 105-Motorcycling / 1414-Gasoline / \n", "1-Vehicle / 4-Car / 11-Motorsport / 493-Rim / \n", "67-Cycling / 69-Bicycle / 83-Skateboarding / 221-Skatepark / 377-BMX bike / \n", "7-Musician / 22-Nature / 46-Choir / 86-Plant / \n", "1604-Satan / \n", "0-Games / 2-Video game / 8-Football / 21-Stadium / 73-Ball / 79-American football / 90-Sports game / 561-Madden NFL / \n", "9-Music video / \n", "121-Photography / 155-Camera / 566-Digital camera / 629-Camera lens / 721-Digital SLR / \n", "0-Games / 49-School / 60-Basketball / 168-Basketball moves / 306-High school / \n", "12-Food / 26-Cooking / 32-Recipe / 52-Dish / 274-Meat / 688-Beef / 752-Pasta / 1230-Italian food / 2562-Meatball / \n", "5-Dance / 259-Fire / 350-Circus / 788-Juggling / \n", "0-Games / 59-Winter sport / 81-Athlete / 112-Ice skating / 142-Hockey / 312-Ice rink / \n", "9-Music video / \n", "1-Vehicle / 18-Outdoor recreation / 67-Cycling / 69-Bicycle / 220-Mountain bike / 645-Road bicycle / \n", "0-Games / 2-Video game / 6-Animation / 15-Cartoon / 123-Naruto / 285-Sasuke Uchiha / 591-Naruto: Ultimate Ninja / \n", "85-Combat / 117-Boxing / \n", "256-Tennis / 974-Racket / \n", "3-Concert / 16-Performance art / \n", "3-Concert / 7-Musician / 13-Musical ensemble / 38-Orchestra / \n", "3-Concert / 9-Music video / \n", "5-Dance / 49-School / \n", "1-Vehicle / 27-Motorcycle / \n", "1-Vehicle / 25-Toy / 62-Train / 3396-Ressha Sentai ToQger / \n", "28-Fashion / \n", "1082-Knot / 1369-Necktie / \n", "28-Fashion / 45-Cosmetics / 235-Lipstick / \n", "380-Assassin's Creed / 1522-Assassin's Creed II / \n", "22-Nature / \n", "3-Concert / 7-Musician / 13-Musical ensemble / \n", "0-Games / 2-Video game / 8-Football / 21-Stadium / 54-Highlight film / 73-Ball / 90-Sports game / 453-Pro Evolution Soccer / \n", "9-Music video / \n", "9-Music video / 308-Ship / \n", "0-Games / 60-Basketball / 168-Basketball moves / 255-Slam dunk / \n", "754-Gramophone record / \n", "23-Mobile phone / 29-Smartphone / 37-Gadget / 553-Sony Xperia / 2770-Sony Xperia Z1 / \n", "12-Food / 26-Cooking / 32-Recipe / 52-Dish / 58-Cuisine / 109-Cooking show / 226-Dessert / \n", "365-Fireworks / \n", "75-Wrestling / 85-Combat / \n", "260-Prayer / \n", "75-Wrestling / \n", "0-Games / 1000-AdventureQuest Worlds / 1178-Quest / \n", "2-Video game / 192-Battlefield / 374-Battlefield 3 / \n", "1-Vehicle / 50-Aircraft / 82-Airplane / 129-Aviation / 151-Landing / 157-Jet aircraft / 187-Takeoff / 388-Fighter aircraft / \n", "3-Concert / 7-Musician / 13-Musical ensemble / \n", "3-Concert / 7-Musician / 13-Musical ensemble / 16-Performance art / \n", "1411-ESPN / \n", "67-Cycling / 69-Bicycle / \n", "9-Music video / \n", "3-Concert / 5-Dance / 16-Performance art / \n", "6-Animation / 15-Cartoon / 2948-Dogfight / 2974-Macross / \n", "0-Games / 8-Football / \n", "10-Animal / 18-Outdoor recreation / 42-Fishing / 91-Fish / 147-River / 268-Recreational fishing / 478-Fishing rod / 698-Fisherman / 844-Bass / 3292-Rock fishing / \n", "5-Dance / 95-Talent show / 263-Light / \n", "1-Vehicle / 4-Car / \n", "28-Fashion / 390-Model / \n", "18-Outdoor recreation / 83-Skateboarding / 108-Skateboard / 221-Skatepark / \n", "1-Vehicle / 50-Aircraft / 82-Airplane / 129-Aviation / 929-Aerobatics / \n", "6-Animation / 15-Cartoon / \n", "18-Outdoor recreation / 22-Nature / 164-Beach / 223-Weather / 233-Ocean / 423-Coast / 502-Sky / \n", "25-Toy / 35-Minecraft / 140-LEGO / 264-Star Wars / 852-Lego Star Wars / 1452-BIONICLE / 1478-Lego Ninjago / 3181-Lego Legends of Chima / \n", "1-Vehicle / 4-Car / 11-Motorsport / 17-Racing / 57-Race track / 651-Stock car racing / \n", "162-Medicine / 209-University / \n", "3-Concert / 7-Musician / 14-Guitar / 24-String instrument / 39-Drums / \n", "1-Vehicle / 4-Car / 11-Motorsport / 40-Road / 74-Truck / 167-Four-wheel drive / 198-Off-road vehicle / 327-Mud / 375-Pickup truck / 565-Mud bogging / 4251-Chevrolet K5 Blazer / \n", "567-Compact disc / \n", "3-Concert / 7-Musician / 16-Performance art / \n", "3-Concert / 133-Nightclub / \n", "1-Vehicle / 11-Motorsport / 124-Tractor / 156-Agriculture / 252-Tractor pulling / 459-Plough / \n", "28-Fashion / \n", "5-Dance / 95-Talent show / \n", "1-Vehicle / 4-Car / 1163-Carpet / \n", "8-Football / \n", "9-Music video / \n", "219-Accordion / 655-Diatonic button accordion / \n", "0-Games / 33-Weapon / 1394-Alliance of Valiant Arms / \n", "0-Games / 6-Animation / 2750-Mii / \n", "78-Wedding / 190-Bride / \n", "1-Vehicle / 4-Car / 27-Motorcycle / 40-Road / \n", "3-Concert / 5-Dance / 16-Performance art / 46-Choir / \n", "2224-Sherlock Holmes / \n", "12-Food / 86-Plant / \n", "0-Games / 19-PC game / 35-Minecraft / \n", "3-Concert / 5-Dance / 28-Fashion / 98-Festival / 785-Jeans / \n", "432-Kingdom Hearts / 1730-Roxas / 3635-Kingdom Hearts 358/2 Days / \n", "5-Dance / 115-Ballet / \n", "0-Games / 8-Football / 21-Stadium / 65-Kick / 76-Ball / 81-Athlete / \n", "1-Vehicle / 4-Car / 11-Motorsport / 41-Sports car / 135-Drifting / \n", "3-Concert / 36-Piano / 53-Keyboard / 107-Musical keyboard / \n", "1-Vehicle / 4-Car / 70-Driving / 484-Subwoofer / \n", "3-Concert / 16-Performance art / \n", "59-Winter sport / 112-Ice skating / 142-Hockey / \n", "3-Concert / 7-Musician / 44-Drums / \n", "322-Mixtape / \n", "12-Food / 1088-Infomercial / 1162-Dietary supplement / \n", "1-Vehicle / 62-Train / 119-Locomotive / 143-Railroad car / 161-Train station / 2824-Diesel locomotive / \n", "1-Vehicle / 62-Train / 64-Transport / 103-Rail transport / 119-Locomotive / 143-Railroad car / 152-Track / 161-Train station / \n", "12-Food / \n", "0-Games / 2-Video game / 951-Teenage Mutant Ninja Turtles / \n", "2080-Cup / \n", "10-Animal / 18-Outdoor recreation / 42-Fishing / 91-Fish / 147-River / 225-Lake / 510-Fishing lure / 896-Largemouth bass / \n", "25-Toy / 148-Painting / 303-Paint / 873-Acrylic paint / \n", "117-Boxing / \n", "86-Plant / 136-Gardening / 154-Wheel / 158-Tire / 279-Garden / \n", "273-News program / \n", "1-Vehicle / 1879-Hot air balloon / \n", "122-Weight training / 125-Television / \n", "196-Eye / 2120-Tears / \n", "1-Vehicle / 4-Car / 25-Toy / 27-Motorcycle / 40-Road / 105-Motorcycling / 154-Wheel / 207-Exhaust system / 1380-Kawasaki motorcycles / 1951-Sport bike / \n", "9-Music video / 84-Snow / \n", "1-Vehicle / 4-Car / 11-Motorsport / 3839-Volkswagen Golf Mk5 / \n", "402-Jewellery / 458-Gold / 521-Metal / 731-Coin / 912-Silver / 1001-Gemstone / 1300-Earrings / \n", "31-Disc jockey / \n", "3-Concert / \n", "59-Winter sport / 83-Skateboarding / 112-Ice skating / 703-Roller skating / 1783-Roller skates / \n", "3-Concert / 7-Musician / 16-Performance art / \n", "0-Games / 2-Video game / 411-MapleStory / \n", "8-Football / \n", "75-Wrestling / 85-Combat / \n", "1909-Analog synthesizer / \n", "1-Vehicle / 4-Car / 1134-Taxicab / \n", "0-Games / 1-Vehicle / 2-Video game / 4-Car / 34-Action-adventure game / 111-PlayStation 3 / 118-Grand Theft Auto V / \n", "3-Concert / \n", "12-Food / 26-Cooking / 52-Dish / 99-Christmas / \n", "0-Games / 85-Combat / 211-RuneScape / \n", "1-Vehicle / 50-Aircraft / 92-Radio-controlled model / 149-Model aircraft / 160-Radio-controlled aircraft / 266-Helicopter / 392-Radio-controlled helicopter / 393-Unmanned aerial vehicle / 455-Unidentified flying object / 604-Quadcopter / \n", "144-Amusement park / \n", "0-Games / 8-Football / 21-Stadium / 65-Kick / 77-Arena / 1247-Penalty kick / \n", "28-Fashion / 676-Handbag / 835-Backpack / \n", "1054-Vacuum cleaner / \n", "0-Games / 2-Video game / 6-Animation / 20-Trailer / 111-PlayStation 3 / 176-PlayStation 4 / 849-Tales / 3883-Tales of Xillia / \n", "5-Dance / 16-Performance art / 350-Circus / \n", "28-Fashion / 153-Dress / \n", "0-Games / 43-Call of Duty / \n", "0-Games / 2-Video game / 116-Pokémon / 186-Pokémon / \n", "99-Christmas / 293-Slide show / 558-Santa Claus / \n", "5-Dance / 13-Musical ensemble / 250-Parade / \n", "0-Games / 8-Football / 21-Stadium / 65-Kick / 73-Ball / 76-Ball / 81-Athlete / \n", "0-Games / 2-Video game / 241-Fighting game / 311-Street Fighter / 461-Street Fighter IV / 496-Super Street Fighter IV / 1732-Oni / \n", "0-Games / 1-Vehicle / 4-Car / 11-Motorsport / 17-Racing / \n", "89-Comics / 185-Dragon Ball / 244-Goku / 418-Gohan / \n", "28-Fashion / 282-Skin / \n", "3-Concert / 7-Musician / 13-Musical ensemble / 38-Orchestra / 39-Drums / \n", "0-Games / 2-Video game / 19-PC game / 33-Weapon / 43-Call of Duty / 111-PlayStation 3 / 422-Call of Duty: Ghosts / \n", "2-Video game / 6-Animation / 15-Cartoon / 20-Trailer / 123-Naruto / 591-Naruto: Ultimate Ninja / \n", "0-Games / 51-Strategy video game / 831-Plants vs. Zombies / \n", "9-Music video / 110-Album / \n", "5-Dance / 16-Performance art / \n", "20-Trailer / 434-Ninja / 951-Teenage Mutant Ninja Turtles / \n", "1-Vehicle / 67-Cycling / 69-Bicycle / 197-Trail / 220-Mountain bike / 284-Mountain biking / 645-Road bicycle / \n", "1-Vehicle / 50-Aircraft / 82-Airplane / \n", "0-Games / 8-Football / 54-Highlight film / 79-American football / \n", "33-Weapon / 992-Rocket / \n", "9-Music video / \n", "7-Musician / 44-Drums / 410-Teacher / 1442-Tabla / \n", "8-Football / 259-Fire / \n", "1-Vehicle / 2-Video game / 4-Car / 34-Action-adventure game / 118-Grand Theft Auto V / 158-Tire / \n", "0-Games / 49-School / 60-Basketball / \n", "45-Cosmetics / 56-Hair / \n", "337-Advertising / 512-Printing / 1757-Screen printing / \n", "131-Computer / 141-Microsoft Windows / \n", "3-Concert / \n", "49-School / 163-Running / \n", "6-Animation / 712-Outer space / \n", "0-Games / 2-Video game / 19-PC game / 96-Soldier / 192-Battlefield / 374-Battlefield 3 / \n", "0-Games / 59-Winter sport / 77-Arena / 81-Athlete / 112-Ice skating / 142-Hockey / 312-Ice rink / \n", "5-Dance / 31-Disc jockey / \n", "5-Dance / 49-School / \n", "18-Outdoor recreation / 287-Human swimming / 710-Handball / 2360-Water polo / \n", "23-Mobile phone / 29-Smartphone / 37-Gadget / 47-Personal computer / 131-Computer / 141-Microsoft Windows / \n", "2493-The Hobbit / \n", "3-Concert / 7-Musician / \n", "10-Animal / 80-Horse / 247-Stallion / 289-Dressage / \n", "5-Dance / 16-Performance art / \n", "5-Dance / 31-Disc jockey / \n", "2-Video game / \n", "5-Dance / \n", "18-Outdoor recreation / 59-Winter sport / 84-Snow / 120-Winter / 177-Skiing / 305-Ski / \n", "5-Dance / \n", "31-Disc jockey / \n", "0-Games / 2-Video game / 116-Pokémon / 186-Pokémon / 1204-Pokémon Ruby and Sapphire / 2933-Pokémon Trainer / \n", "0-Games / 8-Football / \n", "3-Concert / \n", "0-Games / 49-School / 54-Highlight film / 60-Basketball / 76-Ball / 1101-Point guard / \n", "0-Games / 8-Football / 65-Kick / 73-Ball / 76-Ball / \n", "2-Video game / 25-Toy / 140-LEGO / 884-Lego minifigure / 1452-BIONICLE / \n", "7-Musician / 13-Musical ensemble / 49-School / 203-Marching band / \n", "3-Concert / 5-Dance / 133-Nightclub / \n", "5-Dance / \n", "219-Accordion / \n", "1-Vehicle / 4-Car / 70-Driving / 114-Engine / \n", "0-Games / \n", "25-Toy / 1154-Anpanman / \n", "10-Animal / 132-Bird / 404-Livestock / 424-Chicken / 1136-Poultry / 1428-Rooster / \n", "9-Music video / \n", "162-Medicine / \n", "1-Vehicle / 4-Car / 154-Wheel / 158-Tire / 327-Mud / 3828-Amphibious ATV / \n", "1-Vehicle / 4-Car / 11-Motorsport / 17-Racing / 41-Sports car / 918-Brake / 953-Gran Turismo 5 / 983-Lexus / \n", "14-Guitar / 24-String instrument / \n", "28-Fashion / 206-Shoe / 681-Leather / 757-Boot / \n", "28-Fashion / 45-Cosmetics / 173-Eye shadow / 193-Eye liner / 195-Mascara / 196-Eye / \n", "1-Vehicle / 4-Car / 27-Motorcycle / 154-Wheel / \n", "3-Concert / 5-Dance / 16-Performance art / 1546-Line / \n", "0-Games / 1-Vehicle / 2-Video game / 4-Car / 34-Action-adventure game / 118-Grand Theft Auto V / 201-Grand Theft Auto: San Andreas / 297-Grand Theft Auto IV / 413-PlayStation 2 / 858-Carl Johnson / \n", "3-Concert / \n", "3-Concert / 7-Musician / 14-Guitar / 361-Flamenco / \n", "5-Dance / \n", "9-Music video / \n", "10-Animal / 15-Cartoon / 48-Pet / 71-Dog / 166-Cat / 346-Kitten / 2997-Talking Tom and Friends / \n", "0-Games / 8-Football / 21-Stadium / 79-American football / 81-Athlete / \n", "3-Concert / 5-Dance / 16-Performance art / \n", "1106-Character / 2368-Marvel Universe / \n", "28-Fashion / 206-Shoe / 333-Nike; Inc. / 368-Sneakers / 763-Snake / 4171-Air Force 1 / \n", "1-Vehicle / 50-Aircraft / 82-Airplane / 129-Aviation / 149-Model aircraft / 693-Sword / 778-Gliding / 1317-Glider / \n", "6-Animation / 15-Cartoon / \n", "28-Fashion / 56-Hair / 1359-Hair iron / 1827-Hair straightening / \n", "12-Food / 25-Toy / 42-Fishing / 1154-Anpanman / \n", "8-Football / \n", "6-Animation / 9-Music video / 15-Cartoon / \n", "1-Vehicle / 4-Car / 40-Road / 259-Fire / 371-Firefighter / \n", "93-Comedy / 355-Comedian / \n", "6-Animation / 61-Art / 97-Drawing / 148-Painting / 628-Portrait / \n", "1149-Castle / 2033-Princess Peach / 2708-Paper Mario / \n", "10-Animal / 48-Pet / 71-Dog / \n", "89-Comics / 781-Pachinko / \n", "0-Games / 2-Video game / 33-Weapon / 192-Battlefield / 383-Battlefield 4 / \n", "3-Concert / 30-Drummer / \n", "5-Dance / 95-Talent show / \n", "1288-Marvel vs. Capcom / 1505-Ultimate Marvel vs. Capcom 3 / 3109-Clint Barton / \n", "3-Concert / 7-Musician / 13-Musical ensemble / 16-Performance art / 46-Choir / 49-School / \n", "250-Parade / \n", "3-Concert / \n", "1-Vehicle / 11-Motorsport / \n", "3-Concert / 7-Musician / 13-Musical ensemble / 46-Choir / 318-Church / \n", "7-Musician / 46-Choir / 307-Hand / \n", "3-Concert / 7-Musician / 13-Musical ensemble / 14-Guitar / 24-String instrument / 594-Music festival / \n", "24-String instrument / 842-Harp / \n", "1-Vehicle / 11-Motorsport / 17-Racing / 27-Motorcycle / 172-Motocross / \n", "223-Weather / 490-Rain / 499-Cloud / 502-Sky / 958-Thunderstorm / \n", "7-Musician / 13-Musical ensemble / 46-Choir / 78-Wedding / 209-University / \n", "22-Nature / 42-Fishing / 275-Diving / 288-Underwater / \n", "1-Vehicle / 50-Aircraft / 139-Airport / 151-Landing / \n", "0-Games / 148-Painting / 878-Composer / \n", "79-American football / \n", "3-Concert / 7-Musician / 13-Musical ensemble / 38-Orchestra / \n", "9-Music video / \n", "0-Games / 165-League of Legends / \n", "6-Animation / 455-Unidentified flying object / \n", "25-Toy / 140-LEGO / 884-Lego minifigure / \n", "0-Games / 1109-Counter-Strike Online / \n", "8-Football / \n", "238-Nail / 300-Nail art / \n", "9-Music video / \n", "1-Vehicle / 62-Train / 64-Transport / 103-Rail transport / 143-Railroad car / \n", "10-Animal / 25-Toy / 329-Swimming pool / 524-Littlest Pet Shop / \n", "1-Vehicle / 4-Car / 74-Truck / 158-Tire / 234-Ford / 484-Subwoofer / 3684-Ford Bronco / \n", "0-Games / 831-Plants vs. Zombies / 1185-Plasticine / \n", "2-Video game / 6-Animation / \n", "59-Winter sport / 112-Ice skating / 316-Figure skating / 632-Ice dancing / \n", "359-Harry Potter (Literary Series) / 1436-Draco Malfoy / 1856-Hogwarts School of Witchcraft and Wizardry / \n", "0-Games / 462-Table / 544-Table tennis / 1615-Pong / \n", "23-Mobile phone / 29-Smartphone / 37-Gadget / 101-Telephone / 802-Google Nexus / \n", "45-Cosmetics / 193-Eye liner / \n", "1-Vehicle / 62-Train / 64-Transport / 103-Rail transport / 119-Locomotive / 143-Railroad car / 152-Track / 161-Train station / \n", "8-Football / 259-Fire / \n", "0-Games / 2-Video game / 3747-Wario / \n", "0-Games / 2-Video game / 411-MapleStory / \n", "0-Games / 2-Video game / 20-Trailer / 127-Xbox 360 / 181-Xbox / 1425-Gears of War 2 / \n", "3-Concert / \n", "6-Animation / 15-Cartoon / 116-Pokémon / 1104-Ash Ketchum / 3347-Misty / \n", "28-Fashion / 206-Shoe / 1300-Earrings / \n", "794-Bridge / \n", "117-Boxing / \n", "9-Music video / 390-Model / \n", "6-Animation / 183-Manga / 2065-Tsubasa: Reservoir Chronicle / \n", "1-Vehicle / 4-Car / 20-Trailer / 245-BMW / \n", "0-Games / 2-Video game / 55-Video game console / 2034-Dreamcast / \n", "3-Concert / 7-Musician / 13-Musical ensemble / 38-Orchestra / 230-Brass instrument / 427-Trumpet / \n", "567-Compact disc / \n", "98-Festival / \n", "3-Concert / 7-Musician / 30-Drummer / \n", "10-Animal / 125-Television / 384-Horse racing / \n", "75-Wrestling / 1042-Jumbotron / \n", "6-Animation / 89-Comics / 413-PlayStation 2 / \n", "49-School / \n", "863-Body piercing / \n", "3151-Kermit the Frog / \n", "1-Vehicle / 4-Car / 683-Renault / 1242-Car wash / 1678-Foam / \n", "1-Vehicle / 4-Car / 11-Motorsport / 40-Road / 64-Transport / 286-Bus / \n", "1-Vehicle / 4-Car / 114-Engine / 360-Jeep / 982-V8 engine / 1406-Jeep Cherokee / 1984-Jeep Grand Cherokee / \n", "8-Football / \n", "3-Concert / 16-Performance art / \n", "516-Headphones / \n", "117-Boxing / \n", "9-Music video / \n", "1-Vehicle / 4-Car / 40-Road / 70-Driving / 94-Dashcam / 212-Highway / 800-Mountain pass / \n", "12-Food / 32-Recipe / 282-Skin / 385-Face / 641-Mask / 784-Honey / 3404-Oatmeal / \n", "20-Trailer / \n", "122-Weight training / 214-Muscle / 390-Model / \n", "18-Outdoor recreation / 22-Nature / 267-Mountain / 283-Rock / 614-Hiking / 1808-National park / \n", "7-Musician / 30-Drummer / 39-Drums / 44-Drums / 1033-Drill / \n", "170-Home improvement / 208-Wood / 246-Furniture / 462-Table / \n", "56-Hair / 607-Lace / \n", "1-Vehicle / 4-Car / 11-Motorsport / 27-Motorcycle / 64-Transport / 105-Motorcycling / 336-Scooter / 482-Asphalt / \n", "0-Games / 2-Video game / 89-Comics / \n", "28-Fashion / 469-Bag / 676-Handbag / 681-Leather / \n", "2-Video game / 23-Mobile phone / 29-Smartphone / 37-Gadget / 72-iPhone / 189-iPad / 272-iPod / 331-iPod touch / 554-iPhone 4 / 675-iPhone 4S / 1009-iPhone 3G / 1155-iPhone 3GS / \n", "2-Video game / 6-Animation / 477-Transformers / 2186-Megatron / 4445-Transformers: Beast Wars / \n", "31-Disc jockey / 133-Nightclub / \n", "0-Games / 2-Video game / 19-PC game / 43-Call of Duty / 422-Call of Duty: Ghosts / \n", "3-Concert / \n", "9-Music video / \n", "1995-Hammer / \n", "23-Mobile phone / 29-Smartphone / 37-Gadget / 101-Telephone / 553-Sony Xperia / 3019-Sony Ericsson Xperia X10 / \n", "5-Dance / 747-Retail / 1328-Grocery store / \n", "0-Games / 2-Video game / 2237-Kantai Collection / \n", "402-Jewellery / 457-Knitting / 829-Bead / 1082-Knot / 2961-Button / \n", "8-Football / \n", "23-Mobile phone / 29-Smartphone / 178-Samsung Galaxy / 1727-Samsung Galaxy Note 3 / \n", "66-Bollywood / \n", "3-Concert / \n", "5-Dance / 78-Wedding / 190-Bride / 421-Gown / \n", "0-Games / 2-Video game / 243-Sonic the Hedgehog / 251-Sonic the Hedgehog / 4019-Sonic the Hedgehog CD / \n", "233-Ocean / 248-Hotel / \n", "3-Concert / \n", "791-Magic: The Gathering / \n", "61-Art / 148-Painting / 303-Paint / 541-Black-and-white / 2124-Spray painting / 2853-Aerosol paint / \n", "0-Games / 15-Cartoon / 759-Icing / \n", "20-Trailer / \n", "0-Games / 8-Football / \n", "1-Vehicle / 62-Train / 64-Transport / 103-Rail transport / 119-Locomotive / 143-Railroad car / 161-Train station / 224-Rapid transit / \n", "6-Animation / 15-Cartoon / 183-Manga / \n", "0-Games / 2-Video game / 33-Weapon / 34-Action-adventure game / 380-Assassin's Creed / 1048-Assassin's Creed / \n", "66-Bollywood / \n", "9-Music video / \n", "3-Concert / \n", "1-Vehicle / 4-Car / 11-Motorsport / 17-Racing / 25-Toy / 41-Sports car / 92-Radio-controlled model / 135-Drifting / 202-Radio-controlled car / 569-Lamborghini / 926-Traxxas / 4631-Lamborghini Diablo / \n", "0-Games / 2-Video game / 34-Action-adventure game / 118-Grand Theft Auto V / \n", "3-Concert / 7-Musician / \n", "1-Vehicle / 147-River / 794-Bridge / \n", "2641-Radha / \n", "1-Vehicle / 27-Motorcycle / 158-Tire / 806-Chopper / \n", "5-Dance / 16-Performance art / 49-School / 95-Talent show / \n", "0-Games / 8-Football / 21-Stadium / 65-Kick / 73-Ball / 76-Ball / 81-Athlete / \n", "20-Trailer / \n", "0-Games / 2-Video game / 55-Video game console / 199-Wii / 320-Super Smash Bros. / 335-Nintendo 3DS / 443-Wii U / 697-Super Smash Bros. for Nintendo 3DS and Wii U / \n", "164-Beach / 248-Hotel / 369-Resort / \n", "49-School / 949-Rowing / \n", "110-Album / \n", "22-Nature / 490-Rain / 958-Thunderstorm / \n", "3-Concert / \n", "22-Nature / 423-Coast / \n", "419-Earth / 593-Moon / 1686-Orbit / \n", "0-Games / 1976-Nostale / \n", "8-Football / 21-Stadium / 77-Arena / \n", "9-Music video / 2461-DatPiff / \n", "85-Combat / 209-University / \n", "0-Games / 1-Vehicle / 2-Video game / 11-Motorsport / 17-Racing / 55-Video game console / 57-Race track / 199-Wii / 456-Go-kart / 518-Mario Kart / 3346-Super Mario Kart / \n", "0-Games / 8-Football / 90-Sports game / 180-FIFA 15 / 480-FIFA 13 / \n", "0-Games / 2-Video game / 55-Video game console / 199-Wii / \n", "3-Concert / \n", "1-Vehicle / 74-Truck / 259-Fire / 371-Firefighter / 796-Fire engine / 2414-Firefighting / \n", "9-Music video / \n", "36-Piano / 53-Keyboard / \n", "23-Mobile phone / 29-Smartphone / 487-Samsung Electronics / \n", "1-Vehicle / 4-Car / 449-Honda / 552-Honda Civic / \n", "6-Animation / 15-Cartoon / 3600-The Jungle Book / \n", "0-Games / 2-Video game / 19-PC game / 33-Weapon / 43-Call of Duty / 96-Soldier / 127-Xbox 360 / 194-Call of Duty: Modern Warfare 2 / 302-Call of Duty 4: Modern Warfare / 464-First-person Shooter / \n", "31-Disc jockey / \n", "5-Dance / 49-School / \n", "1-Vehicle / 4-Car / 27-Motorcycle / 1124-Yamaha YZF-R1 / \n", "3-Concert / 5-Dance / 350-Circus / \n", "7-Musician / 14-Guitar / 24-String instrument / 63-Acoustic guitar / \n", "0-Games / 2-Video game / 34-Action-adventure game / 118-Grand Theft Auto V / \n", "0-Games / 2-Video game / 159-World of Warcraft / 1356-Paladin / \n", "7-Musician / 14-Guitar / 99-Christmas / 2252-Steel guitar / 2809-Pedal steel guitar / \n", "3-Concert / 9-Music video / \n", "144-Amusement park / 321-Roller coaster / 334-Amusement ride / \n", "3-Concert / 7-Musician / 13-Musical ensemble / 16-Performance art / 30-Drummer / \n", "0-Games / 59-Winter sport / 112-Ice skating / 142-Hockey / 312-Ice rink / \n", "118-Grand Theft Auto V / 120-Winter / \n", "3-Concert / 7-Musician / 13-Musical ensemble / 24-String instrument / 38-Orchestra / 150-Violin / 401-Cello / 450-Viola / 545-Quartet (ensemble) / 1002-String quartet / \n", "6-Animation / 20-Trailer / 298-Television advertisement / \n", "10-Animal / 22-Nature / 80-Horse / 290-Pony / 956-Glasses / \n", "3-Concert / \n", "0-Games / 2-Video game / 23-Mobile phone / 29-Smartphone / 37-Gadget / 101-Telephone / \n", "359-Harry Potter (Literary Series) / 4489-Harry Potter and the Philosopher's Stone / \n", "1-Vehicle / 4-Car / 11-Motorsport / 17-Racing / 135-Drifting / \n", "7-Musician / 13-Musical ensemble / 14-Guitar / 24-String instrument / 100-Electric guitar / \n", "85-Combat / \n", "10-Animal / 18-Outdoor recreation / 80-Horse / 404-Livestock / 438-Mare / 523-Cattle / 1586-Cowboy / \n", "1-Vehicle / 4-Car / 18-Outdoor recreation / 167-Four-wheel drive / 198-Off-road vehicle / 277-Sport utility vehicle / 360-Jeep / \n", "1-Vehicle / 4-Car / 50-Aircraft / 114-Engine / 129-Aviation / 151-Landing / 451-Cockpit / \n", "463-Cricket / \n", "14-Guitar / \n", "0-Games / 2-Video game / 241-Fighting game / 764-The King of Fighters / \n", "1-Vehicle / 11-Motorsport / 27-Motorcycle / 105-Motorcycling / 197-Trail / 213-Forest / 220-Mountain bike / 256-Tennis / 491-Enduro / \n", "0-Games / 2-Video game / 55-Video game console / 125-Television / 413-PlayStation 2 / 581-PlayStation / 1349-USB flash drive / \n", "573-Paragliding / 765-Parachute / \n", "7-Musician / 13-Musical ensemble / 14-Guitar / 24-String instrument / 639-Fiddle / 670-Banjo / \n", "0-Games / 8-Football / 90-Sports game / 180-FIFA 15 / \n", "0-Games / 2-Video game / 102-Building / 1027-Terraria / \n", "33-Weapon / 121-Photography / 3258-Leica Camera / 4195-Binoculars / \n", "0-Games / 2-Video game / 75-Wrestling / 85-Combat / 90-Sports game / 3764-EA Sports UFC / \n", "49-School / 209-University / \n", "14-Guitar / 24-String instrument / 100-Electric guitar / \n", "0-Games / 8-Football / 21-Stadium / 65-Kick / 76-Ball / \n", "9-Music video / \n", "9-Music video / \n", "1-Vehicle / 4-Car / 11-Motorsport / 17-Racing / 27-Motorcycle / 57-Race track / 1448-Vespa / \n", "1-Vehicle / 11-Motorsport / 17-Racing / 18-Outdoor recreation / 27-Motorcycle / 105-Motorcycling / 172-Motocross / 197-Trail / 213-Forest / 220-Mountain bike / 491-Enduro / \n", "3-Concert / 5-Dance / \n", "23-Mobile phone / 29-Smartphone / 37-Gadget / 72-iPhone / \n", "3-Concert / 5-Dance / \n", "1345-Dean Winchester / \n", "0-Games / 2-Video game / 35-Minecraft / 55-Video game console / 111-PlayStation 3 / 127-Xbox 360 / 181-Xbox / \n", "2062-Kidney / \n", "23-Mobile phone / \n", "200-Doll / 4237-Ball-jointed doll / \n", "6-Animation / 15-Cartoon / 61-Art / 97-Drawing / 148-Painting / 290-Pony / \n", "7-Musician / 13-Musical ensemble / \n", "5-Dance / \n", "1-Vehicle / 4-Car / 11-Motorsport / 17-Racing / 40-Road / 212-Highway / 286-Bus / \n", "15-Cartoon / \n", "0-Games / 8-Football / \n", "20-Trailer / \n", "5-Dance / \n", "14-Guitar / \n", "2-Video game / 192-Battlefield / 374-Battlefield 3 / \n", "12-Food / 295-Chocolate / 442-Bread / \n", "1-Vehicle / 4-Car / 360-Jeep / 395-Camping / 707-Roof / 914-Tent / \n", "0-Games / 2-Video game / 34-Action-adventure game / 380-Assassin's Creed / 1522-Assassin's Creed II / \n", "1-Vehicle / 4-Car / 41-Sports car / \n", "0-Games / 35-Minecraft / 3175-Hide-and-seek / \n", "3-Concert / 5-Dance / 16-Performance art / 31-Disc jockey / 3458-DMC World DJ Championships / \n", "142-Hockey / 288-Underwater / \n", "992-Rocket / 1508-Spacecraft / 1779-Satellite / 4397-Rocket launch / \n", "1-Vehicle / 50-Aircraft / 3752-Airship / \n", "12-Food / 367-Drink / 904-Cocktail / 1669-Vodka / \n", "293-Slide show / \n", "7-Musician / \n", "2-Video game / 33-Weapon / 474-Destiny / 1130-Armour / \n", "75-Wrestling / 85-Combat / 122-Weight training / 204-Gym / 1584-Kettlebell / \n", "420-The Elder Scrolls V: Skyrim / 441-The Elder Scrolls / \n", "1-Vehicle / 50-Aircraft / 82-Airplane / 129-Aviation / 1200-Ultralight aviation / \n", "1-Vehicle / 8-Football / 286-Bus / \n", "1-Vehicle / 4-Car / 11-Motorsport / 17-Racing / 41-Sports car / 135-Drifting / 2718-Gran Turismo 4 / \n", "12-Food / 26-Cooking / 32-Recipe / 52-Dish / 58-Cuisine / 109-Cooking show / 301-Meal / 442-Bread / 966-Sandwich / \n", "14-Guitar / 24-String instrument / 63-Acoustic guitar / 100-Electric guitar / \n", "0-Games / 2-Video game / 910-Dragon Quest / \n", "5-Dance / \n", "5-Dance / \n", "9-Music video / \n", "1-Vehicle / 88-Machine / 124-Tractor / 156-Agriculture / \n", "0-Games / 2-Video game / 412-Monster / \n", "1088-Infomercial / \n", "6-Animation / 15-Cartoon / 123-Naruto / 183-Manga / 541-Black-and-white / \n", "59-Winter sport / 83-Skateboarding / 281-Snowboarding / \n", "1-Vehicle / 4-Car / 11-Motorsport / 17-Racing / 41-Sports car / 57-Race track / 135-Drifting / \n", "67-Cycling / 69-Bicycle / 221-Skatepark / 377-BMX bike / \n", "9-Music video / \n", "7-Musician / 14-Guitar / 24-String instrument / 63-Acoustic guitar / \n", "0-Games / 51-Strategy video game / 373-Clash of Clans / 924-Trophy / \n", "6-Animation / 10-Animal / 15-Cartoon / 1428-Rooster / 2832-Looney Tunes / \n", "14-Guitar / \n", "3-Concert / 7-Musician / 13-Musical ensemble / 46-Choir / \n", "209-University / \n", "1-Vehicle / 4-Car / 11-Motorsport / 124-Tractor / 252-Tractor pulling / \n", "0-Games / 19-PC game / 33-Weapon / 596-Monster Hunter / 3910-Mackerel / \n", "3-Concert / 7-Musician / 14-Guitar / 16-Performance art / \n", "28-Fashion / 56-Hair / 106-Hairstyle / 340-Afro-textured hair / \n", "78-Wedding / 293-Slide show / 590-Wedding dress / \n", "1-Vehicle / 40-Road / 64-Transport / 74-Truck / \n", "6-Animation / 1866-Katara / \n", "402-Jewellery / 1082-Knot / 2507-Macramé / \n", "9-Music video / \n", "125-Television / \n", "5-Dance / 130-Gymnastics / 892-Aerobics / 2228-Sport aerobics / \n", "3-Concert / 7-Musician / 16-Performance art / 77-Arena / \n", "6-Animation / 15-Cartoon / 755-Hetalia: Axis Powers / \n", "3-Concert / 1143-Marimba / 1734-Xylophone / 2878-Vibraphone / \n", "433-Construction / 976-Demolition / \n", "3-Concert / 1680-Seafight / \n", "99-Christmas / 415-Sewing / 476-Textile / 964-Christmas tree / 997-Sewing machine / 2329-Felt / 2499-Christmas ornament / \n", "3-Concert / 7-Musician / 13-Musical ensemble / 38-Orchestra / 46-Choir / 296-Saxophone / \n", "144-Amusement park / 321-Roller coaster / 334-Amusement ride / \n", "359-Harry Potter (Literary Series) / 3719-Harry Potter / \n", "3-Concert / \n", "1-Vehicle / 4-Car / 11-Motorsport / 94-Dashcam / \n", "45-Cosmetics / 173-Eye shadow / 193-Eye liner / 195-Mascara / 196-Eye / 218-Eyelash / 235-Lipstick / 291-Rouge / 436-Concealer / 1012-Lip gloss / \n", "0-Games / 1-Vehicle / 2-Video game / 199-Wii / 443-Wii U / 456-Go-kart / 518-Mario Kart / 685-Luigi / 1831-Mario Kart 8 / \n", "162-Medicine / \n", "2-Video game / 6-Animation / 506-Garry's Mod / \n", "5-Dance / \n", "0-Games / 2-Video game / 51-Strategy video game / 125-Television / 165-League of Legends / \n", "779-Point Blank / \n", "298-Television advertisement / 337-Advertising / \n", "694-Tai chi / \n", "85-Combat / 1623-Shaolin Kung Fu / \n", "3-Concert / 7-Musician / 3208-Lowlands / \n", "0-Games / 8-Football / 21-Stadium / 54-Highlight film / 65-Kick / 76-Ball / \n", "1-Vehicle / 11-Motorsport / 17-Racing / 27-Motorcycle / 105-Motorcycling / 172-Motocross / 491-Enduro / \n", "1-Vehicle / 18-Outdoor recreation / 42-Fishing / 87-Boat / 225-Lake / 753-Sailboat / 917-Sail / \n", "12-Food / 26-Cooking / 32-Recipe / 109-Cooking show / 226-Dessert / 232-Cake / 295-Chocolate / 985-Strawberry / 1208-Coconut / \n", "0-Games / 1138-Tower of Saviors / \n", "42-Fishing / \n", "7-Musician / 13-Musical ensemble / 44-Drums / 203-Marching band / 230-Brass instrument / 250-Parade / \n", "110-Album / 754-Gramophone record / \n", "47-Personal computer / 141-Microsoft Windows / 145-Tablet computer / 228-Laptop / 513-Computer keyboard / 631-Cube / \n", "0-Games / 8-Football / 21-Stadium / 54-Highlight film / 79-American football / \n", "6-Animation / 15-Cartoon / 61-Art / 97-Drawing / 396-One Piece / 408-Sketch / \n", "1873-Crossbow / \n", "6-Animation / 15-Cartoon / \n", "170-Home improvement / 574-Elevator / \n", "23-Mobile phone / 29-Smartphone / 72-iPhone / 760-iPhone 5s / 1774-iPhone 5c / \n", "10-Animal / 25-Toy / 48-Pet / 132-Bird / 598-Parrot / \n", "3-Concert / \n", "31-Disc jockey / \n", "12-Food / 26-Cooking / 32-Recipe / 52-Dish / 58-Cuisine / 210-Cookware and bakeware / 1186-Seafood / \n", "0-Games / 8-Football / 21-Stadium / 65-Kick / 73-Ball / \n", "2-Video game / 176-PlayStation 4 / 223-Weather / 441-The Elder Scrolls / 1188-Uncharted / 1401-The Witcher 3: Wild Hunt / 1919-The Elder Scrolls Online / \n", "1-Vehicle / 4-Car / 11-Motorsport / 270-Sedan / 901-Mitsubishi / 1122-Mitsubishi Lancer Evolution / 1413-Mitsubishi Lancer / \n", "22-Nature / 164-Beach / 233-Ocean / 423-Coast / \n", "410-Teacher / \n", "25-Toy / 140-LEGO / 264-Star Wars / 852-Lego Star Wars / 884-Lego minifigure / \n", "31-Disc jockey / \n", "0-Games / 51-Strategy video game / \n", "9-Music video / \n", "20-Trailer / 121-Photography / 633-Photograph / \n", "4591-Texas Longhorns / \n", "10-Animal / 18-Outdoor recreation / 217-Hunting / 636-Rabbit / 2135-Bowhunting / \n", "1-Vehicle / 18-Outdoor recreation / 22-Nature / 67-Cycling / 69-Bicycle / 197-Trail / 220-Mountain bike / 284-Mountain biking / \n", "78-Wedding / 121-Photography / 190-Bride / 293-Slide show / 590-Wedding dress / 2269-Wedding photography / \n", "5-Dance / 13-Musical ensemble / \n", "9-Music video / \n", "3-Concert / \n", "9-Music video / \n", "49-School / 410-Teacher / 829-Bead / \n", "12-Food / 232-Cake / 607-Lace / 759-Icing / 1102-Cake decorating / 1365-Fondant icing / 1541-Buttercream / 2255-Stencil / 2941-Wedding cake / \n", "10-Animal / 22-Nature / 215-Eating / 540-Insect / 642-Leaf / 1733-Ant / \n", "18-Outdoor recreation / 287-Human swimming / 949-Rowing / 2542-Cross / \n", "1-Vehicle / 4-Car / \n", "59-Winter sport / 112-Ice skating / 142-Hockey / 4454-Hockey stick / \n", "28-Fashion / 45-Cosmetics / 1572-Husband / \n", "0-Games / 2-Video game / 55-Video game console / 111-PlayStation 3 / 3605-Puppetry / \n", "75-Wrestling / \n", "7-Musician / 14-Guitar / 24-String instrument / 100-Electric guitar / \n", "7-Musician / 13-Musical ensemble / 46-Choir / \n", "102-Building / \n", "0-Games / 8-Football / 65-Kick / 76-Ball / 81-Athlete / 1037-Grass / \n", "3-Concert / \n", "465-Microphone / 1954-Power supply / \n", "6-Animation / 15-Cartoon / \n", "322-Mixtape / \n", "20-Trailer / \n", "12-Food / 348-Fruit / 2000-Mango / \n", "8-Football / 90-Sports game / 180-FIFA 15 / \n", "1-Vehicle / 22-Nature / 27-Motorcycle / 40-Road / 245-BMW / 2763-BMW R1200GS / \n", "18-Outdoor recreation / 21-Stadium / 448-Planet / \n", "0-Games / 2-Video game / 432-Kingdom Hearts / 1908-Kingdom Hearts Birth by Sleep / 2934-Zack Fair / \n", "12-Food / 25-Toy / 26-Cooking / 179-Kitchen / 1154-Anpanman / \n", "10-Animal / 48-Pet / 71-Dog / 881-German Shepherd / \n", "2693-Puma SE / \n", "0-Games / 211-RuneScape / \n", "38-Orchestra / 1358-Canyon / \n", "1203-Backpacking / \n", "1-Vehicle / 4-Car / 574-Elevator / 1391-Kia / \n", "2-Video game / 47-Personal computer / 131-Computer / 309-Computer hardware / 602-Central processing unit / 621-Asus / 1045-Computer case / \n", "22-Nature / \n", "0-Games / 2-Video game / 34-Action-adventure game / 297-Grand Theft Auto IV / 658-Grand Theft Auto: The Lost and Damned / \n", "110-Album / \n", "12-Food / 1550-Newspaper / \n", "205-Newscaster / 1920-Copper / \n", "0-Games / 2-Video game / 33-Weapon / 474-Destiny / 1578-Red Dead Redemption / \n", "0-Games / 2-Video game / 774-Galaxy / 1397-Super Mario Galaxy / \n", "25-Toy / 435-Action figure / 2120-Tears / \n", "5-Dance / 28-Fashion / \n", "0-Games / 6-Animation / \n", "6-Animation / 15-Cartoon / \n", "3540-Violet / \n", "10-Animal / 48-Pet / 71-Dog / 1252-Dog agility / 2083-Border Collie / 2087-Collie / \n", "7-Musician / 14-Guitar / 24-String instrument / 100-Electric guitar / \n", "113-House / 136-Gardening / 170-Home improvement / \n", "9-Music video / \n", "1-Vehicle / 11-Motorsport / 27-Motorcycle / \n", "122-Weight training / \n", "3-Concert / 7-Musician / \n", "66-Bollywood / \n", "439-Barbie / \n", "0-Games / 8-Football / 21-Stadium / 65-Kick / 73-Ball / \n", "31-Disc jockey / \n", "9-Music video / \n", "12-Food / 26-Cooking / 32-Recipe / \n", "5-Dance / 16-Performance art / \n", "9-Music video / \n", "402-Jewellery / \n", "9-Music video / \n", "208-Wood / 259-Fire / 1739-Fireplace / \n", "20-Trailer / \n", "0-Games / 6-Animation / 88-Machine / 240-Arcade game / 535-Slot machine / \n", "10-Animal / 18-Outdoor recreation / 80-Horse / 1434-Saddle / \n", "12-Food / 26-Cooking / 32-Recipe / 52-Dish / 500-Rice / 1227-Sushi / 1569-Steamed rice / \n", "3-Concert / 7-Musician / 30-Drummer / \n", "3861-Toner / \n", "0-Games / 2-Video game / 3366-Amnesia: The Dark Descent / \n", "67-Cycling / 69-Bicycle / 220-Mountain bike / 520-Bicycle frame / 645-Road bicycle / \n", "46-Choir / \n", "3-Concert / 7-Musician / 13-Musical ensemble / 38-Orchestra / \n", "10-Animal / 91-Fish / 258-Aquarium / 1378-Multiplayer online battle arena / \n", "28-Fashion / 49-School / 56-Hair / 106-Hairstyle / 405-Braid / 1064-French braid / \n", "1-Vehicle / 42-Fishing / 87-Boat / \n", "20-Trailer / 1551-The Phantom of the Opera / \n", "0-Games / 54-Highlight film / \n", "3-Concert / 5-Dance / \n", "6-Animation / 20-Trailer / 126-The Walt Disney Company / \n", "56-Hair / 106-Hairstyle / 1271-Beauty salon / \n", "3-Concert / 7-Musician / 36-Piano / 104-Pianist / 150-Violin / \n", "116-Pokémon / 186-Pokémon / 372-Playing card / 690-Pokémon X and Y / 1021-Booster pack / 1049-Collectible card game / 1062-Pokémon Trading Card Game / 2911-Charizard / \n", "23-Mobile phone / 29-Smartphone / 141-Microsoft Windows / \n", "0-Games / 2-Video game / 535-Slot machine / \n", "31-Disc jockey / \n", "1-Vehicle / 11-Motorsport / 17-Racing / 92-Radio-controlled model / 167-Four-wheel drive / 202-Radio-controlled car / \n", "6-Animation / 15-Cartoon / \n", "75-Wrestling / 85-Combat / \n", "36-Piano / 1023-Rose / \n", "372-Playing card / 746-Card manipulation / \n", "117-Boxing / \n", "2-Video game / 85-Combat / 89-Comics / 294-Batman / 2146-Bane / \n", "10-Animal / 48-Pet / 71-Dog / 182-Puppy / 2626-Shih Tzu / \n", "7-Musician / 13-Musical ensemble / \n", "1-Vehicle / 4-Car / 88-Machine / 207-Exhaust system / 1661-Volkswagen Gol / \n", "20-Trailer / 748-Iron Man / \n", "3-Concert / 7-Musician / 13-Musical ensemble / 30-Drummer / 38-Orchestra / 39-Drums / 44-Drums / \n", "460-Yu-Gi-Oh! Trading Card Game / \n", "3-Concert / 30-Drummer / 39-Drums / \n", "18-Outdoor recreation / 22-Nature / \n", "14-Guitar / 100-Electric guitar / \n", "31-Disc jockey / \n", "20-Trailer / \n", "0-Games / 1-Vehicle / 2-Video game / 3252-Burnout Paradise / \n", "0-Games / 2-Video game / 264-Star Wars / 714-Jedi / \n", "0-Games / 2-Video game / 19-PC game / 43-Call of Duty / 138-Call of Duty: Black Ops II / \n", "3-Concert / 7-Musician / 13-Musical ensemble / 38-Orchestra / 46-Choir / 716-Coral / \n", "1-Vehicle / 4-Car / 154-Wheel / 493-Rim / \n", "1-Vehicle / 62-Train / 479-Graffiti / \n", "3-Concert / 68-Lighting / \n", "123-Naruto / 272-iPod / \n", "31-Disc jockey / \n", "242-Paper / 415-Sewing / 468-Biology / 1525-Wallet / 2950-Zipper / \n", "1-Vehicle / \n", "0-Games / 2-Video game / 19-PC game / 1091-Saints Row / 2695-Saints Row IV / \n", "3-Concert / \n", "1-Vehicle / 50-Aircraft / 82-Airplane / 129-Aviation / 139-Airport / 151-Landing / 157-Jet aircraft / 175-Airliner / 187-Takeoff / 188-Airline / \n", "3-Concert / 7-Musician / 13-Musical ensemble / 16-Performance art / 30-Drummer / \n", "0-Games / 35-Minecraft / \n", "2625-Gabrielle / \n", "10-Animal / 91-Fish / \n", "0-Games / 2-Video game / 85-Combat / 317-PlayStation / 511-Mortal Kombat / \n", "28-Fashion / \n", "3-Concert / 7-Musician / 13-Musical ensemble / \n", "6-Animation / 801-Inuyasha / \n", "1-Vehicle / 12-Food / 124-Tractor / 156-Agriculture / 174-Farm / 249-Heavy equipment / 660-Rural area / 984-Silage / 1473-Forage harvester / \n", "122-Weight training / \n", "376-Flute / \n", "2454-Notebook / \n", "0-Games / 35-Minecraft / \n", "48-Pet / 2087-Collie / \n", "1-Vehicle / 4-Car / 349-Chevrolet / 406-Chevrolet / 3190-Chevrolet Cruze / \n", "6-Animation / 15-Cartoon / 183-Manga / \n", "8-Football / 21-Stadium / 77-Arena / \n", "3-Concert / 7-Musician / 13-Musical ensemble / 230-Brass instrument / \n", "955-Police officer / \n", "240-Arcade game / 781-Pachinko / \n", "14-Guitar / 24-String instrument / 63-Acoustic guitar / \n", "23-Mobile phone / 29-Smartphone / 72-iPhone / 178-Samsung Galaxy / 3648-Samsung Galaxy S III Mini / \n", "9-Music video / \n", "0-Games / 734-DayZ / 1216-DayZ / \n", "61-Art / 242-Paper / 370-Origami / \n", "3-Concert / 5-Dance / 16-Performance art / \n", "15-Cartoon / 286-Bus / \n", "61-Art / 2856-Decoupage / \n", "88-Machine / 208-Wood / 261-Tool / 400-Woodturning / 972-Lathe / 1384-Greenhouse / \n", "1-Vehicle / 11-Motorsport / 2875-Quadracycle / \n", "6-Animation / 15-Cartoon / 251-Sonic the Hedgehog / \n", "3-Concert / 5-Dance / 133-Nightclub / 718-Duck / \n", "7-Musician / 14-Guitar / 24-String instrument / 63-Acoustic guitar / \n", "1-Vehicle / 4-Car / 11-Motorsport / 17-Racing / 57-Race track / 1400-rFactor / \n", "3-Concert / 5-Dance / 30-Drummer / \n", "12-Food / 26-Cooking / 32-Recipe / 52-Dish / 58-Cuisine / 109-Cooking show / 301-Meal / 542-Potato / 557-Salad / 1024-Culinary art / \n", "1-Vehicle / 4-Car / 245-BMW / 1709-BMW 3 Series (E46) / 2736-BMW 5 Series (E39) / \n", "0-Games / 2-Video game / 43-Call of Duty / 134-Call of Duty: Black Ops / 138-Call of Duty: Black Ops II / \n", "0-Games / 2-Video game / 34-Action-adventure game / 398-The Legend of Zelda / 692-Link / 2423-The Legend of Zelda: Skyward Sword / \n", "1-Vehicle / 4-Car / 74-Truck / 154-Wheel / 158-Tire / 3601-Fender / \n", "12-Food / 1085-Barbecue grill / \n", "10-Animal / 22-Nature / 132-Bird / 1905-Owl / \n", "20-Trailer / \n", "3-Concert / 7-Musician / 30-Drummer / 39-Drums / \n", "1-Vehicle / 4-Car / 11-Motorsport / 17-Racing / 41-Sports car / 57-Race track / 169-Rallying / 680-Subaru / 1041-Subaru Impreza / 1286-Subaru / \n", "1-Vehicle / 23-Mobile phone / 37-Gadget / 47-Personal computer / 145-Tablet computer / \n", "3-Concert / 7-Musician / 13-Musical ensemble / 16-Performance art / 30-Drummer / \n", "358-Rugby football / \n", "14-Guitar / 24-String instrument / 63-Acoustic guitar / 1313-Strum / 2241-Plectrum / \n", "0-Games / 2-Video game / 19-PC game / 33-Weapon / 192-Battlefield / 383-Battlefield 4 / 1267-Battlefield Hardline / \n", "1-Vehicle / 4-Car / 552-Honda Civic / \n", "3-Concert / 7-Musician / 14-Guitar / 16-Performance art / \n", "1-Vehicle / 4-Car / 806-Chopper / \n", "3-Concert / 7-Musician / \n", "625-Factory / \n", "1-Vehicle / 4-Car / 11-Motorsport / 17-Racing / 57-Race track / 651-Stock car racing / 728-Dirt track racing / \n", "0-Games / 8-Football / 90-Sports game / \n", "3-Concert / \n", "28-Fashion / 45-Cosmetics / 153-Dress / 206-Shoe / 785-Jeans / 793-Shirt / \n", "1-Vehicle / 18-Outdoor recreation / 67-Cycling / 69-Bicycle / 197-Trail / 220-Mountain bike / 278-GoPro / 284-Mountain biking / \n", "23-Mobile phone / 29-Smartphone / 101-Telephone / \n", "110-Album / \n", "1-Vehicle / 4-Car / 11-Motorsport / 70-Driving / \n", "9-Music video / 260-Prayer / \n", "30-Drummer / 39-Drums / 44-Drums / 128-Cymbal / 146-Snare drum / \n", "1-Vehicle / 62-Train / 64-Transport / 103-Rail transport / 119-Locomotive / 143-Railroad car / 152-Track / 161-Train station / 224-Rapid transit / \n", "0-Games / 240-Arcade game / 241-Fighting game / 764-The King of Fighters / \n", "163-Running / 206-Shoe / 368-Sneakers / \n", "1-Vehicle / 4-Car / 11-Motorsport / 17-Racing / 124-Tractor / 252-Tractor pulling / \n", "0-Games / 2-Video game / 6-Animation / 15-Cartoon / 55-Video game console / 973-Nintendo 64 / 1343-Super Robot Wars / \n", "1030-Trampoline / \n", "20-Trailer / \n", "20-Trailer / \n", "10-Animal / 48-Pet / 132-Bird / 215-Eating / 307-Hand / 598-Parrot / 1341-Parakeet / \n", "31-Disc jockey / 861-Mixing console / 3024-Numark Industries / \n", "6-Animation / 15-Cartoon / 89-Comics / 243-Sonic the Hedgehog / 251-Sonic the Hedgehog / 437-Cream / 1268-Tails / \n", "25-Toy / 140-LEGO / 1478-Lego Ninjago / \n", "3-Concert / 10-Animal / \n", "0-Games / 2-Video game / 8-Football / 90-Sports game / 180-FIFA 15 / 480-FIFA 13 / 848-FIFA 12 / \n", "9-Music video / 121-Photography / \n", "0-Games / 8-Football / 21-Stadium / 54-Highlight film / 65-Kick / \n", "261-Tool / 315-Knife / 2017-Sharpening / 3880-Knife sharpening / \n", "347-Climbing / 828-Rock climbing / \n", "6-Animation / 89-Comics / 396-One Piece / \n", "130-Gymnastics / \n", "12-Food / 367-Drink / 1065-Bartender / \n", "2-Video game / 445-Guitar Hero / \n", "36-Piano / 53-Keyboard / 104-Pianist / \n", "0-Games / 2-Video game / 20-Trailer / 686-Diablo III / \n", "2-Video game / 20-Trailer / \n", "81-Athlete / \n", "6-Animation / 15-Cartoon / 185-Dragon Ball / 244-Goku / \n", "45-Cosmetics / 385-Face / \n", "8-Football / \n", "1-Vehicle / 4-Car / 11-Motorsport / 17-Racing / 40-Road / \n", "56-Hair / 106-Hairstyle / \n", "6-Animation / 4221-Saiyuki / \n", "30-Drummer / 39-Drums / 44-Drums / \n", "159-World of Warcraft / 171-Warcraft / 994-Cue stick / \n", "3-Concert / \n", "121-Photography / 155-Camera / \n", "9-Music video / \n", "10-Animal / 22-Nature / 86-Plant / 136-Gardening / 540-Insect / 775-Bee / 784-Honey / 1121-Beehive / \n", "0-Games / 2-Video game / 34-Action-adventure game / 241-Fighting game / 1638-Soulcalibur / 2633-Soulcalibur V / \n", "10-Animal / 12-Food / 26-Cooking / 32-Recipe / 58-Cuisine / 91-Fish / 893-Fish as food / \n", "0-Games / 2-Video game / 6-Animation / 34-Action-adventure game / 858-Carl Johnson / \n", "9-Music video / \n", "20-Trailer / 66-Bollywood / \n", "1-Vehicle / 4-Car / 27-Motorcycle / 114-Engine / \n", "1-Vehicle / 4-Car / 270-Sedan / 515-Dodge / \n", "5-Dance / 49-School / \n", "18-Outdoor recreation / 42-Fishing / 902-Fishing bait / 1306-Carp fishing / \n", "1-Vehicle / 62-Train / 64-Transport / 103-Rail transport / 119-Locomotive / 143-Railroad car / 152-Track / 161-Train station / 224-Rapid transit / \n", "238-Nail / 300-Nail art / 304-Nail polish / 307-Hand / 324-Finger / 338-Manicure / 1325-Gel / \n", "5-Dance / \n", "458-Gold / 912-Silver / \n", "20-Trailer / \n", "0-Games / 6-Animation / 1934-Hunter × Hunter / \n", "15-Cartoon / 294-Batman / \n", "37-Gadget / 47-Personal computer / 228-Laptop / 900-MacBook / 1086-MacBook Pro / 1657-MacBook Air / \n", "0-Games / 35-Minecraft / \n", "66-Bollywood / 260-Prayer / 1019-Christ / \n", "67-Cycling / 69-Bicycle / \n", "12-Food / \n", "31-Disc jockey / 44-Drums / \n", "1-Vehicle / 74-Truck / 259-Fire / 371-Firefighter / 796-Fire engine / 1020-Emergency vehicle / \n", "121-Photography / \n", "0-Games / 51-Strategy video game / 373-Clash of Clans / \n", "23-Mobile phone / 29-Smartphone / 37-Gadget / 72-iPhone / 101-Telephone / \n", "8-Football / 54-Highlight film / 79-American football / \n", "9-Music video / \n", "9-Music video / \n", "5-Dance / 16-Performance art / 95-Talent show / 115-Ballet / \n", "14-Guitar / \n", "66-Bollywood / \n", "3-Concert / \n", "47-Personal computer / 309-Computer hardware / 700-Hard disk drive / \n", "0-Games / 2-Video game / 89-Comics / 185-Dragon Ball / 244-Goku / 1304-Cell / 2144-Dragon Ball: Raging Blast 2 / 2384-Dragon Ball: Raging Blast / \n", "8-Football / 54-Highlight film / \n", "59-Winter sport / 112-Ice skating / 142-Hockey / \n", "7-Musician / 24-String instrument / 36-Piano / 150-Violin / \n", "102-Building / 113-House / 170-Home improvement / 276-Room / 353-Wall / 580-Bedroom / \n", "75-Wrestling / 85-Combat / \n", "219-Accordion / \n", "8-Football / 76-Ball / \n", "322-Mixtape / \n", "6-Animation / 749-Figurine / \n", "3-Concert / \n", "3-Concert / 16-Performance art / \n", "1-Vehicle / 4-Car / \n", "86-Plant / 136-Gardening / 137-Tree / 279-Garden / \n", "3-Concert / 16-Performance art / \n", "0-Games / 2-Video game / 33-Weapon / 43-Call of Duty / 96-Soldier / 194-Call of Duty: Modern Warfare 2 / 302-Call of Duty 4: Modern Warfare / \n", "12-Food / 18-Outdoor recreation / 86-Plant / 124-Tractor / 156-Agriculture / 459-Plough / 2039-Soybean / \n", "0-Games / 2-Video game / 1091-Saints Row / 1695-Saints Row: The Third / \n", "14-Guitar / \n", "170-Home improvement / 179-Kitchen / 488-Door / 536-Glass / 1272-Cabinetry / \n", "246-Furniture / 462-Table / 888-Living room / \n", "1-Vehicle / 4-Car / 11-Motorsport / 2517-Chevrolet Opala / \n", "0-Games / 1-Vehicle / 4-Car / 11-Motorsport / 17-Racing / 40-Road / 70-Driving / 84-Snow / 120-Winter / 135-Drifting / 245-BMW / 874-BMW M3 / \n", "10-Animal / 48-Pet / 71-Dog / 182-Puppy / \n", "2531-Dojo / \n", "7-Musician / 14-Guitar / 24-String instrument / 63-Acoustic guitar / \n", "1-Vehicle / 342-Windows Media Video / 2194-Wind power / \n", "78-Wedding / 190-Bride / 293-Slide show / \n", "2-Video game / 111-PlayStation 3 / \n", "47-Personal computer / 72-iPhone / 131-Computer / 145-Tablet computer / 189-iPad / 228-Laptop / 513-Computer keyboard / 1150-iPad 2 / \n", "3-Concert / \n", "121-Photography / 155-Camera / 633-Photograph / \n", "1-Vehicle / 4-Car / 11-Motorsport / 17-Racing / 27-Motorcycle / 40-Road / 2788-Volkswagen Golf Mk2 / 3762-PSA HDi engine / 3909-Hyundai Elantra / \n", "6-Animation / 9-Music video / 25-Toy / 200-Doll / 412-Monster / 503-Monster High / \n", "5-Dance / 16-Performance art / \n", "3-Concert / 7-Musician / 13-Musical ensemble / 38-Orchestra / 230-Brass instrument / 634-Trombone / \n", "0-Games / 2-Video game / 19-PC game / 192-Battlefield / 383-Battlefield 4 / 3005-MicroVolts / \n", "5-Dance / 16-Performance art / 126-The Walt Disney Company / \n", "0-Games / 1-Vehicle / 2-Video game / 62-Train / \n", "66-Bollywood / \n", "0-Games / 2-Video game / 90-Sports game / 561-Madden NFL / 1899-E-book / \n", "1125-God of War / 2754-God of War II / \n", "122-Weight training / 214-Muscle / 1250-Biceps / \n", "1283-Pancake / \n", "31-Disc jockey / \n", "22-Nature / 86-Plant / 156-Agriculture / 217-Hunting / 1229-Magnet / \n", "1-Vehicle / 4-Car / 34-Action-adventure game / 118-Grand Theft Auto V / 299-Volkswagen Passenger Cars / 659-Van / 2486-Scooby-Doo / \n", "18-Outdoor recreation / 59-Winter sport / 112-Ice skating / 316-Figure skating / 632-Ice dancing / \n", "3-Concert / 31-Disc jockey / \n", "1-Vehicle / 4-Car / 11-Motorsport / 17-Racing / 27-Motorcycle / 57-Race track / 105-Motorcycling / 172-Motocross / \n", "1-Vehicle / 62-Train / \n", "2176-Fist of the North Star / \n", "0-Games / 6-Animation / 15-Cartoon / 2412-Fate/stay night / \n", "14-Guitar / 100-Electric guitar / \n", "1-Vehicle / 4-Car / 34-Action-adventure game / 201-Grand Theft Auto: San Andreas / \n", "6-Animation / 10-Animal / 15-Cartoon / \n", "211-RuneScape / 366-Dragon / \n", "429-Website / \n", "1-Vehicle / 4-Car / 5-Dance / \n", "12-Food / 612-Beer / \n", "7-Musician / 14-Guitar / 63-Acoustic guitar / 99-Christmas / \n", "60-Basketball / \n", "2-Video game / \n", "0-Games / 2-Video game / 37-Gadget / 55-Video game console / 362-PlayStation Portable / 581-PlayStation / \n", "141-Microsoft Windows / \n", "1-Vehicle / 62-Train / 507-Thomas the Tank Engine / 837-Roblox / \n", "31-Disc jockey / \n", "999-Elephant / \n", "0-Games / 8-Football / \n", "20-Trailer / \n", "9-Music video / 122-Weight training / 204-Gym / 214-Muscle / \n", "402-Jewellery / 907-Necklace / 3012-Jilbāb / \n", "3-Concert / 7-Musician / 13-Musical ensemble / \n", "6-Animation / 123-Naruto / \n", "6-Animation / 9-Music video / 15-Cartoon / \n", "0-Games / 60-Basketball / 77-Arena / 81-Athlete / 168-Basketball moves / 255-Slam dunk / \n", "8-Football / 21-Stadium / 77-Arena / \n", "5-Dance / 9-Music video / \n", "10-Animal / 48-Pet / 71-Dog / 182-Puppy / 1599-Bone / 1723-Chihuahua / 2311-Human tooth / \n", "3-Concert / 7-Musician / 13-Musical ensemble / 16-Performance art / 30-Drummer / \n", "0-Games / 1-Vehicle / 2-Video game / 4-Car / 11-Motorsport / 17-Racing / 41-Sports car / 352-Need for Speed / 875-Need for Speed: Most Wanted / \n", "0-Games / 2-Video game / 19-PC game / 35-Minecraft / \n", "0-Games / 59-Winter sport / 81-Athlete / 112-Ice skating / 142-Hockey / 312-Ice rink / 4285-The Sports Network / \n", "911-Studio / \n", "1-Vehicle / 4-Car / 11-Motorsport / 17-Racing / 1040-Street racing / \n", "308-Ship / 605-Cruise ship / \n", "8-Football / \n", "1-Vehicle / 4-Car / 40-Road / 70-Driving / 94-Dashcam / 212-Highway / \n", "3-Concert / 5-Dance / 16-Performance art / \n", "66-Bollywood / \n", "1-Vehicle / 10-Animal / 50-Aircraft / 132-Bird / 471-Wing / \n", "8-Football / \n", "1-Vehicle / 4-Car / 70-Driving / 167-Four-wheel drive / 270-Sedan / 277-Sport utility vehicle / 901-Mitsubishi / 3896-Mitsubishi Outlander / \n", "22-Nature / 86-Plant / \n", "28-Fashion / 45-Cosmetics / 173-Eye shadow / 193-Eye liner / 195-Mascara / 196-Eye / 218-Eyelash / 291-Rouge / \n", "99-Christmas / \n", "31-Disc jockey / \n", "14-Guitar / 529-Squat / \n", "1-Vehicle / 4-Car / 2690-Škoda Octavia / \n", "56-Hair / 954-Dreadlocks / \n", "10-Animal / 22-Nature / 636-Rabbit / 1047-Easter egg / \n", "1-Vehicle / 4-Car / 11-Motorsport / 344-Coupé / 1074-Hyundai / 3316-Hyundai Genesis / \n", "1-Vehicle / 4-Car / 11-Motorsport / 17-Racing / 57-Race track / 651-Stock car racing / 728-Dirt track racing / \n", "3-Concert / 7-Musician / 13-Musical ensemble / 46-Choir / \n", "14-Guitar / 24-String instrument / 63-Acoustic guitar / 100-Electric guitar / \n", "3-Concert / 7-Musician / 13-Musical ensemble / 46-Choir / \n", "0-Games / 8-Football / 21-Stadium / 77-Arena / \n", "96-Soldier / 1573-United States Army / \n", "12-Food / 26-Cooking / 389-Restaurant / \n", "28-Fashion / 56-Hair / 106-Hairstyle / 340-Afro-textured hair / 405-Braid / 857-Afro / 954-Dreadlocks / 1812-Hair twists / \n", "130-Gymnastics / \n", "0-Games / 2-Video game / 6-Animation / 1618-Rayman / 4058-Rayman Origins / \n", "20-Trailer / 1014-Movieclips / 1756-Predator / 2011-Alien / \n", "3-Concert / 5-Dance / 253-Carnival / \n", "10-Animal / 18-Outdoor recreation / 80-Horse / 247-Stallion / 289-Dressage / 351-Jumping / \n", "695-Balloon / 1972-Toy balloon / \n", "56-Hair / \n", "863-Body piercing / 2118-Tongue / 4123-Tongue piercing / \n", "0-Games / 35-Minecraft / 395-Camping / \n", "7-Musician / 13-Musical ensemble / 24-String instrument / \n", "20-Trailer / \n", "3-Concert / 7-Musician / 13-Musical ensemble / \n", "144-Amusement park / 334-Amusement ride / 3576-Iguana / \n", "31-Disc jockey / 342-Windows Media Video / \n", "3-Concert / 7-Musician / 13-Musical ensemble / 16-Performance art / 30-Drummer / \n", "296-Saxophone / \n", "45-Cosmetics / 193-Eye liner / 218-Eyelash / \n", "4108-Orangutan / \n", "310-The Sims / 564-The Sims 3 / \n", "5-Dance / \n", "1-Vehicle / 4-Car / 34-Action-adventure game / 41-Sports car / 201-Grand Theft Auto: San Andreas / 297-Grand Theft Auto IV / \n", "66-Bollywood / \n", "102-Building / 433-Construction / \n", "6-Animation / 15-Cartoon / 116-Pokémon / \n", "2-Video game / 20-Trailer / 34-Action-adventure game / 264-Star Wars / 3999-Star Wars: The Force Unleashed II / \n", "10-Animal / 2356-Comb / \n", "0-Games / 2-Video game / 19-PC game / 33-Weapon / 43-Call of Duty / 194-Call of Duty: Modern Warfare 2 / \n", "98-Festival / \n", "1-Vehicle / 4-Car / 11-Motorsport / 74-Truck / 92-Radio-controlled model / 135-Drifting / 167-Four-wheel drive / 202-Radio-controlled car / \n", "434-Ninja / \n", "1043-Router / \n", "455-Unidentified flying object / \n", "9-Music video / \n", "1-Vehicle / 62-Train / 286-Bus / \n", "3-Concert / 16-Performance art / \n", "298-Television advertisement / 337-Advertising / \n", "3-Concert / 7-Musician / 13-Musical ensemble / 30-Drummer / 38-Orchestra / 44-Drums / \n", "0-Games / 8-Football / 21-Stadium / 54-Highlight film / \n", "0-Games / 6-Animation / 372-Playing card / 494-Card game / \n", "248-Hotel / 916-Spa / \n", "3-Concert / 7-Musician / 36-Piano / 104-Pianist / 115-Ballet / \n", "10-Animal / 18-Outdoor recreation / 42-Fishing / 91-Fish / 147-River / 268-Recreational fishing / 478-Fishing rod / 510-Fishing lure / 698-Fisherman / 844-Bass / 896-Largemouth bass / 902-Fishing bait / \n", "22-Nature / \n", "12-Food / 26-Cooking / 32-Recipe / 58-Cuisine / 109-Cooking show / 226-Dessert / 232-Cake / 236-Baking / 1088-Infomercial / \n", "1058-Volcano / \n", "1-Vehicle / 59-Winter sport / 144-Amusement park / 321-Roller coaster / 1829-Sledding / 4446-Bobsleigh / \n", "0-Games / 2-Video game / 34-Action-adventure game / 1447-Skylanders / 2321-Skylanders: Spyro's Adventure / \n", "1-Vehicle / 4-Car / 519-Recreational vehicle / 659-Van / \n", "6-Animation / 15-Cartoon / 185-Dragon Ball / \n", "5-Dance / \n", "8-Football / \n", "9-Music video / \n", "1-Vehicle / 4-Car / \n", "1-Vehicle / 18-Outdoor recreation / 50-Aircraft / 82-Airplane / 1200-Ultralight aviation / \n", "23-Mobile phone / 29-Smartphone / 37-Gadget / \n", "22-Nature / \n", "1-Vehicle / 25-Toy / 50-Aircraft / 92-Radio-controlled model / 149-Model aircraft / 160-Radio-controlled aircraft / 227-Water / 266-Helicopter / 392-Radio-controlled helicopter / \n", "1-Vehicle / 4-Car / 11-Motorsport / 17-Racing / 70-Driving / 449-Honda / 552-Honda Civic / \n", "14-Guitar / \n", "567-Compact disc / \n", "23-Mobile phone / 29-Smartphone / 37-Gadget / 47-Personal computer / 72-iPhone / 145-Tablet computer / 189-iPad / 1504-iPad 3 / \n", "1-Vehicle / 27-Motorcycle / \n", "313-Book / \n", "3-Concert / \n", "0-Games / 10-Animal / 80-Horse / \n", "59-Winter sport / 84-Snow / 281-Snowboarding / 501-Snowboard / \n", "31-Disc jockey / \n", "47-Personal computer / 131-Computer / 1045-Computer case / 1954-Power supply / \n", "12-Food / 1363-Candle / 3170-Bath & Body Works / \n", "5-Dance / \n", "148-Painting / 408-Sketch / \n", "23-Mobile phone / 29-Smartphone / 37-Gadget / 802-Google Nexus / 2264-Nexus 7 / \n", "23-Mobile phone / 29-Smartphone / \n", "971-Alphabet / \n", "99-Christmas / 1784-Goblin / \n", "1-Vehicle / 11-Motorsport / 17-Racing / 27-Motorcycle / 40-Road / 57-Race track / 105-Motorcycling / 718-Duck / 2133-Road racing / \n", "165-League of Legends / \n", "6-Animation / 15-Cartoon / 123-Naruto / 2462-Neji Hyuga / 4333-Kiba Inuzuka / \n", "8-Football / \n", "12-Food / 227-Water / 238-Nail / 300-Nail art / \n", "0-Games / 6-Animation / \n", "121-Photography / 155-Camera / 566-Digital camera / 1117-Camcorder / \n", "5-Dance / 16-Performance art / 361-Flamenco / \n", "10-Animal / 48-Pet / 166-Cat / 346-Kitten / \n", "1-Vehicle / 18-Outdoor recreation / 42-Fishing / 87-Boat / 481-Kayak / \n", "3-Concert / 2795-Root / \n", "476-Textile / \n", "1-Vehicle / 62-Train / 64-Transport / 103-Rail transport / 119-Locomotive / 143-Railroad car / 161-Train station / 224-Rapid transit / 854-Diesel engine / \n", "10-Animal / 91-Fish / 227-Water / 258-Aquarium / 2090-Siamese fighting fish / 3018-Betta / \n", "12-Food / 26-Cooking / 944-Hamburger / \n", "67-Cycling / 69-Bicycle / \n", "15-Cartoon / 1930-Rurouni Kenshin / \n", "5-Dance / 16-Performance art / \n", "94-Dashcam / \n", "3-Concert / 7-Musician / 13-Musical ensemble / 16-Performance art / 38-Orchestra / \n", "0-Games / 54-Highlight film / 60-Basketball / 77-Arena / 81-Athlete / 168-Basketball moves / \n", "2-Video game / 34-Action-adventure game / 118-Grand Theft Auto V / 395-Camping / \n", "18-Outdoor recreation / 83-Skateboarding / 108-Skateboard / 1031-Fisheye lens / \n", "20-Trailer / \n", "5-Dance / 98-Festival / \n", "6-Animation / \n", "578-Cookie / \n", "2-Video game / 176-PlayStation 4 / 314-Xbox One / \n", "1-Vehicle / 50-Aircraft / 82-Airplane / 129-Aviation / 139-Airport / 151-Landing / 157-Jet aircraft / 175-Airliner / 187-Takeoff / 188-Airline / 254-Runway / 471-Wing / 472-Boeing 737 / \n", "521-Metal / 731-Coin / 959-Metal detector / \n", "22-Nature / 42-Fishing / 164-Beach / 233-Ocean / 423-Coast / \n", "3-Concert / 7-Musician / 13-Musical ensemble / \n", "30-Drummer / 39-Drums / 44-Drums / 128-Cymbal / 146-Snare drum / \n", "1-Vehicle / 4-Car / 50-Aircraft / 82-Airplane / 129-Aviation / 139-Airport / 151-Landing / 254-Runway / 918-Brake / 2296-Joystick / \n", "0-Games / 1043-Router / \n", "1-Vehicle / 50-Aircraft / 129-Aviation / 157-Jet aircraft / 175-Airliner / 188-Airline / 388-Fighter aircraft / 758-Air force / \n", "21-Stadium / 59-Winter sport / 77-Arena / 112-Ice skating / 142-Hockey / \n", "6-Animation / 10-Animal / 15-Cartoon / 20-Trailer / 71-Dog / 1164-Cartoon Network / \n", "3-Concert / 68-Lighting / \n", "9-Music video / \n", "616-The Bible / \n", "3-Concert / 7-Musician / 13-Musical ensemble / 46-Choir / \n", "7-Musician / 14-Guitar / 100-Electric guitar / \n", "88-Machine / 208-Wood / \n", "0-Games / 2-Video game / 19-PC game / 34-Action-adventure game / 47-Personal computer / 118-Grand Theft Auto V / \n", "8-Football / \n", "12-Food / 86-Plant / 279-Garden / 671-Tomato / \n", "28-Fashion / 45-Cosmetics / 56-Hair / 106-Hairstyle / \n", "3-Concert / 5-Dance / 16-Performance art / \n", "22-Nature / 86-Plant / \n", "0-Games / 2-Video game / 20-Trailer / 34-Action-adventure game / 656-Metal Gear / \n", "1-Vehicle / 11-Motorsport / 17-Racing / 18-Outdoor recreation / 27-Motorcycle / 172-Motocross / 2014-Vans / \n", "3-Concert / \n", "14-Guitar / \n", "2-Video game / 34-Action-adventure game / 1275-Watch Dogs / \n", "5-Dance / \n", "45-Cosmetics / 235-Lipstick / \n", "6-Animation / 15-Cartoon / 2098-Tom and Jerry / 2112-Jerry Mouse / \n", "293-Slide show / \n", "0-Games / 2-Video game / 19-PC game / 257-Counter-Strike / \n", "23-Mobile phone / 47-Personal computer / 72-iPhone / 145-Tablet computer / 228-Laptop / 700-Hard disk drive / \n", "1-Vehicle / 4-Car / \n", "3-Concert / 7-Musician / 13-Musical ensemble / \n", "1-Vehicle / 4-Car / 64-Transport / 234-Ford / \n", "3-Concert / \n", "1-Vehicle / 27-Motorcycle / 1575-Clutch / \n", "5-Dance / 16-Performance art / \n", "5-Dance / 16-Performance art / \n", "3-Concert / 5-Dance / \n", "14-Guitar / 24-String instrument / \n", "5-Dance / 16-Performance art / \n", "23-Mobile phone / 29-Smartphone / 72-iPhone / 554-iPhone 4 / 675-iPhone 4S / \n", "0-Games / 1911-StepMania / \n", "1-Vehicle / 4-Car / 154-Wheel / 158-Tire / 167-Four-wheel drive / 277-Sport utility vehicle / 1688-Suzuki Jimny / \n", "3-Concert / 7-Musician / 13-Musical ensemble / 24-String instrument / 36-Piano / 150-Violin / \n", "9-Music video / 83-Skateboarding / \n", "6-Animation / 20-Trailer / \n", "8-Football / 21-Stadium / \n", "6-Animation / 15-Cartoon / \n", "205-Newscaster / \n", "5-Dance / 325-Vampire / \n", "15-Cartoon / 126-The Walt Disney Company / \n", "1-Vehicle / 4-Car / 11-Motorsport / 17-Racing / 40-Road / 41-Sports car / 57-Race track / 135-Drifting / \n", "6-Animation / 20-Trailer / \n", "9-Music video / \n", "9-Music video / \n", "10-Animal / 48-Pet / 71-Dog / \n", "33-Weapon / 242-Paper / \n", "981-Sculpture / \n", "1-Vehicle / 62-Train / 64-Transport / 161-Train station / 224-Rapid transit / \n", "5-Dance / 39-Drums / 44-Drums / 1825-Cajón / \n", "0-Games / 8-Football / 21-Stadium / \n", "14-Guitar / 425-City / \n", "1-Vehicle / 4-Car / 62-Train / 64-Transport / 103-Rail transport / 119-Locomotive / 143-Railroad car / 161-Train station / \n", "3-Concert / 7-Musician / 46-Choir / \n", "12-Food / 466-Gift / \n", "49-School / 845-Classroom / \n", "0-Games / 2-Video game / 257-Counter-Strike / 351-Jumping / \n", "3-Concert / \n", "0-Games / 6-Animation / \n", "0-Games / 2-Video game / 62-Train / 229-Halo / 470-Halo 3 / 1161-Halo 2 / \n", "0-Games / 2-Video game / 19-PC game / 33-Weapon / 43-Call of Duty / 96-Soldier / 134-Call of Duty: Black Ops / 138-Call of Duty: Black Ops II / 422-Call of Duty: Ghosts / \n", "9-Music video / \n", "0-Games / 3149-Melty Blood / \n", "9-Music video / \n", "1-Vehicle / 4-Car / 88-Machine / 114-Engine / \n", "2-Video game / 23-Mobile phone / 29-Smartphone / 37-Gadget / 121-Photography / 155-Camera / 566-Digital camera / 2248-Zoom lens / \n", "0-Games / 2-Video game / 1166-Dark Souls II / \n", "1-Vehicle / 3-Concert / 5-Dance / 11-Motorsport / 49-School / 99-Christmas / 135-Drifting / \n", "28-Fashion / 56-Hair / \n", "6-Animation / 15-Cartoon / 1826-Higurashi When They Cry / \n", "14-Guitar / 24-String instrument / 63-Acoustic guitar / \n", "12-Food / \n", "1-Vehicle / 50-Aircraft / 139-Airport / \n", "551-British Broadcasting Corporation / 2224-Sherlock Holmes / \n", "54-Highlight film / 79-American football / 81-Athlete / 163-Running / \n", "10-Animal / 80-Horse / \n", "1-Vehicle / 4-Car / 11-Motorsport / 17-Racing / 40-Road / 70-Driving / 169-Rallying / \n", "8-Football / \n", "66-Bollywood / \n", "18-Outdoor recreation / 83-Skateboarding / 108-Skateboard / 221-Skatepark / \n", "12-Food / 353-Wall / 1456-Clothes dryer / 2067-Duct / \n", "6-Animation / 15-Cartoon / 849-Tales / \n", "28-Fashion / 45-Cosmetics / 66-Bollywood / 78-Wedding / 153-Dress / 190-Bride / \n", "97-Drawing / 148-Painting / 1579-Circle / \n", "2-Video game / 9-Music video / 237-Final Fantasy / 730-Final Fantasy VII / \n", "3-Concert / 7-Musician / 16-Performance art / 95-Talent show / 99-Christmas / \n", "20-Trailer / 313-Book / \n", "0-Games / 51-Strategy video game / 373-Clash of Clans / \n", "2-Video game / 55-Video game console / 398-The Legend of Zelda / 2659-Universe of The Legend of Zelda / \n", "67-Cycling / 69-Bicycle / 570-Road bicycle racing / 645-Road bicycle / \n", "3-Concert / 16-Performance art / \n", "3-Concert / 7-Musician / 13-Musical ensemble / 46-Choir / 99-Christmas / \n", "7-Musician / 14-Guitar / 24-String instrument / 63-Acoustic guitar / \n", "18-Outdoor recreation / 42-Fishing / 225-Lake / 268-Recreational fishing / 478-Fishing rod / 510-Fishing lure / 844-Bass / 3037-Micropterus / \n", "23-Mobile phone / 29-Smartphone / 37-Gadget / 72-iPhone / 101-Telephone / 178-Samsung Galaxy / 802-Google Nexus / 3089-Nokia N8 / 4059-Nokia N900 / 4128-Nokia N97 / 4424-Nexus One / \n", "7-Musician / 13-Musical ensemble / 38-Orchestra / 125-Television / \n", "573-Paragliding / 765-Parachute / \n", "1-Vehicle / 4-Car / 11-Motorsport / 198-Off-road vehicle / 3386-Hot tub / \n", "5-Dance / 78-Wedding / \n", "56-Hair / 871-Hair coloring / 1269-Human hair color / \n", "2449-Chuck E. Cheese's / \n", "3-Concert / 16-Performance art / 68-Lighting / \n", "6-Animation / 3791-Magi: The Labyrinth of Magic / \n", "10-Animal / 709-Bull / 2091-Destroyer / \n", "8-Football / 73-Ball / \n", "46-Choir / 102-Building / 318-Church / 397-Organ / 646-Pipe organ / 870-Chapel / \n", "2-Video game / 3468-Heroes of Might and Magic / \n", "3-Concert / \n", "317-PlayStation / 1196-Dragon Age / 1735-Dragon Age: Inquisition / 2297-Elf / \n", "93-Comedy / \n", "56-Hair / \n", "5-Dance / 16-Performance art / \n", "42-Fishing / 1008-Kite / \n", "0-Games / 60-Basketball / \n", "3-Concert / 7-Musician / 13-Musical ensemble / 14-Guitar / 16-Performance art / 30-Drummer / \n", "337-Advertising / \n", "0-Games / 2-Video game / 19-PC game / 43-Call of Duty / 134-Call of Duty: Black Ops / 138-Call of Duty: Black Ops II / \n", "0-Games / 2-Video game / 123-Naruto / \n", "159-World of Warcraft / 171-Warcraft / \n", "3-Concert / 7-Musician / 13-Musical ensemble / 38-Orchestra / \n", "0-Games / 2-Video game / 924-Trophy / 1137-Crash Bandicoot / \n", "0-Games / 2-Video game / 125-Television / 345-Loudspeaker / 484-Subwoofer / \n", "18-Outdoor recreation / 83-Skateboarding / 108-Skateboard / 221-Skatepark / 556-Skateboarding trick / \n", "843-The Doctor / \n", "61-Art / 479-Graffiti / \n", "837-Roblox / \n", "0-Games / 8-Football / 21-Stadium / \n", "8-Football / 21-Stadium / \n", "8-Football / 21-Stadium / 77-Arena / \n", "0-Games / 2-Video game / 9-Music video / \n", "0-Games / 2-Video game / 19-PC game / 43-Call of Duty / 134-Call of Duty: Black Ops / 138-Call of Duty: Black Ops II / \n", "3-Concert / 7-Musician / 9-Music video / 13-Musical ensemble / 14-Guitar / \n", "1-Vehicle / 4-Car / 11-Motorsport / 17-Racing / 245-BMW / \n", "7-Musician / 14-Guitar / 24-String instrument / 63-Acoustic guitar / \n", "0-Games / 2-Video game / 1566-Doom / 2352-Doom II: Hell on Earth / \n", "3-Concert / 5-Dance / 16-Performance art / 68-Lighting / \n", "323-Toddler / \n", "616-The Bible / \n", "31-Disc jockey / \n", "5-Dance / 49-School / 95-Talent show / \n", "36-Piano / \n", "22-Nature / 86-Plant / 3177-Tulip / \n", "1-Vehicle / 4-Car / 114-Engine / \n", "3-Concert / 5-Dance / 16-Performance art / \n", "9-Music video / 96-Soldier / 1882-Warrior / \n", "6-Animation / 4221-Saiyuki / \n", "5-Dance / 16-Performance art / \n", "5-Dance / \n", "995-Attack on Titan / \n", "7-Musician / 13-Musical ensemble / 14-Guitar / 24-String instrument / 150-Violin / 639-Fiddle / \n", "1-Vehicle / 8-Football / 350-Circus / 788-Juggling / \n", "0-Games / 2-Video game / 33-Weapon / 43-Call of Duty / 127-Xbox 360 / 134-Call of Duty: Black Ops / 138-Call of Duty: Black Ops II / 394-Map / 704-Call of Duty: World at War / \n", "49-School / 410-Teacher / 845-Classroom / \n", "125-Television / \n", "0-Games / 6-Animation / 1647-Kidō Senshi Gundam: Senjō no Kizuna / 3289-MS-06 Zaku II / \n", "20-Trailer / \n", "10-Animal / 20-Trailer / 48-Pet / 71-Dog / 298-Television advertisement / 1723-Chihuahua / \n", "3-Concert / 68-Lighting / \n", "0-Games / 8-Football / 21-Stadium / \n", "0-Games / 1-Vehicle / 2-Video game / 4-Car / 34-Action-adventure game / 201-Grand Theft Auto: San Andreas / 858-Carl Johnson / \n", "0-Games / 54-Highlight film / 60-Basketball / 81-Athlete / 168-Basketball moves / \n", "83-Skateboarding / 108-Skateboard / \n", "8-Football / \n", "28-Fashion / 45-Cosmetics / 173-Eye shadow / 193-Eye liner / 195-Mascara / 196-Eye / 218-Eyelash / 235-Lipstick / 291-Rouge / \n", "61-Art / 97-Drawing / 479-Graffiti / \n", "2-Video game / \n", "3-Concert / \n", "0-Games / 2-Video game / 19-PC game / 33-Weapon / 43-Call of Duty / 96-Soldier / 422-Call of Duty: Ghosts / \n", "1-Vehicle / 67-Cycling / 69-Bicycle / \n", "0-Games / 2-Video game / 34-Action-adventure game / 213-Forest / 380-Assassin's Creed / 1241-Assassin's Creed IV: Black Flag / \n", "59-Winter sport / 177-Skiing / 355-Comedian / \n", "0-Games / 6-Animation / 241-Fighting game / 486-M.U.G.E.N / 4231-Slayers / \n", "1-Vehicle / 4-Car / 84-Snow / 154-Wheel / 158-Tire / \n", "886-James Bond / \n", "45-Cosmetics / 56-Hair / \n", "615-Poker / \n", "1162-Dietary supplement / \n", "473-Ford Mustang / \n", "6-Animation / 15-Cartoon / 97-Drawing / 183-Manga / 1556-Dōjinshi / \n", "0-Games / 2-Video game / 2265-Subway Surfers / \n", "12-Food / \n", "25-Toy / \n", "5-Dance / 16-Performance art / 98-Festival / 306-High school / \n", "0-Games / 19-PC game / 271-Tank / 409-World of Tanks / \n", "10-Animal / 80-Horse / 247-Stallion / 438-Mare / 3772-Paso Fino / \n", "9-Music video / \n", "3-Concert / 16-Performance art / \n", "12-Food / 224-Rapid transit / \n", "12-Food / 26-Cooking / 32-Recipe / 210-Cookware and bakeware / 2563-Gravy / \n", "25-Toy / 140-LEGO / 884-Lego minifigure / 1882-Warrior / \n", "75-Wrestling / \n", "2-Video game / 61-Art / \n", "199-Wii / 298-Television advertisement / \n", "39-Drums / \n", "10-Animal / 91-Fish / \n", "0-Games / 1-Vehicle / 2-Video game / 4-Car / 34-Action-adventure game / 118-Grand Theft Auto V / 176-PlayStation 4 / 330-Money / 601-Shark / \n", "1-Vehicle / 4-Car / 11-Motorsport / 17-Racing / 41-Sports car / 57-Race track / 651-Stock car racing / \n", "0-Games / 2-Video game / 19-PC game / 35-Minecraft / \n", "3-Concert / 5-Dance / 16-Performance art / 38-Orchestra / \n", "444-Family / \n", "20-Trailer / \n", "10-Animal / 48-Pet / 71-Dog / 182-Puppy / 1723-Chihuahua / \n", "5-Dance / 78-Wedding / 153-Dress / 190-Bride / 421-Gown / \n", "1-Vehicle / 4-Car / 114-Engine / 536-Glass / \n", "20-Trailer / 298-Television advertisement / 337-Advertising / 1172-VHS / \n", "0-Games / 2-Video game / 366-Dragon / \n", "6-Animation / 15-Cartoon / 851-Digimon / \n", "0-Games / 35-Minecraft / \n", "13-Musical ensemble / 14-Guitar / 24-String instrument / 63-Acoustic guitar / \n", "125-Television / 298-Television advertisement / 337-Advertising / \n", "1-Vehicle / 4-Car / 11-Motorsport / 17-Racing / 27-Motorcycle / 57-Race track / 105-Motorcycling / 927-Supermoto / 2133-Road racing / 3553-Track racing / \n", "10-Animal / 540-Insect / 775-Bee / 1110-Cell / \n", "12-Food / 26-Cooking / 32-Recipe / 52-Dish / 58-Cuisine / 109-Cooking show / 179-Kitchen / 215-Eating / 262-Ice / 295-Chocolate / 437-Cream / 684-Ice cream / 770-Banana / 2699-Chocolate chip / \n", "0-Games / 8-Football / 21-Stadium / 65-Kick / 73-Ball / 76-Ball / \n", "117-Boxing / \n", "0-Games / 2-Video game / 328-Simulation video game / 4187-MechWarrior Online / \n", "45-Cosmetics / 148-Painting / 303-Paint / 385-Face / 855-Hulk / \n", "1-Vehicle / 86-Plant / 124-Tractor / 156-Agriculture / 174-Farm / 459-Plough / 741-Mower / 984-Silage / 1037-Grass / \n", "14-Guitar / 63-Acoustic guitar / 121-Photography / 722-Waterfall / \n", "93-Comedy / \n", "0-Games / 6-Animation / 1338-Devil / 1527-Pretty Cure / \n", "860-Christian Church / \n", "0-Games / 2-Video game / 35-Minecraft / 55-Video game console / 111-PlayStation 3 / 127-Xbox 360 / 181-Xbox / 314-Xbox One / \n", "0-Games / 1-Vehicle / 2-Video game / 4-Car / 17-Racing / 34-Action-adventure game / 118-Grand Theft Auto V / 297-Grand Theft Auto IV / 787-Monster truck / 1077-Lightning McQueen / \n", "774-Galaxy / \n", "356-Star / \n", "0-Games / 2-Video game / 116-Pokémon / \n", "1-Vehicle / 27-Motorcycle / \n", "12-Food / 282-Skin / 385-Face / 641-Mask / \n", "0-Games / 2-Video game / \n", "2808-Printed circuit board / \n", "20-Trailer / \n", "1-Vehicle / 4-Car / \n", "150-Violin / \n", "0-Games / 51-Strategy video game / 373-Clash of Clans / \n", "13-Musical ensemble / \n", "416-Apartment / \n", "5-Dance / 98-Festival / \n", "10-Animal / 48-Pet / 71-Dog / \n", "1-Vehicle / 1699-Cart / 3573-Golf cart / \n", "0-Games / 2-Video game / 6-Animation / 3426-Inazuma Eleven GO / \n", "3-Concert / 5-Dance / 31-Disc jockey / 68-Lighting / 133-Nightclub / \n", "93-Comedy / \n", "7-Musician / 30-Drummer / 39-Drums / 44-Drums / 296-Saxophone / \n", "6-Animation / 15-Cartoon / 755-Hetalia: Axis Powers / 4671-Little Busters! / \n", "6-Animation / 15-Cartoon / \n", "10-Animal / 80-Horse / 290-Pony / \n", "10-Animal / 80-Horse / \n", "2009-Electromagnetic coil / 3806-Tesla coil / \n", "10-Animal / 18-Outdoor recreation / 22-Nature / 71-Dog / 217-Hunting / 280-Wildlife / 2673-Moose / \n", "1-Vehicle / 4-Car / 11-Motorsport / 17-Racing / 1040-Street racing / \n", "6-Animation / 15-Cartoon / 126-The Walt Disney Company / \n", "0-Games / 159-World of Warcraft / 171-Warcraft / 1213-World of Warcraft: Wrath of the Lich King / \n", "45-Cosmetics / 195-Mascara / \n", "47-Personal computer / 131-Computer / 497-Macintosh / 900-MacBook / \n", "93-Comedy / \n", "0-Games / 8-Football / 21-Stadium / 65-Kick / 73-Ball / \n", "122-Weight training / 214-Muscle / \n", "9-Music video / \n", "5-Dance / 95-Talent show / \n", "5-Dance / 16-Performance art / 227-Water / \n", "1151-Temple / \n", "9-Music video / \n", "0-Games / 2-Video game / 6-Animation / 123-Naruto / 244-Goku / 591-Naruto: Ultimate Ninja / \n", "0-Games / 79-American football / 90-Sports game / 561-Madden NFL / 2916-Madden NFL 12 / \n", "8-Football / 425-City / \n", "357-DVD / \n", "3-Concert / 7-Musician / 9-Music video / 13-Musical ensemble / 46-Choir / \n", "0-Games / 2-Video game / 2708-Paper Mario / \n", "10-Animal / \n", "2-Video game / 31-Disc jockey / \n", "0-Games / 1-Vehicle / 2-Video game / 4-Car / 11-Motorsport / 17-Racing / 41-Sports car / 231-Supercar / 352-Need for Speed / 1770-Need for Speed: Hot Pursuit / 2545-Need for Speed III: Hot Pursuit / \n", "3-Concert / \n", "0-Games / 9-Music video / 35-Minecraft / 321-Roller coaster / 334-Amusement ride / \n", "1-Vehicle / 22-Nature / 87-Boat / 164-Beach / 233-Ocean / 423-Coast / \n", "209-University / \n", "293-Slide show / \n", "25-Toy / 126-The Walt Disney Company / 200-Doll / 417-Play-Doh / 996-Princess / 1219-Ariel / \n", "3-Concert / \n", "1-Vehicle / 4-Car / 11-Motorsport / 167-Four-wheel drive / \n", "1-Vehicle / 124-Tractor / 156-Agriculture / 174-Farm / 660-Rural area / \n", "12-Food / 26-Cooking / 32-Recipe / 274-Meat / 589-Grilling / 649-Oven / 688-Beef / 1256-Smoking / \n", "1-Vehicle / 4-Car / 70-Driving / 74-Truck / \n", "6-Animation / 15-Cartoon / 4344-Gakuen Alice / \n", "28-Fashion / 45-Cosmetics / \n", "75-Wrestling / 1080-WWE 2K15 / 1582-WWE 2K14 / \n", "9-Music video / \n", "6-Animation / 15-Cartoon / 183-Manga / \n", "10-Animal / 130-Gymnastics / 132-Bird / \n", "23-Mobile phone / 29-Smartphone / 37-Gadget / 72-iPhone / \n", "5-Dance / 44-Drums / 1579-Circle / \n", "0-Games / \n", "38-Orchestra / 3263-Shadow of the Colossus / \n", "5-Dance / 49-School / 306-High school / \n", "1-Vehicle / 4-Car / \n", "1-Vehicle / 4-Car / 395-Camping / 519-Recreational vehicle / \n", "12-Food / 26-Cooking / 2164-Silk / 2552-Dyeing / \n", "1-Vehicle / 18-Outdoor recreation / 22-Nature / 67-Cycling / 69-Bicycle / 197-Trail / 220-Mountain bike / 284-Mountain biking / \n", "170-Home improvement / 353-Wall / 879-Ceiling / 1350-Plaster / 1671-Drywall / \n", "3-Concert / 16-Performance art / \n", "6-Animation / 15-Cartoon / \n", "809-Logo / \n", "18-Outdoor recreation / 67-Cycling / 69-Bicycle / 83-Skateboarding / 108-Skateboard / 221-Skatepark / 377-BMX bike / \n", "263-Light / \n", "1-Vehicle / 50-Aircraft / 151-Landing / 254-Runway / 262-Ice / \n", "1-Vehicle / 4-Car / 94-Dashcam / 683-Renault / \n", "0-Games / 2-Video game / 6-Animation / \n", "0-Games / 96-Soldier / 264-Star Wars / 714-Jedi / 1050-Star Wars: The Old Republic / \n", "117-Boxing / \n", "3-Concert / 7-Musician / 14-Guitar / 24-String instrument / \n", "28-Fashion / \n", "75-Wrestling / \n", "0-Games / 2-Video game / 8-Football / 90-Sports game / 180-FIFA 15 / \n", "9-Music video / \n", "3-Concert / 7-Musician / 13-Musical ensemble / 3060-Chameleon / \n", "1-Vehicle / 4-Car / 11-Motorsport / 17-Racing / 41-Sports car / 57-Race track / \n", "0-Games / 8-Football / 21-Stadium / 54-Highlight film / 65-Kick / 76-Ball / 81-Athlete / 1037-Grass / \n", "25-Toy / 200-Doll / 4237-Ball-jointed doll / \n", "9-Music video / 1396-Lara Croft / \n", "2-Video game / 34-Action-adventure game / 380-Assassin's Creed / 1241-Assassin's Creed IV: Black Flag / \n", "5-Dance / \n", "12-Food / 26-Cooking / 32-Recipe / 52-Dish / 210-Cookware and bakeware / 239-Roasting / 274-Meat / 454-Sauce / 589-Grilling / 688-Beef / 720-Wine / 862-Pork / 1115-Steak / 2125-Fillet / \n", "3-Concert / 7-Musician / 46-Choir / \n", "12-Food / 120-Winter / 137-Tree / 348-Fruit / \n", "0-Games / 116-Pokémon / 186-Pokémon / 690-Pokémon X and Y / \n", "8-Football / \n", "1-Vehicle / 27-Motorcycle / 278-GoPro / \n", "121-Photography / 633-Photograph / \n", "0-Games / 2-Video game / 686-Diablo III / 872-Diablo / \n", "0-Games / 2-Video game / 141-Microsoft Windows / \n", "14-Guitar / 63-Acoustic guitar / 361-Flamenco / \n", "20-Trailer / \n", "3-Concert / 7-Musician / 13-Musical ensemble / 30-Drummer / 38-Orchestra / \n", "12-Food / 52-Dish / 58-Cuisine / 1539-Ramen / 2230-Buffet / \n", "0-Games / 2-Video game / 257-Counter-Strike / 425-City / 702-Counter-Strike: Source / \n", "170-Home improvement / 276-Room / 308-Ship / 605-Cruise ship / 652-Bathroom / 1402-Suite / 2729-Balcony / \n", "9-Music video / \n", "3-Concert / \n", "12-Food / 26-Cooking / 32-Recipe / 52-Dish / 58-Cuisine / 109-Cooking show / 179-Kitchen / 1108-Mushroom / 1230-Italian food / \n", "10-Animal / 22-Nature / 280-Wildlife / 537-Zoo / \n", "0-Games / 319-Robot / 1885-Robotics / \n", "12-Food / 26-Cooking / 32-Recipe / 52-Dish / 557-Salad / 1748-Carrot / 2588-Mayonnaise / 3205-Beetroot / \n", "122-Weight training / 214-Muscle / 1162-Dietary supplement / \n", "60-Basketball / \n", "242-Paper / 370-Origami / \n", "615-Poker / \n", "0-Games / 8-Football / 21-Stadium / 65-Kick / 73-Ball / \n", "10-Animal / 91-Fish / 258-Aquarium / 2090-Siamese fighting fish / 3018-Betta / \n", "0-Games / 25-Toy / 489-Board game / 922-Tabletop game / \n", "1-Vehicle / 4-Car / 11-Motorsport / \n", "259-Fire / 616-The Bible / \n", "3-Concert / 7-Musician / 13-Musical ensemble / 30-Drummer / 38-Orchestra / \n", "457-Knitting / 696-T-shirt / 907-Necklace / 1096-Scarf / \n", "9-Music video / \n", "5-Dance / \n", "9-Music video / 20-Trailer / 925-Brain / \n", "0-Games / 2-Video game / 159-World of Warcraft / 171-Warcraft / \n", "1-Vehicle / 4-Car / 11-Motorsport / 41-Sports car / 790-Nissan GT-R / 952-Chevrolet Corvette / \n", "313-Book / \n", "1-Vehicle / 4-Car / 11-Motorsport / 17-Racing / 2244-Autocross / \n", "0-Games / 1-Vehicle / 2-Video game / 4-Car / 11-Motorsport / 17-Racing / 40-Road / 41-Sports car / 57-Race track / 352-Need for Speed / 621-Asus / 875-Need for Speed: Most Wanted / 1577-Need for Speed: Most Wanted / \n", "3-Concert / \n", "56-Hair / 106-Hairstyle / 340-Afro-textured hair / 405-Braid / 798-Shampoo / 1812-Hair twists / \n", "10-Animal / 48-Pet / 71-Dog / 182-Puppy / 1013-Walking / 2019-Dental braces / \n", "36-Piano / 53-Keyboard / 104-Pianist / \n", "12-Food / 26-Cooking / 32-Recipe / 52-Dish / 542-Potato / 2955-Waffle / 4415-Mashed potato / \n", "8-Football / 21-Stadium / 60-Basketball / \n", "28-Fashion / \n", "8-Football / \n", "20-Trailer / \n", "68-Lighting / 92-Radio-controlled model / 202-Radio-controlled car / 263-Light / \n", "1-Vehicle / 20-Trailer / 74-Truck / 412-Monster / 787-Monster truck / 1014-Movieclips / \n", "3-Concert / \n", "6-Animation / 15-Cartoon / 126-The Walt Disney Company / 1172-VHS / \n", "12-Food / 26-Cooking / 32-Recipe / 58-Cuisine / 239-Roasting / 1295-Stir frying / \n", "18-Outdoor recreation / 42-Fishing / 478-Fishing rod / 510-Fishing lure / 844-Bass / 3204-Cod / \n", "1-Vehicle / 4-Car / 40-Road / 70-Driving / 517-Audi / \n", "0-Games / 8-Football / 21-Stadium / 54-Highlight film / 65-Kick / 76-Ball / \n", "1642-Cigarette / \n", "489-Board game / 799-Chess / \n", "45-Cosmetics / 780-Fat / \n", "10-Animal / 22-Nature / 48-Pet / 136-Gardening / 588-Reptile / 763-Snake / \n", "3-Concert / \n", "10-Animal / 48-Pet / 71-Dog / 2049-Police dog / 2758-Malinois / \n", "3-Concert / 5-Dance / 16-Performance art / \n", "0-Games / 2-Video game / 34-Action-adventure game / 2140-Tony Hawk's / \n", "358-Rugby football / \n", "10-Animal / 48-Pet / 166-Cat / \n", "3-Concert / \n", "0-Games / 2-Video game / 8-Football / 21-Stadium / 90-Sports game / 97-Drawing / 180-FIFA 15 / 848-FIFA 12 / \n", "6-Animation / 123-Naruto / 183-Manga / 622-Sakura Haruno / \n", "6-Animation / 15-Cartoon / 123-Naruto / 285-Sasuke Uchiha / \n", "0-Games / 8-Football / \n", "10-Animal / 132-Bird / \n", "1-Vehicle / 4-Car / 40-Road / 78-Wedding / \n", "0-Games / 2-Video game / 33-Weapon / 43-Call of Duty / 85-Combat / 134-Call of Duty: Black Ops / 138-Call of Duty: Black Ops II / 1970-Axe / \n", "18-Outdoor recreation / 33-Weapon / 42-Fishing / 227-Water / \n", "1-Vehicle / 4-Car / \n", "9-Music video / \n", "79-American football / \n", "28-Fashion / 768-Diamond / \n", "10-Animal / 48-Pet / 132-Bird / 598-Parrot / \n", "463-Cricket / 1190-Batting / \n", "1-Vehicle / 4-Car / 11-Motorsport / 17-Racing / 57-Race track / \n", "322-Mixtape / \n", "444-Family / \n", "56-Hair / 106-Hairstyle / \n", "9-Music video / 14-Guitar / \n", "361-Flamenco / \n", "1-Vehicle / 11-Motorsport / 50-Aircraft / 82-Airplane / 129-Aviation / 139-Airport / 151-Landing / 157-Jet aircraft / 175-Airliner / 187-Takeoff / 188-Airline / 472-Boeing 737 / \n", "1-Vehicle / 4-Car / 11-Motorsport / 17-Racing / \n", "0-Games / 8-Football / \n", "125-Television / \n", "0-Games / 2-Video game / 19-PC game / 33-Weapon / 43-Call of Duty / 96-Soldier / 704-Call of Duty: World at War / 1248-Call of Duty: Zombies / \n", "5-Dance / \n", "20-Trailer / \n", "0-Games / 985-Strawberry / 1211-Transformice / \n", "0-Games / 241-Fighting game / 486-M.U.G.E.N / 1618-Rayman / \n", "12-Food / 215-Eating / \n", "5-Dance / 430-Ballroom dance / \n", "1-Vehicle / 4-Car / 11-Motorsport / 17-Racing / 84-Snow / 262-Ice / \n", "14-Guitar / 24-String instrument / 100-Electric guitar / \n", "28-Fashion / 757-Boot / \n" ] } ], "source": [ "i = 0\n", "for item in train_labels.iterrows():\n", " labels = item[1][\"labels\"]\n", " for l in sorted(map(int, labels.split())):\n", " print \"%d-%s\" % (l, label[l]), \"/\",\n", " print \"\"\n", " i += 1\n", " if i >= 10000:\n", " break" ] }, { "cell_type": "markdown", "metadata": {}, "source": [ "### statistics for label-sample coverage" ] }, { "cell_type": "code", "execution_count": 11, "metadata": { "collapsed": false }, "outputs": [], "source": [ "i = 0\n", "label_sample_set = {}\n", "for item in train_labels.iterrows():\n", " labels = item[1][\"labels\"]\n", " for l in sorted(map(int, labels.split())):\n", " if l not in label_sample_set:\n", " label_sample_set[l] = set()\n", " label_sample_set[l].add(i)\n", " i += 1" ] }, { "cell_type": "code", "execution_count": 14, "metadata": { "collapsed": true }, "outputs": [], "source": [ "x = range(4716)\n", "y = [0] * 4716\n", "s = set()\n", "for l in x:\n", " s.update(label_sample_set[l])\n", " y[l] = len(s)" ] }, { "cell_type": "code", "execution_count": 19, "metadata": { "collapsed": false }, "outputs": [ { "data": { "image/png": "iVBORw0KGgoAAAANSUhEUgAAAhIAAAFkCAYAAAB1rtL+AAAABHNCSVQICAgIfAhkiAAAAAlwSFlz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\nZf8fJizUt/VvYvkAAAAASUVORK5CYII=\n", "text/plain": [ "" ] }, "metadata": {}, "output_type": "display_data" } ], "source": [ "max_y = float(y[-1])\n", "norm_y = [item/max_y for item in y]\n", "plt.plot(x, norm_y)\n", "plt.show()" ] }, { "cell_type": "code", "execution_count": 22, "metadata": { "collapsed": false }, "outputs": [ { "data": { "text/plain": [ "0.9011726918107226" ] }, "execution_count": 22, "metadata": {}, "output_type": "execute_result" } ], "source": [ "norm_y[200]" ] }, { "cell_type": "markdown", "metadata": {}, "source": [ "## Validate labels" ] }, { "cell_type": "code", "execution_count": 12, "metadata": { "collapsed": false }, "outputs": [ { "data": { "text/html": [ "
\n", "\n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", "
idlabels
0--h7o0bAOpk0 2 581 111 55 317
1--O5HgNtvSc208 1 3073 1294 1073
2--q4Vg_nqEw0 656 2 19 2677
3--wpV2HK8zQ676 28
4--4eRn5OaI83105
5---FI9JLZRg0 2944
6--vnrH7mF1k1 484 345
7--7Z6cyWKmk361 3
8--JGGtGgoqI0 51 373
9--jKa62hkw05
10--Yiihx8DQM1 659 4 1828 519
11--AeBuy3cNY1 4 94 575
12--hqmmJ2fqw298 37 6 15
13--_Ch7tD8xg200 232 126
14--q5_kUt7Qo624 3 5
15--jHGbHxyOo18 83 108
16--s_zBYzhgk0 8 76
17--uMOMKVso40 599
18--_fsPVHB6I8
19--nokr3akJM593
20--b5lu-49ik0 2 1220
21--P7RAx3lrw289 10 80 18 404 438 247
22--fK3UFck7c1 1107 308 1715
23--W1zGEL5gU3
24--mgF8UcvTU68 2703
25--tvS5H063k0 3649 2 34 199
26--2B4d4lQhs104 36 53
27--QhOJxZyjc46 13 38 7
28--nec1RQWNY2062
29--0gptpMDmY66
.........
1401798zzm2O40aPJc8 54 79
1401799zzpLddu603Q101 23 1009 2834 1207 29
1401800zz5irKkXBdM1 1025 42 11 4349 114 147 1076 87 2282
1401801zzJpVCsXhZQ148 61 125
1401802zzuDlhu9sLQ56 890 28 1215
1401803zzBMXp4D3m41 4 135 1169 41 11 17
1401804zzL_IeqdcLA1 13
1401805zzQZvZNaf3E1742
1401806zzwpk_6QnTA3
1401807zzmxIhDgd8o4268 13 726 7
1401808zzZT4OM26ls110 14 7
1401809zzsBXUxAR0g512 228 1063 2634 525 47
1401810zzw1p2KHi1A429 23
1401811zz5LQRMqDTE48 10 636
1401812zzPUcw7_RBg291 235 45 436 56 28
1401813zztZPYsbrvo0 2 19 506 1382
1401814zzyuSRyfNcw1 42 308 605 87
1401815zzCz1UtVRVg160 1 129 82 50 149 92
1401816zzW5krTlh_U32 1223 232 1322 295 12
1401817zzAzChc2wBw0 8 21 54 79
1401818zz8Ah6Yo7Gw1376 10 2479 440 1591 280 217
1401819zzG3TcgGYNA89 387 269 4422 855
1401820zzRUfWjGz9M0 171 159
1401821zzVTEl2QJBw0 2 34 118 3797
1401822zzFGqSCp3t483 108 221
1401823zzpJaaj0FIg3
1401824zz-3UjTglRY20
1401825zz9LRC0P8Os0 25 2 23
1401826zz4DceTfj-c1946 15
1401827zzvd6yEiOmk0 1 2 388 50
\n", "

1401828 rows × 2 columns

\n", "
" ], "text/plain": [ " id labels\n", "0 --h7o0bAOpk 0 2 581 111 55 317\n", "1 --O5HgNtvSc 208 1 3073 1294 1073\n", "2 --q4Vg_nqEw 0 656 2 19 2677\n", "3 --wpV2HK8zQ 676 28\n", "4 --4eRn5OaI8 3105\n", "5 ---FI9JLZRg 0 2944\n", "6 --vnrH7mF1k 1 484 345\n", "7 --7Z6cyWKmk 361 3\n", "8 --JGGtGgoqI 0 51 373\n", "9 --jKa62hkw0 5\n", "10 --Yiihx8DQM 1 659 4 1828 519\n", "11 --AeBuy3cNY 1 4 94 575\n", "12 --hqmmJ2fqw 298 37 6 15\n", "13 --_Ch7tD8xg 200 232 126\n", "14 --q5_kUt7Qo 624 3 5\n", "15 --jHGbHxyOo 18 83 108\n", "16 --s_zBYzhgk 0 8 76\n", "17 --uMOMKVso4 0 599\n", "18 --_fsPVHB6I 8\n", "19 --nokr3akJM 593\n", "20 --b5lu-49ik 0 2 1220\n", "21 --P7RAx3lrw 289 10 80 18 404 438 247\n", "22 --fK3UFck7c 1 1107 308 1715\n", "23 --W1zGEL5gU 3\n", "24 --mgF8UcvTU 68 2703\n", "25 --tvS5H063k 0 3649 2 34 199\n", "26 --2B4d4lQhs 104 36 53\n", "27 --QhOJxZyjc 46 13 38 7\n", "28 --nec1RQWNY 2062\n", "29 --0gptpMDmY 66\n", "... ... ...\n", "1401798 zzm2O40aPJc 8 54 79\n", "1401799 zzpLddu603Q 101 23 1009 2834 1207 29\n", "1401800 zz5irKkXBdM 1 1025 42 11 4349 114 147 1076 87 2282\n", "1401801 zzJpVCsXhZQ 148 61 125\n", "1401802 zzuDlhu9sLQ 56 890 28 1215\n", "1401803 zzBMXp4D3m4 1 4 135 1169 41 11 17\n", "1401804 zzL_IeqdcLA 1 13\n", "1401805 zzQZvZNaf3E 1742\n", "1401806 zzwpk_6QnTA 3\n", "1401807 zzmxIhDgd8o 4268 13 726 7\n", "1401808 zzZT4OM26ls 110 14 7\n", "1401809 zzsBXUxAR0g 512 228 1063 2634 525 47\n", "1401810 zzw1p2KHi1A 429 23\n", "1401811 zz5LQRMqDTE 48 10 636\n", "1401812 zzPUcw7_RBg 291 235 45 436 56 28\n", "1401813 zztZPYsbrvo 0 2 19 506 1382\n", "1401814 zzyuSRyfNcw 1 42 308 605 87\n", "1401815 zzCz1UtVRVg 160 1 129 82 50 149 92\n", "1401816 zzW5krTlh_U 32 1223 232 1322 295 12\n", "1401817 zzAzChc2wBw 0 8 21 54 79\n", "1401818 zz8Ah6Yo7Gw 1376 10 2479 440 1591 280 217\n", "1401819 zzG3TcgGYNA 89 387 269 4422 855\n", "1401820 zzRUfWjGz9M 0 171 159\n", "1401821 zzVTEl2QJBw 0 2 34 118 3797\n", "1401822 zzFGqSCp3t4 83 108 221\n", "1401823 zzpJaaj0FIg 3\n", "1401824 zz-3UjTglRY 20\n", "1401825 zz9LRC0P8Os 0 25 2 23\n", "1401826 zz4DceTfj-c 1946 15\n", "1401827 zzvd6yEiOmk 0 1 2 388 50\n", "\n", "[1401828 rows x 2 columns]" ] }, "execution_count": 12, "metadata": {}, "output_type": "execute_result" } ], "source": [ "validate_labels" ] }, { "cell_type": "code", "execution_count": 13, "metadata": { "collapsed": false }, "outputs": [ { "name": "stdout", "output_type": "stream", "text": [ "0-Games / 2-Video game / 55-Video game console / 111-PlayStation 3 / 317-PlayStation / 581-PlayStation / \n", "1-Vehicle / 208-Wood / 1073-Chainsaw / 1294-Logging / 3073-Log splitter / \n", "0-Games / 2-Video game / 19-PC game / 656-Metal Gear / 2677-Metal Gear Online / \n", "28-Fashion / 676-Handbag / \n", "3105-Numerical digit / \n", "0-Games / 2944-La Tale / \n", "1-Vehicle / 345-Loudspeaker / 484-Subwoofer / \n", "3-Concert / 361-Flamenco / \n", "0-Games / 51-Strategy video game / 373-Clash of Clans / \n", "5-Dance / \n", "1-Vehicle / 4-Car / 519-Recreational vehicle / 659-Van / 1828-Campervan / \n", "1-Vehicle / 4-Car / 94-Dashcam / 575-Traffic / \n", "6-Animation / 15-Cartoon / 37-Gadget / 298-Television advertisement / \n", "126-The Walt Disney Company / 200-Doll / 232-Cake / \n", "3-Concert / 5-Dance / 624-Number / \n", "18-Outdoor recreation / 83-Skateboarding / 108-Skateboard / \n", "0-Games / 8-Football / 76-Ball / \n", "0-Games / 599-Dofus / \n", "8-Football / \n", "593-Moon / \n", "0-Games / 2-Video game / 1220-New Super Mario Bros / \n", "10-Animal / 18-Outdoor recreation / 80-Horse / 247-Stallion / 289-Dressage / 404-Livestock / 438-Mare / \n", "1-Vehicle / 308-Ship / 1107-United States Navy / 1715-Aircraft carrier / \n", "3-Concert / \n", "68-Lighting / 2703-Light fixture / \n", "0-Games / 2-Video game / 34-Action-adventure game / 199-Wii / 3649-Metroid Prime / \n", "36-Piano / 53-Keyboard / 104-Pianist / \n", "7-Musician / 13-Musical ensemble / 38-Orchestra / 46-Choir / \n", "2062-Kidney / \n", "66-Bollywood / \n", "273-News program / \n", "49-School / \n", "10-Animal / 18-Outdoor recreation / 42-Fishing / 91-Fish / 3292-Rock fishing / \n", "56-Hair / 106-Hairstyle / 871-Hair coloring / 1269-Human hair color / \n", "13-Musical ensemble / 31-Disc jockey / \n", "5-Dance / 399-Kickboxing / \n", "9-Music video / \n", "3-Concert / 7-Musician / 13-Musical ensemble / 46-Choir / \n", "3-Concert / 14-Guitar / \n", "45-Cosmetics / 282-Skin / 291-Rouge / \n", "9-Music video / \n", "45-Cosmetics / 193-Eye liner / 195-Mascara / 218-Eyelash / 235-Lipstick / 291-Rouge / 436-Concealer / \n", "28-Fashion / 276-Room / \n", "28-Fashion / 205-Newscaster / \n", "14-Guitar / 24-String instrument / 100-Electric guitar / \n", "0-Games / 2-Video game / 33-Weapon / 948-Far Cry / 3618-Far Cry / \n", "1-Vehicle / 4-Car / 11-Motorsport / 135-Drifting / 201-Grand Theft Auto: San Andreas / \n", "1-Vehicle / 4-Car / 297-Grand Theft Auto IV / \n", "242-Paper / 1642-Cigarette / \n", "0-Games / 2-Video game / 1288-Marvel vs. Capcom / 3081-Marvel vs. Capcom 2: New Age of Heroes / \n", "23-Mobile phone / 29-Smartphone / 72-iPhone / 272-iPod / 331-iPod touch / \n", "5-Dance / 769-Red Bull / \n", "359-Harry Potter (Literary Series) / \n", "0-Games / 2-Video game / 25-Toy / 140-LEGO / 199-Wii / 443-Wii U / 1079-Lego City / 4016-Lego City Undercover / \n", "28-Fashion / 45-Cosmetics / 173-Eye shadow / 193-Eye liner / 195-Mascara / 196-Eye / 218-Eyelash / 235-Lipstick / \n", "30-Drummer / 39-Drums / 44-Drums / 128-Cymbal / 146-Snare drum / \n", "164-Beach / 369-Resort / \n", "0-Games / 8-Football / 54-Highlight film / 76-Ball / \n", "12-Food / 170-Home improvement / 246-Furniture / 303-Paint / 1272-Cabinetry / \n", "45-Cosmetics / 56-Hair / 235-Lipstick / \n", "1880-Marinera / \n", "6-Animation / 15-Cartoon / \n", "36-Piano / 53-Keyboard / \n", "3-Concert / 7-Musician / \n", "0-Games / 2-Video game / 241-Fighting game / 764-The King of Fighters / 3818-The King of Fighters '98 / \n", "0-Games / 2-Video game / 34-Action-adventure game / 380-Assassin's Creed / \n", "1182-Softball / \n", "3-Concert / 5-Dance / \n", "3-Concert / \n", "75-Wrestling / \n", "5-Dance / 16-Performance art / 49-School / 66-Bollywood / \n", "959-Metal detector / \n", "1-Vehicle / 4-Car / 11-Motorsport / 17-Racing / 1040-Street racing / \n", "67-Cycling / 69-Bicycle / \n", "67-Cycling / \n", "230-Brass instrument / 427-Trumpet / \n", "28-Fashion / 45-Cosmetics / 1352-Airbrush / \n", "0-Games / 264-Star Wars / 714-Jedi / \n", "2-Video game / 6-Animation / 199-Wii / 1004-Samurai / \n", "10-Animal / 71-Dog / \n", "1-Vehicle / 27-Motorcycle / 245-BMW / 2763-BMW R1200GS / 2975-BMW Motorrad / \n", "12-Food / 86-Plant / \n", "5-Dance / 99-Christmas / 558-Santa Claus / \n", "20-Trailer / 643-Hero / 1632-Villain / \n", "8-Football / 21-Stadium / 77-Arena / \n", "28-Fashion / 45-Cosmetics / 56-Hair / 173-Eye shadow / 195-Mascara / 218-Eyelash / \n", "5-Dance / 1139-Belle / 1265-Kyle Kingson / \n", "5-Dance / 16-Performance art / \n", "1-Vehicle / 4-Car / \n", "227-Water / 640-Bottle / 1058-Volcano / \n", "56-Hair / \n", "3-Concert / 7-Musician / 13-Musical ensemble / 46-Choir / \n", "455-Unidentified flying object / \n", "5-Dance / \n", "0-Games / 8-Football / 21-Stadium / 65-Kick / 73-Ball / 76-Ball / 77-Arena / 81-Athlete / 1247-Penalty kick / \n", "60-Basketball / \n", "3-Concert / 7-Musician / \n", "12-Food / \n", "1782-Supermarket / \n", "15-Cartoon / 467-Chipmunk / \n", "20-Trailer / \n", "0-Games / 1-Vehicle / 33-Weapon / 51-Strategy video game / 271-Tank / 409-World of Tanks / \n", "1-Vehicle / 18-Outdoor recreation / 147-River / 481-Kayak / \n", "14-Guitar / 63-Acoustic guitar / \n", "1-Vehicle / 67-Cycling / 69-Bicycle / 220-Mountain bike / \n", "1-Vehicle / 4-Car / 114-Engine / \n", "66-Bollywood / \n", "2-Video game / 47-Personal computer / 131-Computer / 309-Computer hardware / 700-Hard disk drive / \n", "0-Games / 2-Video game / 51-Strategy video game / 165-League of Legends / \n", "10-Animal / 12-Food / 26-Cooking / 32-Recipe / 52-Dish / 58-Cuisine / 91-Fish / 109-Cooking show / 239-Roasting / 391-Chicken meat / 454-Sauce / 893-Fish as food / 2125-Fillet / \n", "3-Concert / 7-Musician / 13-Musical ensemble / 46-Choir / \n", "3-Concert / 16-Performance art / 68-Lighting / \n", "0-Games / 2-Video game / 43-Call of Duty / 55-Video game console / 111-PlayStation 3 / 216-Call of Duty: Modern Warfare 3 / \n", "93-Comedy / \n", "3-Concert / \n", "257-Counter-Strike / 1226-Metalworking / \n", "36-Piano / 53-Keyboard / 397-Organ / \n", "7-Musician / 13-Musical ensemble / 46-Choir / \n", "3-Concert / 16-Performance art / \n", "1-Vehicle / 4-Car / 683-Renault / 4672-Renault 5 / \n", "5-Dance / 78-Wedding / 153-Dress / \n", "10-Animal / 22-Nature / 42-Fishing / 718-Duck / 814-Pond / \n", "12-Food / 52-Dish / 58-Cuisine / 86-Plant / 500-Rice / 1438-Street food / 2037-Straw / \n", "0-Games / 1644-ArcheAge / 2840-Engraving / \n", "9-Music video / \n", "12-Food / 49-School / \n", "0-Games / 2-Video game / 638-Gears of War / 773-Gears of War / \n", "0-Games / 35-Minecraft / \n", "20-Trailer / \n", "8-Football / 54-Highlight film / 65-Kick / \n", "6-Animation / 123-Naruto / \n", "3-Concert / 7-Musician / 13-Musical ensemble / 46-Choir / \n", "444-Family / \n", "9-Music video / \n", "66-Bollywood / \n", "9-Music video / \n", "18-Outdoor recreation / 22-Nature / 283-Rock / 347-Climbing / 351-Jumping / 548-Parachuting / 669-Extreme sport / \n", "7-Musician / 14-Guitar / 24-String instrument / 30-Drummer / 39-Drums / \n", "0-Games / 1-Vehicle / 2-Video game / 4-Car / 19-PC game / 34-Action-adventure game / 118-Grand Theft Auto V / 1275-Watch Dogs / \n", "78-Wedding / 121-Photography / 190-Bride / \n", "0-Games / 2-Video game / 47-Personal computer / 345-Loudspeaker / 484-Subwoofer / \n", "25-Toy / 269-Comic book / \n", "9-Music video / \n", "164-Beach / 502-Sky / 573-Paragliding / 1561-Paramotor / \n", "3-Concert / \n", "3-Concert / \n", "12-Food / 26-Cooking / 32-Recipe / 227-Water / \n", "102-Building / 885-Bell / \n", "238-Nail / 304-Nail polish / 1498-Key / \n", "6-Animation / 15-Cartoon / 185-Dragon Ball / 244-Goku / 418-Gohan / 876-Freeza / \n", "6-Animation / 183-Manga / \n", "0-Games / 8-Football / 21-Stadium / 65-Kick / 73-Ball / 76-Ball / 77-Arena / 81-Athlete / \n", "14-Guitar / \n", "12-Food / 26-Cooking / 32-Recipe / 109-Cooking show / 236-Baking / 442-Bread / \n", "36-Piano / 53-Keyboard / 379-Touhou Project / 3528-Imperishable Night / \n", "1-Vehicle / 11-Motorsport / 124-Tractor / 156-Agriculture / 174-Farm / 252-Tractor pulling / \n", "3-Concert / \n", "18-Outdoor recreation / 83-Skateboarding / 108-Skateboard / 221-Skatepark / \n", "10-Animal / 22-Nature / 132-Bird / 137-Tree / \n", "18-Outdoor recreation / 81-Athlete / 287-Human swimming / 2605-Medley swimming / \n", "3-Concert / 7-Musician / 13-Musical ensemble / 14-Guitar / 30-Drummer / 39-Drums / 44-Drums / \n", "0-Games / 2-Video game / 43-Call of Duty / 138-Call of Duty: Black Ops II / \n", "13-Musical ensemble / \n", "14-Guitar / \n", "8-Football / \n", "0-Games / 1-Vehicle / 2-Video game / 34-Action-adventure game / 102-Building / 118-Grand Theft Auto V / \n", "8-Football / \n", "18-Outdoor recreation / 147-River / 794-Bridge / \n", "116-Pokémon / \n", "0-Games / 2-Video game / 14-Guitar / 445-Guitar Hero / 3558-Guitar Hero: Metallica / \n", "9-Music video / \n", "8-Football / \n", "1678-Foam / \n", "1-Vehicle / 27-Motorcycle / 207-Exhaust system / 1380-Kawasaki motorcycles / 2404-Kawasaki Ninja / \n", "20-Trailer / 390-Model / \n", "102-Building / 113-House / \n", "1-Vehicle / 4-Car / 84-Snow / 120-Winter / \n", "3-Concert / \n", "3-Concert / 5-Dance / 483-Costume / \n", "3169-Diva / \n", "10-Animal / 18-Outdoor recreation / 22-Nature / 86-Plant / 213-Forest / 217-Hunting / 280-Wildlife / 706-Deer / 709-Bull / 2804-Elk / \n", "616-The Bible / \n", "13-Musical ensemble / 28-Fashion / 153-Dress / \n", "5-Dance / 16-Performance art / 1422-Statue / \n", "1-Vehicle / 50-Aircraft / 139-Airport / 157-Jet aircraft / 175-Airliner / 188-Airline / 1340-Airbus A380 / 2790-British Airways / \n", "131-Computer / 141-Microsoft Windows / \n", "47-Personal computer / 131-Computer / 414-Computer monitor / \n", "1-Vehicle / 62-Train / 64-Transport / 103-Rail transport / 119-Locomotive / 143-Railroad car / 152-Track / 161-Train station / 691-Cargo / \n", "10-Animal / 91-Fish / 258-Aquarium / 366-Dragon / \n", "410-Teacher / \n", "96-Soldier / \n", "6-Animation / \n", "6-Animation / 15-Cartoon / \n", "0-Games / 1-Vehicle / 2-Video game / 4-Car / 11-Motorsport / 17-Racing / 57-Race track / 1169-Hot Wheels / \n", "3-Concert / 7-Musician / 14-Guitar / 49-School / 98-Festival / \n", "0-Games / 51-Strategy video game / 373-Clash of Clans / \n", "10-Animal / 48-Pet / 71-Dog / 881-German Shepherd / \n", "3-Concert / 7-Musician / 14-Guitar / 24-String instrument / 63-Acoustic guitar / 150-Violin / 726-Clarinet / \n", "3-Concert / 6-Animation / \n", "1-Vehicle / 50-Aircraft / 82-Airplane / 92-Radio-controlled model / 149-Model aircraft / 157-Jet aircraft / 160-Radio-controlled aircraft / \n", "7-Musician / 14-Guitar / 24-String instrument / 219-Accordion / 2185-PRS Guitars / \n", "2-Video game / 55-Video game console / 118-Grand Theft Auto V / \n", "223-Weather / 225-Lake / \n", "1-Vehicle / 4-Car / 40-Road / 94-Dashcam / 575-Traffic / 663-Street / 1960-Traffic light / \n", "7-Musician / 14-Guitar / \n", "6-Animation / 15-Cartoon / \n", "744-Window / 4448-Window cleaner / \n", "1-Vehicle / 4-Car / 509-Classic car / 1321-Škoda Auto / \n", "0-Games / 2-Video game / 237-Final Fantasy / 1131-Final Fantasy XIII / 2506-Final Fantasy XIII-2 / \n", "5-Dance / 7-Musician / 13-Musical ensemble / 49-School / 203-Marching band / 3669-Oceanic dolphin / \n", "7-Musician / 9-Music video / \n", "1-Vehicle / 11-Motorsport / 1025-Personal water craft / 1336-Jet Ski / \n", "3-Concert / 7-Musician / 13-Musical ensemble / 46-Choir / \n", "31-Disc jockey / \n", "1-Vehicle / 4-Car / 4115-Nissan JUKE / \n", "3-Concert / 9-Music video / 178-Samsung Galaxy / \n", "6-Animation / 15-Cartoon / \n", "0-Games / 2-Video game / 19-PC game / 33-Weapon / 192-Battlefield / 464-First-person Shooter / 1267-Battlefield Hardline / \n", "5-Dance / \n", "67-Cycling / 69-Bicycle / 520-Bicycle frame / \n", "0-Games / 530-Cue sports / 568-Pool / 994-Cue stick / 1334-Eight-ball / 1364-Nine-ball / \n", "3-Concert / \n", "5-Dance / \n", "6-Animation / 15-Cartoon / \n", "444-Family / \n", "223-Weather / \n", "38-Orchestra / \n", "3-Concert / 5-Dance / 30-Drummer / \n", "36-Piano / 53-Keyboard / \n", "6-Animation / 801-Inuyasha / \n", "28-Fashion / 45-Cosmetics / 466-Gift / 4258-Raiden / \n", "6-Animation / 13-Musical ensemble / \n", "8-Football / \n", "45-Cosmetics / 173-Eye shadow / 193-Eye liner / 195-Mascara / 196-Eye / 218-Eyelash / \n", "3-Concert / 7-Musician / 13-Musical ensemble / 46-Choir / \n", "0-Games / 326-Metin2 / \n", "148-Painting / \n", "1-Vehicle / 4-Car / 11-Motorsport / 17-Racing / 378-Drag racing / 473-Ford Mustang / \n", "1-Vehicle / 4-Car / 234-Ford / 473-Ford Mustang / 918-Brake / \n", "1-Vehicle / 4-Car / 389-Restaurant / \n", "3-Concert / 5-Dance / 16-Performance art / \n", "1151-Temple / \n", "1449-Rubber band / \n", "1-Vehicle / 117-Boxing / \n", "20-Trailer / \n", "3-Concert / 5-Dance / 16-Performance art / \n", "1-Vehicle / 62-Train / 64-Transport / 103-Rail transport / 119-Locomotive / 143-Railroad car / 161-Train station / 224-Rapid transit / \n", "1-Vehicle / 4-Car / 11-Motorsport / 918-Brake / \n", "938-Steel / 3515-Gas tungsten arc welding / \n", "1939-Logic Pro / \n", "1-Vehicle / 87-Boat / 308-Ship / 2007-Intermodal container / \n", "0-Games / 326-Metin2 / \n", "1-Vehicle / 4-Car / 11-Motorsport / 17-Racing / 27-Motorcycle / 105-Motorcycling / \n", "9-Music video / \n", "115-Ballet / 2339-Swan / \n", "5-Dance / 16-Performance art / \n", "0-Games / 2-Video game / 19-PC game / 192-Battlefield / 266-Helicopter / 374-Battlefield 3 / 3513-Split screen / \n", "0-Games / 2-Video game / 55-Video game console / 116-Pokémon / 1118-Kirby / 1133-Kirby / 1446-Game Boy / 1637-GameCube / \n", "113-House / 170-Home improvement / \n", "1-Vehicle / 87-Boat / 308-Ship / 3163-Tugboat / \n", "1-Vehicle / 4-Car / \n", "5-Dance / 68-Lighting / 133-Nightclub / \n", "5-Dance / 410-Teacher / \n", "0-Games / 2-Video game / 229-Halo / 644-Halo: Reach / \n", "246-Furniture / 276-Room / 416-Apartment / \n", "0-Games / 8-Football / 54-Highlight film / 79-American football / \n", "248-Hotel / 369-Resort / \n", "3-Concert / 68-Lighting / \n", "0-Games / 60-Basketball / \n", "208-Wood / 261-Tool / 315-Knife / 400-Woodturning / 2017-Sharpening / 4309-Plane / \n", "0-Games / 2-Video game / 20-Trailer / 192-Battlefield / 374-Battlefield 3 / 383-Battlefield 4 / \n", "3-Concert / 5-Dance / 16-Performance art / \n", "0-Games / 75-Wrestling / 977-WWE '13 / \n", "0-Games / 2-Video game / 6-Animation / 506-Garry's Mod / \n", "1-Vehicle / 27-Motorcycle / 105-Motorcycling / 114-Engine / 207-Exhaust system / \n", "1-Vehicle / 59-Winter sport / 84-Snow / 120-Winter / 177-Skiing / 281-Snowboarding / 305-Ski / 501-Snowboard / \n", "315-Knife / \n", "1-Vehicle / 4-Car / 114-Engine / 2377-Cylinder / \n", "0-Games / 271-Tank / 409-World of Tanks / \n", "15-Cartoon / 750-Sailor Moon / \n", "144-Amusement park / 3794-Legoland / \n", "1-Vehicle / 11-Motorsport / 17-Racing / 18-Outdoor recreation / 87-Boat / 147-River / \n", "0-Games / 8-Football / 77-Arena / 90-Sports game / 180-FIFA 15 / \n", "0-Games / 2-Video game / 90-Sports game / 180-FIFA 15 / 480-FIFA 13 / \n", "8-Football / 21-Stadium / 77-Arena / \n", "0-Games / 75-Wrestling / \n", "444-Family / \n", "10-Animal / 48-Pet / 166-Cat / 346-Kitten / \n", "3-Concert / \n", "3-Concert / 7-Musician / 13-Musical ensemble / 39-Drums / \n", "85-Combat / \n", "14-Guitar / 24-String instrument / 100-Electric guitar / \n", "6-Animation / 15-Cartoon / 61-Art / 97-Drawing / 4248-Spacetoon / \n", "1-Vehicle / 4-Car / 114-Engine / 945-MINI Cooper / 1221-MINI / \n", "10-Animal / 48-Pet / 71-Dog / 132-Bird / 598-Parrot / \n", "512-Printing / 525-Printer / 551-British Broadcasting Corporation / \n", "5-Dance / 16-Performance art / \n", "1-Vehicle / 610-Snowmobile / \n", "2-Video game / 159-World of Warcraft / \n", "10-Animal / 523-Cattle / 1457-Herd / \n", "12-Food / 232-Cake / 1175-Soap / \n", "0-Games / 2-Video game / 19-PC game / 686-Diablo III / 872-Diablo / \n", "1-Vehicle / 4-Car / 84-Snow / 158-Tire / 167-Four-wheel drive / 198-Off-road vehicle / 360-Jeep / 1205-Jeep Wrangler / \n", "361-Flamenco / \n", "6-Animation / 10-Animal / 15-Cartoon / 3332-Hippopotamus / \n", "0-Games / 51-Strategy video game / 174-Farm / 373-Clash of Clans / \n", "18-Outdoor recreation / 84-Snow / 120-Winter / 137-Tree / 347-Climbing / 395-Camping / 1203-Backpacking / 3065-Hammock / \n", "3-Concert / 7-Musician / 30-Drummer / 39-Drums / \n", "45-Cosmetics / 56-Hair / \n", "9-Music video / \n", "8-Football / \n", "12-Food / \n", "25-Toy / 140-LEGO / 264-Star Wars / \n", "1105-Fingerboard / \n", "1-Vehicle / 62-Train / 64-Transport / 94-Dashcam / 103-Rail transport / 119-Locomotive / 143-Railroad car / 152-Track / 691-Cargo / \n", "36-Piano / 53-Keyboard / \n", "110-Album / \n", "12-Food / 26-Cooking / 32-Recipe / 52-Dish / 58-Cuisine / 274-Meat / 391-Chicken meat / 572-Indian cuisine / 589-Grilling / 620-Barbecue / 663-Street / 1438-Street food / \n", "12-Food / 26-Cooking / 32-Recipe / 226-Dessert / 232-Cake / 236-Baking / 295-Chocolate / 437-Cream / 560-Sugar / 578-Cookie / 684-Ice cream / 759-Icing / 795-Cupcake / 966-Sandwich / 1408-Snack / 1541-Buttercream / 4319-Oreo / \n", "122-Weight training / 516-Headphones / 529-Squat / \n", "10-Animal / 48-Pet / 71-Dog / 1278-Labrador Retriever / \n", "8-Football / \n", "6-Animation / 15-Cartoon / 637-Five Nights at Freddy's / \n", "0-Games / 2-Video game / 51-Strategy video game / 1570-Command & Conquer / 4526-Command & Conquer: Generals / \n", "0-Games / 2-Video game / 1027-Terraria / 1130-Armour / \n", "3-Concert / 7-Musician / 13-Musical ensemble / 38-Orchestra / \n", "8-Football / 21-Stadium / \n", "9-Music video / \n", "0-Games / 51-Strategy video game / 165-League of Legends / \n", "1066-Aikido / 2531-Dojo / \n", "0-Games / 8-Football / 79-American football / \n", "5-Dance / 133-Nightclub / 342-Windows Media Video / \n", "93-Comedy / 355-Comedian / \n", "43-Call of Duty / 302-Call of Duty 4: Modern Warfare / \n", "14-Guitar / \n", "6-Animation / 15-Cartoon / 2784-Sponge / \n", "3-Concert / 7-Musician / \n", "5-Dance / 95-Talent show / \n", "0-Games / 2-Video game / 19-PC game / 33-Weapon / 441-The Elder Scrolls / 1919-The Elder Scrolls Online / 3569-Scrolls / \n", "1-Vehicle / 4-Car / 6-Animation / 2769-Decal / \n", "0-Games / 2-Video game / 2491-Spec Ops: The Line / \n", "30-Drummer / 36-Piano / 53-Keyboard / 107-Musical keyboard / 397-Organ / \n", "88-Machine / 3515-Gas tungsten arc welding / 3929-Gas metal arc welding / \n", "10-Animal / \n", "609-Tattoo / \n", "43-Call of Duty / 194-Call of Duty: Modern Warfare 2 / \n", "0-Games / 2-Video game / 366-Dragon / 412-Monster / \n", "12-Food / 26-Cooking / 32-Recipe / 232-Cake / 485-Dough / 560-Sugar / 578-Cookie / \n", "110-Album / \n", "3-Concert / \n", "3-Concert / 14-Guitar / 16-Performance art / \n", "117-Boxing / \n", "1-Vehicle / 4-Car / 11-Motorsport / 17-Racing / 27-Motorcycle / 40-Road / 57-Race track / 105-Motorcycling / 2133-Road racing / \n", "3-Concert / 30-Drummer / \n", "0-Games / 2-Video game / 19-PC game / 33-Weapon / 43-Call of Duty / 96-Soldier / 134-Call of Duty: Black Ops / 138-Call of Duty: Black Ops II / 216-Call of Duty: Modern Warfare 3 / \n", "3-Concert / 7-Musician / 24-String instrument / 39-Drums / \n", "75-Wrestling / \n", "72-iPhone / 189-iPad / \n", "14-Guitar / 24-String instrument / 63-Acoustic guitar / 183-Manga / \n", "0-Games / 2-Video game / 33-Weapon / 43-Call of Duty / \n", "3-Concert / 7-Musician / 14-Guitar / \n", "9-Music video / 92-Radio-controlled model / 149-Model aircraft / 155-Camera / 160-Radio-controlled aircraft / 278-GoPro / 393-Unmanned aerial vehicle / 604-Quadcopter / \n", "66-Bollywood / 355-Comedian / \n", "12-Food / 26-Cooking / 52-Dish / 58-Cuisine / 121-Photography / 301-Meal / 2040-Korean food / \n", "60-Basketball / 1399-Streetball / \n", "10-Animal / 22-Nature / 80-Horse / \n", "1-Vehicle / 4-Car / 11-Motorsport / 40-Road / 277-Sport utility vehicle / 1391-Kia / 1415-Kia Motors / 4398-Kia Sorento / \n", "3-Concert / 7-Musician / 13-Musical ensemble / 14-Guitar / 30-Drummer / 44-Drums / 1533-Lead guitar / \n", "248-Hotel / 369-Resort / \n", "1524-Zee Bangla / \n", "10-Animal / 22-Nature / 71-Dog / 217-Hunting / \n", "1-Vehicle / 11-Motorsport / 27-Motorcycle / 105-Motorcycling / 806-Chopper / \n", "0-Games / 2-Video game / 12-Food / 713-Egg / 1047-Easter egg / 1069-Easter egg / \n", "6-Animation / 15-Cartoon / 849-Tales / 2195-Tales of Symphonia / \n", "512-Printing / \n", "1-Vehicle / 40-Road / 938-Steel / \n", "6-Animation / 15-Cartoon / 2403-Crayon Shin-chan / \n", "22-Nature / 86-Plant / 102-Building / 657-Autumn / \n", "102-Building / 113-House / 170-Home improvement / 179-Kitchen / 205-Newscaster / 276-Room / 720-Wine / 2055-Mansion / \n", "1-Vehicle / 4-Car / 84-Snow / 120-Winter / 1768-Snow blower / \n", "5-Dance / 16-Performance art / 430-Ballroom dance / 678-Latin dance / \n", "30-Drummer / 39-Drums / 44-Drums / 128-Cymbal / 146-Snare drum / \n", "0-Games / 2-Video game / 34-Action-adventure game / 176-PlayStation 4 / 656-Metal Gear / 2212-Metal Gear Solid V: The Phantom Pain / 4027-Metal Gear Solid V: Ground Zeroes / \n", "3-Concert / 7-Musician / 13-Musical ensemble / 14-Guitar / 16-Performance art / 24-String instrument / \n", "0-Games / 362-PlayStation Portable / \n", "0-Games / 2-Video game / 1138-Tower of Saviors / \n", "12-Food / 915-Town / \n", "199-Wii / \n", "18-Outdoor recreation / 83-Skateboarding / 108-Skateboard / 1031-Fisheye lens / \n", "1-Vehicle / 4-Car / 87-Boat / 114-Engine / \n", "7-Musician / 14-Guitar / 24-String instrument / \n", "1-Vehicle / 4-Car / 11-Motorsport / 70-Driving / 1056-Porsche 911 / \n", "356-Star / \n", "3-Concert / 5-Dance / \n", "9-Music video / \n", "9-Music video / 22-Nature / 267-Mountain / 283-Rock / \n", "22-Nature / 86-Plant / 121-Photography / \n", "0-Games / 2-Video game / 271-Tank / 1167-Tanki Online / \n", "10-Animal / 12-Food / 91-Fish / 1018-Salmon / 1227-Sushi / \n", "7-Musician / 46-Choir / \n", "269-Comic book / 357-DVD / 950-Blu-ray disc / \n", "548-Parachuting / 573-Paragliding / 765-Parachute / \n", "0-Games / 2-Video game / 15-Cartoon / 89-Comics / 185-Dragon Ball / 244-Goku / \n", "0-Games / 2-Video game / 35-Minecraft / 47-Personal computer / 145-Tablet computer / \n", "0-Games / 2-Video game / 928-Borderlands 2 / 2613-Borderlands: The Pre-Sequel! / \n", "15-Cartoon / 183-Manga / \n", "708-Mermaid / 1219-Ariel / \n", "23-Mobile phone / 29-Smartphone / 37-Gadget / \n", "59-Winter sport / 112-Ice skating / 142-Hockey / \n", "20-Trailer / 4211-Walter White / \n", "0-Games / 2-Video game / 8-Football / 79-American football / 90-Sports game / 561-Madden NFL / \n", "83-Skateboarding / 108-Skateboard / 556-Skateboarding trick / 1315-Kickflip / 1454-Stairs / 4097-Nollie / \n", "6-Animation / 25-Toy / 701-Beyblade / \n", "45-Cosmetics / 173-Eye shadow / 193-Eye liner / 195-Mascara / 196-Eye / 218-Eyelash / \n", "1-Vehicle / 4-Car / 11-Motorsport / 17-Racing / 169-Rallying / \n", "1-Vehicle / 22-Nature / 124-Tractor / 156-Agriculture / 249-Heavy equipment / 617-Maize / \n", "3-Concert / 5-Dance / 16-Performance art / 68-Lighting / \n", "7-Musician / 36-Piano / \n", "3-Concert / 46-Choir / \n", "6-Animation / \n", "6-Animation / 15-Cartoon / 1211-Transformice / \n", "0-Games / 8-Football / 77-Arena / 2985-Chunichi Dragons Doala / \n", "269-Comic book / \n", "75-Wrestling / \n", "0-Games / 2-Video game / 19-PC game / 33-Weapon / 2554-F.E.A.R. / \n", "0-Games / 35-Minecraft / \n", "54-Highlight film / 79-American football / \n", "0-Games / 8-Football / 79-American football / 358-Rugby football / \n", "22-Nature / 36-Piano / \n", "13-Musical ensemble / 17-Racing / 203-Marching band / \n", "49-School / 286-Bus / 1013-Walking / 2199-School bus / \n", "5-Dance / 3902-Just Dance 3 / 4437-Just Dance 2 / \n", "1-Vehicle / 11-Motorsport / 198-Off-road vehicle / 213-Forest / 327-Mud / 565-Mud bogging / \n", "1-Vehicle / 221-Skatepark / 336-Scooter / \n", "5-Dance / 78-Wedding / 2653-Wedding ring / \n", "93-Comedy / \n", "3-Concert / \n", "31-Disc jockey / \n", "350-Circus / \n", "18-Outdoor recreation / 42-Fishing / 233-Ocean / 275-Diving / 381-Underwater diving / 2618-Lobster / \n", "9-Music video / \n", "75-Wrestling / 1582-WWE 2K14 / \n", "5-Dance / \n", "14-Guitar / \n", "9-Music video / 66-Bollywood / \n", "9-Music video / 14-Guitar / 488-Door / \n", "6-Animation / 15-Cartoon / 971-Alphabet / \n", "1-Vehicle / 62-Train / 64-Transport / 103-Rail transport / 119-Locomotive / 143-Railroad car / 152-Track / 161-Train station / 224-Rapid transit / 1547-Passenger / \n", "0-Games / 739-Guild Wars / \n", "0-Games / 2-Video game / 33-Weapon / 43-Call of Duty / 96-Soldier / 422-Call of Duty: Ghosts / 3522-ProSieben / \n", "5-Dance / 66-Bollywood / \n", "9-Music video / \n", "3-Concert / 7-Musician / 30-Drummer / \n", "30-Drummer / 39-Drums / 44-Drums / 128-Cymbal / 146-Snare drum / \n", "0-Games / 211-RuneScape / \n", "110-Album / \n", "3-Concert / \n", "1-Vehicle / 4-Car / 11-Motorsport / 41-Sports car / 231-Supercar / 790-Nissan GT-R / 1301-Nissan Skyline / 2247-Nissan Skyline GT-R / \n", "0-Games / 4467-Jade Dynasty / \n", "3-Concert / \n", "22-Nature / 657-Autumn / \n", "1-Vehicle / 50-Aircraft / 129-Aviation / 139-Airport / 151-Landing / 187-Takeoff / 451-Cockpit / \n", "10-Animal / 20-Trailer / 80-Horse / \n", "3-Concert / 16-Performance art / \n", "3-Concert / 5-Dance / 49-School / 776-Kindergarten / \n", "3-Concert / 7-Musician / 14-Guitar / 98-Festival / 670-Banjo / \n", "59-Winter sport / 112-Ice skating / 142-Hockey / \n", "0-Games / 2-Video game / 586-Nintendo Entertainment System / \n", "1-Vehicle / 4-Car / 17-Racing / 40-Road / 57-Race track / 94-Dashcam / 245-BMW / 254-Runway / 874-BMW M3 / 1607-BMW 3 Series (E36) / \n", "5-Dance / 59-Winter sport / 112-Ice skating / 316-Figure skating / \n", "3-Concert / 5-Dance / 31-Disc jockey / 133-Nightclub / 559-Night / \n", "3-Concert / 7-Musician / 13-Musical ensemble / 46-Choir / \n", "24-String instrument / 150-Violin / 450-Viola / \n", "1-Vehicle / 11-Motorsport / 27-Motorcycle / 154-Wheel / 158-Tire / \n", "3-Concert / 7-Musician / 13-Musical ensemble / 38-Orchestra / 46-Choir / \n", "6-Animation / 15-Cartoon / 1771-The Pink Panther / 2520-Pink Panther / \n", "89-Comics / \n", "1-Vehicle / 11-Motorsport / 17-Racing / 27-Motorcycle / 105-Motorcycling / 172-Motocross / 363-Soil / 669-Extreme sport / \n", "1-Vehicle / 4-Car / 10-Animal / 12-Food / 26-Cooking / 274-Meat / 723-Pig / 862-Pork / 1123-Domestic pig / 1557-Bacon / 1843-Ham / 2395-Ribs / \n", "0-Games / 2-Video game / 6-Animation / 89-Comics / 123-Naruto / 285-Sasuke Uchiha / 591-Naruto: Ultimate Ninja / 1421-Naruto: Ultimate Ninja Storm / 2158-Naruto Shippuden: Ultimate Ninja Storm Generations / \n", "45-Cosmetics / 56-Hair / 155-Camera / \n", "1-Vehicle / 4-Car / 64-Transport / 178-Samsung Galaxy / 899-Samsung Galaxy S III / 4290-Engadget / \n", "1-Vehicle / 4-Car / 41-Sports car / 349-Chevrolet / 509-Classic car / 650-Convertible / 952-Chevrolet Corvette / \n", "93-Comedy / \n", "9-Music video / \n", "0-Games / 2-Video game / 72-iPhone / \n", "3-Concert / \n", "115-Ballet / 350-Circus / 2358-Contortion / \n", "1-Vehicle / 4-Car / 207-Exhaust system / 245-BMW / \n", "66-Bollywood / \n", "0-Games / 2-Video game / 1777-Ratchet / 1824-Clank / 1859-Ratchet & Clank / \n", "1-Vehicle / 18-Outdoor recreation / 67-Cycling / 69-Bicycle / 197-Trail / 220-Mountain bike / 284-Mountain biking / \n", "3-Concert / 7-Musician / 13-Musical ensemble / 14-Guitar / 24-String instrument / 39-Drums / 44-Drums / 296-Saxophone / 1771-The Pink Panther / \n", "121-Photography / 1046-Remote control / \n", "10-Animal / 48-Pet / 71-Dog / 182-Puppy / \n", "259-Fire / \n", "0-Games / 2-Video game / 20-Trailer / 2056-Battleship / \n", "122-Weight training / 214-Muscle / 295-Chocolate / 841-Butter / 1162-Dietary supplement / 1814-Peanut / 2197-Peanut butter / 2681-Marshmallow / 3498-Whey protein / \n", "3-Concert / 7-Musician / 30-Drummer / 39-Drums / 44-Drums / \n", "3-Concert / 7-Musician / 14-Guitar / 99-Christmas / \n", "116-Pokémon / \n", "66-Bollywood / \n", "9-Music video / \n", "445-Guitar Hero / \n", "45-Cosmetics / 56-Hair / 235-Lipstick / \n", "0-Games / 2-Video game / 19-PC game / 35-Minecraft / 80-Horse / \n", "6-Animation / 269-Comic book / \n", "6-Animation / 15-Cartoon / \n", "3-Concert / 68-Lighting / 4687-Hollywood Palladium / \n", "12-Food / 391-Chicken meat / 1408-Snack / 2080-Cup / 2943-KFC / \n", "0-Games / 2-Video game / 34-Action-adventure game / 118-Grand Theft Auto V / 176-PlayStation 4 / \n", "260-Prayer / 1944-Vatican City / \n", "5-Dance / 2551-Para Para / \n", "1-Vehicle / 27-Motorcycle / 105-Motorcycling / \n", "0-Games / 2-Video game / 19-PC game / 1852-Final Fantasy XI / \n", "0-Games / 8-Football / 79-American football / \n", "10-Animal / 132-Bird / 627-Clock / \n", "695-Balloon / 2099-Balloon modelling / \n", "9-Music video / \n", "0-Games / 19-PC game / 51-Strategy video game / \n", "0-Games / 411-MapleStory / \n", "1-Vehicle / 4-Car / 625-Factory / \n", "3-Concert / \n", "121-Photography / 155-Camera / 566-Digital camera / 629-Camera lens / 1348-Panasonic Corporation / \n", "1-Vehicle / 4-Car / 27-Motorcycle / 40-Road / 64-Transport / 96-Soldier / 105-Motorcycling / \n", "916-Spa / \n", "0-Games / 181-Xbox / 386-Xbox / 429-Website / \n", "6-Animation / 15-Cartoon / 183-Manga / 1312-Dynasty Warriors / 3874-Dynasty Warriors 8 / \n", "0-Games / 5-Dance / 2985-Chunichi Dragons Doala / \n", "125-Television / 205-Newscaster / 1572-Husband / \n", "1288-Marvel vs. Capcom / 1499-Marvel vs. Capcom 3: Fate of Two Worlds / 1505-Ultimate Marvel vs. Capcom 3 / 2172-Deadpool / \n", "102-Building / 339-Architecture / 425-City / 1214-Skyscraper / \n", "12-Food / 26-Cooking / 354-Egg / 713-Egg / \n", "14-Guitar / 100-Electric guitar / \n", "47-Personal computer / 145-Tablet computer / \n", "1-Vehicle / 4-Car / 263-Light / 403-Toyota / 1194-Automotive lighting / 2512-Toyota Prius / \n", "5-Dance / \n", "232-Cake / 1322-Biscuit / \n", "308-Ship / 2056-Battleship / \n", "7-Musician / 13-Musical ensemble / 46-Choir / 99-Christmas / \n", "1-Vehicle / 4-Car / 113-House / \n", "260-Prayer / \n", "10-Animal / 17-Racing / 48-Pet / 71-Dog / 80-Horse / \n", "0-Games / 530-Cue sports / 568-Pool / 994-Cue stick / 1334-Eight-ball / \n", "0-Games / \n", "10-Animal / 132-Bird / 404-Livestock / 424-Chicken / 1136-Poultry / 1428-Rooster / \n", "0-Games / 2-Video game / 229-Halo / \n", "2-Video game / 43-Call of Duty / 96-Soldier / \n", "122-Weight training / 214-Muscle / \n", "1-Vehicle / 11-Motorsport / 17-Racing / 1025-Personal water craft / 1336-Jet Ski / \n", "1-Vehicle / 4-Car / 11-Motorsport / 17-Racing / 57-Race track / 109-Cooking show / \n", "1-Vehicle / 11-Motorsport / 167-Four-wheel drive / 2742-Mini 4WD / \n", "0-Games / 1-Vehicle / 25-Toy / 62-Train / 119-Locomotive / 507-Thomas the Tank Engine / 1279-Plarail / 1685-Toy train / \n", "3-Concert / 7-Musician / 13-Musical ensemble / 38-Orchestra / 46-Choir / \n", "0-Games / 51-Strategy video game / 373-Clash of Clans / \n", "14-Guitar / 24-String instrument / 63-Acoustic guitar / \n", "6-Animation / 116-Pokémon / 1104-Ash Ketchum / \n", "0-Games / 8-Football / 21-Stadium / 65-Kick / 81-Athlete / 1247-Penalty kick / \n", "126-The Walt Disney Company / 144-Amusement park / 576-Walt Disney World / \n", "1-Vehicle / 4-Car / 299-Volkswagen Passenger Cars / 668-Volkswagen Golf / 2147-Fiber / \n", "9-Music video / \n", "6-Animation / \n", "36-Piano / 53-Keyboard / 104-Pianist / 107-Musical keyboard / \n", "5-Dance / 9-Music video / \n", "22-Nature / 86-Plant / 137-Tree / 155-Camera / 213-Forest / \n", "0-Games / 382-Dota 2 / \n", "28-Fashion / 206-Shoe / 333-Nike; Inc. / 368-Sneakers / 737-Joker / \n", "1-Vehicle / 4-Car / 27-Motorcycle / 40-Road / \n", "83-Skateboarding / 108-Skateboard / 555-Longboard / 667-Longboarding / 1654-Peter Pan / \n", "0-Games / 59-Winter sport / 112-Ice skating / 142-Hockey / 312-Ice rink / \n", "1144-Castlevania / \n", "75-Wrestling / 85-Combat / \n", "130-Gymnastics / 1417-Balance beam / \n", "5-Dance / 49-School / \n", "0-Games / 8-Football / 21-Stadium / 65-Kick / \n", "0-Games / 35-Minecraft / 102-Building / 113-House / 339-Architecture / \n", "20-Trailer / \n", "0-Games / 8-Football / 21-Stadium / 77-Arena / \n", "5-Dance / \n", "5-Dance / 115-Ballet / \n", "0-Games / 2-Video game / 6-Animation / 185-Dragon Ball / 244-Goku / 418-Gohan / 2144-Dragon Ball: Raging Blast 2 / 2384-Dragon Ball: Raging Blast / \n", "2414-Firefighting / \n", "51-Strategy video game / 271-Tank / 409-World of Tanks / \n", "0-Games / 2-Video game / 6-Animation / 15-Cartoon / 20-Trailer / 294-Batman / 630-Superman / 803-DC Comics / 1164-Cartoon Network / 2766-Lex Luthor / \n", "23-Mobile phone / 29-Smartphone / 37-Gadget / 47-Personal computer / 72-iPhone / 101-Telephone / 145-Tablet computer / 178-Samsung Galaxy / 487-Samsung Electronics / 1015-Samsung Galaxy Note series / 1629-Samsung Galaxy Note / \n", "1-Vehicle / 4-Car / 11-Motorsport / 17-Racing / 27-Motorcycle / 158-Tire / 378-Drag racing / \n", "18-Outdoor recreation / 144-Amusement park / 321-Roller coaster / 334-Amusement ride / \n", "8-Football / \n", "0-Games / 8-Football / 21-Stadium / 125-Television / \n", "9-Music video / \n", "0-Games / 8-Football / 76-Ball / \n", "544-Table tennis / 1615-Pong / \n", "10-Animal / 42-Fishing / 91-Fish / \n", "7-Musician / 46-Choir / 78-Wedding / \n", "6-Animation / 15-Cartoon / \n", "20-Trailer / \n", "6-Animation / 15-Cartoon / \n", "185-Dragon Ball / 244-Goku / 869-Dragon Ball Z: Budokai Tenkaichi / \n", "3-Concert / \n", "6-Animation / 25-Toy / 477-Transformers / 990-Optimus Prime / \n", "6-Animation / 15-Cartoon / 3791-Magi: The Labyrinth of Magic / \n", "0-Games / 2-Video game / 19-PC game / 33-Weapon / 43-Call of Duty / 138-Call of Duty: Black Ops II / \n", "20-Trailer / \n", "2357-Vampire Knight / \n", "28-Fashion / 469-Bag / 676-Handbag / \n", "3-Concert / \n", "3-Concert / 1854-Centella / \n", "49-School / 56-Hair / \n", "1-Vehicle / 11-Motorsport / 17-Racing / \n", "0-Games / 1290-Dragon Nest / 2043-Mabinogi / \n", "15-Cartoon / 61-Art / 97-Drawing / 382-Dota 2 / 408-Sketch / 643-Hero / \n", "1-Vehicle / 4-Car / 40-Road / 70-Driving / 94-Dashcam / 120-Winter / 725-Human / 915-Town / \n", "3-Concert / 7-Musician / 13-Musical ensemble / 30-Drummer / 38-Orchestra / 44-Drums / 1240-Mambo / \n", "1-Vehicle / 4-Car / 11-Motorsport / 17-Racing / 25-Toy / 92-Radio-controlled model / 202-Radio-controlled car / \n", "1-Vehicle / 4-Car / 23-Mobile phone / 29-Smartphone / 37-Gadget / 72-iPhone / 245-BMW / \n", "0-Games / 2-Video game / 20-Trailer / 3078-Metro: Last Light / \n", "3-Concert / \n", "6-Animation / 15-Cartoon / 123-Naruto / 285-Sasuke Uchiha / \n", "3-Concert / \n", "14-Guitar / 24-String instrument / 63-Acoustic guitar / 1044-Coca-Cola / 1224-Cola / \n", "661-Stretching / 2582-Split / \n", "3-Concert / 7-Musician / \n", "115-Ballet / 980-Solo dance / \n", "6-Animation / 15-Cartoon / 1737-Dance Dance Revolution / 1911-StepMania / \n", "1-Vehicle / 4-Car / 11-Motorsport / 17-Racing / 92-Radio-controlled model / 202-Radio-controlled car / 1040-Street racing / \n", "0-Games / 739-Guild Wars / \n", "9-Music video / \n", "3-Concert / 68-Lighting / \n", "0-Games / 60-Basketball / \n", "10-Animal / 48-Pet / \n", "6-Animation / 123-Naruto / 285-Sasuke Uchiha / \n", "18-Outdoor recreation / 347-Climbing / 828-Rock climbing / \n", "10-Animal / 18-Outdoor recreation / 22-Nature / 132-Bird / 217-Hunting / 280-Wildlife / 2333-Wild turkey / 4351-Turkey / \n", "356-Star / 3096-Hiru TV / \n", "1-Vehicle / 4-Car / 11-Motorsport / 17-Racing / \n", "6-Animation / 10-Animal / 987-Sheep / \n", "0-Games / 6-Animation / \n", "208-Wood / 261-Tool / 1226-Metalworking / 2035-Forge / \n", "0-Games / 2-Video game / 72-iPhone / 189-iPad / \n", "0-Games / 1-Vehicle / 2-Video game / 4-Car / 2249-Driver / \n", "1237-Baby carriage / \n", "5-Dance / 49-School / 95-Talent show / \n", "162-Medicine / \n", "1-Vehicle / 4-Car / 11-Motorsport / 17-Racing / 57-Race track / 378-Drag racing / \n", "5-Dance / 25-Toy / 115-Ballet / 323-Toddler / \n", "0-Games / \n", "0-Games / 2-Video game / \n", "2-Video game / 34-Action-adventure game / 483-Costume / \n", "248-Hotel / \n", "434-Ninja / \n", "3-Concert / 31-Disc jockey / \n", "25-Toy / 323-Toddler / \n", "1-Vehicle / 4-Car / 11-Motorsport / 27-Motorcycle / 40-Road / 67-Cycling / 69-Bicycle / 94-Dashcam / 105-Motorcycling / 212-Highway / 608-Wheelie / \n", "248-Hotel / \n", "0-Games / 8-Football / 21-Stadium / 54-Highlight film / 65-Kick / 81-Athlete / \n", "18-Outdoor recreation / 33-Weapon / 217-Hunting / 1873-Crossbow / \n", "8-Football / 880-Trousers / \n", "5-Dance / 679-Mickey Mouse / 1234-Mouse / 1327-Minnie Mouse / \n", "0-Games / 2-Video game / 19-PC game / 33-Weapon / 43-Call of Duty / 111-PlayStation 3 / 704-Call of Duty: World at War / \n", "1-Vehicle / 18-Outdoor recreation / 22-Nature / 67-Cycling / 69-Bicycle / 197-Trail / 220-Mountain bike / 284-Mountain biking / \n", "5-Dance / 16-Performance art / 95-Talent show / 115-Ballet / \n", "20-Trailer / 325-Vampire / \n", "0-Games / 8-Football / 21-Stadium / 54-Highlight film / 79-American football / 993-Quarterback / \n", "7-Musician / 14-Guitar / \n", "359-Harry Potter (Literary Series) / \n", "3-Concert / 7-Musician / 14-Guitar / 30-Drummer / \n", "5-Dance / 16-Performance art / 95-Talent show / \n", "10-Animal / 80-Horse / 247-Stallion / 289-Dressage / \n", "373-Clash of Clans / \n", "6-Animation / 15-Cartoon / \n", "3-Concert / 5-Dance / 16-Performance art / \n", "6-Animation / 25-Toy / 140-LEGO / 940-Clay animation / \n", "0-Games / 8-Football / 21-Stadium / 54-Highlight film / 65-Kick / 73-Ball / 76-Ball / 77-Arena / 81-Athlete / \n", "10-Animal / 1568-Koi / \n", "9-Music video / \n", "0-Games / 2-Video game / 33-Weapon / 43-Call of Duty / 96-Soldier / 194-Call of Duty: Modern Warfare 2 / \n", "6-Animation / 15-Cartoon / \n", "0-Games / 60-Basketball / 77-Arena / 168-Basketball moves / 255-Slam dunk / \n", "3-Concert / \n", "116-Pokémon / 690-Pokémon X and Y / \n", "5-Dance / 16-Performance art / 115-Ballet / \n", "59-Winter sport / 177-Skiing / 789-Alpine skiing / \n", "28-Fashion / 45-Cosmetics / 195-Mascara / \n", "61-Art / 242-Paper / 2424-Greeting card / \n", "0-Games / 2-Video game / 35-Minecraft / 174-Farm / 424-Chicken / \n", "67-Cycling / 69-Bicycle / 377-BMX bike / \n", "9-Music video / \n", "3-Concert / 68-Lighting / \n", "86-Plant / 137-Tree / \n", "45-Cosmetics / 238-Nail / 300-Nail art / 304-Nail polish / 338-Manicure / \n", "159-World of Warcraft / 171-Warcraft / \n", "5-Dance / \n", "8-Football / 21-Stadium / 77-Arena / \n", "30-Drummer / 39-Drums / 44-Drums / 3066-Timbales / \n", "1013-Walking / \n", "1-Vehicle / 4-Car / 27-Motorcycle / 245-BMW / \n", "0-Games / 2-Video game / 19-PC game / 33-Weapon / 96-Soldier / 192-Battlefield / 383-Battlefield 4 / \n", "3-Concert / 5-Dance / 16-Performance art / \n", "3-Concert / 7-Musician / 13-Musical ensemble / 16-Performance art / \n", "3-Concert / 7-Musician / 30-Drummer / \n", "0-Games / 211-RuneScape / \n", "23-Mobile phone / 29-Smartphone / 72-iPhone / 189-iPad / 272-iPod / 331-iPod touch / \n", "6-Animation / 9-Music video / 781-Pachinko / \n", "25-Toy / 587-Clay / 2499-Christmas ornament / \n", "8-Football / 21-Stadium / 1488-Mosaic / \n", "9-Music video / \n", "1-Vehicle / 62-Train / 64-Transport / 103-Rail transport / 119-Locomotive / 143-Railroad car / 152-Track / 161-Train station / 224-Rapid transit / \n", "0-Games / 1-Vehicle / 4-Car / 11-Motorsport / 17-Racing / 27-Motorcycle / 294-Batman / \n", "1-Vehicle / 4-Car / 11-Motorsport / 50-Aircraft / 82-Airplane / 129-Aviation / 151-Landing / 223-Weather / 451-Cockpit / \n", "75-Wrestling / 269-Comic book / \n", "159-World of Warcraft / 171-Warcraft / 1356-Paladin / \n", "47-Personal computer / 131-Computer / 141-Microsoft Windows / 228-Laptop / 309-Computer hardware / 621-Asus / 1344-Netbook / \n", "97-Drawing / 567-Compact disc / 1010-Ring / \n", "3-Concert / 7-Musician / 13-Musical ensemble / 46-Choir / \n", "56-Hair / 106-Hairstyle / \n", "98-Festival / 144-Amusement park / 348-Fruit / \n", "56-Hair / 106-Hairstyle / 609-Tattoo / 863-Body piercing / \n", "153-Dress / \n", "93-Comedy / \n", "9-Music video / \n", "12-Food / 26-Cooking / 32-Recipe / 52-Dish / 58-Cuisine / 239-Roasting / 454-Sauce / 671-Tomato / \n", "12-Food / 26-Cooking / 32-Recipe / 52-Dish / 391-Chicken meat / 424-Chicken / 597-Soup / \n", "0-Games / 2-Video game / 43-Call of Duty / 47-Personal computer / 111-PlayStation 3 / 134-Call of Duty: Black Ops / 394-Map / \n", "3-Concert / 7-Musician / 30-Drummer / 68-Lighting / \n", "0-Games / 1138-Tower of Saviors / \n", "5-Dance / 95-Talent show / \n", "12-Food / 26-Cooking / 32-Recipe / 58-Cuisine / 1218-Japanese Cuisine / 1227-Sushi / \n", "0-Games / 8-Football / 54-Highlight film / 79-American football / \n", "0-Games / 2-Video game / 34-Action-adventure game / 1559-Spyro / \n", "6-Animation / 15-Cartoon / \n", "25-Toy / 89-Comics / 269-Comic book / 1676-Aikatsu! / \n", "23-Mobile phone / 29-Smartphone / 37-Gadget / 178-Samsung Galaxy / 1366-LG Optimus series / 3962-LG G series / 4407-LG Optimus G / \n", "0-Games / 35-Minecraft / \n", "0-Games / 75-Wrestling / 777-WWE 2K / \n", "66-Bollywood / \n", "1-Vehicle / 4-Car / 11-Motorsport / 17-Racing / 84-Snow / 262-Ice / \n", "1-Vehicle / 4-Car / 114-Engine / \n", "86-Plant / 136-Gardening / 184-Vegetable / 279-Garden / 1381-Garlic / \n", "208-Wood / \n", "2531-Dojo / \n", "0-Games / 8-Football / 65-Kick / \n", "3-Concert / 68-Lighting / \n", "36-Piano / 53-Keyboard / 4443-Jekyll & Hyde / \n", "1-Vehicle / 4-Car / 365-Fireworks / \n", "0-Games / 1-Vehicle / 25-Toy / 62-Train / 140-LEGO / 161-Train station / \n", "7-Musician / 14-Guitar / 24-String instrument / \n", "5-Dance / \n", "0-Games / \n", "126-The Walt Disney Company / 996-Princess / \n", "222-Surfing / 815-Surfboard / 1667-Standup paddleboarding / \n", "12-Food / \n", "51-Strategy video game / 831-Plants vs. Zombies / 1120-Tower defense / \n", "1-Vehicle / \n", "419-Earth / 448-Planet / \n", "0-Games / 33-Weapon / 257-Counter-Strike / \n", "10-Animal / 48-Pet / 71-Dog / 85-Combat / 1123-Domestic pig / 1215-Wild boar / \n", "9-Music video / \n", "356-Star / 419-Earth / 448-Planet / 654-Sun / 2761-Venus / \n", "0-Games / 2-Video game / 43-Call of Duty / 138-Call of Duty: Black Ops II / \n", "0-Games / 2-Video game / 257-Counter-Strike / 315-Knife / \n", "61-Art / 457-Knitting / 587-Clay / \n", "88-Machine / \n", "5-Dance / 16-Performance art / 66-Bollywood / \n", "1-Vehicle / 4-Car / 683-Renault / \n", "131-Computer / 141-Microsoft Windows / \n", "6-Animation / 15-Cartoon / 185-Dragon Ball / 244-Goku / \n", "0-Games / 2-Video game / 19-PC game / 33-Weapon / 43-Call of Duty / 96-Soldier / 127-Xbox 360 / 216-Call of Duty: Modern Warfare 3 / 2491-Spec Ops: The Line / \n", "1-Vehicle / 4-Car / 49-School / 62-Train / 64-Transport / 286-Bus / 2199-School bus / \n", "0-Games / 2-Video game / 6-Animation / 34-Action-adventure game / 243-Sonic the Hedgehog / 4019-Sonic the Hedgehog CD / \n", "6-Animation / 123-Naruto / 622-Sakura Haruno / 4333-Kiba Inuzuka / \n", "0-Games / 8-Football / 21-Stadium / 65-Kick / 73-Ball / 76-Ball / \n", "9-Music video / \n", "23-Mobile phone / 29-Smartphone / 37-Gadget / 178-Samsung Galaxy / 899-Samsung Galaxy S III / 1078-Installation art / \n", "5-Dance / \n", "3-Concert / \n", "4646-Tweety / \n", "1-Vehicle / 4-Car / 11-Motorsport / 17-Racing / 41-Sports car / 135-Drifting / 245-BMW / \n", "3-Concert / 7-Musician / 30-Drummer / \n", "144-Amusement park / 1731-Haunted house / 4274-Knott's Berry Farm / \n", "12-Food / 26-Cooking / \n", "3-Concert / 7-Musician / 14-Guitar / 24-String instrument / 63-Acoustic guitar / 670-Banjo / \n", "3-Concert / 7-Musician / 14-Guitar / \n", "10-Animal / 48-Pet / 132-Bird / 1799-Domestic canary / \n", "0-Games / 2-Video game / 5-Dance / 199-Wii / 2123-Just Dance / 3902-Just Dance 3 / \n", "155-Camera / 629-Camera lens / 4195-Binoculars / \n", "59-Winter sport / 177-Skiing / \n", "122-Weight training / 214-Muscle / 1162-Dietary supplement / \n", "246-Furniture / \n", "3-Concert / 16-Performance art / \n", "3-Concert / \n", "6-Animation / 15-Cartoon / 750-Sailor Moon / \n", "14-Guitar / 24-String instrument / 63-Acoustic guitar / \n", "3-Concert / 13-Musical ensemble / 16-Performance art / \n", "227-Water / 640-Bottle / \n", "2-Video game / 6-Animation / 362-PlayStation Portable / 1492-Persona / 2544-Megami Tensei / 2896-Shin Megami Tensei: Persona 3 / \n", "102-Building / 248-Hotel / 369-Resort / \n", "1-Vehicle / 4-Car / 336-Scooter / \n", "1-Vehicle / 4-Car / 1074-Hyundai / \n", "0-Games / 2-Video game / 43-Call of Duty / 138-Call of Duty: Black Ops II / 3188-Camouflage / \n", "0-Games / 2-Video game / 43-Call of Duty / \n", "12-Food / 32-Recipe / 109-Cooking show / 534-Milk / \n", "1-Vehicle / 67-Cycling / 69-Bicycle / 570-Road bicycle racing / 645-Road bicycle / \n", "3-Concert / 7-Musician / 13-Musical ensemble / 38-Orchestra / 46-Choir / \n", "1-Vehicle / 11-Motorsport / 17-Racing / 27-Motorcycle / 927-Supermoto / \n", "93-Comedy / 125-Television / \n", "28-Fashion / 390-Model / \n", "0-Games / 2-Video game / 19-PC game / 159-World of Warcraft / 171-Warcraft / \n", "1-Vehicle / 11-Motorsport / 27-Motorcycle / 99-Christmas / 105-Motorcycling / \n", "7-Musician / 14-Guitar / 30-Drummer / 39-Drums / 44-Drums / 146-Snare drum / \n", "1-Vehicle / 4-Car / 92-Radio-controlled model / 114-Engine / 154-Wheel / 202-Radio-controlled car / \n", "102-Building / 353-Wall / \n", "0-Games / 25-Toy / 162-Medicine / \n", "10-Animal / 80-Horse / \n", "9-Music video / \n", "0-Games / 35-Minecraft / \n", "1-Vehicle / 124-Tractor / 156-Agriculture / 174-Farm / 715-Farming Simulator / \n", "6-Animation / 460-Yu-Gi-Oh! Trading Card Game / \n", "20-Trailer / \n", "8-Football / 54-Highlight film / 79-American football / \n", "56-Hair / 61-Art / 97-Drawing / 148-Painting / 162-Medicine / 282-Skin / \n", "0-Games / 2-Video game / 351-Jumping / \n", "12-Food / 26-Cooking / 32-Recipe / 236-Baking / 485-Dough / 600-Flour / 662-Pizza / \n", "10-Animal / 48-Pet / 166-Cat / 346-Kitten / \n", "78-Wedding / \n", "3-Concert / 7-Musician / 13-Musical ensemble / 38-Orchestra / 133-Nightclub / \n", "14-Guitar / 24-String instrument / 63-Acoustic guitar / \n", "126-The Walt Disney Company / \n", "14-Guitar / 24-String instrument / 63-Acoustic guitar / 3812-Dobro / 4377-Resonator guitar / \n", "9-Music video / 1623-Shaolin Kung Fu / \n", "1-Vehicle / 50-Aircraft / 92-Radio-controlled model / 149-Model aircraft / 157-Jet aircraft / 160-Radio-controlled aircraft / 266-Helicopter / 392-Radio-controlled helicopter / \n", "1-Vehicle / 4-Car / 70-Driving / 155-Camera / \n", "3-Concert / 5-Dance / 16-Performance art / \n", "0-Games / 2-Video game / 43-Call of Duty / 96-Soldier / 134-Call of Duty: Black Ops / 138-Call of Duty: Black Ops II / \n", "12-Food / 170-Home improvement / 492-Floor / 666-Concrete / 934-Flooring / 1314-Garage / 2093-Coating / \n", "6-Animation / 25-Toy / 319-Robot / 477-Transformers / 990-Optimus Prime / 2186-Megatron / \n", "1-Vehicle / 62-Train / 64-Transport / 103-Rail transport / 119-Locomotive / 143-Railroad car / 152-Track / 161-Train station / 224-Rapid transit / 1978-New York City Subway / \n", "0-Games / 8-Football / 21-Stadium / 54-Highlight film / 65-Kick / 73-Ball / 77-Arena / \n", "230-Brass instrument / 296-Saxophone / \n", "12-Food / 26-Cooking / 32-Recipe / 52-Dish / 58-Cuisine / 184-Vegetable / 239-Roasting / 274-Meat / 391-Chicken meat / 2178-Chinese food / \n", "12-Food / 26-Cooking / 32-Recipe / 179-Kitchen / 315-Knife / 2017-Sharpening / 3880-Knife sharpening / \n", "10-Animal / 80-Horse / 351-Jumping / \n", "162-Medicine / 925-Brain / 1925-Skull / \n", "122-Weight training / 1162-Dietary supplement / \n", "28-Fashion / \n", "3-Concert / 7-Musician / 13-Musical ensemble / 16-Performance art / 38-Orchestra / \n", "0-Games / 1-Vehicle / 25-Toy / \n", "28-Fashion / \n", "3-Concert / 7-Musician / 13-Musical ensemble / 30-Drummer / 39-Drums / 49-School / \n", "0-Games / 8-Football / \n", "93-Comedy / \n", "96-Soldier / 205-Newscaster / \n", "56-Hair / 106-Hairstyle / \n", "3-Concert / 7-Musician / 13-Musical ensemble / 24-String instrument / 30-Drummer / 38-Orchestra / 39-Drums / \n", "2864-Monogatari Series / \n", "49-School / \n", "12-Food / 26-Cooking / 32-Recipe / 1174-Smoothie / \n", "6-Animation / 2808-Printed circuit board / \n", "68-Lighting / 99-Christmas / 263-Light / 729-Christmas decoration / 1098-Christmas lights / \n", "83-Skateboarding / 108-Skateboard / 555-Longboard / 667-Longboarding / \n", "1-Vehicle / 4-Car / 11-Motorsport / 114-Engine / 349-Chevrolet / 1752-Chevrolet SS / 2061-Chevrolet Chevelle / \n", "9-Music video / \n", "1944-Vatican City / \n", "28-Fashion / 2319-Zara / \n", "0-Games / 8-Football / 21-Stadium / 65-Kick / 76-Ball / 710-Handball / \n", "66-Bollywood / \n", "130-Gymnastics / \n", "30-Drummer / 39-Drums / 44-Drums / 128-Cymbal / 146-Snare drum / 1458-Electronic drum / 1953-Roland V-Drums / \n", "23-Mobile phone / 29-Smartphone / 47-Personal computer / 145-Tablet computer / 178-Samsung Galaxy / 487-Samsung Electronics / 1287-Samsung Galaxy Tab series / \n", "1-Vehicle / 213-Forest / 259-Fire / 371-Firefighter / 796-Fire engine / \n", "0-Games / 2-Video game / 8-Football / 90-Sports game / 180-FIFA 15 / 3235-Monopoly / \n", "273-News program / \n", "5-Dance / 49-School / \n", "22-Nature / \n", "6-Animation / 15-Cartoon / 126-The Walt Disney Company / 708-Mermaid / 1219-Ariel / \n", "1-Vehicle / 4-Car / 790-Nissan GT-R / \n", "6-Animation / 995-Attack on Titan / \n", "61-Art / 97-Drawing / \n", "3-Concert / 7-Musician / 13-Musical ensemble / 46-Choir / \n", "3-Concert / 7-Musician / 30-Drummer / \n", "237-Final Fantasy / \n", "0-Games / 2-Video game / 264-Star Wars / \n", "1-Vehicle / 18-Outdoor recreation / 59-Winter sport / 84-Snow / 120-Winter / 177-Skiing / 281-Snowboarding / \n", "9-Music video / 20-Trailer / \n", "0-Games / 60-Basketball / 81-Athlete / \n", "23-Mobile phone / 29-Smartphone / 56-Hair / 72-iPhone / 272-iPod / 331-iPod touch / \n", "227-Water / 1081-Lemon / 1999-Eggplant / \n", "28-Fashion / \n", "12-Food / 2706-Starbucks / \n", "8-Football / \n", "9-Music video / \n", "242-Paper / 370-Origami / 1023-Rose / \n", "6-Animation / \n", "3-Concert / 16-Performance art / \n", "10-Animal / 2534-Scorpion / \n", "0-Games / 1-Vehicle / 2-Video game / 4-Car / 11-Motorsport / 17-Racing / 34-Action-adventure game / 135-Drifting / 201-Grand Theft Auto: San Andreas / 297-Grand Theft Auto IV / \n", "7-Musician / 14-Guitar / 24-String instrument / 63-Acoustic guitar / \n", "66-Bollywood / \n", "1182-Softball / \n", "12-Food / 720-Wine / \n", "56-Hair / 97-Drawing / 106-Hairstyle / 1827-Hair straightening / \n", "8-Football / 21-Stadium / 595-Champion / \n", "0-Games / 2-Video game / 51-Strategy video game / 165-League of Legends / \n", "6-Animation / 15-Cartoon / 116-Pokémon / 1104-Ash Ketchum / 3347-Misty / \n", "0-Games / 326-Metin2 / 1552-Video lesson / \n", "0-Games / 8-Football / \n", "28-Fashion / 56-Hair / \n", "8-Football / \n", "0-Games / 25-Toy / 140-LEGO / 852-Lego Star Wars / 884-Lego minifigure / 1478-Lego Ninjago / \n", "0-Games / 2-Video game / 1492-Persona / 2139-Dungeon crawl / \n", "1-Vehicle / 4-Car / 3219-Daihatsu / \n", "36-Piano / 53-Keyboard / 107-Musical keyboard / \n", "3-Concert / 68-Lighting / 263-Light / \n", "46-Choir / \n", "6-Animation / 31-Disc jockey / \n", "242-Paper / 370-Origami / 631-Cube / \n", "8-Football / 21-Stadium / \n", "28-Fashion / \n", "0-Games / 60-Basketball / 168-Basketball moves / \n", "23-Mobile phone / 29-Smartphone / 72-iPhone / 932-Microsoft Lumia / \n", "20-Trailer / \n", "1-Vehicle / 50-Aircraft / 451-Cockpit / 856-Airbus A320 family / \n", "3-Concert / 16-Performance art / 133-Nightclub / \n", "0-Games / 2-Video game / 35-Minecraft / \n", "3-Concert / 7-Musician / \n", "6-Animation / \n", "1-Vehicle / 4-Car / 1661-Volkswagen Gol / \n", "0-Games / 599-Dofus / \n", "56-Hair / 106-Hairstyle / 3584-Hair roller / \n", "7-Musician / 13-Musical ensemble / 21-Stadium / 79-American football / 203-Marching band / 209-University / \n", "0-Games / 2-Video game / 6-Animation / 816-Gundam / 1564-Mobile Suit Gundam: Extreme Vs. / \n", "5-Dance / \n", "67-Cycling / 69-Bicycle / 377-BMX bike / \n", "7-Musician / 14-Guitar / 24-String instrument / 63-Acoustic guitar / 1281-Couch / \n", "28-Fashion / \n", "1-Vehicle / 62-Train / 64-Transport / 161-Train station / 224-Rapid transit / \n", "28-Fashion / 45-Cosmetics / \n", "322-Mixtape / \n", "1-Vehicle / 50-Aircraft / 139-Airport / 151-Landing / 157-Jet aircraft / 254-Runway / \n", "9-Music video / 67-Cycling / 69-Bicycle / \n", "60-Basketball / \n", "0-Games / 2-Video game / 33-Weapon / \n", "0-Games / 2-Video game / 33-Weapon / 257-Counter-Strike / \n", "0-Games / 2-Video game / 756-Tibia / \n", "1-Vehicle / 11-Motorsport / 87-Boat / 327-Mud / \n", "3-Concert / 1562-Viola caipira / 2649-Nativity scene / \n", "36-Piano / 53-Keyboard / 107-Musical keyboard / \n", "6-Animation / 10-Animal / 15-Cartoon / \n", "93-Comedy / \n", "1-Vehicle / 4-Car / 517-Audi / \n", "0-Games / 19-PC game / 837-Roblox / \n", "1-Vehicle / 11-Motorsport / 124-Tractor / 156-Agriculture / 252-Tractor pulling / \n", "3-Concert / 7-Musician / \n", "1-Vehicle / 11-Motorsport / 17-Racing / 27-Motorcycle / 105-Motorcycling / 172-Motocross / \n", "47-Personal computer / 228-Laptop / 621-Asus / \n", "6-Animation / 15-Cartoon / \n", "7-Musician / 14-Guitar / 24-String instrument / 100-Electric guitar / \n", "61-Art / 97-Drawing / 479-Graffiti / \n", "12-Food / 26-Cooking / 32-Recipe / 52-Dish / 58-Cuisine / 184-Vegetable / 210-Cookware and bakeware / 239-Roasting / 274-Meat / 688-Beef / 1509-Stew / 3008-Slow cooker / \n", "1-Vehicle / 11-Motorsport / 27-Motorcycle / 1520-Minibike / \n", "5-Dance / \n", "6-Animation / 15-Cartoon / 183-Manga / 4231-Slayers / \n", "0-Games / 2-Video game / 19-PC game / 43-Call of Duty / 431-Call of Duty: Advanced Warfare / 2752-God Mode / \n", "31-Disc jockey / \n", "1604-Satan / \n", "0-Games / 10-Animal / 80-Horse / 310-The Sims / 328-Simulation video game / 564-The Sims 3 / 4083-The Sims 3: Pets / \n", "5-Dance / \n", "1-Vehicle / 11-Motorsport / 17-Racing / 27-Motorcycle / 57-Race track / \n", "1-Vehicle / 74-Truck / 88-Machine / 249-Heavy equipment / 433-Construction / \n", "0-Games / 34-Action-adventure game / \n", "0-Games / 2-Video game / 6-Animation / 15-Cartoon / 89-Comics / 185-Dragon Ball / \n", "5-Dance / \n", "18-Outdoor recreation / 137-Tree / 217-Hunting / 315-Knife / \n", "8-Football / \n", "0-Games / 23-Mobile phone / 72-iPhone / \n", "12-Food / 442-Bread / 1943-Bakery / \n", "43-Call of Duty / 704-Call of Duty: World at War / \n", "7-Musician / 14-Guitar / 24-String instrument / 63-Acoustic guitar / 545-Quartet (ensemble) / \n", "0-Games / 2-Video game / 8-Football / 90-Sports game / 480-FIFA 13 / \n", "0-Games / 2-Video game / 6-Animation / 15-Cartoon / 241-Fighting game / 486-M.U.G.E.N / \n", "0-Games / 6-Animation / 460-Yu-Gi-Oh! Trading Card Game / 494-Card game / \n", "12-Food / 727-Curry / \n", "6-Animation / 15-Cartoon / \n", "83-Skateboarding / 108-Skateboard / \n", "1-Vehicle / 4-Car / 167-Four-wheel drive / 198-Off-road vehicle / 277-Sport utility vehicle / 360-Jeep / \n", "2152-Selfie / \n", "0-Games / 1-Vehicle / 2-Video game / 4-Car / 34-Action-adventure game / 297-Grand Theft Auto IV / \n", "1-Vehicle / 4-Car / 40-Road / 70-Driving / 94-Dashcam / 212-Highway / 575-Traffic / \n", "2-Video game / 584-Super Mario Bros. / 831-Plants vs. Zombies / \n", "14-Guitar / 24-String instrument / 100-Electric guitar / 606-String / 738-Guitar amplifier / \n", "0-Games / 8-Football / 21-Stadium / 54-Highlight film / 77-Arena / 4270-Old Trafford / \n", "204-Gym / \n", "3-Concert / 5-Dance / 16-Performance art / 38-Orchestra / \n", "8-Football / 767-Goalkeeper / \n", "1-Vehicle / 4-Car / \n", "5-Dance / \n", "61-Art / 97-Drawing / 609-Tattoo / 1834-Mehndi / \n", "5-Dance / 8-Football / \n", "66-Bollywood / 78-Wedding / 2429-India TV / \n", "282-Skin / 3579-Mona Lisa / \n", "1-Vehicle / 4-Car / 11-Motorsport / 17-Racing / 27-Motorcycle / 40-Road / 105-Motorcycling / \n", "3-Concert / 93-Comedy / 355-Comedian / \n", "14-Guitar / 63-Acoustic guitar / \n", "5-Dance / 99-Christmas / 729-Christmas decoration / 964-Christmas tree / \n", "3-Concert / 7-Musician / 24-String instrument / \n", "1-Vehicle / 4-Car / 332-Mercedes-Benz / 3826-Mercedes-Benz A-Class / 4364-Mercedes-Benz CLA-Class / \n", "3-Concert / \n", "2-Video game / \n", "1-Vehicle / 4-Car / 27-Motorcycle / 245-BMW / \n", "9-Music video / \n", "0-Games / 412-Monster / 3234-Kixeye; Inc. / \n", "3-Concert / 5-Dance / 16-Performance art / \n", "10-Animal / 86-Plant / 132-Bird / 571-Bear / 1639-Eagle / \n", "0-Games / 8-Football / 21-Stadium / 65-Kick / 73-Ball / 76-Ball / \n", "7-Musician / 12-Food / 14-Guitar / 389-Restaurant / 720-Wine / 3829-Champagne / \n", "1-Vehicle / 4-Car / 20-Trailer / \n", "25-Toy / 264-Star Wars / \n", "3-Concert / 7-Musician / 551-British Broadcasting Corporation / \n", "12-Food / 26-Cooking / 1256-Smoking / \n", "113-House / 225-Lake / \n", "0-Games / 2-Video game / 19-PC game / 34-Action-adventure game / 43-Call of Duty / 194-Call of Duty: Modern Warfare 2 / 297-Grand Theft Auto IV / 1923-Niko Bellic / \n", "14-Guitar / \n", "1-Vehicle / 11-Motorsport / 27-Motorcycle / 105-Motorcycling / 198-Off-road vehicle / \n", "444-Family / \n", "0-Games / 19-PC game / \n", "18-Outdoor recreation / 22-Nature / 67-Cycling / 69-Bicycle / 137-Tree / 197-Trail / 213-Forest / 220-Mountain bike / 284-Mountain biking / 2092-Coal / \n", "23-Mobile phone / 29-Smartphone / 72-iPhone / 189-iPad / 508-iPhone 5 / \n", "0-Games / 3643-Starbound / \n", "113-House / 170-Home improvement / 208-Wood / 276-Room / 488-Door / \n", "1-Vehicle / 4-Car / 11-Motorsport / 17-Racing / 787-Monster truck / \n", "0-Games / 2-Video game / 15-Cartoon / 3747-Wario / \n", "45-Cosmetics / 173-Eye shadow / 193-Eye liner / 195-Mascara / 196-Eye / 218-Eyelash / 235-Lipstick / 291-Rouge / 436-Concealer / \n", "3-Concert / 7-Musician / 13-Musical ensemble / 38-Orchestra / 203-Marching band / 820-Bagpipes / \n", "0-Games / 1-Vehicle / 2-Video game / 4-Car / 11-Motorsport / 17-Racing / 41-Sports car / 57-Race track / 169-Rallying / 581-PlayStation / 3500-Colin McRae: Dirt / \n", "1-Vehicle / 766-President of the United States / \n", "1-Vehicle / 59-Winter sport / 1829-Sledding / 4641-Toboggan / \n", "122-Weight training / 214-Muscle / \n", "8-Football / \n", "8-Football / \n", "468-Biology / 1544-Laboratory / \n", "0-Games / 60-Basketball / 255-Slam dunk / \n", "13-Musical ensemble / \n", "44-Drums / 98-Festival / 657-Autumn / \n", "10-Animal / 208-Wood / 512-Printing / 2840-Engraving / 2967-Printmaking / \n", "130-Gymnastics / 163-Running / 1335-Trampolining / \n", "93-Comedy / \n", "23-Mobile phone / 29-Smartphone / 72-iPhone / 148-Painting / 272-iPod / 331-iPod touch / \n", "6-Animation / 15-Cartoon / \n", "1-Vehicle / 4-Car / 64-Transport / 74-Truck / 158-Tire / 249-Heavy equipment / 822-Crane / \n", "3-Concert / 7-Musician / 13-Musical ensemble / 24-String instrument / 38-Orchestra / \n", "3-Concert / 16-Performance art / 3988-Talk box / \n", "3129-Sauna / \n", "79-American football / \n", "2006-BlazBlue / 2936-BlazBlue: Calamity Trigger / \n", "0-Games / 2-Video game / 19-PC game / 739-Guild Wars / 839-Guild Wars 2 / \n", "5-Dance / 1549-Mime artist / \n", "6-Animation / 15-Cartoon / \n", "1-Vehicle / 4-Car / 167-Four-wheel drive / 1540-Station Wagon / \n", "0-Games / 1-Vehicle / 2-Video game / 4-Car / 118-Grand Theft Auto V / \n", "0-Games / 2-Video game / 19-PC game / 1196-Dragon Age / 1735-Dragon Age: Inquisition / 2700-Dragon Age II / \n", "1-Vehicle / 12-Food / 88-Machine / 666-Concrete / \n", "10-Animal / 22-Nature / 137-Tree / 280-Wildlife / 846-Squirrel / 3104-Tree squirrel / \n", "0-Games / 2-Video game / 55-Video game console / 116-Pokémon / 292-Handheld game console / 1153-Pokémon HeartGold and SoulSilver / \n", "5-Dance / 133-Nightclub / 613-Bar / 1065-Bartender / \n", "0-Games / 2-Video game / \n", "1669-Vodka / 3505-Wellington boot / \n", "12-Food / 26-Cooking / 32-Recipe / 236-Baking / 442-Bread / \n", "110-Album / \n", "0-Games / 1-Vehicle / 2-Video game / 4-Car / 34-Action-adventure game / 118-Grand Theft Auto V / 1790-Gold medal / \n", "810-Village / \n", "75-Wrestling / \n", "9-Music video / 237-Final Fantasy / 2691-Tifa Lockhart / \n", "3-Concert / 7-Musician / 46-Choir / \n", "5-Dance / 1110-Cell / \n", "78-Wedding / \n", "12-Food / 720-Wine / \n", "222-Surfing / 2415-Big wave surfing / \n", "10-Animal / 48-Pet / 71-Dog / 166-Cat / \n", "0-Games / 2-Video game / 413-PlayStation 2 / 1492-Persona / \n", "3579-Mona Lisa / \n", "5-Dance / 7-Musician / 13-Musical ensemble / 38-Orchestra / 115-Ballet / 200-Doll / \n", "6-Animation / 15-Cartoon / 979-Star Trek / \n", "20-Trailer / 4603-Frankenstein / \n", "130-Gymnastics / \n", "0-Games / 792-DarkOrbit / \n", "12-Food / 28-Fashion / 56-Hair / 106-Hairstyle / \n", "0-Games / 1-Vehicle / 124-Tractor / 174-Farm / 252-Tractor pulling / 715-Farming Simulator / \n", "0-Games / 1816-Fire Emblem / \n", "45-Cosmetics / 173-Eye shadow / 193-Eye liner / 195-Mascara / \n", "1-Vehicle / 4-Car / 11-Motorsport / 135-Drifting / \n", "5-Dance / 18-Outdoor recreation / 287-Human swimming / 329-Swimming pool / 1711-Synchronized swimming / \n", "6-Animation / 15-Cartoon / 183-Manga / 750-Sailor Moon / 1342-Sailor Moon / \n", "0-Games / 756-Tibia / \n", "1-Vehicle / 4-Car / 11-Motorsport / 17-Racing / 40-Road / 41-Sports car / 449-Honda / 482-Asphalt / 552-Honda Civic / 3697-Honda K engine / \n", "0-Games / 59-Winter sport / 112-Ice skating / 142-Hockey / 262-Ice / 312-Ice rink / 1973-Hockey puck / \n", "0-Games / 8-Football / 73-Ball / 90-Sports game / 180-FIFA 15 / \n", "1-Vehicle / 62-Train / 64-Transport / 103-Rail transport / 119-Locomotive / 143-Railroad car / 152-Track / 161-Train station / 224-Rapid transit / \n", "66-Bollywood / \n", "0-Games / 2-Video game / 19-PC game / 33-Weapon / 43-Call of Duty / 96-Soldier / 138-Call of Duty: Black Ops II / 431-Call of Duty: Advanced Warfare / \n", "0-Games / 2-Video game / 192-Battlefield / 374-Battlefield 3 / \n", "47-Personal computer / 145-Tablet computer / 357-DVD / \n", "12-Food / 88-Machine / 512-Printing / 625-Factory / \n", "0-Games / 19-PC game / 33-Weapon / 43-Call of Duty / \n", "1-Vehicle / 11-Motorsport / 124-Tractor / 156-Agriculture / 249-Heavy equipment / 252-Tractor pulling / 363-Soil / 459-Plough / \n", "2038-Credit card / \n", "3-Concert / 5-Dance / 16-Performance art / \n", "3-Concert / \n", "5-Dance / 79-American football / 191-Cheerleading / 203-Marching band / \n", "1-Vehicle / 27-Motorcycle / \n", "1-Vehicle / 4-Car / 68-Lighting / 2580-Capacitor / \n", "648-Clown / \n", "0-Games / 1-Vehicle / 2-Video game / 4-Car / 34-Action-adventure game / 201-Grand Theft Auto: San Andreas / 2402-Grand Theft Auto / \n", "10-Animal / 217-Hunting / 280-Wildlife / 636-Rabbit / 3022-Hare / \n", "1-Vehicle / 4-Car / 11-Motorsport / 17-Racing / \n", "527-Battery / \n", "0-Games / 2-Video game / 19-PC game / 51-Strategy video game / 382-Dota 2 / 549-Defense of the Ancients / \n", "0-Games / 8-Football / 21-Stadium / 65-Kick / 73-Ball / 76-Ball / \n", "3-Concert / 7-Musician / 13-Musical ensemble / 14-Guitar / 30-Drummer / \n", "12-Food / 59-Winter sport / 112-Ice skating / 142-Hockey / 4454-Hockey stick / \n", "9-Music video / \n", "978-Love song / \n", "6-Animation / 88-Machine / 319-Robot / 619-Manufacturing / \n", "10-Animal / 144-Amusement park / 537-Zoo / 1135-Dolphin / \n", "1-Vehicle / 4-Car / 40-Road / 64-Transport / 94-Dashcam / 212-Highway / \n", "205-Newscaster / \n", "7-Musician / 13-Musical ensemble / 230-Brass instrument / \n", "12-Food / 539-Home appliance / 1439-Refrigerator / \n", "18-Outdoor recreation / \n", "1-Vehicle / 4-Car / 84-Snow / 120-Winter / \n", "3-Concert / 13-Musical ensemble / 253-Carnival / \n", "9-Music video / 61-Art / \n", "3-Concert / 7-Musician / 219-Accordion / 1994-Garmon / \n", "22-Nature / \n", "363-Soil / \n", "1-Vehicle / 27-Motorcycle / \n", "1-Vehicle / 11-Motorsport / 27-Motorcycle / 172-Motocross / \n", "6-Animation / 102-Building / 113-House / 170-Home improvement / 339-Architecture / 416-Apartment / \n", "12-Food / \n", "0-Games / 2-Video game / 33-Weapon / 43-Call of Duty / 111-PlayStation 3 / 704-Call of Duty: World at War / \n", "3-Concert / 7-Musician / 14-Guitar / \n", "20-Trailer / \n", "4349-Neptune / \n", "42-Fishing / \n", "28-Fashion / 45-Cosmetics / 173-Eye shadow / 193-Eye liner / 195-Mascara / 235-Lipstick / \n", "1-Vehicle / 4-Car / 72-iPhone / 212-Highway / 675-iPhone 4S / \n", "36-Piano / 53-Keyboard / 104-Pianist / 107-Musical keyboard / \n", "432-Kingdom Hearts / 3407-Kingdom Hearts III / \n", "0-Games / 1-Vehicle / 2-Video game / 4-Car / 34-Action-adventure game / 118-Grand Theft Auto V / \n", "83-Skateboarding / 108-Skateboard / 555-Longboard / 667-Longboarding / \n", "5-Dance / 49-School / 95-Talent show / \n", "3-Concert / 8-Football / \n", "1-Vehicle / 4-Car / 124-Tractor / 234-Ford / \n", "88-Machine / 319-Robot / 521-Metal / 619-Manufacturing / 625-Factory / 938-Steel / 2745-Beam / \n", "28-Fashion / 45-Cosmetics / \n", "3-Concert / 144-Amusement park / \n", "0-Games / 2-Video game / 19-PC game / 734-DayZ / 1216-DayZ / \n", "12-Food / 26-Cooking / 32-Recipe / 274-Meat / 862-Pork / 1256-Smoking / 1557-Bacon / \n", "23-Mobile phone / 29-Smartphone / 37-Gadget / 72-iPhone / 189-iPad / 272-iPod / 331-iPod touch / \n", "0-Games / 159-World of Warcraft / 171-Warcraft / \n", "121-Photography / \n", "9-Music video / \n", "230-Brass instrument / 296-Saxophone / \n", "209-University / 913-President / \n", "593-Moon / 654-Sun / \n", "0-Games / 2-Video game / 35-Minecraft / 394-Map / \n", "5-Dance / \n", "5-Dance / 16-Performance art / \n", "355-Comedian / \n", "28-Fashion / 206-Shoe / \n", "9-Music video / 1866-Katara / 3862-Toph Beifong / \n", "5-Dance / 10-Animal / 80-Horse / \n", "3-Concert / 7-Musician / 13-Musical ensemble / 30-Drummer / 545-Quartet (ensemble) / \n", "278-GoPro / 1358-Canyon / \n", "20-Trailer / \n", "0-Games / 19-PC game / 51-Strategy video game / 756-Tibia / \n", "9-Music video / 310-The Sims / 543-The Sims 2 / \n", "3-Concert / 7-Musician / 14-Guitar / 30-Drummer / 39-Drums / 44-Drums / \n", "78-Wedding / 190-Bride / 293-Slide show / 421-Gown / 2399-Collage / \n", "5-Dance / 16-Performance art / 1039-Tiger / \n", "0-Games / \n", "401-Cello / \n", "20-Trailer / \n", "0-Games / 2-Video game / 264-Star Wars / 2117-Star Wars Battlefront / 2340-Star Wars: Battlefront II / 2891-Star Wars: Battlefront / \n", "1-Vehicle / 50-Aircraft / 92-Radio-controlled model / 149-Model aircraft / 160-Radio-controlled aircraft / 393-Unmanned aerial vehicle / 604-Quadcopter / \n", "0-Games / 8-Football / \n", "12-Food / 26-Cooking / 32-Recipe / 52-Dish / 58-Cuisine / 391-Chicken meat / \n", "28-Fashion / 1038-Hijab / 1096-Scarf / 2448-Shawl / \n", "60-Basketball / 255-Slam dunk / \n", "0-Games / 2-Video game / 19-PC game / 464-First-person Shooter / \n", "275-Diving / 381-Underwater diving / 407-Scuba diving / \n", "35-Minecraft / \n", "6-Animation / 185-Dragon Ball / 244-Goku / 876-Freeza / \n", "0-Games / 2-Video game / 19-PC game / 330-Money / 1027-Terraria / \n", "1-Vehicle / 50-Aircraft / 82-Airplane / 129-Aviation / 139-Airport / 151-Landing / 175-Airliner / 188-Airline / 254-Runway / 451-Cockpit / \n", "8-Football / \n", "23-Mobile phone / 29-Smartphone / 37-Gadget / \n", "2-Video game / 55-Video game console / 314-Xbox One / \n", "258-Aquarium / 288-Underwater / 601-Shark / \n", "10-Animal / 91-Fish / 258-Aquarium / 1368-Gallon / \n", "0-Games / 664-CrossFire / \n", "0-Games / 2-Video game / 90-Sports game / 561-Madden NFL / \n", "567-Compact disc / \n", "0-Games / 2-Video game / 19-PC game / 43-Call of Duty / 134-Call of Duty: Black Ops / \n", "0-Games / 1-Vehicle / 2-Video game / 4-Car / 2999-Sleeping Dogs / \n", "13-Musical ensemble / 38-Orchestra / \n", "59-Winter sport / 112-Ice skating / 115-Ballet / 262-Ice / 316-Figure skating / 632-Ice dancing / 833-Ice skate / \n", "7-Musician / 9-Music video / \n", "0-Games / 2-Video game / 51-Strategy video game / 165-League of Legends / \n", "10-Animal / 48-Pet / 132-Bird / \n", "3-Concert / 14-Guitar / 24-String instrument / 150-Violin / \n", "707-Roof / 794-Bridge / \n", "1-Vehicle / 4-Car / 11-Motorsport / 17-Racing / 57-Race track / 651-Stock car racing / 728-Dirt track racing / \n", "5-Dance / 98-Festival / \n", "8-Football / 21-Stadium / \n", "9-Music video / \n", "3-Concert / 7-Musician / 13-Musical ensemble / 38-Orchestra / 46-Choir / \n", "1-Vehicle / 4-Car / 11-Motorsport / 17-Racing / 40-Road / 57-Race track / 1400-rFactor / \n", "1-Vehicle / 4-Car / 11-Motorsport / 17-Racing / 40-Road / 57-Race track / 135-Drifting / 482-Asphalt / 3636-Toyota JZ engine / \n", "7-Musician / 14-Guitar / \n", "1-Vehicle / 4-Car / 234-Ford / \n", "458-Gold / \n", "0-Games / 2-Video game / 6-Animation / 19-PC game / 460-Yu-Gi-Oh! Trading Card Game / 4280-Jaden Yuki / \n", "14-Guitar / \n", "61-Art / 155-Camera / 479-Graffiti / 1385-Street art / 1846-Mural / 2898-Canon EF lens mount / \n", "6-Animation / 15-Cartoon / 547-Bleach / \n", "5-Dance / 95-Talent show / \n", "264-Star Wars / 717-Darth Vader / 4117-Leia Organa / \n", "0-Games / 2-Video game / 1566-Doom / \n", "0-Games / 2-Video game / 35-Minecraft / \n", "5-Dance / \n", "0-Games / 2-Video game / 116-Pokémon / 186-Pokémon / \n", "46-Choir / \n", "18-Outdoor recreation / 147-River / \n", "8-Football / 333-Nike; Inc. / 757-Boot / \n", "0-Games / 19-PC game / 35-Minecraft / \n", "76-Ball / \n", "3-Concert / 7-Musician / 13-Musical ensemble / 30-Drummer / 38-Orchestra / \n", "0-Games / 2-Video game / 6-Animation / 2431-Metal Slug / 2597-Sprite / 3217-Neo Geo / \n", "5-Dance / 204-Gym / \n", "0-Games / 2-Video game / 19-PC game / 33-Weapon / 664-CrossFire / \n", "1-Vehicle / 11-Motorsport / 62-Train / 64-Transport / 103-Rail transport / 119-Locomotive / 143-Railroad car / 161-Train station / \n", "0-Games / 8-Football / \n", "6-Animation / 15-Cartoon / 366-Dragon / \n", "36-Piano / 53-Keyboard / 446-Sheet music / \n", "14-Guitar / 63-Acoustic guitar / \n", "3-Concert / \n", "94-Dashcam / \n", "5-Dance / 95-Talent show / \n", "31-Disc jockey / 2371-Hot dog / \n", "156-Agriculture / \n", "12-Food / 295-Chocolate / 367-Drink / \n", "31-Disc jockey / \n", "6-Animation / 9-Music video / 995-Attack on Titan / \n", "8-Football / \n", "102-Building / 113-House / 170-Home improvement / 2055-Mansion / \n", "6-Animation / 15-Cartoon / \n", "1-Vehicle / 4-Car / 245-BMW / \n", "3-Concert / 5-Dance / 16-Performance art / 95-Talent show / \n", "12-Food / 26-Cooking / 32-Recipe / 52-Dish / 58-Cuisine / 109-Cooking show / \n", "5-Dance / 16-Performance art / 115-Ballet / \n", "2-Video game / 2628-Sly Cooper / \n", "5-Dance / 130-Gymnastics / 350-Circus / 2940-Trapeze / \n", "0-Games / 256-Tennis / \n", "76-Ball / 350-Circus / 788-Juggling / \n", "0-Games / 51-Strategy video game / 373-Clash of Clans / \n", "0-Games / 2-Video game / 51-Strategy video game / 165-League of Legends / \n", "102-Building / 113-House / 339-Architecture / 416-Apartment / 939-Condominium / \n", "3-Concert / 68-Lighting / \n", "10-Animal / 80-Horse / 247-Stallion / 289-Dressage / 438-Mare / \n", "5-Dance / 16-Performance art / 95-Talent show / \n", "5-Dance / 115-Ballet / \n", "31-Disc jockey / 133-Nightclub / \n", "5-Dance / 16-Performance art / 95-Talent show / \n", "3-Concert / 7-Musician / 30-Drummer / \n", "516-Headphones / \n", "5-Dance / \n", "0-Games / 2-Video game / 19-PC game / 33-Weapon / 43-Call of Duty / 134-Call of Duty: Black Ops / 138-Call of Duty: Black Ops II / \n", "0-Games / 2-Video game / 19-PC game / 33-Weapon / 96-Soldier / 948-Far Cry / 1683-Far Cry 4 / \n", "31-Disc jockey / \n", "3-Concert / 7-Musician / 13-Musical ensemble / 24-String instrument / 38-Orchestra / 150-Violin / 401-Cello / 450-Viola / \n", "10-Animal / 48-Pet / 132-Bird / 598-Parrot / 1341-Parakeet / \n", "191-Cheerleading / \n", "121-Photography / 155-Camera / 566-Digital camera / 629-Camera lens / 721-Digital SLR / 3328-Canon EOS 7D / \n", "162-Medicine / 1110-Cell / \n", "10-Animal / 48-Pet / 71-Dog / 182-Puppy / 1839-Rottweiler / \n", "3-Concert / \n", "5-Dance / 78-Wedding / 153-Dress / 190-Bride / 421-Gown / 590-Wedding dress / \n", "12-Food / 26-Cooking / 32-Recipe / 52-Dish / 58-Cuisine / 239-Roasting / 671-Tomato / \n", "9-Music video / \n", "3-Concert / \n", "5-Dance / 1544-Laboratory / \n", "3-Concert / \n", "3-Concert / \n", "0-Games / 2-Video game / \n", "6-Animation / 10-Animal / 20-Trailer / 91-Fish / \n", "5-Dance / \n", "12-Food / 28-Fashion / 56-Hair / 106-Hairstyle / \n", "66-Bollywood / \n", "14-Guitar / 24-String instrument / 63-Acoustic guitar / \n", "2065-Tsubasa: Reservoir Chronicle / \n", "7-Musician / 14-Guitar / 24-String instrument / 63-Acoustic guitar / \n", "1-Vehicle / 4-Car / 154-Wheel / 234-Ford / 918-Brake / 2899-Ford Explorer / \n", "22-Nature / 23-Mobile phone / 29-Smartphone / 223-Weather / 1379-HTC One / 2328-HTC One (M8) / \n", "3-Concert / 38-Orchestra / \n", "0-Games / 59-Winter sport / 81-Athlete / 112-Ice skating / 142-Hockey / 312-Ice rink / \n", "0-Games / 2-Video game / 55-Video game console / 320-Super Smash Bros. / 443-Wii U / 697-Super Smash Bros. for Nintendo 3DS and Wii U / \n", "12-Food / 179-Kitchen / 2153-Sink / 2243-Stainless steel / \n", "12-Food / 227-Water / \n", "2-Video game / 23-Mobile phone / 29-Smartphone / 37-Gadget / 101-Telephone / 1379-HTC One / \n", "6-Animation / 15-Cartoon / 61-Art / 89-Comics / 123-Naruto / 183-Manga / 285-Sasuke Uchiha / \n", "25-Toy / 140-LEGO / \n", "3-Concert / 7-Musician / \n", "15-Cartoon / 126-The Walt Disney Company / 2058-Snowman / 2145-Olaf / \n", "3-Concert / 68-Lighting / \n", "1-Vehicle / 42-Fishing / 82-Airplane / 778-Gliding / 1263-Glider / \n", "0-Games / 2-Video game / 6-Animation / 362-PlayStation Portable / 3813-Dynasty Warriors: Gundam / \n", "3-Concert / 7-Musician / \n", "6-Animation / 616-The Bible / \n", "2-Video game / 55-Video game console / 127-Xbox 360 / 181-Xbox / \n", "313-Book / \n", "6-Animation / 89-Comics / 183-Manga / \n", "3-Concert / 7-Musician / 13-Musical ensemble / 30-Drummer / 39-Drums / 670-Banjo / \n", "3-Concert / 7-Musician / 13-Musical ensemble / 38-Orchestra / 46-Choir / \n", "12-Food / 215-Eating / 323-Toddler / \n", "916-Spa / \n", "5-Dance / 95-Talent show / 98-Festival / \n", "115-Ballet / \n", "0-Games / 2-Video game / 685-Luigi / 887-King Koopa / 4464-Mario & Luigi: Bowser's Inside Story / \n", "12-Food / 32-Recipe / 226-Dessert / 232-Cake / 236-Baking / 295-Chocolate / 1189-Torte / \n", "6-Animation / 15-Cartoon / 1118-Kirby / 2266-Potato chip / \n", "5-Dance / 16-Performance art / 95-Talent show / \n", "0-Games / 39-Drums / 356-Star / 445-Guitar Hero / 724-Rock Band / \n", "1-Vehicle / 27-Motorcycle / 40-Road / 94-Dashcam / 158-Tire / \n", "55-Video game console / 111-PlayStation 3 / 414-Computer monitor / \n", "12-Food / 238-Nail / \n", "3-Concert / 7-Musician / 16-Performance art / 30-Drummer / \n", "3-Concert / 7-Musician / 13-Musical ensemble / 133-Nightclub / \n", "0-Games / 8-Football / 21-Stadium / 77-Arena / \n", "28-Fashion / 164-Beach / \n", "12-Food / 26-Cooking / 32-Recipe / 109-Cooking show / 179-Kitchen / 232-Cake / 1189-Torte / \n", "0-Games / 2-Video game / 19-PC game / 35-Minecraft / \n", "1-Vehicle / 87-Boat / 308-Ship / 917-Sail / 1309-Sailing ship / \n", "122-Weight training / 204-Gym / \n", "0-Games / 2-Video game / 19-PC game / 34-Action-adventure game / 35-Minecraft / \n", "209-University / 2015-News broadcasting / \n", "9-Music video / \n", "47-Personal computer / 131-Computer / 141-Microsoft Windows / \n", "9-Music video / \n", "0-Games / 2-Video game / 19-PC game / 739-Guild Wars / 839-Guild Wars 2 / \n", "0-Games / 8-Football / 21-Stadium / 65-Kick / 73-Ball / 81-Athlete / \n", "86-Plant / 136-Gardening / 137-Tree / 1023-Rose / \n", "9-Music video / \n", "23-Mobile phone / 29-Smartphone / 37-Gadget / 72-iPhone / 101-Telephone / 3089-Nokia N8 / \n", "10-Animal / 80-Horse / 351-Jumping / \n", "5-Dance / 16-Performance art / \n", "419-Earth / \n", "25-Toy / 140-LEGO / 749-Figurine / 855-Hulk / 884-Lego minifigure / 2276-Lego Marvel Super Heroes / \n", "6-Animation / 89-Comics / 123-Naruto / 444-Family / \n", "0-Games / 2-Video game / 33-Weapon / 43-Call of Duty / 96-Soldier / 422-Call of Duty: Ghosts / \n", "5-Dance / 16-Performance art / 361-Flamenco / \n", "9-Music video / 110-Album / \n", "46-Choir / \n", "22-Nature / 42-Fishing / 267-Mountain / 573-Paragliding / 800-Mountain pass / \n", "1-Vehicle / 50-Aircraft / 92-Radio-controlled model / 137-Tree / 149-Model aircraft / 160-Radio-controlled aircraft / 266-Helicopter / 392-Radio-controlled helicopter / \n", "10-Animal / 280-Wildlife / \n", "0-Games / 1029-Aion: The Tower of Eternity / \n", "3-Concert / \n", "9-Music video / 78-Wedding / 190-Bride / 590-Wedding dress / \n", "61-Art / 148-Painting / 329-Swimming pool / \n", "1-Vehicle / 4-Car / 74-Truck / 154-Wheel / 167-Four-wheel drive / 234-Ford / 375-Pickup truck / \n", "9-Music video / \n", "3-Concert / 7-Musician / 13-Musical ensemble / \n", "3-Concert / 5-Dance / \n", "9-Music video / \n", "23-Mobile phone / 2038-Credit card / \n", "9-Music video / \n", "1-Vehicle / 4-Car / 27-Motorcycle / 40-Road / 212-Highway / \n", "0-Games / 8-Football / \n", "61-Art / 148-Painting / \n", "12-Food / 26-Cooking / 32-Recipe / 348-Fruit / 985-Strawberry / \n", "28-Fashion / 59-Winter sport / 281-Snowboarding / \n", "3-Concert / 7-Musician / 13-Musical ensemble / 1019-Christ / \n", "1-Vehicle / 17-Racing / 67-Cycling / 69-Bicycle / 570-Road bicycle racing / \n", "31-Disc jockey / \n", "3-Concert / \n", "9-Music video / \n", "1-Vehicle / 25-Toy / 62-Train / 507-Thomas the Tank Engine / 1154-Anpanman / 1685-Toy train / \n", "1-Vehicle / 4-Car / 25-Toy / 198-Off-road vehicle / 360-Jeep / \n", "656-Metal Gear / 2698-Metal Gear Rising: Revengeance / \n", "12-Food / 205-Newscaster / \n", "10-Animal / 174-Farm / 1116-Goat / \n", "18-Outdoor recreation / 222-Surfing / 233-Ocean / 4412-Alcatraz Island / \n", "3-Concert / 68-Lighting / 133-Nightclub / \n", "12-Food / 367-Drink / 1325-Gel / 2731-Aloe vera / \n", "200-Doll / 587-Clay / \n", "0-Games / 2-Video game / 90-Sports game / 127-Xbox 360 / \n", "10-Animal / 80-Horse / 1323-Helmet camera / \n", "1-Vehicle / 527-Battery / 2755-Lithium polymer battery / \n", "3-Concert / 7-Musician / 13-Musical ensemble / 14-Guitar / 30-Drummer / \n", "22-Nature / 42-Fishing / 147-River / 283-Rock / 825-Stream / \n", "17-Racing / 67-Cycling / 69-Bicycle / 570-Road bicycle racing / \n", "31-Disc jockey / \n", "93-Comedy / \n", "30-Drummer / 39-Drums / 44-Drums / 128-Cymbal / 146-Snare drum / \n", "0-Games / 1-Vehicle / 2-Video game / 4-Car / \n", "12-Food / 1545-Floristry / \n", "22-Nature / 260-Prayer / \n", "209-University / \n", "0-Games / 2-Video game / 34-Action-adventure game / 243-Sonic the Hedgehog / 251-Sonic the Hedgehog / 1197-Hedgehog / 1299-Shadow the Hedgehog / 1512-Sonic Generations / 2028-Sonic World Adventure / 3321-Shadow the Hedgehog / \n", "6-Animation / 15-Cartoon / \n", "3-Concert / 5-Dance / 2032-Pepsi / \n", "1-Vehicle / 4-Car / \n", "162-Medicine / \n", "6-Animation / 15-Cartoon / 183-Manga / \n", "191-Cheerleading / \n", "12-Food / \n", "3-Concert / 7-Musician / 13-Musical ensemble / 38-Orchestra / 46-Choir / 1942-Oboe / \n", "1-Vehicle / 8-Football / 27-Motorcycle / \n", "5-Dance / \n", "20-Trailer / 66-Bollywood / \n", "457-Knitting / \n", "9-Music video / 1950-The CW / 3068-Green Arrow / \n", "5-Dance / 16-Performance art / 95-Talent show / \n", "0-Games / 8-Football / 21-Stadium / 65-Kick / 73-Ball / 76-Ball / 81-Athlete / \n", "587-Clay / \n", "0-Games / 2-Video game / 116-Pokémon / 186-Pokémon / \n", "61-Art / 97-Drawing / 148-Painting / \n", "3-Concert / 7-Musician / 13-Musical ensemble / 16-Performance art / \n", "1-Vehicle / 64-Transport / 74-Truck / 158-Tire / \n", "1-Vehicle / 4-Car / \n", "66-Bollywood / \n", "1-Vehicle / 4-Car / 259-Fire / 371-Firefighter / \n", "290-Pony / 315-Knife / \n", "3-Concert / 5-Dance / \n", "2128-Pollution / \n", "6-Animation / 25-Toy / \n", "59-Winter sport / 281-Snowboarding / 1008-Kite / 4321-Mountainboarding / \n", "326-Metin2 / \n", "3-Concert / 5-Dance / 16-Performance art / 38-Orchestra / \n", "1-Vehicle / 85-Combat / 271-Tank / \n", "1-Vehicle / 4-Car / 27-Motorcycle / 40-Road / 74-Truck / 137-Tree / 197-Trail / 363-Soil / \n", "67-Cycling / 69-Bicycle / 669-Extreme sport / 769-Red Bull / \n", "47-Personal computer / 131-Computer / 309-Computer hardware / 602-Central processing unit / \n", "0-Games / 411-MapleStory / \n", "0-Games / 960-Badminton / \n", "0-Games / 1-Vehicle / 2-Video game / 4-Car / 19-PC game / 34-Action-adventure game / 201-Grand Theft Auto: San Andreas / 413-PlayStation 2 / 858-Carl Johnson / \n", "3-Concert / 7-Musician / 13-Musical ensemble / 24-String instrument / 150-Violin / 401-Cello / 450-Viola / \n", "1-Vehicle / 87-Boat / \n", "1-Vehicle / 4-Car / 11-Motorsport / 17-Racing / 169-Rallying / \n", "0-Games / 1-Vehicle / 2-Video game / 4-Car / 11-Motorsport / 41-Sports car / \n", "0-Games / 35-Minecraft / 488-Door / \n", "1-Vehicle / 4-Car / 11-Motorsport / 569-Lamborghini / 1132-Lamborghini Aventador / \n", "0-Games / 2-Video game / 6-Animation / 15-Cartoon / 89-Comics / 123-Naruto / 591-Naruto: Ultimate Ninja / \n", "7-Musician / 36-Piano / 53-Keyboard / 104-Pianist / 107-Musical keyboard / \n", "36-Piano / 53-Keyboard / 107-Musical keyboard / 116-Pokémon / 592-Synthesia / \n", "5-Dance / \n", "5-Dance / 49-School / 191-Cheerleading / 306-High school / \n", "0-Games / 2-Video game / 19-PC game / 43-Call of Duty / 431-Call of Duty: Advanced Warfare / \n", "1061-Marriage proposal / \n", "13-Musical ensemble / 14-Guitar / 24-String instrument / 1533-Lead guitar / \n", "3-Concert / 5-Dance / 16-Performance art / \n", "14-Guitar / \n", "1-Vehicle / 4-Car / 27-Motorcycle / 336-Scooter / 1448-Vespa / \n", "12-Food / 227-Water / \n", "419-Earth / \n", "3-Concert / 7-Musician / 13-Musical ensemble / 16-Performance art / 38-Orchestra / \n", "75-Wrestling / 641-Mask / \n", "12-Food / 26-Cooking / 179-Kitchen / \n", "1-Vehicle / 87-Boat / 308-Ship / \n", "3-Concert / 16-Performance art / \n", "9-Music video / 110-Album / \n", "0-Games / 2-Video game / 51-Strategy video game / 171-Warcraft / 382-Dota 2 / 549-Defense of the Ancients / 1378-Multiplayer online battle arena / \n", "2767-Captain Hook / \n", "402-Jewellery / 859-Beadwork / 1300-Earrings / \n", "1-Vehicle / 59-Winter sport / 84-Snow / 610-Snowmobile / 1201-Ski-Doo / \n", "3-Concert / 7-Musician / 16-Performance art / \n", "12-Food / \n", "1-Vehicle / 4-Car / 11-Motorsport / 17-Racing / 135-Drifting / \n", "6-Animation / 25-Toy / 200-Doll / 439-Barbie / \n", "1-Vehicle / 84-Snow / 120-Winter / 610-Snowmobile / \n", "1007-Portal / \n", "66-Bollywood / \n", "1057-Embroidery / 1451-Ribbon / \n", "0-Games / 2-Video game / 34-Action-adventure game / 1447-Skylanders / 2321-Skylanders: Spyro's Adventure / \n", "9-Music video / \n", "3-Concert / 5-Dance / 16-Performance art / \n", "9-Music video / 541-Black-and-white / 3154-Pajamas / \n", "8-Football / 54-Highlight film / \n", "0-Games / 837-Roblox / \n", "574-Elevator / \n", "979-Star Trek / 2981-Spock / \n", "1-Vehicle / 4-Car / 11-Motorsport / 17-Racing / 57-Race track / 169-Rallying / \n", "23-Mobile phone / 29-Smartphone / 72-iPhone / \n", "163-Running / \n", "9-Music video / \n", "726-Clarinet / \n", "67-Cycling / 69-Bicycle / 147-River / \n", "6-Animation / 20-Trailer / 159-World of Warcraft / 171-Warcraft / \n", "3-Concert / \n", "0-Games / \n", "0-Games / 2-Video game / 51-Strategy video game / 165-League of Legends / 356-Star / \n", "1-Vehicle / 4-Car / 11-Motorsport / 17-Racing / 40-Road / 41-Sports car / 57-Race track / 378-Drag racing / 482-Asphalt / 552-Honda Civic / \n", "0-Games / 13-Musical ensemble / 79-American football / 209-University / \n", "6-Animation / 15-Cartoon / 183-Manga / \n", "223-Weather / 490-Rain / \n", "7-Musician / 14-Guitar / 24-String instrument / \n", "0-Games / 8-Football / \n", "3-Concert / 7-Musician / 13-Musical ensemble / 46-Choir / 318-Church / 1868-Preacher / \n", "1-Vehicle / 4-Car / 88-Machine / \n", "0-Games / 6-Animation / 19-PC game / \n", "15-Cartoon / 61-Art / 97-Drawing / \n", "7-Musician / 13-Musical ensemble / 14-Guitar / \n", "18-Outdoor recreation / 59-Winter sport / 112-Ice skating / 312-Ice rink / 316-Figure skating / 632-Ice dancing / \n", "992-Rocket / 3117-Match / \n", "47-Personal computer / 145-Tablet computer / 189-iPad / \n", "0-Games / 19-PC game / 1983-Infestation: Survivor Stories / \n", "86-Plant / 136-Gardening / 1023-Rose / 2316-Floral design / 3177-Tulip / \n", "0-Games / 8-Football / 65-Kick / 81-Athlete / \n", "6-Animation / \n", "31-Disc jockey / \n", "12-Food / 26-Cooking / 32-Recipe / 52-Dish / 58-Cuisine / 227-Water / 557-Salad / \n", "2-Video game / 55-Video game console / 111-PlayStation 3 / 145-Tablet computer / 3238-GameStop / \n", "3200-Morgana / \n", "28-Fashion / 390-Model / 495-Runway / \n", "512-Printing / \n", "0-Games / 8-Football / 21-Stadium / 54-Highlight film / 65-Kick / 73-Ball / \n", "6-Animation / \n", "3-Concert / 126-The Walt Disney Company / 144-Amusement park / \n", "1-Vehicle / 25-Toy / 62-Train / 140-LEGO / 507-Thomas the Tank Engine / \n", "3-Concert / 28-Fashion / 45-Cosmetics / 153-Dress / 206-Shoe / \n", "13-Musical ensemble / 98-Festival / 253-Carnival / \n", "0-Games / 2-Video game / 23-Mobile phone / 29-Smartphone / 37-Gadget / 178-Samsung Galaxy / \n", "1-Vehicle / 4-Car / 70-Driving / 94-Dashcam / \n", "0-Games / 2-Video game / 19-PC game / 20-Trailer / 43-Call of Duty / 55-Video game console / 138-Call of Duty: Black Ops II / 422-Call of Duty: Ghosts / \n", "1-Vehicle / 4-Car / 27-Motorcycle / 64-Transport / 969-Volkswagen Beetle / \n", "122-Weight training / \n", "2-Video game / 8-Football / 90-Sports game / 127-Xbox 360 / 181-Xbox / 1199-Kinect / \n", "78-Wedding / \n", "9-Music video / \n", "5-Dance / 49-School / \n", "31-Disc jockey / 85-Combat / 1623-Shaolin Kung Fu / \n", "9-Music video / \n", "10-Animal / 166-Cat / 1347-Coffeehouse / \n", "28-Fashion / 45-Cosmetics / 56-Hair / 195-Mascara / 235-Lipstick / 291-Rouge / 436-Concealer / \n", "3-Concert / 7-Musician / 13-Musical ensemble / 38-Orchestra / \n", "3-Concert / 356-Star / \n", "0-Games / 2-Video game / 17-Racing / \n", "0-Games / 35-Minecraft / \n", "1684-Hillsong Church / \n", "23-Mobile phone / 29-Smartphone / 37-Gadget / 101-Telephone / 178-Samsung Galaxy / 487-Samsung Electronics / 1015-Samsung Galaxy Note series / 1727-Samsung Galaxy Note 3 / \n", "3-Concert / \n", "18-Outdoor recreation / 83-Skateboarding / 108-Skateboard / 1031-Fisheye lens / \n", "12-Food / 26-Cooking / 32-Recipe / 52-Dish / 58-Cuisine / 109-Cooking show / 348-Fruit / 1175-Soap / \n", "10-Animal / 48-Pet / 166-Cat / 182-Puppy / \n", "35-Minecraft / 141-Microsoft Windows / \n", "1143-Marimba / \n", "232-Cake / \n", "5-Dance / \n", "75-Wrestling / 85-Combat / 117-Boxing / 399-Kickboxing / \n", "78-Wedding / \n", "1-Vehicle / 4-Car / 74-Truck / 1242-Car wash / \n", "0-Games / 241-Fighting game / 486-M.U.G.E.N / 1506-Akuma / \n", "5-Dance / 95-Talent show / \n", "1-Vehicle / 4-Car / 158-Tire / 167-Four-wheel drive / 198-Off-road vehicle / 277-Sport utility vehicle / 901-Mitsubishi / 2432-Mitsubishi Pajero / \n", "147-River / \n", "8-Football / 73-Ball / \n", "415-Sewing / 546-Stitch / 2961-Button / \n", "1-Vehicle / 4-Car / 2583-Limousine / \n", "5-Dance / \n", "6-Animation / 20-Trailer / 123-Naruto / \n", "353-Wall / 1350-Plaster / 2164-Silk / \n", "0-Games / 1876-Beatmania IIDX / 2063-Beatmania / \n", "6-Animation / 15-Cartoon / \n", "0-Games / 6-Animation / 15-Cartoon / \n", "75-Wrestling / \n", "1-Vehicle / 4-Car / 154-Wheel / 299-Volkswagen Passenger Cars / 493-Rim / \n", "3-Concert / \n", "59-Winter sport / 112-Ice skating / 262-Ice / 905-Flash Video / \n", "12-Food / 226-Dessert / \n", "439-Barbie / \n", "14-Guitar / 24-String instrument / \n", "61-Art / 457-Knitting / 476-Textile / 546-Stitch / \n", "6-Animation / 89-Comics / \n", "1-Vehicle / 4-Car / 533-Oil / 2303-Motor oil / \n", "42-Fishing / 573-Paragliding / \n", "8-Football / 21-Stadium / \n", "1604-Satan / 1868-Preacher / \n", "0-Games / 8-Football / 21-Stadium / 65-Kick / 76-Ball / 81-Athlete / \n", "99-Christmas / \n", "1-Vehicle / 4-Car / 3896-Mitsubishi Outlander / \n", "0-Games / 1492-Persona / \n", "0-Games / 138-Call of Duty: Black Ops II / 194-Call of Duty: Modern Warfare 2 / \n", "0-Games / 2-Video game / 6-Animation / 240-Arcade game / 514-Tekken / 797-Tekken / \n", "0-Games / 2-Video game / 20-Trailer / 34-Action-adventure game / 55-Video game console / 127-Xbox 360 / 2863-Max Payne 3 / \n", "0-Games / 2-Video game / 1412-Half-Life / \n", "126-The Walt Disney Company / \n", "3-Concert / \n", "5-Dance / 191-Cheerleading / \n", "3-Concert / 7-Musician / 30-Drummer / \n", "955-Police officer / \n", "1-Vehicle / 4-Car / 11-Motorsport / 158-Tire / 207-Exhaust system / 947-Muffler / 1321-Škoda Auto / 2690-Škoda Octavia / \n", "42-Fishing / 125-Television / 844-Bass / \n", "0-Games / 8-Football / 90-Sports game / 180-FIFA 15 / 480-FIFA 13 / \n", "10-Animal / 18-Outdoor recreation / 80-Horse / 247-Stallion / 289-Dressage / 404-Livestock / 438-Mare / \n", "10-Animal / 709-Bull / \n", "67-Cycling / 69-Bicycle / \n", "56-Hair / 106-Hairstyle / 340-Afro-textured hair / \n", "9-Music video / \n", "31-Disc jockey / 133-Nightclub / \n", "5-Dance / \n", "18-Outdoor recreation / 283-Rock / 548-Parachuting / 669-Extreme sport / \n", "0-Games / 2-Video game / 6-Animation / 15-Cartoon / 89-Comics / 185-Dragon Ball / 244-Goku / 418-Gohan / 869-Dragon Ball Z: Budokai Tenkaichi / 876-Freeza / \n", "121-Photography / 155-Camera / 566-Digital camera / 629-Camera lens / 721-Digital SLR / \n", "9-Music video / \n", "49-School / \n", "9-Music video / \n", "0-Games / 2-Video game / 6-Animation / 34-Action-adventure game / 116-Pokémon / 186-Pokémon / 1387-Pokémon Battle Revolution / \n", "2337-Passport / \n", "10-Animal / 12-Food / 88-Machine / 3571-Lollipop / \n", "28-Fashion / 696-T-shirt / 793-Shirt / 1525-Wallet / \n", "5-Dance / 365-Fireworks / 440-Lion / \n", "9-Music video / \n", "0-Games / 8-Football / \n", "46-Choir / \n", "2-Video game / 37-Gadget / 55-Video game console / 111-PlayStation 3 / 127-Xbox 360 / 181-Xbox / 317-PlayStation / 677-Game controller / \n", "12-Food / \n", "326-Metin2 / \n", "3-Concert / 5-Dance / 16-Performance art / \n", "5-Dance / 18-Outdoor recreation / 59-Winter sport / 112-Ice skating / 262-Ice / 316-Figure skating / 632-Ice dancing / \n", "110-Album / \n", "1-Vehicle / 4-Car / 11-Motorsport / 17-Racing / 57-Race track / 74-Truck / 124-Tractor / 252-Tractor pulling / \n", "9-Music video / \n", "12-Food / 18-Outdoor recreation / 22-Nature / 86-Plant / 137-Tree / 208-Wood / 213-Forest / 673-Saw / 1073-Chainsaw / \n", "0-Games / 2-Video game / 362-PlayStation Portable / \n", "0-Games / 1-Vehicle / 2-Video game / 4-Car / 19-PC game / 118-Grand Theft Auto V / 176-PlayStation 4 / 1275-Watch Dogs / \n", "5-Dance / 85-Combat / \n", "0-Games / 116-Pokémon / 186-Pokémon / \n", "0-Games / 2-Video game / 739-Guild Wars / 839-Guild Wars 2 / 1130-Armour / \n", "24-String instrument / 670-Banjo / \n", "1042-Jumbotron / \n", "0-Games / 59-Winter sport / 81-Athlete / 112-Ice skating / 142-Hockey / 312-Ice rink / \n", "1-Vehicle / 42-Fishing / 87-Boat / \n", "1-Vehicle / 10-Animal / 197-Trail / \n", "9-Music video / \n", "9-Music video / \n", "12-Food / 26-Cooking / 32-Recipe / 52-Dish / 58-Cuisine / 771-Noodle / \n", "44-Drums / 227-Water / \n", "1-Vehicle / 4-Car / 11-Motorsport / 158-Tire / 198-Off-road vehicle / 360-Jeep / 1205-Jeep Wrangler / \n", "3-Concert / 16-Performance art / \n", "13-Musical ensemble / \n", "45-Cosmetics / 173-Eye shadow / 193-Eye liner / 195-Mascara / 196-Eye / 218-Eyelash / \n", "1-Vehicle / 4-Car / \n", "10-Animal / 48-Pet / 71-Dog / 182-Puppy / 2083-Border Collie / \n", "1-Vehicle / 4-Car / 27-Motorcycle / 40-Road / \n", "31-Disc jockey / \n", "0-Games / 8-Football / 90-Sports game / 180-FIFA 15 / \n", "3-Concert / 7-Musician / 13-Musical ensemble / 230-Brass instrument / 427-Trumpet / 634-Trombone / \n", "1-Vehicle / 50-Aircraft / 82-Airplane / 139-Airport / 151-Landing / 157-Jet aircraft / 175-Airliner / 187-Takeoff / 188-Airline / 254-Runway / 923-Boeing 777 / 2391-American Airlines / 3705-Singapore Airlines / \n", "22-Nature / 694-Tai chi / \n", "10-Animal / 48-Pet / 132-Bird / \n", "0-Games / 2-Video game / 6-Animation / 15-Cartoon / 89-Comics / 185-Dragon Ball / 244-Goku / 418-Gohan / 876-Freeza / 2144-Dragon Ball: Raging Blast 2 / \n", "9-Music video / \n", "1989-Oculus Rift / \n", "0-Games / 2-Video game / 60-Basketball / 176-PlayStation 4 / 314-Xbox One / 991-NBA 2K15 / \n", "0-Games / 8-Football / 54-Highlight film / 79-American football / 993-Quarterback / 3987-Tight end / \n", "164-Beach / 423-Coast / \n", "10-Animal / 141-Microsoft Windows / \n", "25-Toy / 140-LEGO / 264-Star Wars / 852-Lego Star Wars / \n", "0-Games / 8-Football / 21-Stadium / 54-Highlight film / 65-Kick / 77-Arena / 81-Athlete / \n", "42-Fishing / \n", "1-Vehicle / 4-Car / 11-Motorsport / 17-Racing / 41-Sports car / 135-Drifting / \n", "38-Orchestra / \n", "18-Outdoor recreation / 83-Skateboarding / 108-Skateboard / \n", "95-Talent show / 130-Gymnastics / \n", "0-Games / 2-Video game / 366-Dragon / 420-The Elder Scrolls V: Skyrim / 441-The Elder Scrolls / \n", "1-Vehicle / 18-Outdoor recreation / 40-Road / 67-Cycling / 69-Bicycle / 197-Trail / 220-Mountain bike / 284-Mountain biking / 520-Bicycle frame / 645-Road bicycle / \n", "5-Dance / 13-Musical ensemble / \n", "315-Knife / \n", "5-Dance / \n", "1-Vehicle / 4-Car / 94-Dashcam / \n", "2-Video game / 237-Final Fantasy / \n", "6-Animation / 25-Toy / 200-Doll / \n", "10-Animal / 86-Plant / 113-House / 280-Wildlife / \n", "402-Jewellery / 1010-Ring / 1089-Wire / \n", "31-Disc jockey / \n", "0-Games / 411-MapleStory / \n", "6-Animation / 15-Cartoon / 89-Comics / 750-Sailor Moon / 1342-Sailor Moon / \n", "18-Outdoor recreation / 22-Nature / 164-Beach / 233-Ocean / 423-Coast / \n", "0-Games / 8-Football / 21-Stadium / 77-Arena / \n", "117-Boxing / \n", "1-Vehicle / 4-Car / 125-Television / 205-Newscaster / \n", "920-Chair / 1454-Stairs / \n", "3-Concert / 7-Musician / 13-Musical ensemble / 46-Choir / \n", "3-Concert / \n", "1-Vehicle / 4-Car / 11-Motorsport / 27-Motorcycle / \n", "6-Animation / 15-Cartoon / 126-The Walt Disney Company / 1139-Belle / 1265-Kyle Kingson / \n", "6-Animation / 61-Art / 97-Drawing / 148-Painting / 408-Sketch / 528-Pencil / 628-Portrait / \n", "8-Football / \n", "6-Animation / \n", "0-Games / 2-Video game / 241-Fighting game / 1291-Mortal Kombat / \n", "1-Vehicle / 4-Car / 118-Grand Theft Auto V / 127-Xbox 360 / 176-PlayStation 4 / 181-Xbox / 314-Xbox One / 330-Money / 2212-Metal Gear Solid V: The Phantom Pain / \n", "0-Games / 2-Video game / 34-Action-adventure game / 85-Combat / 111-PlayStation 3 / 380-Assassin's Creed / 1435-Ezio Auditore da Firenze / 1485-Assassin's Creed: Brotherhood / \n", "79-American football / \n", "0-Games / 2-Video game / 55-Video game console / 366-Dragon / 1559-Spyro / \n", "6-Animation / 15-Cartoon / \n", "0-Games / 2-Video game / 6-Animation / 241-Fighting game / 514-Tekken / 797-Tekken / 1470-Tekken Tag Tournament 2 / \n", "0-Games / 8-Football / 21-Stadium / 65-Kick / \n", "0-Games / 2-Video game / 51-Strategy video game / 165-League of Legends / \n", "0-Games / 2-Video game / 2379-Tom Clancy's Ghost Recon / \n", "1-Vehicle / 124-Tractor / 249-Heavy equipment / \n", "85-Combat / 117-Boxing / 399-Kickboxing / 699-Kick / \n", "10-Animal / 48-Pet / 71-Dog / 162-Medicine / 1845-Pug / \n", "14-Guitar / 24-String instrument / \n", "0-Games / 2-Video game / 55-Video game console / 243-Sonic the Hedgehog / 251-Sonic the Hedgehog / 961-Sega Genesis / 1268-Tails / 1475-Knuckles the Echidna / 2327-Sonic the Hedgehog / \n", "0-Games / 51-Strategy video game / 603-StarCraft II: Wings of Liberty / 2353-Heroes of the Storm / 2647-StarCraft II: Heart of the Swarm / 4353-Sarah Kerrigan / \n", "0-Games / 2-Video game / 34-Action-adventure game / 201-Grand Theft Auto: San Andreas / \n", "0-Games / 2-Video game / 55-Video game console / \n", "3-Concert / \n", "10-Animal / 22-Nature / 280-Wildlife / 4253-Koala / \n", "0-Games / 2-Video game / 19-PC game / 51-Strategy video game / 165-League of Legends / \n", "121-Photography / \n", "3-Concert / 7-Musician / 16-Performance art / \n", "1-Vehicle / 22-Nature / 86-Plant / 136-Gardening / \n", "93-Comedy / \n", "0-Games / 54-Highlight film / 60-Basketball / 81-Athlete / 168-Basketball moves / 255-Slam dunk / \n", "66-Bollywood / \n", "20-Trailer / 1265-Kyle Kingson / 4301-Beauty and the Beast / \n", "0-Games / 33-Weapon / 1938-Rust / \n", "1-Vehicle / 50-Aircraft / 82-Airplane / 129-Aviation / 139-Airport / 151-Landing / 157-Jet aircraft / 175-Airliner / 719-Boeing 747 / 856-Airbus A320 family / 1316-Airbus A330 / 3671-Instrument landing system / \n", "0-Games / 2-Video game / 20-Trailer / 111-PlayStation 3 / 127-Xbox 360 / 366-Dragon / 2621-GameTrailers / 3675-Dragon's Dogma / \n", "293-Slide show / \n", "9-Music video / \n", "6-Animation / 18-Outdoor recreation / 22-Nature / \n", "1-Vehicle / 4-Car / 11-Motorsport / 17-Racing / 27-Motorcycle / 40-Road / 158-Tire / \n", "12-Food / 32-Recipe / 52-Dish / 215-Eating / 557-Salad / \n", "3-Concert / 7-Musician / 24-String instrument / \n", "1-Vehicle / 4-Car / 40-Road / 64-Transport / 74-Truck / 114-Engine / 259-Fire / 371-Firefighter / 796-Fire engine / 1020-Emergency vehicle / 1168-Neighbourhood / 1245-Alarm device / \n", "12-Food / 26-Cooking / 32-Recipe / 52-Dish / 58-Cuisine / 109-Cooking show / 184-Vegetable / 210-Cookware and bakeware / 301-Meal / 557-Salad / 2614-Lettuce / \n", "3-Concert / \n", "6-Animation / 15-Cartoon / 185-Dragon Ball / 244-Goku / 418-Gohan / \n", "162-Medicine / \n", "0-Games / 2-Video game / 43-Call of Duty / 96-Soldier / 176-PlayStation 4 / 422-Call of Duty: Ghosts / \n", "12-Food / \n", "3-Concert / 7-Musician / 30-Drummer / 39-Drums / 44-Drums / 128-Cymbal / \n", "2-Video game / 1792-Router / \n", "5-Dance / 16-Performance art / 430-Ballroom dance / \n", "25-Toy / 323-Toddler / 354-Egg / 417-Play-Doh / 611-Kinder Surprise / \n", "3-Concert / 7-Musician / 30-Drummer / 44-Drums / \n", "12-Food / 867-Tea / \n", "130-Gymnastics / 1417-Balance beam / \n", "6-Animation / 15-Cartoon / \n", "230-Brass instrument / 296-Saxophone / \n", "3-Concert / 7-Musician / 30-Drummer / \n", "3-Concert / 5-Dance / \n", "0-Games / 2-Video game / 55-Video game console / \n", "0-Games / 8-Football / 21-Stadium / 76-Ball / \n", "5-Dance / \n", "1-Vehicle / 4-Car / 84-Snow / 120-Winter / 158-Tire / \n", "8-Football / 28-Fashion / 206-Shoe / 333-Nike; Inc. / 1800-Football boot / \n", "28-Fashion / 1178-Quest / \n", "0-Games / 2-Video game / \n", "7-Musician / 14-Guitar / \n", "0-Games / 2-Video game / 14-Guitar / 445-Guitar Hero / \n", "0-Games / 18-Outdoor recreation / 278-GoPro / \n", "2-Video game / 55-Video game console / 176-PlayStation 4 / 181-Xbox / 282-Skin / 317-PlayStation / 677-Game controller / 2592-DualShock / 2769-Decal / \n", "1-Vehicle / 4-Car / 40-Road / 70-Driving / 94-Dashcam / 212-Highway / \n", "162-Medicine / \n", "0-Games / 2-Video game / 1630-Warframe / \n", "61-Art / 97-Drawing / 148-Painting / 303-Paint / 873-Acrylic paint / 936-Watercolor paint / \n", "12-Food / 88-Machine / 2196-Boiler / \n", "5-Dance / 16-Performance art / 95-Talent show / 115-Ballet / \n", "223-Weather / 502-Sky / 3942-Meteor shower / \n", "3-Concert / 7-Musician / 13-Musical ensemble / 14-Guitar / 24-String instrument / 842-Harp / 878-Composer / \n", "0-Games / 8-Football / 59-Winter sport / 112-Ice skating / 142-Hockey / \n", "402-Jewellery / 907-Necklace / \n", "122-Weight training / 204-Gym / 214-Muscle / 541-Black-and-white / \n", "93-Comedy / \n", "1-Vehicle / 4-Car / 27-Motorcycle / 336-Scooter / 2686-Electric motorcycles and scooters / \n", "18-Outdoor recreation / 721-Digital SLR / \n", "7-Musician / 38-Orchestra / \n", "31-Disc jockey / \n", "752-Pasta / 2362-Porcelain / 3340-Cold porcelain / \n", "7-Musician / 13-Musical ensemble / 14-Guitar / 39-Drums / 44-Drums / \n", "1-Vehicle / 88-Machine / 124-Tractor / 249-Heavy equipment / \n", "325-Vampire / \n", "8-Football / 85-Combat / 358-Rugby football / \n", "9-Music video / \n", "22-Nature / 288-Underwater / 381-Underwater diving / 1135-Dolphin / \n", "1-Vehicle / 62-Train / 64-Transport / 103-Rail transport / 119-Locomotive / 143-Railroad car / 152-Track / 161-Train station / 840-Museum / \n", "9-Music video / \n", "293-Slide show / \n", "10-Animal / 48-Pet / 132-Bird / 598-Parrot / \n", "882-Pen / \n", "0-Games / 2-Video game / 445-Guitar Hero / \n", "13-Musical ensemble / 1532-Military band / \n", "0-Games / 2-Video game / 55-Video game console / 292-Handheld game console / 320-Super Smash Bros. / 782-Super Smash Bros. Melee / \n", "468-Biology / \n", "0-Games / 8-Football / 54-Highlight film / 79-American football / \n", "0-Games / 34-Action-adventure game / \n", "1-Vehicle / 124-Tractor / 156-Agriculture / 249-Heavy equipment / 617-Maize / \n", "309-Computer hardware / 516-Headphones / \n", "0-Games / 2-Video game / 43-Call of Duty / 302-Call of Duty 4: Modern Warfare / \n", "9-Music video / \n", "66-Bollywood / \n", "56-Hair / \n", "31-Disc jockey / \n", "31-Disc jockey / \n", "0-Games / 2-Video game / 35-Minecraft / \n", "5-Dance / 356-Star / \n", "6-Animation / 15-Cartoon / \n", "32-Recipe / 162-Medicine / 3770-Toothpaste / \n", "0-Games / 1138-Tower of Saviors / \n", "36-Piano / 53-Keyboard / 104-Pianist / 107-Musical keyboard / \n", "1-Vehicle / 11-Motorsport / 17-Racing / 135-Drifting / 2785-Slot car / \n", "7-Musician / 13-Musical ensemble / 203-Marching band / 1532-Military band / \n", "6-Animation / \n", "75-Wrestling / 2547-Staples Center / \n", "136-Gardening / 428-Rail transport modelling / 1037-Grass / \n", "1-Vehicle / 4-Car / 27-Motorcycle / 40-Road / 105-Motorcycling / \n", "23-Mobile phone / 29-Smartphone / 72-iPhone / 189-iPad / 272-iPod / \n", "6-Animation / 123-Naruto / \n", "0-Games / 2-Video game / 33-Weapon / 834-Combat Arms / \n", "20-Trailer / 1014-Movieclips / \n", "0-Games / 2-Video game / 19-PC game / 159-World of Warcraft / 171-Warcraft / 1213-World of Warcraft: Wrath of the Lich King / 2139-Dungeon crawl / \n", "10-Animal / 22-Nature / 48-Pet / 86-Plant / 136-Gardening / 166-Cat / 217-Hunting / 346-Kitten / 2924-Persian cat / \n", "1-Vehicle / 4-Car / 124-Tractor / 156-Agriculture / 174-Farm / 617-Maize / 984-Silage / \n", "3-Concert / 7-Musician / \n", "10-Animal / 36-Piano / 53-Keyboard / 104-Pianist / 107-Musical keyboard / 550-Dinosaur / \n", "20-Trailer / 51-Strategy video game / 1072-Total War / 2810-Total War: Rome II / \n", "1-Vehicle / 4-Car / 249-Heavy equipment / 1093-Loader / \n", "31-Disc jockey / \n", "6-Animation / 9-Music video / 15-Cartoon / 781-Pachinko / \n", "3-Concert / \n", "31-Disc jockey / 1240-Mambo / \n", "5-Dance / 44-Drums / 3465-Davul / \n", "0-Games / 59-Winter sport / 112-Ice skating / 142-Hockey / \n", "14-Guitar / 24-String instrument / 1971-Epiphone / \n", "5-Dance / \n", "3-Concert / \n", "5-Dance / \n", "3-Concert / 6-Animation / 98-Festival / \n", "0-Games / 6-Animation / 781-Pachinko / \n", "3-Concert / \n", "0-Games / 2-Video game / 199-Wii / \n", "5-Dance / \n", "0-Games / 1-Vehicle / 2-Video game / 11-Motorsport / 199-Wii / 443-Wii U / 456-Go-kart / 518-Mario Kart / 1831-Mario Kart 8 / 4447-Rosalina / \n", "0-Games / 8-Football / 21-Stadium / 358-Rugby football / \n", "3380-Horror comedy / \n", "0-Games / 8-Football / 21-Stadium / 65-Kick / 73-Ball / 81-Athlete / \n", "205-Newscaster / 330-Money / 2642-CNBC / \n", "1-Vehicle / 4-Car / 11-Motorsport / 64-Transport / \n", "309-Computer hardware / 700-Hard disk drive / \n", "2-Video game / 35-Minecraft / \n", "20-Trailer / \n", "6-Animation / 995-Attack on Titan / \n", "110-Album / \n", "0-Games / 2-Video game / 20-Trailer / 2277-The Lord of the Rings Online / \n", "1-Vehicle / 4-Car / 691-Cargo / 707-Roof / 2355-Baggage / \n", "3-Concert / \n", "2-Video game / 17-Racing / 127-Xbox 360 / 181-Xbox / 386-Xbox / 677-Game controller / 705-Forza Motorsport / \n", "487-Samsung Electronics / \n", "22-Nature / 42-Fishing / 164-Beach / 222-Surfing / 223-Weather / 233-Ocean / 423-Coast / \n", "1-Vehicle / 18-Outdoor recreation / \n", "0-Games / 1-Vehicle / 124-Tractor / 156-Agriculture / 174-Farm / 617-Maize / 715-Farming Simulator / 984-Silage / \n", "12-Food / 26-Cooking / 32-Recipe / 226-Dessert / 232-Cake / 236-Baking / 578-Cookie / 1024-Culinary art / 1230-Italian food / 1322-Biscuit / 3735-Tiramisu / \n", "66-Bollywood / \n", "66-Bollywood / 78-Wedding / \n", "10-Animal / 48-Pet / 71-Dog / 179-Kitchen / 182-Puppy / 770-Banana / 1839-Rottweiler / \n", "3-Concert / \n", "3-Concert / 7-Musician / 16-Performance art / \n", "8-Football / 623-Coach / \n", "0-Games / 2-Video game / 19-PC game / 1630-Warframe / \n", "319-Robot / \n", "1017-Yo-yo / \n", "0-Games / 2-Video game / 19-PC game / 34-Action-adventure game / 3059-Dead Rising 2 / 3308-Dead Rising / \n", "1-Vehicle / 4-Car / 41-Sports car / 234-Ford / 531-Hatchback / 1634-Ford Fiesta / \n", "0-Games / 51-Strategy video game / 373-Clash of Clans / \n", "42-Fishing / 164-Beach / 423-Coast / 3145-Spinnerbait / \n", "46-Choir / \n", "28-Fashion / 45-Cosmetics / 49-School / 4632-School supplies / \n", "22-Nature / 1183-Karaoke box / \n", "1-Vehicle / 4-Car / 11-Motorsport / 84-Snow / 135-Drifting / 680-Subaru / 1286-Subaru / \n", "12-Food / 442-Bread / \n", "0-Games / 2-Video game / 33-Weapon / \n", "12-Food / 795-Cupcake / \n", "260-Prayer / \n", "14-Guitar / 63-Acoustic guitar / \n", "3-Concert / 7-Musician / 30-Drummer / 39-Drums / 44-Drums / \n", "6-Animation / 15-Cartoon / \n", "10-Animal / \n", "0-Games / 2-Video game / 55-Video game console / 413-PlayStation 2 / 4201-Jikkyō Powerful Pro Yakyū series / \n", "66-Bollywood / \n", "0-Games / 1-Vehicle / 2-Video game / 4-Car / 34-Action-adventure game / 40-Road / 118-Grand Theft Auto V / 297-Grand Theft Auto IV / \n", "3-Concert / 7-Musician / 30-Drummer / \n", "0-Games / 8-Football / 21-Stadium / 65-Kick / 73-Ball / 81-Athlete / \n", "0-Games / 60-Basketball / 168-Basketball moves / 1399-Streetball / \n", "1-Vehicle / 2-Video game / 4-Car / \n", "6-Animation / 15-Cartoon / 123-Naruto / 185-Dragon Ball / 396-One Piece / \n", "3172-Lyre / \n", "8-Football / 21-Stadium / 77-Arena / \n", "8-Football / \n", "0-Games / 51-Strategy video game / 165-League of Legends / \n", "23-Mobile phone / 29-Smartphone / 37-Gadget / 47-Personal computer / 145-Tablet computer / 621-Asus / \n", "42-Fishing / 147-River / 1542-Catfish / \n", "0-Games / 8-Football / 358-Rugby football / \n", "9-Music video / \n", "209-University / \n", "0-Games / 2-Video game / 3717-Suikoden / \n", "59-Winter sport / 83-Skateboarding / 108-Skateboard / 281-Snowboarding / 4321-Mountainboarding / \n", "162-Medicine / 475-Cheese / 600-Flour / 867-Tea / 1224-Cola / \n", "6-Animation / 15-Cartoon / 31-Disc jockey / 290-Pony / 343-My Little Pony / \n", "25-Toy / 140-LEGO / 4341-Anti-aircraft warfare / \n", "1-Vehicle / 4-Car / 27-Motorcycle / \n", "3-Concert / 7-Musician / 16-Performance art / \n", "9-Music video / 78-Wedding / 153-Dress / 190-Bride / \n", "9-Music video / \n", "1-Vehicle / 336-Scooter / \n", "59-Winter sport / 112-Ice skating / 142-Hockey / 259-Fire / \n", "23-Mobile phone / 29-Smartphone / 37-Gadget / 101-Telephone / \n", "0-Games / \n", "28-Fashion / 894-Jacket / 2188-Coat / \n", "465-Microphone / \n", "28-Fashion / 574-Elevator / 804-Shopping mall / \n", "0-Games / 8-Football / 21-Stadium / \n", "25-Toy / \n", "5-Dance / 6-Animation / 10-Animal / 15-Cartoon / 71-Dog / \n", "3-Concert / 7-Musician / \n", "0-Games / 2-Video game / 6-Animation / 55-Video game console / 320-Super Smash Bros. / 697-Super Smash Bros. for Nintendo 3DS and Wii U / \n", "0-Games / 85-Combat / 117-Boxing / \n", "0-Games / 2-Video game / 20-Trailer / 55-Video game console / 111-PlayStation 3 / 127-Xbox 360 / 176-PlayStation 4 / 314-Xbox One / \n", "93-Comedy / \n", "3-Concert / 7-Musician / 13-Musical ensemble / \n", "0-Games / 837-Roblox / \n", "2-Video game / 47-Personal computer / 4549-Fruit Ninja / \n", "0-Games / 8-Football / 21-Stadium / \n", "3-Concert / 7-Musician / 14-Guitar / 24-String instrument / \n", "25-Toy / 3151-Kermit the Frog / \n", "1-Vehicle / 25-Toy / 74-Truck / 140-LEGO / 822-Crane / \n", "47-Personal computer / 131-Computer / 228-Laptop / 504-Web page / 525-Printer / \n", "5-Dance / 16-Performance art / 95-Talent show / \n", "47-Personal computer / 53-Keyboard / 309-Computer hardware / 513-Computer keyboard / \n", "3-Concert / 7-Musician / 13-Musical ensemble / 38-Orchestra / 46-Choir / \n", "30-Drummer / 39-Drums / 44-Drums / 128-Cymbal / 146-Snare drum / \n", "1-Vehicle / 11-Motorsport / 17-Racing / 87-Boat / 498-Motorboat / 2732-Offshore powerboat racing / \n", "163-Running / 466-Gift / \n", "9-Music video / \n", "0-Games / 8-Football / 21-Stadium / 54-Highlight film / \n", "0-Games / 2-Video game / 10-Animal / 1116-Goat / \n", "0-Games / 2-Video game / 34-Action-adventure game / 111-PlayStation 3 / 118-Grand Theft Auto V / 127-Xbox 360 / 176-PlayStation 4 / 314-Xbox One / \n", "492-Floor / \n", "3-Concert / 46-Choir / 318-Church / \n", "1-Vehicle / 62-Train / 64-Transport / 103-Rail transport / 119-Locomotive / 152-Track / 161-Train station / \n", "6-Animation / 15-Cartoon / \n", "1-Vehicle / 50-Aircraft / 273-News program / \n", "3-Concert / \n", "47-Personal computer / 131-Computer / \n", "0-Games / 2-Video game / 6-Animation / 514-Tekken / 797-Tekken / 1273-Tekken 6 / \n", "0-Games / 54-Highlight film / 60-Basketball / 255-Slam dunk / \n", "0-Games / 2-Video game / 1613-Crazyracing Kartrider / \n", "25-Toy / 140-LEGO / 1007-Portal / 3597-GLaDOS / \n", "1868-Preacher / \n", "330-Money / \n", "36-Piano / 53-Keyboard / \n", "66-Bollywood / \n", "1-Vehicle / 9-Music video / \n", "162-Medicine / \n", "0-Games / 2-Video game / 159-World of Warcraft / 171-Warcraft / \n", "0-Games / 2-Video game / 34-Action-adventure game / 1447-Skylanders / 2321-Skylanders: Spyro's Adventure / 2560-Skylanders: Swap Force / 2688-Skylanders: Giants / \n", "20-Trailer / \n", "1-Vehicle / 50-Aircraft / 82-Airplane / 129-Aviation / 151-Landing / \n", "238-Nail / 300-Nail art / 304-Nail polish / 307-Hand / 324-Finger / 338-Manicure / \n", "10-Animal / 18-Outdoor recreation / 42-Fishing / 80-Horse / 287-Human swimming / \n", "36-Piano / 53-Keyboard / 104-Pianist / 107-Musical keyboard / 592-Synthesia / \n", "23-Mobile phone / 29-Smartphone / 37-Gadget / 101-Telephone / \n", "130-Gymnastics / \n", "10-Animal / 48-Pet / 71-Dog / 182-Puppy / 636-Rabbit / 824-Terrier / \n", "10-Animal / 747-Retail / \n", "0-Games / 411-MapleStory / \n", "8-Football / 21-Stadium / 77-Arena / \n", "12-Food / 227-Water / 499-Cloud / \n", "93-Comedy / \n", "126-The Walt Disney Company / \n", "1-Vehicle / \n", "1656-Projector / \n", "60-Basketball / 1608-Injury / \n", "0-Games / 2-Video game / 387-Spider-Man / \n", "9-Music video / \n", "45-Cosmetics / 56-Hair / 106-Hairstyle / 173-Eye shadow / 196-Eye / 218-Eyelash / \n", "1-Vehicle / 4-Car / 41-Sports car / \n", "2152-Selfie / \n", "1-Vehicle / 9-Music video / 27-Motorcycle / 121-Photography / \n", "3-Concert / \n", "0-Games / 25-Toy / 455-Unidentified flying object / 1668-Claw crane / \n", "1-Vehicle / 4-Car / 11-Motorsport / 3229-Opel Kadett / \n", "12-Food / 26-Cooking / 52-Dish / 369-Resort / 389-Restaurant / \n", "843-The Doctor / \n", "3-Concert / 7-Musician / 3489-Rock and Roll Hall of Fame / \n", "641-Mask / 3240-Matryoshka doll / \n", "59-Winter sport / 112-Ice skating / 142-Hockey / \n", "1-Vehicle / 4-Car / 70-Driving / 121-Photography / 2248-Zoom lens / 2487-Canon PowerShot / \n", "22-Nature / 998-Landscape / \n", "122-Weight training / 204-Gym / 214-Muscle / 385-Face / 529-Squat / 532-Barbell / 1055-Back / \n", "7-Musician / 24-String instrument / 150-Violin / 670-Banjo / \n", "1-Vehicle / 25-Toy / 92-Radio-controlled model / 149-Model aircraft / 160-Radio-controlled aircraft / 266-Helicopter / 392-Radio-controlled helicopter / \n", "14-Guitar / 63-Acoustic guitar / \n", "3-Concert / 95-Talent show / \n", "126-The Walt Disney Company / 144-Amusement park / 605-Cruise ship / \n", "3-Concert / 7-Musician / 13-Musical ensemble / \n", "1-Vehicle / 4-Car / 74-Truck / 94-Dashcam / \n", "415-Sewing / 525-Printer / \n", "1-Vehicle / 4-Car / 11-Motorsport / 17-Racing / 74-Truck / 167-Four-wheel drive / 327-Mud / 565-Mud bogging / \n", "66-Bollywood / \n", "45-Cosmetics / 56-Hair / \n", "93-Comedy / \n", "0-Games / 2-Video game / 43-Call of Duty / 216-Call of Duty: Modern Warfare 3 / \n", "1-Vehicle / 4-Car / 11-Motorsport / 17-Racing / 57-Race track / 378-Drag racing / \n", "121-Photography / \n", "1-Vehicle / 4-Car / 11-Motorsport / 17-Racing / 40-Road / 41-Sports car / 57-Race track / 74-Truck / \n", "1-Vehicle / 4-Car / 64-Transport / 259-Fire / 371-Firefighter / \n", "3-Concert / 7-Musician / \n", "0-Games / 2-Video game / 34-Action-adventure game / 1496-LittleBigPlanet 2 / 1672-LittleBigPlanet / \n", "13-Musical ensemble / 30-Drummer / 44-Drums / 2376-Steelpan / \n", "1-Vehicle / 2-Video game / 4-Car / 189-iPad / \n", "9-Music video / \n", "7-Musician / 14-Guitar / 24-String instrument / \n", "12-Food / 26-Cooking / 32-Recipe / 52-Dish / 58-Cuisine / 215-Eating / 301-Meal / 663-Street / 1438-Street food / \n", "3-Concert / 7-Musician / 13-Musical ensemble / 594-Music festival / \n", "12-Food / 86-Plant / 1720-Grape / 2473-Vineyard / \n", "5-Dance / 16-Performance art / 59-Winter sport / 112-Ice skating / 316-Figure skating / 632-Ice dancing / \n", "30-Drummer / \n", "0-Games / 2-Video game / 19-PC game / 33-Weapon / 43-Call of Duty / 422-Call of Duty: Ghosts / \n", "2-Video game / 55-Video game console / 181-Xbox / 314-Xbox One / 386-Xbox / \n", "0-Games / 2-Video game / 237-Final Fantasy / \n", "102-Building / 998-Landscape / \n", "8-Football / 20-Trailer / \n", "3-Concert / 5-Dance / \n", "1087-Donkey Kong / \n", "0-Games / 60-Basketball / \n", "1-Vehicle / 4-Car / 11-Motorsport / 70-Driving / 277-Sport utility vehicle / 1916-Buick / \n", "0-Games / 1-Vehicle / 2-Video game / 33-Weapon / 34-Action-adventure game / 201-Grand Theft Auto: San Andreas / 858-Carl Johnson / 3331-Quarry / \n", "45-Cosmetics / 2183-Stardoll / \n", "0-Games / 8-Football / 21-Stadium / \n", "47-Personal computer / 131-Computer / 228-Laptop / \n", "22-Nature / 85-Combat / 86-Plant / \n", "0-Games / 2-Video game / 20-Trailer / 159-World of Warcraft / 171-Warcraft / \n", "4323-Cricket / \n", "10-Animal / 48-Pet / 132-Bird / 2573-Domestic pigeon / \n", "28-Fashion / 45-Cosmetics / 56-Hair / 193-Eye liner / 195-Mascara / 218-Eyelash / 235-Lipstick / \n", "3-Concert / 7-Musician / 77-Arena / \n", "14-Guitar / 100-Electric guitar / \n", "322-Mixtape / 1721-Tetris / \n", "7-Musician / 14-Guitar / 24-String instrument / 63-Acoustic guitar / \n", "3-Concert / \n", "83-Skateboarding / 108-Skateboard / \n", "3-Concert / 7-Musician / 13-Musical ensemble / 78-Wedding / \n", "28-Fashion / 153-Dress / \n", "8-Football / \n", "1-Vehicle / 62-Train / 119-Locomotive / 161-Train station / \n", "20-Trailer / \n", "8-Football / 21-Stadium / \n", "273-News program / \n", "1-Vehicle / 9-Music video / \n", "122-Weight training / 204-Gym / 214-Muscle / \n", "3-Concert / 7-Musician / 13-Musical ensemble / 16-Performance art / \n", "7-Musician / 36-Piano / 150-Violin / 401-Cello / 450-Viola / \n", "0-Games / 35-Minecraft / 99-Christmas / \n", "1-Vehicle / 4-Car / \n", "78-Wedding / 190-Bride / 590-Wedding dress / \n", "1-Vehicle / 40-Road / 94-Dashcam / 663-Street / \n", "10-Animal / 42-Fishing / 91-Fish / \n", "3-Concert / 7-Musician / 13-Musical ensemble / 24-String instrument / 38-Orchestra / 150-Violin / \n", "7-Musician / 14-Guitar / 24-String instrument / 1231-Mandolin / \n", "10-Animal / 15-Cartoon / 61-Art / 91-Fish / 97-Drawing / \n", "0-Games / 8-Football / \n", "9-Music video / \n", "6-Animation / 1551-The Phantom of the Opera / \n", "1-Vehicle / 4-Car / 11-Motorsport / 17-Racing / 57-Race track / 728-Dirt track racing / \n", "12-Food / 28-Fashion / 56-Hair / 106-Hairstyle / 340-Afro-textured hair / 857-Afro / \n", "117-Boxing / \n", "0-Games / 2-Video game / 19-PC game / 460-Yu-Gi-Oh! Trading Card Game / 494-Card game / \n", "0-Games / 2-Video game / 23-Mobile phone / 37-Gadget / 55-Video game console / 159-World of Warcraft / 171-Warcraft / 189-iPad / \n", "457-Knitting / 546-Stitch / 653-Thread / 1497-Wool / 1536-Woven fabric / 1712-Sweater / \n", "12-Food / 26-Cooking / 32-Recipe / 52-Dish / 58-Cuisine / 662-Pizza / \n", "8-Football / 21-Stadium / 77-Arena / 1247-Penalty kick / \n", "0-Games / 2-Video game / 35-Minecraft / 113-House / \n", "1-Vehicle / 4-Car / 114-Engine / 449-Honda / 552-Honda Civic / \n", "6-Animation / 15-Cartoon / 2339-Swan / \n", "10-Animal / 164-Beach / 3123-Jellyfish / \n", "31-Disc jockey / \n", "1516-Belt / \n", "1-Vehicle / 67-Cycling / 69-Bicycle / 520-Bicycle frame / 645-Road bicycle / \n", "23-Mobile phone / 29-Smartphone / 37-Gadget / 72-iPhone / 101-Telephone / 1009-iPhone 3G / 1032-BlackBerry / \n", "23-Mobile phone / 29-Smartphone / 37-Gadget / 101-Telephone / 178-Samsung Galaxy / 487-Samsung Electronics / 1015-Samsung Galaxy Note series / 1349-USB flash drive / 1727-Samsung Galaxy Note 3 / \n", "10-Animal / 48-Pet / 71-Dog / 4166-Rescue dog / \n", "0-Games / 2-Video game / 1496-LittleBigPlanet 2 / \n", "102-Building / 113-House / 170-Home improvement / 179-Kitchen / 580-Bedroom / 3716-Mobile home / \n", "56-Hair / 1359-Hair iron / \n", "60-Basketball / 3490-Eiffel Tower / \n", "0-Games / 2-Video game / 116-Pokémon / 186-Pokémon / \n", "1-Vehicle / 67-Cycling / 69-Bicycle / 158-Tire / 220-Mountain bike / 659-Van / \n", "59-Winter sport / 112-Ice skating / 316-Figure skating / 632-Ice dancing / \n", "28-Fashion / 2114-Goth subculture / \n", "12-Food / 102-Building / 208-Wood / 259-Fire / 279-Garden / 585-Lawn / 666-Concrete / 2258-Patio / \n", "1-Vehicle / 18-Outdoor recreation / 22-Nature / 42-Fishing / 87-Boat / 147-River / 225-Lake / \n", "12-Food / 720-Wine / 1720-Grape / \n", "8-Football / \n", "1-Vehicle / 208-Wood / 1294-Logging / 3073-Log splitter / \n", "3-Concert / 7-Musician / 30-Drummer / 39-Drums / \n", "425-City / \n", "3-Concert / 5-Dance / 16-Performance art / 1535-Prize / \n", "419-Earth / 448-Planet / 774-Galaxy / \n", "0-Games / 2-Video game / 6-Animation / 15-Cartoon / 768-Diamond / \n", "376-Flute / 2375-Ney / \n", "25-Toy / 126-The Walt Disney Company / 417-Play-Doh / \n", "3-Concert / 7-Musician / 24-String instrument / 356-Star / \n", "743-Edward Cullen / \n", "8-Football / \n", "1-Vehicle / 62-Train / 64-Transport / 94-Dashcam / 103-Rail transport / 152-Track / 161-Train station / \n", "3-Concert / 7-Musician / 16-Performance art / \n", "6-Animation / \n", "75-Wrestling / 777-WWE 2K / \n", "31-Disc jockey / \n", "23-Mobile phone / 29-Smartphone / 101-Telephone / 553-Sony Xperia / 3019-Sony Ericsson Xperia X10 / 3867-Sony Ericsson Xperia arc / 4635-Sony Ericsson Xperia X10 Mini / \n", "1-Vehicle / 4-Car / 891-Mercedes-AMG / \n", "1-Vehicle / 4-Car / 94-Dashcam / \n", "12-Food / 26-Cooking / 32-Recipe / 52-Dish / 58-Cuisine / 109-Cooking show / 179-Kitchen / 210-Cookware and bakeware / \n", "14-Guitar / 24-String instrument / 1754-Rocksmith / \n", "5-Dance / 16-Performance art / 115-Ballet / \n", "35-Minecraft / 242-Paper / 3410-Paper model / \n", "3-Concert / \n", "3-Concert / 7-Musician / 13-Musical ensemble / 14-Guitar / 30-Drummer / 39-Drums / 44-Drums / 1117-Camcorder / \n", "46-Choir / \n", "10-Animal / 48-Pet / 132-Bird / 598-Parrot / 1341-Parakeet / \n", "242-Paper / 370-Origami / 1403-Flower bouquet / \n", "5-Dance / \n", "31-Disc jockey / 322-Mixtape / \n", "9-Music video / 14-Guitar / \n", "163-Running / \n", "0-Games / 141-Microsoft Windows / \n", "0-Games / 8-Football / 25-Toy / \n", "36-Piano / 376-Flute / \n", "12-Food / \n", "1-Vehicle / 4-Car / 757-Boot / 1221-MINI / \n", "293-Slide show / \n", "0-Games / 1-Vehicle / 2-Video game / 4-Car / 11-Motorsport / 17-Racing / 41-Sports car / 413-PlayStation 2 / 581-PlayStation / 3510-Ridge Racer / \n", "6-Animation / \n", "8-Football / 76-Ball / 905-Flash Video / \n", "12-Food / 619-Manufacturing / 720-Wine / \n", "1-Vehicle / 4-Car / 27-Motorcycle / 1380-Kawasaki motorcycles / 1951-Sport bike / 2494-Ninja ZX-6R / \n", "12-Food / \n", "7-Musician / 14-Guitar / 24-String instrument / 63-Acoustic guitar / \n", "6-Animation / 4555-Amagami / \n", "3-Concert / \n", "10-Animal / 279-Garden / \n", "5-Dance / 16-Performance art / 95-Talent show / \n", "0-Games / 489-Board game / 799-Chess / 922-Tabletop game / \n", "256-Tennis / 469-Bag / 974-Racket / \n", "1-Vehicle / 50-Aircraft / 82-Airplane / 129-Aviation / 139-Airport / 151-Landing / 157-Jet aircraft / 175-Airliner / 188-Airline / 254-Runway / 472-Boeing 737 / \n", "217-Hunting / 1716-Rat / \n", "0-Games / 1-Vehicle / 4-Car / 34-Action-adventure game / 40-Road / 41-Sports car / 297-Grand Theft Auto IV / 2511-Pontiac / \n", "113-House / 3779-Craigslist Inc. / \n", "3-Concert / 68-Lighting / \n", "1-Vehicle / 88-Machine / 124-Tractor / 156-Agriculture / \n", "3-Concert / 5-Dance / 16-Performance art / 49-School / \n", "6-Animation / 15-Cartoon / 325-Vampire / \n", "1-Vehicle / 18-Outdoor recreation / 42-Fishing / 87-Boat / 114-Engine / 2338-Inflatable boat / \n", "8-Football / \n", "12-Food / \n", "1-Vehicle / 4-Car / 27-Motorcycle / 94-Dashcam / \n", "0-Games / 2-Video game / 111-PlayStation 3 / \n", "18-Outdoor recreation / 42-Fishing / 147-River / 478-Fishing rod / 698-Fisherman / 1917-Fishing tackle / \n", "1-Vehicle / 4-Car / 11-Motorsport / 17-Racing / 41-Sports car / 1301-Nissan Skyline / \n", "6-Animation / 15-Cartoon / \n", "10-Animal / 80-Horse / 290-Pony / \n", "5-Dance / \n", "0-Games / 2-Video game / 33-Weapon / 1412-Half-Life / \n", "0-Games / 240-Arcade game / 687-Taiko no Tatsujin / \n", "36-Piano / 53-Keyboard / 107-Musical keyboard / 265-Electronic keyboard / \n", "14-Guitar / 63-Acoustic guitar / \n", "7-Musician / 14-Guitar / 24-String instrument / 63-Acoustic guitar / \n", "1-Vehicle / 11-Motorsport / 17-Racing / 27-Motorcycle / 57-Race track / \n", "0-Games / 2-Video game / 1027-Terraria / \n", "15-Cartoon / 61-Art / 97-Drawing / 99-Christmas / 466-Gift / 528-Pencil / 538-Illustration / \n", "93-Comedy / \n", "3-Concert / \n", "0-Games / 2-Video game / 310-The Sims / 328-Simulation video game / 1553-The Sims 4 / 1713-The Sims / \n", "0-Games / 2-Video game / \n", "1-Vehicle / 50-Aircraft / 82-Airplane / 139-Airport / 151-Landing / 175-Airliner / 188-Airline / 472-Boeing 737 / \n", "1-Vehicle / 4-Car / 11-Motorsport / 18-Outdoor recreation / 27-Motorcycle / 67-Cycling / 69-Bicycle / 172-Motocross / 1520-Minibike / \n", "0-Games / 2-Video game / 33-Weapon / 834-Combat Arms / \n", "31-Disc jockey / \n", "75-Wrestling / 85-Combat / \n", "347-Climbing / 828-Rock climbing / 1416-Cliff / \n", "9-Music video / \n", "36-Piano / 53-Keyboard / \n", "0-Games / 2-Video game / 19-PC game / 33-Weapon / 43-Call of Duty / 96-Soldier / 138-Call of Duty: Black Ops II / \n", "12-Food / 26-Cooking / 32-Recipe / 475-Cheese / 534-Milk / \n", "126-The Walt Disney Company / 144-Amusement park / 576-Walt Disney World / 1222-Fountain / \n", "0-Games / 2-Video game / 35-Minecraft / \n", "0-Games / 1876-Beatmania IIDX / 2063-Beatmania / \n", "287-Human swimming / \n", "3-Concert / 7-Musician / 14-Guitar / \n", "1-Vehicle / 67-Cycling / 69-Bicycle / \n", "1-Vehicle / 4-Car / 41-Sports car / 70-Driving / 1763-Lotus Cars / \n", "6-Animation / 15-Cartoon / \n", "1-Vehicle / 4-Car / 11-Motorsport / 27-Motorcycle / 40-Road / 105-Motorcycling / 482-Asphalt / 608-Wheelie / \n", "223-Weather / 419-Earth / 448-Planet / 593-Moon / 654-Sun / 1918-Comet / \n", "0-Games / 2-Video game / 237-Final Fantasy / 1178-Quest / 2667-Final Fantasy IV / \n", "5-Dance / 49-School / \n", "766-President of the United States / \n", "0-Games / 8-Football / 54-Highlight film / 79-American football / \n", "14-Guitar / 24-String instrument / 63-Acoustic guitar / \n", "12-Food / 26-Cooking / 32-Recipe / 52-Dish / 58-Cuisine / 109-Cooking show / 179-Kitchen / 210-Cookware and bakeware / 226-Dessert / 236-Baking / 437-Cream / 968-Pie / 1208-Coconut / 1753-Mixer / \n", "1-Vehicle / 4-Car / 11-Motorsport / 17-Racing / 20-Trailer / 40-Road / 41-Sports car / 135-Drifting / \n", "18-Outdoor recreation / 163-Running / 327-Mud / \n", "20-Trailer / \n", "3-Concert / 7-Musician / 13-Musical ensemble / \n", "0-Games / 8-Football / 54-Highlight film / 65-Kick / \n", "28-Fashion / 78-Wedding / 153-Dress / 190-Bride / 590-Wedding dress / \n", "5-Dance / \n", "3-Concert / 7-Musician / 13-Musical ensemble / \n", "3-Concert / 133-Nightclub / \n", "3-Concert / 7-Musician / 13-Musical ensemble / 1431-Madison Square Garden / \n", "18-Outdoor recreation / 163-Running / 674-Marathon / \n", "365-Fireworks / \n", "0-Games / 2-Video game / 19-PC game / 51-Strategy video game / 1687-Heroes of Newerth / \n", "1-Vehicle / 62-Train / 64-Transport / 161-Train station / \n", "0-Games / 2-Video game / 35-Minecraft / 55-Video game console / 111-PlayStation 3 / 127-Xbox 360 / 176-PlayStation 4 / 181-Xbox / 317-PlayStation / 488-Door / \n", "10-Animal / 709-Bull / 1373-Torero / \n", "0-Games / 1187-Ragnarok Online / \n", "92-Radio-controlled model / 149-Model aircraft / 160-Radio-controlled aircraft / 365-Fireworks / 393-Unmanned aerial vehicle / \n", "0-Games / 2-Video game / \n", "0-Games / 60-Basketball / 77-Arena / 168-Basketball moves / \n", "0-Games / 2-Video game / 264-Star Wars / 714-Jedi / 1050-Star Wars: The Old Republic / 1882-Warrior / \n", "22-Nature / 35-Minecraft / 1149-Castle / \n", "3-Concert / 5-Dance / 227-Water / \n", "25-Toy / 61-Art / 78-Wedding / 153-Dress / 590-Wedding dress / \n", "6-Animation / 183-Manga / 325-Vampire / 2357-Vampire Knight / \n", "3-Concert / \n", "3-Concert / 16-Performance art / \n", "1-Vehicle / 4-Car / 64-Transport / 74-Truck / 94-Dashcam / \n", "60-Basketball / 85-Combat / \n", "1-Vehicle / 4-Car / 11-Motorsport / 17-Racing / 57-Race track / 135-Drifting / \n", "3-Concert / 30-Drummer / 68-Lighting / \n", "14-Guitar / 24-String instrument / 4165-Washburn Guitars / \n", "86-Plant / 136-Gardening / 1460-Herb / 2795-Root / \n", "3-Concert / 5-Dance / 68-Lighting / 263-Light / 842-Harp / 1375-Laser lighting display / \n", "0-Games / 2-Video game / 34-Action-adventure game / 269-Comic book / 335-Nintendo 3DS / 1092-Game Boy Advance / \n", "12-Food / 26-Cooking / 32-Recipe / 52-Dish / 58-Cuisine / 109-Cooking show / 184-Vegetable / 301-Meal / 454-Sauce / 2187-Toast / \n", "0-Games / 2-Video game / 19-PC game / 33-Weapon / 96-Soldier / 3567-SOCOM / \n", "0-Games / 2912-Flappy Bird / \n", "3-Concert / 7-Musician / \n", "3-Concert / 7-Musician / 13-Musical ensemble / 1003-Tuba / \n", "6-Animation / 15-Cartoon / 434-Ninja / \n", "12-Food / 26-Cooking / 49-School / 301-Meal / \n", "28-Fashion / 2881-Wardrobe / \n", "3-Concert / 7-Musician / 14-Guitar / \n", "9-Music video / 78-Wedding / 190-Bride / \n", "0-Games / 2476-Maze / \n", "20-Trailer / 223-Weather / \n", "12-Food / 227-Water / 1329-Blender / \n", "6-Animation / \n", "9-Music video / \n", "0-Games / 2-Video game / 6-Animation / 514-Tekken / 2812-Kazuya Mishima / 3302-Tekken 5 / 3460-Tekken 5: Dark Resurrection / \n", "1-Vehicle / 11-Motorsport / 27-Motorcycle / 105-Motorcycling / 197-Trail / 213-Forest / 363-Soil / \n", "25-Toy / 89-Comics / 387-Spider-Man / 435-Action figure / 2368-Marvel Universe / \n", "0-Games / 2-Video game / 1137-Crash Bandicoot / 3823-Crash Bandicoot: Warped / \n", "1-Vehicle / 4-Car / 11-Motorsport / 124-Tractor / 252-Tractor pulling / \n", "25-Toy / 140-LEGO / 294-Batman / 884-Lego minifigure / \n", "18-Outdoor recreation / 83-Skateboarding / 108-Skateboard / \n", "1-Vehicle / 4-Car / 270-Sedan / 332-Mercedes-Benz / 1310-Mercedes-Benz C-Class / 3826-Mercedes-Benz A-Class / \n", "253-Carnival / \n", "7-Musician / 14-Guitar / 24-String instrument / \n", "99-Christmas / 558-Santa Claus / \n", "3-Concert / \n", "0-Games / 23-Mobile phone / 29-Smartphone / 37-Gadget / 72-iPhone / 189-iPad / 272-iPod / 331-iPod touch / \n", "18-Outdoor recreation / 67-Cycling / 69-Bicycle / 197-Trail / 220-Mountain bike / 283-Rock / 284-Mountain biking / \n", "1017-Yo-yo / \n", "5-Dance / \n", "46-Choir / \n", "0-Games / 317-PlayStation / \n", "9-Music video / \n", "743-Edward Cullen / \n", "6-Animation / 15-Cartoon / \n", "12-Food / 52-Dish / \n", "5-Dance / 59-Winter sport / 81-Athlete / 112-Ice skating / 316-Figure skating / \n", "0-Games / 1-Vehicle / 2-Video game / 4-Car / 11-Motorsport / 17-Racing / 41-Sports car / 70-Driving / 135-Drifting / 344-Coupé / 352-Need for Speed / 1600-Need for Speed: World / 2420-Audi TT / \n", "36-Piano / 53-Keyboard / 104-Pianist / 107-Musical keyboard / 1710-Pirates of the Caribbean / \n", "1-Vehicle / 64-Transport / 74-Truck / 1611-Garbage truck / \n", "0-Games / 2-Video game / 55-Video game console / 413-PlayStation 2 / 2142-Sega Saturn / \n", "0-Games / 2-Video game / 19-PC game / 33-Weapon / 229-Halo / 464-First-person Shooter / \n", "12-Food / 32-Recipe / 354-Egg / 367-Drink / 534-Milk / 786-Juice / \n", "2-Video game / 14-Guitar / 176-PlayStation 4 / \n", "25-Toy / 68-Lighting / 140-LEGO / 263-Light / 1375-Laser lighting display / \n", "0-Games / 2-Video game / 19-PC game / 51-Strategy video game / \n", "55-Video game console / 414-Computer monitor / 1637-GameCube / 2166-Gamepad / \n", "5-Dance / 49-School / 95-Talent show / 191-Cheerleading / 306-High school / \n", "12-Food / 88-Machine / 812-Plastic / 2031-Molding / \n", "3-Concert / 5-Dance / 16-Performance art / 38-Orchestra / \n", "3-Concert / 7-Musician / 13-Musical ensemble / 14-Guitar / \n", "88-Machine / 400-Woodturning / 462-Table / \n", "25-Toy / 89-Comics / 637-Five Nights at Freddy's / 1094-Puppet / \n", "1-Vehicle / 4-Car / 27-Motorcycle / 105-Motorcycling / 336-Scooter / \n", "12-Food / 26-Cooking / 32-Recipe / 52-Dish / 58-Cuisine / 109-Cooking show / 179-Kitchen / 239-Roasting / 391-Chicken meat / \n", "1-Vehicle / 4-Car / 618-Fiat Automobiles / \n", "0-Games / 8-Football / \n", "36-Piano / 53-Keyboard / 107-Musical keyboard / 265-Electronic keyboard / 446-Sheet music / 592-Synthesia / \n", "3-Concert / 7-Musician / 30-Drummer / \n", "3-Concert / 7-Musician / 1659-Walmart / \n", "1-Vehicle / 4-Car / 3353-Chevrolet Monte Carlo / \n", "2-Video game / 55-Video game console / 677-Game controller / 2166-Gamepad / 2296-Joystick / \n", "5-Dance / \n", "6-Animation / \n", "0-Games / 54-Highlight film / 142-Hockey / 1610-Field hockey / \n", "54-Highlight film / 79-American football / \n", "49-School / 657-Autumn / \n", "1-Vehicle / 82-Airplane / \n", "5-Dance / 79-American football / \n", "0-Games / 59-Winter sport / 77-Arena / 112-Ice skating / 142-Hockey / 312-Ice rink / \n", "49-School / 776-Kindergarten / \n", "162-Medicine / \n", "28-Fashion / 45-Cosmetics / 56-Hair / 173-Eye shadow / 193-Eye liner / 195-Mascara / 196-Eye / 218-Eyelash / \n", "14-Guitar / 63-Acoustic guitar / \n", "164-Beach / \n", "162-Medicine / \n", "365-Fireworks / \n", "1-Vehicle / 87-Boat / 949-Rowing / 2678-Dragon boat / \n", "36-Piano / \n", "23-Mobile phone / 29-Smartphone / 37-Gadget / 141-Microsoft Windows / \n", "14-Guitar / \n", "0-Games / 2-Video game / 34-Action-adventure game / 398-The Legend of Zelda / 1228-The Legend of Zelda: Ocarina of Time / \n", "3-Concert / 7-Musician / 13-Musical ensemble / 30-Drummer / 39-Drums / 44-Drums / \n", "6-Animation / 419-Earth / 448-Planet / 593-Moon / 712-Outer space / 1181-Mars / 1508-Spacecraft / 2942-Saturn / \n", "3-Concert / 7-Musician / 13-Musical ensemble / 38-Orchestra / 46-Choir / \n", "9-Music video / \n", "60-Basketball / \n", "14-Guitar / \n", "1-Vehicle / 4-Car / 11-Motorsport / 74-Truck / 1518-Chevrolet Silverado / \n", "14-Guitar / \n", "0-Games / 2-Video game / 19-PC game / 3840-The Binding of Isaac / \n", "1-Vehicle / 62-Train / 64-Transport / 103-Rail transport / 119-Locomotive / 143-Railroad car / 152-Track / 161-Train station / 691-Cargo / 821-Rail freight transport / 2413-Union Pacific Railroad / \n", "3-Concert / 7-Musician / \n", "22-Nature / \n", "1-Vehicle / 50-Aircraft / 92-Radio-controlled model / 149-Model aircraft / 160-Radio-controlled aircraft / 266-Helicopter / 392-Radio-controlled helicopter / \n", "0-Games / 2-Video game / 326-Metin2 / \n", "0-Games / 8-Football / \n", "1-Vehicle / 4-Car / 11-Motorsport / 17-Racing / 41-Sports car / 135-Drifting / 705-Forza Motorsport / 790-Nissan GT-R / 1301-Nissan Skyline / 2105-Forza Horizon / 2247-Nissan Skyline GT-R / \n", "18-Outdoor recreation / 22-Nature / 155-Camera / 1855-Video camera / \n", "1369-Necktie / \n", "1-Vehicle / 11-Motorsport / \n", "5-Dance / 16-Performance art / \n", "0-Games / 2-Video game / 90-Sports game / 180-FIFA 15 / \n", "1-Vehicle / 42-Fishing / 87-Boat / \n", "6-Animation / 15-Cartoon / 1164-Cartoon Network / 3478-Meme / \n", "448-Planet / 593-Moon / 2756-Full moon / \n", "0-Games / 2-Video game / 887-King Koopa / 1170-Super Mario World / \n", "0-Games / 8-Football / 79-American football / \n", "296-Saxophone / \n", "20-Trailer / \n", "10-Animal / 18-Outdoor recreation / 22-Nature / 71-Dog / 1603-Fox / \n", "122-Weight training / \n", "12-Food / 86-Plant / 348-Fruit / \n", "3-Concert / 68-Lighting / \n", "37-Gadget / 47-Personal computer / 131-Computer / 145-Tablet computer / 228-Laptop / 414-Computer monitor / \n", "18-Outdoor recreation / 365-Fireworks / \n", "5-Dance / 16-Performance art / \n", "1-Vehicle / 11-Motorsport / 17-Racing / 27-Motorcycle / 67-Cycling / 69-Bicycle / \n", "3-Concert / 5-Dance / 98-Festival / \n", "5-Dance / \n", "5-Dance / 49-School / \n", "1-Vehicle / 6-Animation / 25-Toy / 62-Train / 507-Thomas the Tank Engine / \n", "12-Food / 342-Windows Media Video / 1088-Infomercial / \n", "10-Animal / 166-Cat / \n", "0-Games / 2-Video game / 34-Action-adventure game / \n", "0-Games / 2-Video game / 43-Call of Duty / 134-Call of Duty: Black Ops / 138-Call of Duty: Black Ops II / \n", "7-Musician / 13-Musical ensemble / 98-Festival / \n", "5-Dance / 9-Music video / \n", "78-Wedding / \n", "1-Vehicle / 50-Aircraft / 82-Airplane / 149-Model aircraft / 223-Weather / 778-Gliding / 1263-Glider / 1317-Glider / \n", "5-Dance / 9-Music video / 16-Performance art / \n", "1601-Social media / \n", "9-Music video / \n", "85-Combat / 117-Boxing / \n", "3-Concert / \n", "7-Musician / 115-Ballet / \n", "0-Games / 791-Magic: The Gathering / 1701-French fries / \n", "259-Fire / 371-Firefighter / \n", "3-Concert / 7-Musician / \n", "0-Games / 2-Video game / 1543-Crysis / \n", "5-Dance / 115-Ballet / \n", "28-Fashion / 45-Cosmetics / 835-Backpack / \n", "46-Choir / \n", "1726-Harlem Shake (meme) / \n", "0-Games / 1-Vehicle / 2-Video game / 4-Car / 11-Motorsport / 17-Racing / 41-Sports car / 57-Race track / 111-PlayStation 3 / \n", "28-Fashion / 206-Shoe / 333-Nike; Inc. / 368-Sneakers / 1423-Nike Air Max / \n", "45-Cosmetics / 173-Eye shadow / 193-Eye liner / 235-Lipstick / 291-Rouge / \n", "0-Games / 2-Video game / 35-Minecraft / 748-Iron Man / 4401-Iron Man's armor / \n", "9-Music video / \n", "14-Guitar / 24-String instrument / 2738-Fender Jazz Bass / \n", "0-Games / 19-PC game / 35-Minecraft / 321-Roller coaster / \n", "1-Vehicle / 50-Aircraft / 82-Airplane / 129-Aviation / 157-Jet aircraft / 388-Fighter aircraft / 758-Air force / \n", "14-Guitar / \n", "6-Animation / 25-Toy / 816-Gundam / \n", "8-Football / \n", "28-Fashion / 153-Dress / 495-Runway / \n", "136-Gardening / 1863-Fertilizer / \n", "287-Human swimming / 1711-Synchronized swimming / \n", "0-Games / 8-Football / 21-Stadium / 54-Highlight film / 65-Kick / 73-Ball / \n", "0-Games / 85-Combat / 412-Monster / \n", "355-Comedian / \n", "23-Mobile phone / 101-Telephone / \n", "979-Star Trek / \n", "467-Chipmunk / 846-Squirrel / \n", "1-Vehicle / 11-Motorsport / 27-Motorcycle / \n", "144-Amusement park / 287-Human swimming / 329-Swimming pool / 1053-Water park / 2360-Water polo / \n", "12-Food / 25-Toy / 582-Candy / 3571-Lollipop / \n", "3-Concert / 68-Lighting / \n", "31-Disc jockey / \n", "14-Guitar / 24-String instrument / 63-Acoustic guitar / \n", "0-Games / 2-Video game / 6-Animation / 20-Trailer / 664-CrossFire / \n", "5-Dance / \n", "0-Games / 2-Video game / 6-Animation / 15-Cartoon / 89-Comics / 123-Naruto / 591-Naruto: Ultimate Ninja / \n", "12-Food / 31-Disc jockey / 348-Fruit / 1193-Orange / \n", "0-Games / 1-Vehicle / 2-Video game / 1613-Crazyracing Kartrider / \n", "9-Music video / \n", "126-The Walt Disney Company / \n", "45-Cosmetics / 282-Skin / \n", "298-Television advertisement / 337-Advertising / \n", "121-Photography / 155-Camera / 721-Digital SLR / \n", "31-Disc jockey / \n", "42-Fishing / 469-Bag / \n", "523-Cattle / \n", "25-Toy / 242-Paper / 324-Finger / 370-Origami / 1094-Puppet / \n", "8-Football / \n", "93-Comedy / 364-Sitcom / \n", "0-Games / 54-Highlight film / 60-Basketball / 168-Basketball moves / \n", "12-Food / 26-Cooking / 32-Recipe / 52-Dish / 58-Cuisine / 109-Cooking show / 210-Cookware and bakeware / 239-Roasting / 354-Egg / 500-Rice / 1295-Stir frying / 2343-Caridean Shrimp / \n", "0-Games / 2-Video game / 19-PC game / 159-World of Warcraft / 171-Warcraft / \n", "920-Chair / \n", "122-Weight training / \n", "8-Football / 1016-Glove / \n", "60-Basketball / \n", "1-Vehicle / 4-Car / 67-Cycling / 69-Bicycle / 608-Wheelie / \n", "366-Dragon / 756-Tibia / \n", "7-Musician / 14-Guitar / 63-Acoustic guitar / \n", "3304-Look-alike / \n", "14-Guitar / \n", "3-Concert / 14-Guitar / \n", "3-Concert / 444-Family / \n", "7-Musician / 13-Musical ensemble / 38-Orchestra / 150-Violin / 230-Brass instrument / \n", "6-Animation / 15-Cartoon / 183-Manga / 708-Mermaid / 996-Princess / \n", "3-Concert / \n", "125-Television / 1348-Panasonic Corporation / 2786-3D television / \n", "567-Compact disc / \n", "23-Mobile phone / 29-Smartphone / 37-Gadget / 178-Samsung Galaxy / 487-Samsung Electronics / 1207-Samsung Galaxy S5 / 1525-Wallet / 3408-Samsung Galaxy S4 Mini / \n", "3-Concert / \n", "6-Animation / 25-Toy / 200-Doll / \n", "1-Vehicle / 4-Car / 11-Motorsport / 27-Motorcycle / \n", "25-Toy / 417-Play-Doh / 684-Ice cream / 723-Pig / \n", "0-Games / 2-Video game / 25-Toy / 140-LEGO / 1302-The Lego Group / 1478-Lego Ninjago / \n", "3-Concert / \n", "0-Games / 8-Football / 21-Stadium / 54-Highlight film / 73-Ball / \n", "12-Food / 26-Cooking / 32-Recipe / 226-Dessert / 232-Cake / 236-Baking / 442-Bread / 454-Sauce / 534-Milk / 560-Sugar / 1208-Coconut / 3003-Roll / 3728-Coconut milk / 4022-Bun / \n", "0-Games / 35-Minecraft / \n", "363-Soil / \n", "14-Guitar / 24-String instrument / 63-Acoustic guitar / \n", "0-Games / 2-Video game / 1966-The Lord of the Rings / 2277-The Lord of the Rings Online / \n", "68-Lighting / 263-Light / 1318-Flashlight / 3117-Match / 3851-Laser pointer / \n", "1-Vehicle / 4-Car / 84-Snow / 459-Plough / 1548-Snowplow / \n", "67-Cycling / 69-Bicycle / 570-Road bicycle racing / \n", "3-Concert / 7-Musician / \n", "0-Games / 2-Video game / 34-Action-adventure game / 243-Sonic the Hedgehog / 251-Sonic the Hedgehog / 1860-Sonic Adventure / \n", "8-Football / \n", "36-Piano / 53-Keyboard / \n", "22-Nature / \n", "9-Music video / \n", "1-Vehicle / 11-Motorsport / 17-Racing / 172-Motocross / \n", "0-Games / 2-Video game / 310-The Sims / 328-Simulation video game / 1553-The Sims 4 / 1713-The Sims / \n", "6-Animation / 61-Art / 1206-Cinema 4D / \n", "10-Animal / \n", "25-Toy / 243-Sonic the Hedgehog / 251-Sonic the Hedgehog / \n", "3-Concert / \n", "1-Vehicle / 18-Outdoor recreation / 59-Winter sport / 84-Snow / 610-Snowmobile / 1210-Sled / \n", "1-Vehicle / 87-Boat / 1514-Crab / \n", "12-Food / 215-Eating / 301-Meal / 323-Toddler / \n", "139-Airport / \n", "1-Vehicle / 50-Aircraft / 151-Landing / 1992-Space Shuttle / \n", "0-Games / 2-Video game / 1605-Second Life / \n", "6-Animation / 15-Cartoon / 1133-Kirby / \n", "0-Games / 51-Strategy video game / 791-Magic: The Gathering / \n", "0-Games / 2-Video game / 19-PC game / 33-Weapon / 96-Soldier / 834-Combat Arms / \n", "67-Cycling / 69-Bicycle / 261-Tool / 1833-Hummer / \n", "5-Dance / 95-Talent show / \n", "30-Drummer / 39-Drums / 44-Drums / 128-Cymbal / 146-Snare drum / \n", "6-Animation / 15-Cartoon / 538-Illustration / \n", "1-Vehicle / 11-Motorsport / 17-Racing / 27-Motorcycle / 57-Race track / 172-Motocross / \n", "22-Nature / 499-Cloud / 654-Sun / \n", "0-Games / 2-Video game / 55-Video game console / 199-Wii / 1118-Kirby / 1133-Kirby / 4434-Kirby's Return to Dream Land / 4449-King Dedede / \n", "14-Guitar / \n", "12-Food / 992-Rocket / 1489-Coffeemaker / 1823-Espresso machine / 2057-Cappuccino / \n", "1-Vehicle / 4-Car / 345-Loudspeaker / 484-Subwoofer / \n", "75-Wrestling / \n", "3-Concert / 7-Musician / 13-Musical ensemble / 38-Orchestra / 46-Choir / \n", "6-Animation / 251-Sonic the Hedgehog / 538-Illustration / \n", "23-Mobile phone / 29-Smartphone / 131-Computer / 1032-BlackBerry / \n", "0-Games / \n", "10-Animal / 20-Trailer / 48-Pet / 588-Reptile / 1444-Lizard / 3576-Iguana / \n", "7-Musician / 9-Music video / \n", "3-Concert / 7-Musician / 13-Musical ensemble / 38-Orchestra / 49-School / 203-Marching band / 230-Brass instrument / \n", "5-Dance / 16-Performance art / \n", "89-Comics / 269-Comic book / 313-Book / \n", "28-Fashion / 45-Cosmetics / 173-Eye shadow / 193-Eye liner / \n", "88-Machine / 1751-Image scanner / \n", "59-Winter sport / 177-Skiing / 305-Ski / \n", "5-Dance / 9-Music video / \n", "3-Concert / 16-Performance art / \n", "7-Musician / 14-Guitar / 24-String instrument / 63-Acoustic guitar / 2746-Twelve-string guitar / \n", "20-Trailer / \n", "0-Games / 60-Basketball / 168-Basketball moves / \n", "2062-Kidney / \n", "162-Medicine / 1375-Laser lighting display / \n", "0-Games / 2-Video game / 19-PC game / 241-Fighting game / 486-M.U.G.E.N / 764-The King of Fighters / 2714-Iori Yagami / \n", "10-Animal / 18-Outdoor recreation / 217-Hunting / \n", "1-Vehicle / 4-Car / 59-Winter sport / 112-Ice skating / 142-Hockey / \n", "23-Mobile phone / 429-Website / \n", "298-Television advertisement / \n", "322-Mixtape / \n", "1-Vehicle / 11-Motorsport / 124-Tractor / 252-Tractor pulling / \n", "5-Dance / 319-Robot / \n", "1-Vehicle / 11-Motorsport / 27-Motorcycle / 105-Motorcycling / 207-Exhaust system / 1867-Honda CB600F / \n", "1-Vehicle / 4-Car / 198-Off-road vehicle / 277-Sport utility vehicle / 3197-Citroën C4 / 3439-Compact car / \n", "3-Concert / 7-Musician / 13-Musical ensemble / 38-Orchestra / 230-Brass instrument / 1003-Tuba / \n", "6-Animation / 593-Moon / 1744-Black Butler / \n", "1-Vehicle / 4-Car / 12-Food / \n", "6-Animation / 15-Cartoon / 185-Dragon Ball / 244-Goku / \n", "361-Flamenco / \n", "3-Concert / 7-Musician / 14-Guitar / 39-Drums / 44-Drums / \n", "3-Concert / \n", "1-Vehicle / 18-Outdoor recreation / 67-Cycling / 69-Bicycle / 377-BMX bike / \n", "0-Games / 55-Video game console / 292-Handheld game console / \n", "7-Musician / 14-Guitar / 24-String instrument / 63-Acoustic guitar / \n", "5-Dance / 113-House / \n", "0-Games / 60-Basketball / \n", "144-Amusement park / 321-Roller coaster / 334-Amusement ride / \n", "10-Animal / 18-Outdoor recreation / 22-Nature / 42-Fishing / 91-Fish / 147-River / 902-Fishing bait / \n", "0-Games / 51-Strategy video game / 165-League of Legends / 2004-Riot Games / \n", "0-Games / 2-Video game / 8-Football / 21-Stadium / 79-American football / 90-Sports game / 176-PlayStation 4 / 314-Xbox One / 561-Madden NFL / 1813-Madden NFL 13 / \n", "0-Games / 1-Vehicle / 2-Video game / 4-Car / 34-Action-adventure game / 118-Grand Theft Auto V / 1923-Niko Bellic / \n", "13-Musical ensemble / 98-Festival / 203-Marching band / 250-Parade / 253-Carnival / \n", "8-Football / 54-Highlight film / \n", "5-Dance / \n", "1-Vehicle / 4-Car / 3758-Toyota RAV4 / \n", "359-Harry Potter (Literary Series) / \n", "9-Music video / \n", "0-Games / 2-Video game / 229-Halo / 470-Halo 3 / 2240-Halo 5: Guardians / \n", "0-Games / 2-Video game / 51-Strategy video game / 2841-Might and Magic / 3468-Heroes of Might and Magic / \n", "0-Games / 2-Video game / 34-Action-adventure game / 111-PlayStation 3 / 127-Xbox 360 / 176-PlayStation 4 / 314-Xbox One / 948-Far Cry / 1683-Far Cry 4 / 3618-Far Cry / \n", "3-Concert / 7-Musician / 13-Musical ensemble / 14-Guitar / 39-Drums / 44-Drums / \n", "1-Vehicle / 4-Car / 40-Road / 70-Driving / 94-Dashcam / 212-Highway / 1593-Digital video recorder / \n", "1-Vehicle / 27-Motorcycle / \n", "0-Games / 8-Football / 21-Stadium / 65-Kick / \n", "12-Food / 26-Cooking / 32-Recipe / 179-Kitchen / 226-Dessert / 232-Cake / 236-Baking / 437-Cream / 475-Cheese / 1641-Cheesecake / 2964-Cream cheese / \n", "3-Concert / \n", "0-Games / 2-Video game / 47-Personal computer / 961-Sega Genesis / \n", "130-Gymnastics / 191-Cheerleading / \n", "12-Food / 26-Cooking / 32-Recipe / 52-Dish / 58-Cuisine / 109-Cooking show / 179-Kitchen / 184-Vegetable / 210-Cookware and bakeware / 542-Potato / 557-Salad / \n", "248-Hotel / \n", "7-Musician / 36-Piano / 53-Keyboard / 104-Pianist / 107-Musical keyboard / \n", "0-Games / 2-Video game / 25-Toy / 125-Television / 140-LEGO / 199-Wii / 264-Star Wars / \n", "1-Vehicle / 4-Car / 27-Motorcycle / 40-Road / 482-Asphalt / \n", "1-Vehicle / 4-Car / 114-Engine / 332-Mercedes-Benz / \n", "1-Vehicle / 62-Train / 103-Rail transport / 119-Locomotive / 143-Railroad car / 428-Rail transport modelling / 1904-Electric locomotive / \n", "6-Animation / 185-Dragon Ball / 244-Goku / \n", "36-Piano / 53-Keyboard / 911-Studio / \n", "3-Concert / \n", "150-Violin / 3396-Ressha Sentai ToQger / \n", "457-Knitting / 476-Textile / 546-Stitch / 653-Thread / \n", "2425-Farmer / \n", "5-Dance / 7-Musician / 14-Guitar / 16-Performance art / 24-String instrument / 361-Flamenco / \n", "110-Album / \n", "14-Guitar / 24-String instrument / 63-Acoustic guitar / \n", "372-Playing card / 746-Card manipulation / \n", "1545-Floristry / \n", "1-Vehicle / 27-Motorcycle / 3934-Radiator / 4495-Suzuki V-Strom 650 / \n", "0-Games / 1-Vehicle / 2-Video game / 4-Car / 34-Action-adventure game / 297-Grand Theft Auto IV / 658-Grand Theft Auto: The Lost and Damned / \n", "6-Animation / 15-Cartoon / 25-Toy / 200-Doll / 439-Barbie / 1441-Mattel / 2251-Ken / \n", "0-Games / 8-Football / \n", "117-Boxing / \n", "45-Cosmetics / \n", "96-Soldier / \n", "1-Vehicle / 18-Outdoor recreation / 87-Boat / 225-Lake / 481-Kayak / 1090-Canoe / 1305-Canoeing / \n", "5-Dance / 115-Ballet / \n", "467-Chipmunk / \n", "0-Games / 2-Video game / 34-Action-adventure game / 380-Assassin's Creed / 1485-Assassin's Creed: Brotherhood / \n", "10-Animal / 22-Nature / 706-Deer / \n", "1-Vehicle / 4-Car / 327-Mud / \n", "1-Vehicle / 2-Video game / 4-Car / 11-Motorsport / 17-Racing / 41-Sports car / 169-Rallying / \n", "9-Music video / 506-Garry's Mod / \n", "243-Sonic the Hedgehog / 2849-Sonic Heroes / \n", "20-Trailer / \n", "0-Games / 1-Vehicle / 2-Video game / 4-Car / 34-Action-adventure game / 921-Grand Theft Auto: Vice City / \n", "1-Vehicle / 87-Boat / \n", "31-Disc jockey / \n", "0-Games / 1-Vehicle / 2-Video game / 4-Car / 11-Motorsport / 17-Racing / 40-Road / 41-Sports car / 47-Personal computer / 231-Supercar / 352-Need for Speed / 875-Need for Speed: Most Wanted / \n", "1-Vehicle / 4-Car / 126-The Walt Disney Company / 144-Amusement park / 321-Roller coaster / 334-Amusement ride / \n", "1-Vehicle / 4-Car / 11-Motorsport / 64-Transport / 74-Truck / 286-Bus / \n", "0-Games / 2-Video game / 33-Weapon / 1251-The Witcher / \n", "8-Football / \n", "0-Games / 1-Vehicle / 2-Video game / 11-Motorsport / 17-Racing / 27-Motorcycle / 105-Motorcycling / 111-PlayStation 3 / 127-Xbox 360 / 172-Motocross / 2930-MX vs. ATV / \n", "93-Comedy / \n", "0-Games / 1-Vehicle / 4-Car / \n", "1-Vehicle / 11-Motorsport / 17-Racing / 27-Motorcycle / 57-Race track / 172-Motocross / \n", "827-Pump / \n", "1001-Gemstone / \n", "6-Animation / 15-Cartoon / 396-One Piece / \n", "23-Mobile phone / 101-Telephone / \n", "0-Games / 2-Video game / 19-PC game / 176-PlayStation 4 / 948-Far Cry / 1683-Far Cry 4 / \n", "0-Games / 2-Video game / 34-Action-adventure game / 201-Grand Theft Auto: San Andreas / \n", "0-Games / 89-Comics / 1469-X-Men / \n", "3-Concert / \n", "45-Cosmetics / 56-Hair / 106-Hairstyle / \n", "23-Mobile phone / 29-Smartphone / 37-Gadget / 101-Telephone / \n", "0-Games / 2-Video game / 43-Call of Duty / 134-Call of Duty: Black Ops / 138-Call of Duty: Black Ops II / 181-Xbox / \n", "6-Animation / 9-Music video / \n", "3-Concert / 7-Musician / \n", "0-Games / 412-Monster / 3234-Kixeye; Inc. / \n", "0-Games / 2-Video game / 19-PC game / 51-Strategy video game / 165-League of Legends / \n", "36-Piano / 53-Keyboard / 219-Accordion / \n", "23-Mobile phone / 29-Smartphone / 37-Gadget / 72-iPhone / 178-Samsung Galaxy / 553-Sony Xperia / 1207-Samsung Galaxy S5 / 2906-Sony Xperia Z2 / \n", "99-Christmas / 729-Christmas decoration / \n", "264-Star Wars / \n", "1-Vehicle / 18-Outdoor recreation / 67-Cycling / 69-Bicycle / 197-Trail / 220-Mountain bike / 284-Mountain biking / \n", "6-Animation / 15-Cartoon / 1106-Character / \n", "429-Website / 504-Web page / \n", "0-Games / 8-Football / 21-Stadium / 77-Arena / \n", "110-Album / 4686-Nisekoi / \n", "3-Concert / \n", "7-Musician / 230-Brass instrument / 296-Saxophone / \n", "12-Food / 26-Cooking / 32-Recipe / 52-Dish / 184-Vegetable / 557-Salad / 2040-Korean food / 2637-Kimchi / \n", "1-Vehicle / 4-Car / 11-Motorsport / 27-Motorcycle / 70-Driving / 105-Motorcycling / 1124-Yamaha YZF-R1 / 1380-Kawasaki motorcycles / 2494-Ninja ZX-6R / \n", "0-Games / 2-Video game / \n", "37-Gadget / 47-Personal computer / 131-Computer / 141-Microsoft Windows / 145-Tablet computer / \n", "18-Outdoor recreation / 83-Skateboarding / 108-Skateboard / 221-Skatepark / \n", "9-Music video / 483-Costume / \n", "6-Animation / 9-Music video / 10-Animal / 15-Cartoon / 970-Monkey / \n", "0-Games / 2-Video game / 379-Touhou Project / \n", "45-Cosmetics / 235-Lipstick / 1012-Lip gloss / \n", "1-Vehicle / 4-Car / 11-Motorsport / 41-Sports car / 2216-Honda Integra / 2800-Mazda RX-8 / \n", "42-Fishing / 151-Landing / \n", "3-Concert / \n", "10-Animal / 48-Pet / 1116-Goat / 1198-Birth / \n", "910-Dragon Quest / \n", "6-Animation / 15-Cartoon / 877-Hatsune Miku: Project DIVA / \n", "5-Dance / 16-Performance art / 59-Winter sport / 112-Ice skating / 262-Ice / 316-Figure skating / \n", "6-Animation / 15-Cartoon / \n", "0-Games / 1-Vehicle / 50-Aircraft / 82-Airplane / 388-Fighter aircraft / \n", "66-Bollywood / \n", "10-Animal / 48-Pet / 166-Cat / 346-Kitten / \n", "0-Games / 60-Basketball / 81-Athlete / \n", "45-Cosmetics / 173-Eye shadow / 193-Eye liner / 195-Mascara / 196-Eye / \n", "12-Food / 22-Nature / 86-Plant / 137-Tree / 1073-Chainsaw / \n", "0-Games / 2-Video game / 89-Comics / 396-One Piece / \n", "0-Games / 8-Football / 21-Stadium / 65-Kick / 76-Ball / 81-Athlete / \n", "25-Toy / 1676-Aikatsu! / \n", "0-Games / 6-Animation / 816-Gundam / 1564-Mobile Suit Gundam: Extreme Vs. / \n", "5-Dance / \n", "9-Music video / \n", "141-Microsoft Windows / \n", "1-Vehicle / 4-Car / 92-Radio-controlled model / 202-Radio-controlled car / \n", "9-Music video / \n", "6-Animation / 126-The Walt Disney Company / 3161-Tarzan / \n", "1-Vehicle / 4-Car / 11-Motorsport / 70-Driving / 270-Sedan / 552-Honda Civic / \n", "1796-Skate / \n", "3-Concert / 5-Dance / 16-Performance art / \n", "7-Musician / 14-Guitar / 24-String instrument / 63-Acoustic guitar / \n", "0-Games / 2-Video game / 116-Pokémon / 186-Pokémon / 1005-Pokémon Omega Ruby and Alpha Sapphire / \n", "12-Food / 86-Plant / 137-Tree / 1460-Herb / \n", "3-Concert / 16-Performance art / \n", "342-Windows Media Video / 2447-Guzheng / \n", "1-Vehicle / 4-Car / 360-Jeep / \n", "0-Games / 2-Video game / 19-PC game / 241-Fighting game / 311-Street Fighter / 514-Tekken / 2812-Kazuya Mishima / \n", "0-Games / 8-Football / 65-Kick / 73-Ball / 77-Arena / 90-Sports game / 180-FIFA 15 / \n", "1-Vehicle / 4-Car / 11-Motorsport / 17-Racing / 41-Sports car / 57-Race track / 135-Drifting / 378-Drag racing / 651-Stock car racing / \n", "1-Vehicle / 4-Car / 11-Motorsport / 17-Racing / 41-Sports car / 70-Driving / 231-Supercar / \n", "9-Music video / \n", "0-Games / 2-Video game / 3786-Killing Floor / \n", "8-Football / 21-Stadium / 77-Arena / \n", "1-Vehicle / 50-Aircraft / 82-Airplane / 102-Building / 139-Airport / 151-Landing / 157-Jet aircraft / 175-Airliner / 187-Takeoff / 188-Airline / 254-Runway / 472-Boeing 737 / 502-Sky / 2064-Boeing 737 Next Generation / \n", "1-Vehicle / 50-Aircraft / 82-Airplane / 139-Airport / 151-Landing / 502-Sky / 559-Night / 2136-Boeing 757 / 2174-Delta Air Lines / \n", "5-Dance / 16-Performance art / 49-School / 153-Dress / 410-Teacher / \n", "7-Musician / 14-Guitar / 24-String instrument / 63-Acoustic guitar / \n", "23-Mobile phone / 29-Smartphone / 70-Driving / \n", "0-Games / 81-Athlete / \n", "3-Concert / \n", "313-Book / 762-Angel / \n", "3-Concert / 7-Musician / 16-Performance art / \n", "5-Dance / \n", "97-Drawing / \n", "4317-Japanese yen / \n", "880-Trousers / \n", "1-Vehicle / 205-Newscaster / \n", "31-Disc jockey / \n", "6-Animation / 15-Cartoon / 89-Comics / 1469-X-Men / \n", "6-Animation / \n", "0-Games / 2-Video game / 33-Weapon / 262-Ice / 1312-Dynasty Warriors / 4477-Dynasty Warriors Online / \n", "5-Dance / 319-Robot / 483-Costume / 4438-Humanoid robot / \n", "5-Dance / 95-Talent show / \n", "1791-Stargate / \n", "1-Vehicle / 4-Car / 299-Volkswagen Passenger Cars / 668-Volkswagen Golf / \n", "12-Food / 18-Outdoor recreation / 86-Plant / 347-Climbing / 722-Waterfall / \n", "23-Mobile phone / 29-Smartphone / 37-Gadget / 101-Telephone / 1032-BlackBerry / 3732-BlackBerry Bold / \n", "12-Food / 2986-Dishwasher / \n", "0-Games / 506-Garry's Mod / \n", "14-Guitar / 516-Headphones / \n", "1-Vehicle / 4-Car / \n", "6-Animation / 25-Toy / 524-Littlest Pet Shop / \n", "25-Toy / 35-Minecraft / 102-Building / 140-LEGO / 884-Lego minifigure / 3899-Lego Minecraft / \n", "28-Fashion / 3169-Diva / \n", "162-Medicine / \n", "0-Games / 1-Vehicle / 2-Video game / 4-Car / 34-Action-adventure game / 111-PlayStation 3 / 127-Xbox 360 / 176-PlayStation 4 / 181-Xbox / 314-Xbox One / 1275-Watch Dogs / \n", "147-River / 2412-Fate/stay night / \n", "28-Fashion / 206-Shoe / 681-Leather / 757-Boot / \n", "1-Vehicle / 4-Car / 41-Sports car / \n", "0-Games / 2-Video game / 33-Weapon / 43-Call of Duty / \n", "1-Vehicle / 11-Motorsport / 286-Bus / \n", "1-Vehicle / 4-Car / 11-Motorsport / 17-Racing / 456-Go-kart / 522-Helmet / \n", "0-Games / 596-Monster Hunter / 1339-Monster Hunter Freedom Unite / 2994-Monster Hunter Freedom / \n", "2-Video game / 37-Gadget / 47-Personal computer / 145-Tablet computer / \n", "0-Games / 81-Athlete / 98-Festival / \n", "12-Food / \n", "3-Concert / 7-Musician / 16-Performance art / \n", "10-Animal / 80-Horse / 290-Pony / \n", "3-Concert / \n", "25-Toy / \n", "1-Vehicle / 50-Aircraft / \n", "0-Games / 51-Strategy video game / 373-Clash of Clans / 1950-The CW / \n", "689-Harmonica / \n", "0-Games / 6-Animation / 240-Arcade game / \n", "1-Vehicle / 18-Outdoor recreation / 42-Fishing / 87-Boat / 225-Lake / 268-Recreational fishing / 510-Fishing lure / 844-Bass / 902-Fishing bait / \n", "0-Games / 8-Football / 54-Highlight film / 65-Kick / \n", "163-Running / \n", "45-Cosmetics / 193-Eye liner / 196-Eye / 528-Pencil / \n", "0-Games / 2-Video game / 90-Sports game / 180-FIFA 15 / 561-Madden NFL / \n", "6-Animation / \n", "164-Beach / 4596-Flying disc freestyle / \n", "1-Vehicle / 4-Car / 332-Mercedes-Benz / 891-Mercedes-AMG / \n", "1-Vehicle / 87-Boat / 579-Yacht / 4476-Shipyard / \n", "6-Animation / \n", "20-Trailer / \n", "3-Concert / 16-Performance art / 68-Lighting / \n", "14-Guitar / 24-String instrument / 100-Electric guitar / 1198-Birth / \n", "5-Dance / 191-Cheerleading / \n", "0-Games / 8-Football / 65-Kick / 699-Kick / 1623-Shaolin Kung Fu / \n", "0-Games / 6-Animation / 366-Dragon / 460-Yu-Gi-Oh! Trading Card Game / \n", "1-Vehicle / 4-Car / 11-Motorsport / 74-Truck / 114-Engine / 167-Four-wheel drive / 197-Trail / 198-Off-road vehicle / 277-Sport utility vehicle / 327-Mud / 360-Jeep / 565-Mud bogging / \n", "1-Vehicle / 4-Car / 319-Robot / 585-Lawn / 741-Mower / 805-Lawn mower / \n", "86-Plant / 136-Gardening / 279-Garden / 906-Sowing / \n", "5-Dance / \n", "148-Painting / 408-Sketch / 415-Sewing / 476-Textile / 546-Stitch / 653-Thread / 997-Sewing machine / 1057-Embroidery / \n", "10-Animal / 80-Horse / 438-Mare / 523-Cattle / 3212-Team roping / \n", "0-Games / 2-Video game / 34-Action-adventure game / 3059-Dead Rising 2 / \n", "0-Games / 8-Football / 21-Stadium / 65-Kick / 73-Ball / 76-Ball / \n", "12-Food / 56-Hair / 106-Hairstyle / \n", "0-Games / 2-Video game / 34-Action-adventure game / 243-Sonic the Hedgehog / 251-Sonic the Hedgehog / \n", "155-Camera / 1013-Walking / \n", "0-Games / 211-RuneScape / \n", "28-Fashion / 676-Handbag / \n", "10-Animal / 537-Zoo / \n", "25-Toy / 323-Toddler / 3396-Ressha Sentai ToQger / \n", "0-Games / 2-Video game / 1543-Crysis / \n", "548-Parachuting / 765-Parachute / 1483-United States Air Force / \n", "0-Games / 599-Dofus / \n", "1-Vehicle / 4-Car / 70-Driving / 94-Dashcam / 245-BMW / 1607-BMW 3 Series (E36) / \n", "89-Comics / 294-Batman / 1422-Statue / 2082-Catwoman / \n", "49-School / \n", "12-Food / \n", "15-Cartoon / 61-Art / 394-Map / \n", "1-Vehicle / 2-Video game / 4-Car / 11-Motorsport / 17-Racing / 40-Road / 41-Sports car / 57-Race track / 70-Driving / 135-Drifting / 231-Supercar / 705-Forza Motorsport / 790-Nissan GT-R / 1301-Nissan Skyline / \n", "3-Concert / 350-Circus / 1894-Cannon / \n", "12-Food / 86-Plant / 136-Gardening / 137-Tree / 673-Saw / 1073-Chainsaw / \n", "9-Music video / \n", "382-Dota 2 / 549-Defense of the Ancients / \n", "9-Music video / \n", "46-Choir / 49-School / \n", "9-Music video / 61-Art / 408-Sketch / \n", "10-Animal / 80-Horse / 709-Bull / \n", "49-School / 54-Highlight film / 60-Basketball / 168-Basketball moves / \n", "1-Vehicle / 4-Car / 92-Radio-controlled model / 167-Four-wheel drive / 202-Radio-controlled car / \n", "1-Vehicle / 4-Car / 11-Motorsport / 17-Racing / 92-Radio-controlled model / 167-Four-wheel drive / 202-Radio-controlled car / 853-Dune buggy / \n", "0-Games / 2-Video game / 441-The Elder Scrolls / 1130-Armour / 1410-The Elder Scrolls IV: Oblivion / \n", "5-Dance / 7-Musician / 219-Accordion / \n", "1-Vehicle / 50-Aircraft / \n", "1-Vehicle / 245-BMW / \n", "0-Games / 2-Video game / 257-Counter-Strike / \n", "0-Games / 724-Rock Band / \n", "0-Games / 60-Basketball / 168-Basketball moves / \n", "1-Vehicle / 9-Music video / \n", "12-Food / 26-Cooking / 32-Recipe / 52-Dish / 58-Cuisine / 109-Cooking show / 179-Kitchen / 542-Potato / 742-Countertop / 4415-Mashed potato / \n", "1-Vehicle / 4-Car / 1502-Chrysler / \n", "3-Concert / 7-Musician / 38-Orchestra / \n", "3-Concert / \n", "3575-Durarara!! / \n", "0-Games / 2-Video game / 6-Animation / \n", "458-Gold / \n", "10-Animal / 48-Pet / 71-Dog / 1252-Dog agility / \n", "45-Cosmetics / 173-Eye shadow / 193-Eye liner / 195-Mascara / 196-Eye / 218-Eyelash / \n", "1-Vehicle / 87-Boat / \n", "10-Animal / 22-Nature / 213-Forest / 217-Hunting / 706-Deer / \n", "3-Concert / 5-Dance / \n", "9-Music video / \n", "75-Wrestling / \n", "3-Concert / \n", "3-Concert / \n", "144-Amusement park / \n", "1-Vehicle / 4-Car / 11-Motorsport / 74-Truck / 154-Wheel / 158-Tire / 167-Four-wheel drive / 198-Off-road vehicle / 234-Ford / 327-Mud / 565-Mud bogging / 2215-Ford Ranger / \n", "0-Games / 2-Video game / 43-Call of Duty / 216-Call of Duty: Modern Warfare 3 / \n", "1-Vehicle / 4-Car / 11-Motorsport / 41-Sports car / 231-Supercar / 344-Coupé / 1040-Street racing / 1301-Nissan Skyline / 2216-Honda Integra / 3488-Fourth Generation Honda Integra / \n", "0-Games / 1-Vehicle / 2-Video game / 4-Car / 11-Motorsport / 17-Racing / \n", "0-Games / 2-Video game / 19-PC game / 33-Weapon / 928-Borderlands 2 / \n", "3-Concert / 5-Dance / \n", "401-Cello / \n", "0-Games / 2-Video game / 19-PC game / 35-Minecraft / \n", "8-Football / \n", "6-Animation / 61-Art / 479-Graffiti / 1385-Street art / \n", "1-Vehicle / 50-Aircraft / 92-Radio-controlled model / 149-Model aircraft / 160-Radio-controlled aircraft / 266-Helicopter / 392-Radio-controlled helicopter / \n", "59-Winter sport / 112-Ice skating / 142-Hockey / 262-Ice / 312-Ice rink / \n", "10-Animal / 132-Bird / 2339-Swan / \n", "28-Fashion / 676-Handbag / \n", "6-Animation / 15-Cartoon / \n", "5-Dance / 115-Ballet / \n", "25-Toy / 75-Wrestling / 311-Street Fighter / 435-Action figure / 563-Ryu / 749-Figurine / \n", "170-Home improvement / \n", "0-Games / 8-Football / 54-Highlight film / \n", "12-Food / 342-Windows Media Video / 966-Sandwich / \n", "219-Accordion / \n", "23-Mobile phone / 29-Smartphone / 37-Gadget / 47-Personal computer / 145-Tablet computer / 1899-E-book / 2262-E-reader / \n", "0-Games / 8-Football / 21-Stadium / 65-Kick / \n", "208-Wood / 261-Tool / 400-Woodturning / 1700-Lumber / \n", "9-Music video / \n", "10-Animal / 223-Weather / \n", "99-Christmas / \n", "8-Football / 90-Sports game / 453-Pro Evolution Soccer / \n", "9-Music video / \n", "46-Choir / 318-Church / \n", "0-Games / 2-Video game / 553-Sony Xperia / 2770-Sony Xperia Z1 / 2882-Gladiator / \n", "7-Musician / 13-Musical ensemble / 98-Festival / 203-Marching band / 250-Parade / \n", "365-Fireworks / \n", "1-Vehicle / 4-Car / 11-Motorsport / 17-Racing / 41-Sports car / 57-Race track / 135-Drifting / 790-Nissan GT-R / 1810-Live for Speed / \n", "3-Concert / 7-Musician / 14-Guitar / 2393-Didgeridoo / \n", "164-Beach / 248-Hotel / 369-Resort / \n", "1552-Video lesson / \n", "12-Food / 86-Plant / 137-Tree / 1720-Grape / 2473-Vineyard / \n", "0-Games / 2-Video game / 739-Guild Wars / 839-Guild Wars 2 / \n", "1-Vehicle / 11-Motorsport / 27-Motorcycle / \n", "1-Vehicle / 4-Car / 11-Motorsport / 27-Motorcycle / 34-Action-adventure game / 118-Grand Theft Auto V / \n", "14-Guitar / \n", "1-Vehicle / 92-Radio-controlled model / 202-Radio-controlled car / \n", "5-Dance / \n", "2410-Camel / \n", "14-Guitar / 100-Electric guitar / \n", "0-Games / \n", "16-Performance art / 1139-Belle / 1265-Kyle Kingson / \n", "452-Watch / 2168-G-Shock / \n", "0-Games / 2-Video game / 25-Toy / 425-City / \n", "0-Games / 2-Video game / 116-Pokémon / 186-Pokémon / \n", "3-Concert / 7-Musician / 13-Musical ensemble / 14-Guitar / 30-Drummer / \n", "28-Fashion / 2385-Closet / \n", "3-Concert / 16-Performance art / \n", "1-Vehicle / 4-Car / 27-Motorcycle / 40-Road / 64-Transport / 482-Asphalt / 1867-Honda CB600F / \n", "3-Concert / 7-Musician / 13-Musical ensemble / 39-Drums / \n", "3-Concert / \n", "0-Games / 18-Outdoor recreation / 59-Winter sport / 112-Ice skating / 142-Hockey / 262-Ice / 312-Ice rink / 1973-Hockey puck / \n", "10-Animal / 18-Outdoor recreation / 22-Nature / 42-Fishing / 91-Fish / 147-River / 225-Lake / 268-Recreational fishing / 1018-Salmon / 2094-Rainbow trout / \n", "22-Nature / 120-Winter / 339-Architecture / 998-Landscape / \n", "61-Art / 479-Graffiti / \n", "12-Food / 539-Home appliance / 1439-Refrigerator / \n", "1-Vehicle / \n", "5-Dance / 9-Music video / 466-Gift / \n", "0-Games / 51-Strategy video game / 4539-Neighbours from Hell / \n", "3-Concert / 7-Musician / 13-Musical ensemble / \n", "0-Games / 51-Strategy video game / 165-League of Legends / \n", "12-Food / 1175-Soap / 4325-Detergent / \n", "1-Vehicle / 62-Train / 119-Locomotive / 794-Bridge / \n", "0-Games / 530-Cue sports / 568-Pool / 994-Cue stick / 1364-Nine-ball / \n", "5-Dance / \n", "10-Animal / 18-Outdoor recreation / 80-Horse / 247-Stallion / 289-Dressage / 351-Jumping / \n", "8-Football / 76-Ball / \n", "1-Vehicle / 4-Car / \n", "18-Outdoor recreation / 22-Nature / 137-Tree / 213-Forest / 395-Camping / 914-Tent / \n", "2-Video game / 6-Animation / \n", "0-Games / 2-Video game / 240-Arcade game / \n", "2-Video game / 47-Personal computer / 131-Computer / 228-Laptop / 414-Computer monitor / 3000-ThinkPad / \n", "0-Games / 35-Minecraft / \n", "3-Concert / 16-Performance art / \n", "0-Games / 2-Video game / \n", "56-Hair / 106-Hairstyle / \n", "170-Home improvement / 179-Kitchen / 246-Furniture / 652-Bathroom / \n", "273-News program / \n", "85-Combat / 937-Black belt / 2822-Submission wrestling / \n", "3-Concert / 5-Dance / \n", "35-Minecraft / \n", "1-Vehicle / 4-Car / 70-Driving / 733-Washing / 1242-Car wash / \n", "3-Concert / 5-Dance / 68-Lighting / \n", "1-Vehicle / 27-Motorcycle / 207-Exhaust system / 3592-Kawasaki Ninja 650R / \n", "627-Clock / \n", "1492-Persona / 2896-Shin Megami Tensei: Persona 3 / \n", "3-Concert / 7-Musician / \n", "345-Loudspeaker / 484-Subwoofer / \n", "0-Games / 326-Metin2 / \n", "110-Album / \n", "2-Video game / 23-Mobile phone / 29-Smartphone / 37-Gadget / 72-iPhone / 101-Telephone / 272-iPod / 331-iPod touch / \n", "1-Vehicle / 4-Car / 27-Motorcycle / 40-Road / 212-Highway / 245-BMW / 800-Mountain pass / \n", "1-Vehicle / 4-Car / 2991-Ferrari 599 GTB Fiorano / \n", "6-Animation / 9-Music video / 15-Cartoon / 803-DC Comics / 2782-Starfire / \n", "1-Vehicle / 4-Car / 403-Toyota / \n", "1345-Dean Winchester / \n", "23-Mobile phone / 29-Smartphone / 37-Gadget / 47-Personal computer / 72-iPhone / 131-Computer / 145-Tablet computer / 178-Samsung Galaxy / 487-Samsung Electronics / 1015-Samsung Galaxy Note series / 2416-Samsung Galaxy Tab 10.1 / 2802-Samsung Galaxy Note 10.1 / \n", "13-Musical ensemble / 21-Stadium / 77-Arena / 79-American football / 203-Marching band / 209-University / \n", "2-Video game / 1241-Assassin's Creed IV: Black Flag / 1858-Poster / 3238-GameStop / \n", "1-Vehicle / 11-Motorsport / 17-Racing / 57-Race track / 595-Champion / 1609-Auto Race / \n", "20-Trailer / \n", "0-Games / 1-Vehicle / 124-Tractor / 156-Agriculture / 174-Farm / 715-Farming Simulator / \n", "238-Nail / 300-Nail art / 304-Nail polish / 338-Manicure / 1128-Glitter / \n", "0-Games / 302-Call of Duty 4: Modern Warfare / \n", "20-Trailer / \n", "0-Games / 731-Coin / 831-Plants vs. Zombies / 1374-Plants vs. Zombies 2: It's About Time / \n", "12-Food / 348-Fruit / 780-Fat / \n", "49-School / \n", "0-Games / 2-Video game / 34-Action-adventure game / 201-Grand Theft Auto: San Andreas / 2060-Roulette / \n", "10-Animal / 48-Pet / 71-Dog / 84-Snow / 120-Winter / \n", "0-Games / 19-PC game / 211-RuneScape / \n", "7-Musician / \n", "5-Dance / 2071-Indiana Jones / \n", "860-Christian Church / \n", "380-Assassin's Creed / 1241-Assassin's Creed IV: Black Flag / \n", "12-Food / 125-Television / 3044-Sesame / \n", "0-Games / 2-Video game / 34-Action-adventure game / 55-Video game console / \n", "0-Games / 60-Basketball / \n", "23-Mobile phone / 29-Smartphone / 37-Gadget / \n", "2-Video game / 20-Trailer / \n", "0-Games / 60-Basketball / 176-PlayStation 4 / 314-Xbox One / 991-NBA 2K15 / \n", "0-Games / 2-Video game / 19-PC game / 2594-Mount & Blade: Warband / 2622-Mount & Blade / \n", "10-Animal / 71-Dog / \n", "1-Vehicle / 949-Rowing / 1090-Canoe / 1305-Canoeing / \n", "3-Concert / 7-Musician / 13-Musical ensemble / 38-Orchestra / 219-Accordion / \n", "365-Fireworks / \n", "1605-Second Life / \n", "3-Concert / 7-Musician / 13-Musical ensemble / 46-Choir / 716-Coral / \n", "8-Football / 3303-Table football / \n", "425-City / \n", "1-Vehicle / 62-Train / 64-Transport / 84-Snow / 103-Rail transport / 120-Winter / 152-Track / 161-Train station / \n", "7-Musician / 14-Guitar / 24-String instrument / 100-Electric guitar / \n", "3-Concert / 7-Musician / 16-Performance art / \n", "10-Animal / 80-Horse / \n", "0-Games / 90-Sports game / 180-FIFA 15 / \n", "0-Games / 8-Football / 60-Basketball / \n", "1-Vehicle / 22-Nature / 642-Leaf / 657-Autumn / \n", "18-Outdoor recreation / 22-Nature / 395-Camping / 614-Hiking / 835-Backpack / 1013-Walking / 1203-Backpacking / \n", "12-Food / 26-Cooking / 475-Cheese / \n", "23-Mobile phone / 29-Smartphone / 72-iPhone / 1155-iPhone 3GS / \n", "13-Musical ensemble / \n", "1-Vehicle / 4-Car / 25-Toy / 92-Radio-controlled model / 202-Radio-controlled car / \n", "0-Games / 2-Video game / 8-Football / 90-Sports game / 180-FIFA 15 / \n", "1-Vehicle / 4-Car / 114-Engine / 449-Honda / \n", "0-Games / 2-Video game / 19-PC game / 51-Strategy video game / 382-Dota 2 / 549-Defense of the Ancients / \n", "326-Metin2 / 342-Windows Media Video / \n", "1-Vehicle / 62-Train / 64-Transport / 161-Train station / 224-Rapid transit / \n", "0-Games / 23-Mobile phone / 366-Dragon / \n", "12-Food / 26-Cooking / 58-Cuisine / 274-Meat / 589-Grilling / 620-Barbecue / \n", "798-Shampoo / 920-Chair / 1271-Beauty salon / \n", "516-Headphones / 832-Headset / \n", "609-Tattoo / 1834-Mehndi / \n", "99-Christmas / \n", "3-Concert / 7-Musician / 38-Orchestra / \n", "0-Games / 59-Winter sport / 112-Ice skating / 316-Figure skating / \n", "1-Vehicle / 4-Car / 27-Motorcycle / 608-Wheelie / 4133-Yamaha DT125 / \n", "0-Games / 2-Video game / 19-PC game / 739-Guild Wars / 839-Guild Wars 2 / \n", "1-Vehicle / 4-Car / \n", "8-Football / \n", "6-Animation / 10-Animal / 15-Cartoon / 80-Horse / 2155-Unicorn / \n", "0-Games / 1-Vehicle / 2-Video game / 34-Action-adventure game / 118-Grand Theft Auto V / \n", "0-Games / 489-Board game / 799-Chess / 922-Tabletop game / \n", "28-Fashion / 153-Dress / \n", "5-Dance / 115-Ballet / \n", "3-Concert / 7-Musician / 16-Performance art / \n", "0-Games / 43-Call of Duty / 194-Call of Duty: Modern Warfare 2 / \n", "12-Food / \n", "9-Music video / 61-Art / \n", "3-Concert / 7-Musician / 13-Musical ensemble / \n", "200-Doll / 1222-Fountain / 3605-Puppetry / \n", "56-Hair / 106-Hairstyle / 340-Afro-textured hair / 798-Shampoo / 919-Hair conditioner / \n", "1-Vehicle / 50-Aircraft / 82-Airplane / 129-Aviation / 139-Airport / 157-Jet aircraft / 175-Airliner / 187-Takeoff / 188-Airline / 254-Runway / 923-Boeing 777 / \n", "3-Concert / \n", "260-Prayer / \n", "5-Dance / 66-Bollywood / \n", "14-Guitar / 63-Acoustic guitar / \n", "325-Vampire / \n", "3-Concert / \n", "10-Animal / 22-Nature / \n", "1-Vehicle / 4-Car / 683-Renault / 2785-Slot car / \n", "141-Microsoft Windows / \n", "0-Games / 2-Video game / 34-Action-adventure game / 297-Grand Theft Auto IV / 1595-Television set / \n", "5-Dance / 16-Performance art / 99-Christmas / \n", "0-Games / 2-Video game / 51-Strategy video game / 1570-Command & Conquer / \n", "0-Games / 2-Video game / 6-Animation / 15-Cartoon / 89-Comics / 185-Dragon Ball / 244-Goku / 876-Freeza / \n", "9-Music video / 1791-Stargate / \n", "1-Vehicle / 11-Motorsport / 124-Tractor / 156-Agriculture / 249-Heavy equipment / 252-Tractor pulling / 459-Plough / \n", "1-Vehicle / 4-Car / 11-Motorsport / 17-Racing / 40-Road / 41-Sports car / 57-Race track / 231-Supercar / 705-Forza Motorsport / 952-Chevrolet Corvette / 1530-Forza Motorsport 4 / 2885-Chevrolet Corvette (C6) / 4135-Chevrolet Corvette C5 Z06 / \n", "36-Piano / 53-Keyboard / 104-Pianist / 107-Musical keyboard / \n", "28-Fashion / 56-Hair / 793-Shirt / \n", "5-Dance / \n", "6-Animation / \n", "0-Games / 76-Ball / 81-Athlete / 256-Tennis / 974-Racket / 1337-Serve / \n", "131-Computer / 141-Microsoft Windows / \n", "402-Jewellery / 829-Bead / 859-Beadwork / 907-Necklace / \n", "0-Games / 8-Football / 21-Stadium / 65-Kick / 73-Ball / 76-Ball / 77-Arena / 81-Athlete / \n", "5-Dance / \n", "25-Toy / 1063-Ink / \n", "10-Animal / 18-Outdoor recreation / 22-Nature / 91-Fish / \n", "3-Concert / 7-Musician / 13-Musical ensemble / 38-Orchestra / 46-Choir / \n", "1-Vehicle / 50-Aircraft / 82-Airplane / 129-Aviation / 139-Airport / 151-Landing / 157-Jet aircraft / 175-Airliner / 187-Takeoff / 188-Airline / 254-Runway / 1316-Airbus A330 / \n", "9-Music video / \n", "5-Dance / \n", "12-Food / 1408-Snack / \n", "0-Games / 34-Action-adventure game / 201-Grand Theft Auto: San Andreas / \n", "113-House / \n", "3-Concert / \n", "1-Vehicle / 4-Car / 40-Road / 70-Driving / 212-Highway / \n", "843-The Doctor / \n", "1-Vehicle / 50-Aircraft / 471-Wing / 778-Gliding / 1263-Glider / 1317-Glider / \n", "139-Airport / \n", "3-Concert / 7-Musician / 14-Guitar / \n", "110-Album / \n", "3-Concert / \n", "1-Vehicle / 62-Train / 64-Transport / 94-Dashcam / 103-Rail transport / 119-Locomotive / 143-Railroad car / 152-Track / 161-Train station / 224-Rapid transit / 821-Rail freight transport / \n", "960-Badminton / \n", "1-Vehicle / 4-Car / 11-Motorsport / 41-Sports car / 70-Driving / 135-Drifting / 1575-Clutch / \n", "6-Animation / 419-Earth / 701-Beyblade / \n", "3-Concert / 16-Performance art / 68-Lighting / \n", "10-Animal / 588-Reptile / \n", "18-Outdoor recreation / 59-Winter sport / 84-Snow / 120-Winter / 281-Snowboarding / \n", "1-Vehicle / 4-Car / 11-Motorsport / 17-Racing / 94-Dashcam / \n", "0-Games / 2-Video game / 19-PC game / 51-Strategy video game / 85-Combat / 4690-Shadow Fighter / \n", "5-Dance / 49-School / 95-Talent show / 130-Gymnastics / 191-Cheerleading / \n", "1-Vehicle / 4-Car / \n", "8-Football / 54-Highlight film / 79-American football / \n", "1-Vehicle / 50-Aircraft / 82-Airplane / 129-Aviation / 139-Airport / 151-Landing / 157-Jet aircraft / 175-Airliner / 187-Takeoff / 188-Airline / 254-Runway / 471-Wing / 502-Sky / 923-Boeing 777 / 1705-Boeing 767 / 2391-American Airlines / \n", "322-Mixtape / \n", "0-Games / 2-Video game / 19-PC game / 34-Action-adventure game / 201-Grand Theft Auto: San Andreas / 858-Carl Johnson / \n", "25-Toy / 140-LEGO / 387-Spider-Man / \n", "12-Food / 26-Cooking / 32-Recipe / 58-Cuisine / 179-Kitchen / 1725-Custard / \n", "18-Outdoor recreation / 395-Camping / 943-Stove / 1203-Backpacking / 2670-Portable stove / \n", "0-Games / 2-Video game / 33-Weapon / 420-The Elder Scrolls V: Skyrim / 441-The Elder Scrolls / \n", "1-Vehicle / 88-Machine / \n", "2233-Derek Shepherd / \n", "3-Concert / \n", "7-Musician / 24-String instrument / 150-Violin / \n", "0-Games / 2-Video game / 51-Strategy video game / 382-Dota 2 / 549-Defense of the Ancients / \n", "122-Weight training / 204-Gym / 214-Muscle / 532-Barbell / 1250-Biceps / 2045-Biceps curl / \n", "102-Building / 113-House / 170-Home improvement / 276-Room / 279-Garden / 339-Architecture / 2220-Shed / 2291-Log cabin / \n", "6-Animation / 25-Toy / 200-Doll / \n", "6-Animation / 185-Dragon Ball / 244-Goku / 1165-Majin Buu / \n", "98-Festival / 253-Carnival / \n", "0-Games / 2-Video game / 88-Machine / 240-Arcade game / 735-Kamen Rider Series / 2557-Kamen Rider Battle: Ganbaride / \n", "69-Bicycle / 1067-Lock / 2516-Lock picking / 3061-Cylinder / \n", "1-Vehicle / 27-Motorcycle / 522-Helmet / 1243-Mirror / 1903-Motorcycle helmet / \n", "20-Trailer / \n", "1-Vehicle / 4-Car / 11-Motorsport / 17-Racing / 57-Race track / \n", "1-Vehicle / 11-Motorsport / 67-Cycling / 69-Bicycle / 220-Mountain bike / 377-BMX bike / \n", "8-Football / \n", "9-Music video / \n", "0-Games / 2-Video game / 20-Trailer / 55-Video game console / 111-PlayStation 3 / 127-Xbox 360 / 3406-Bayonetta / \n", "420-The Elder Scrolls V: Skyrim / \n", "8-Football / 90-Sports game / 453-Pro Evolution Soccer / \n", "5-Dance / 16-Performance art / \n", "78-Wedding / 190-Bride / \n", "31-Disc jockey / \n", "0-Games / 2-Video game / 34-Action-adventure game / 85-Combat / 111-PlayStation 3 / 127-Xbox 360 / 434-Ninja / 2192-Ninja Gaiden / \n", "3-Concert / 22-Nature / 98-Festival / \n", "1-Vehicle / 4-Car / 11-Motorsport / 92-Radio-controlled model / 202-Radio-controlled car / 327-Mud / 565-Mud bogging / \n", "5-Dance / 16-Performance art / 38-Orchestra / \n", "264-Star Wars / 4205-Motel / \n", "31-Disc jockey / \n", "376-Flute / 1326-Recorder / \n", "0-Games / 2-Video game / 47-Personal computer / 228-Laptop / \n", "0-Games / 411-MapleStory / 3910-Mackerel / \n", "3-Concert / 5-Dance / \n", "1-Vehicle / 50-Aircraft / 82-Airplane / 175-Airliner / 188-Airline / 471-Wing / 923-Boeing 777 / 2587-Air France / \n", "3-Concert / 16-Performance art / \n", "3-Concert / 7-Musician / \n", "3-Concert / \n", "12-Food / 427-Trumpet / \n", "10-Animal / 48-Pet / 71-Dog / 1084-Bulldog / \n", "6-Animation / 15-Cartoon / 750-Sailor Moon / 1342-Sailor Moon / 2761-Venus / 4287-Mercury / \n", "3-Concert / 5-Dance / 16-Performance art / \n", "0-Games / 2-Video game / 35-Minecraft / 2171-The Hunger Games / \n", "0-Games / 2-Video game / 19-PC game / 33-Weapon / 43-Call of Duty / 96-Soldier / 194-Call of Duty: Modern Warfare 2 / \n", "12-Food / \n", "0-Games / 2590-Mahjong / \n", "1-Vehicle / 17-Racing / 18-Outdoor recreation / 67-Cycling / 69-Bicycle / 570-Road bicycle racing / 645-Road bicycle / \n", "6-Animation / 3437-Maid Sama! / \n", "1-Vehicle / 18-Outdoor recreation / 42-Fishing / 87-Boat / 225-Lake / 2836-Walleye / 3788-Crappie / \n", "75-Wrestling / \n", "102-Building / 113-House / 170-Home improvement / \n", "1-Vehicle / 4-Car / 11-Motorsport / 17-Racing / 245-BMW / 874-BMW M3 / \n", "6-Animation / 12-Food / 208-Wood / 246-Furniture / 462-Table / \n", "10-Animal / 48-Pet / 71-Dog / 881-German Shepherd / 3116-Schutzhund / \n", "23-Mobile phone / 29-Smartphone / 72-iPhone / 272-iPod / \n", "1-Vehicle / 4-Car / 484-Subwoofer / 3086-Honda CR-V / \n", "5-Dance / 16-Performance art / \n", "0-Games / 51-Strategy video game / 241-Fighting game / 2320-Guilty Gear / \n", "0-Games / 256-Tennis / 1626-Forehand / \n", "113-House / 179-Kitchen / 276-Room / 1142-Laundry / \n", "3-Concert / 7-Musician / 13-Musical ensemble / 16-Performance art / 98-Festival / \n", "12-Food / 26-Cooking / 179-Kitchen / \n", "10-Animal / 48-Pet / 71-Dog / 1839-Rottweiler / 2049-Police dog / 3116-Schutzhund / \n", "326-Metin2 / \n", "755-Hetalia: Axis Powers / \n", "1-Vehicle / 50-Aircraft / 139-Airport / \n", "3-Concert / 16-Performance art / \n", "45-Cosmetics / 173-Eye shadow / 193-Eye liner / 195-Mascara / 218-Eyelash / 235-Lipstick / \n", "7-Musician / 13-Musical ensemble / 230-Brass instrument / \n", "0-Games / 2-Video game / 43-Call of Duty / 138-Call of Duty: Black Ops II / \n", "0-Games / 2-Video game / 10-Animal / 132-Bird / 847-Angry Birds / 3319-Angry Birds / \n", "3-Concert / 16-Performance art / 133-Nightclub / \n", "14-Guitar / 24-String instrument / 63-Acoustic guitar / 1562-Viola caipira / \n", "6-Animation / 15-Cartoon / \n", "31-Disc jockey / \n", "0-Games / 2-Video game / 10-Animal / 19-PC game / 550-Dinosaur / \n", "3-Concert / 7-Musician / 14-Guitar / 24-String instrument / 63-Acoustic guitar / \n", "9-Music video / 1976-Nostale / \n", "10-Animal / 17-Racing / 80-Horse / 384-Horse racing / \n", "1-Vehicle / 4-Car / 11-Motorsport / 17-Racing / 27-Motorcycle / \n", "0-Games / 2-Video game / 243-Sonic the Hedgehog / 4595-Sonic & All-Stars Racing Transformed / \n", "99-Christmas / 444-Family / 466-Gift / \n", "45-Cosmetics / 282-Skin / \n", "5-Dance / 66-Bollywood / \n", "1-Vehicle / 4-Car / 11-Motorsport / 17-Racing / 27-Motorcycle / 40-Road / 2807-BMW S1000RR / \n", "0-Games / 2-Video game / 162-Medicine / \n", "361-Flamenco / \n", "3-Concert / 7-Musician / 13-Musical ensemble / 16-Performance art / 30-Drummer / 39-Drums / \n", "0-Games / 90-Sports game / 180-FIFA 15 / \n", "0-Games / 2-Video game / 19-PC game / 35-Minecraft / 257-Counter-Strike / 1109-Counter-Strike Online / \n", "93-Comedy / \n", "379-Touhou Project / 2170-The Embodiment of Scarlet Devil / \n", "13-Musical ensemble / \n", "1-Vehicle / 4-Car / 27-Motorcycle / 947-Muffler / \n", "1-Vehicle / 4-Car / 47-Personal computer / 70-Driving / 145-Tablet computer / \n", "28-Fashion / 390-Model / \n", "6-Animation / \n", "36-Piano / 53-Keyboard / 104-Pianist / 107-Musical keyboard / 265-Electronic keyboard / \n", "425-City / \n", "0-Games / 2-Video game / 505-Mega Man / 2474-Mega Man Battle Network / 4482-Mega Man Battle Network 6 / \n", "3-Concert / 7-Musician / 30-Drummer / 39-Drums / \n", "3-Concert / \n", "1-Vehicle / 4-Car / 277-Sport utility vehicle / 4493-Nissan X-Trail / \n", "1-Vehicle / 4-Car / 1321-Škoda Auto / 2690-Škoda Octavia / \n", "33-Weapon / 242-Paper / \n", "10-Animal / \n", "0-Games / 8-Football / 90-Sports game / 180-FIFA 15 / \n", "0-Games / 8-Football / 90-Sports game / 180-FIFA 15 / \n", "20-Trailer / \n", "0-Games / 8-Football / 21-Stadium / 54-Highlight film / 73-Ball / 81-Athlete / \n", "93-Comedy / \n", "9-Music video / \n", "0-Games / 2-Video game / 6-Animation / 2984-.hack / \n", "1-Vehicle / 50-Aircraft / 92-Radio-controlled model / 149-Model aircraft / 160-Radio-controlled aircraft / 223-Weather / 266-Helicopter / 392-Radio-controlled helicopter / 499-Cloud / \n", "3-Concert / 30-Drummer / \n", "3-Concert / 16-Performance art / \n", "150-Violin / \n", "66-Bollywood / \n", "205-Newscaster / 223-Weather / \n", "1-Vehicle / 8-Football / 11-Motorsport / 17-Racing / 172-Motocross / \n", "1-Vehicle / 50-Aircraft / 82-Airplane / 92-Radio-controlled model / 129-Aviation / 149-Model aircraft / 160-Radio-controlled aircraft / 266-Helicopter / 392-Radio-controlled helicopter / \n", "7-Musician / 14-Guitar / 24-String instrument / 376-Flute / 726-Clarinet / \n", "0-Games / 6-Animation / 15-Cartoon / 97-Drawing / 2459-Ceiling fan / \n", "3-Concert / 412-Monster / \n", "9-Music video / \n", "219-Accordion / 655-Diatonic button accordion / \n", "6-Animation / 10-Animal / 132-Bird / \n", "10-Animal / 132-Bird / 217-Hunting / \n", "0-Games / 2-Video game / 637-Five Nights at Freddy's / 1193-Orange / 2500-Pear / \n", "1-Vehicle / 4-Car / 11-Motorsport / 135-Drifting / 447-Sand / 935-Off-road racing / 975-Dune / \n", "359-Harry Potter (Literary Series) / \n", "1-Vehicle / 4-Car / 11-Motorsport / 17-Racing / 57-Race track / 651-Stock car racing / \n", "0-Games / 1628-Cabal Online / 2521-Jupiter / 3790-Granado Espada / \n", "10-Animal / 80-Horse / \n", "0-Games / 159-World of Warcraft / 171-Warcraft / 1213-World of Warcraft: Wrath of the Lich King / \n", "10-Animal / 22-Nature / 42-Fishing / 147-River / 1311-Amazon.com; Inc. / \n", "0-Games / 6-Animation / 15-Cartoon / \n", "42-Fishing / 573-Paragliding / \n", "46-Choir / \n", "14-Guitar / 24-String instrument / 100-Electric guitar / \n", "211-RuneScape / \n", "0-Games / 2-Video game / 19-PC game / 35-Minecraft / \n", "14-Guitar / \n", "6-Animation / 15-Cartoon / 125-Television / \n", "6-Animation / 15-Cartoon / 61-Art / 97-Drawing / 183-Manga / 396-One Piece / 408-Sketch / 528-Pencil / \n", "1-Vehicle / 4-Car / 18-Outdoor recreation / 40-Road / 167-Four-wheel drive / 197-Trail / 198-Off-road vehicle / 277-Sport utility vehicle / 327-Mud / 332-Mercedes-Benz / 3279-Mercedes-Benz G-Class / \n", "3-Concert / 325-Vampire / \n", "1-Vehicle / 50-Aircraft / 151-Landing / 266-Helicopter / \n", "628-Portrait / \n", "0-Games / 1-Vehicle / 2-Video game / 4-Car / 34-Action-adventure game / \n", "28-Fashion / 45-Cosmetics / 56-Hair / \n", "0-Games / 2-Video game / 910-Dragon Quest / 4411-Dragon Quest V: Hand of the Heavenly Bride / \n", "10-Animal / 80-Horse / \n", "12-Food / 26-Cooking / 32-Recipe / 109-Cooking show / 226-Dessert / 437-Cream / 454-Sauce / 475-Cheese / 2964-Cream cheese / \n", "9-Music video / 20-Trailer / 61-Art / 479-Graffiti / \n", "1-Vehicle / 4-Car / \n", "83-Skateboarding / 108-Skateboard / \n", "1-Vehicle / 18-Outdoor recreation / 67-Cycling / 69-Bicycle / 220-Mountain bike / 284-Mountain biking / 332-Mercedes-Benz / \n", "1-Vehicle / 4-Car / 11-Motorsport / 17-Racing / 57-Race track / 651-Stock car racing / 728-Dirt track racing / \n", "0-Games / 8-Football / 21-Stadium / 65-Kick / 73-Ball / 76-Ball / 81-Athlete / \n", "0-Games / 2-Video game / 111-PlayStation 3 / 237-Final Fantasy / 1131-Final Fantasy XIII / 2506-Final Fantasy XIII-2 / \n", "1-Vehicle / 4-Car / 1455-Cadillac / 3681-Cadillac CTS / \n", "429-Website / \n", "25-Toy / 140-LEGO / 477-Transformers / \n", "1-Vehicle / 42-Fishing / 87-Boat / 3108-Fishing vessel / \n", "1-Vehicle / 4-Car / 88-Machine / 114-Engine / 515-Dodge / \n", "0-Games / 35-Minecraft / \n", "9-Music video / 99-Christmas / \n", "0-Games / 326-Metin2 / \n", "2716-The Hunger Games / \n", "31-Disc jockey / \n", "0-Games / 2-Video game / 19-PC game / 35-Minecraft / \n", "0-Games / 8-Football / 65-Kick / 76-Ball / \n", "0-Games / 21-Stadium / 59-Winter sport / 77-Arena / 112-Ice skating / 142-Hockey / 262-Ice / 312-Ice rink / \n", "3-Concert / 5-Dance / \n", "545-Quartet (ensemble) / \n", "970-Monkey / \n", "0-Games / 2-Video game / 33-Weapon / 43-Call of Duty / 96-Soldier / 138-Call of Duty: Black Ops II / 2434-Call of Duty / \n", "0-Games / 8-Football / 21-Stadium / 65-Kick / 73-Ball / 76-Ball / \n", "20-Trailer / \n", "5-Dance / \n", "0-Games / 3-Concert / 724-Rock Band / \n", "5-Dance / \n", "3-Concert / 5-Dance / 16-Performance art / 66-Bollywood / \n", "3-Concert / 68-Lighting / \n", "3-Concert / 7-Musician / 16-Performance art / 30-Drummer / \n", "794-Bridge / \n", "0-Games / 2-Video game / 241-Fighting game / 311-Street Fighter / 4688-Skullgirls / \n", "8-Football / \n", "3-Concert / 7-Musician / 13-Musical ensemble / 14-Guitar / 24-String instrument / 30-Drummer / 39-Drums / 44-Drums / \n", "0-Games / 1737-Dance Dance Revolution / \n", "3-Concert / 7-Musician / 13-Musical ensemble / 36-Piano / 38-Orchestra / 376-Flute / 545-Quartet (ensemble) / \n", "31-Disc jockey / \n", "3-Concert / 7-Musician / 30-Drummer / 98-Festival / \n", "0-Games / 2-Video game / 6-Animation / 15-Cartoon / 55-Video game console / 518-Mario Kart / \n", "12-Food / 26-Cooking / 96-Soldier / 301-Meal / 389-Restaurant / 1107-United States Navy / 1572-Husband / \n", "0-Games / 2-Video game / 19-PC game / 35-Minecraft / \n", "1-Vehicle / 4-Car / 114-Engine / 1321-Škoda Auto / \n", "1146-Wing Chun / \n", "8-Football / \n", "3-Concert / 68-Lighting / \n", "1-Vehicle / 4-Car / 70-Driving / 332-Mercedes-Benz / \n", "8-Football / \n", "0-Games / 8-Football / \n", "5-Dance / 98-Festival / 250-Parade / 253-Carnival / \n", "3-Concert / 7-Musician / 13-Musical ensemble / 14-Guitar / 30-Drummer / 39-Drums / 44-Drums / \n", "12-Food / 1943-Bakery / \n", "20-Trailer / \n", "6-Animation / 15-Cartoon / 97-Drawing / 475-Cheese / \n", "10-Animal / 12-Food / 540-Insect / 775-Bee / 1121-Beehive / \n", "3-Concert / 7-Musician / 13-Musical ensemble / 38-Orchestra / \n", "5-Dance / \n", "69-Bicycle / 1318-Flashlight / \n", "0-Games / 6-Animation / \n", "0-Games / 2-Video game / 1447-Skylanders / 2688-Skylanders: Giants / \n", "1059-Movie theater / 1184-Flag / \n", "257-Counter-Strike / \n", "28-Fashion / \n", "0-Games / 1-Vehicle / 2-Video game / 4-Car / 33-Weapon / 34-Action-adventure game / 118-Grand Theft Auto V / 297-Grand Theft Auto IV / \n", "14-Guitar / 63-Acoustic guitar / \n", "45-Cosmetics / 56-Hair / \n", "0-Games / 1-Vehicle / 4-Car / 62-Train / 837-Roblox / \n", "0-Games / 2-Video game / 6-Animation / 19-PC game / 240-Arcade game / 241-Fighting game / 311-Street Fighter / 461-Street Fighter IV / 563-Ryu / 1331-Ken Masters / 2239-Street Fighter II: The World Warrior / \n", "0-Games / 2-Video game / 33-Weapon / \n", "320-Super Smash Bros. / \n", "121-Photography / 155-Camera / 2487-Canon PowerShot / \n", "49-School / 342-Windows Media Video / \n", "45-Cosmetics / 238-Nail / 300-Nail art / 304-Nail polish / 307-Hand / 324-Finger / 338-Manicure / 823-Artificial nails / \n", "3-Concert / \n", "0-Games / 2-Video game / 34-Action-adventure game / 398-The Legend of Zelda / 692-Link / 1228-The Legend of Zelda: Ocarina of Time / \n", "750-Sailor Moon / 1342-Sailor Moon / \n", "1-Vehicle / 4-Car / 5-Dance / \n", "7-Musician / 14-Guitar / 24-String instrument / 63-Acoustic guitar / \n", "5-Dance / 78-Wedding / 153-Dress / 190-Bride / 3693-Groomsman / \n", "3-Concert / 16-Performance art / 68-Lighting / 1431-Madison Square Garden / \n", "3-Concert / 7-Musician / 14-Guitar / \n", "3-Concert / 7-Musician / 14-Guitar / \n", "425-City / 2128-Pollution / \n", "1-Vehicle / 87-Boat / \n", "102-Building / 113-House / 170-Home improvement / 179-Kitchen / 276-Room / 353-Wall / 492-Floor / 879-Ceiling / 1272-Cabinetry / \n", "359-Harry Potter (Literary Series) / \n", "0-Games / 2-Video game / 264-Star Wars / 1050-Star Wars: The Old Republic / \n", "0-Games / 2-Video game / 19-PC game / 686-Diablo III / 872-Diablo / \n", "296-Saxophone / \n", "36-Piano / 53-Keyboard / \n", "23-Mobile phone / 29-Smartphone / 37-Gadget / 101-Telephone / \n", "1-Vehicle / 92-Radio-controlled model / 94-Dashcam / 149-Model aircraft / 160-Radio-controlled aircraft / 223-Weather / 502-Sky / \n", "0-Games / 2-Video game / 34-Action-adventure game / 126-The Walt Disney Company / 2453-Disney Infinity / \n", "130-Gymnastics / \n", "6-Animation / \n", "5-Dance / 16-Performance art / 66-Bollywood / \n", "0-Games / 2-Video game / \n", "6-Animation / 15-Cartoon / 412-Monster / 503-Monster High / \n", "9-Music video / \n", "18-Outdoor recreation / 137-Tree / 213-Forest / 673-Saw / 1073-Chainsaw / 1294-Logging / \n", "3-Concert / \n", "1-Vehicle / 601-Shark / \n", "10-Animal / 80-Horse / 247-Stallion / 404-Livestock / \n", "1-Vehicle / 4-Car / 158-Tire / 277-Sport utility vehicle / 360-Jeep / \n", "33-Weapon / \n", "56-Hair / 340-Afro-textured hair / 733-Washing / \n", "28-Fashion / \n", "1-Vehicle / 4-Car / 154-Wheel / 158-Tire / 493-Rim / \n", "5-Dance / 762-Angel / \n", "10-Animal / 71-Dog / \n", "0-Games / 54-Highlight film / 256-Tennis / \n", "0-Games / 2-Video game / 55-Video game console / \n", "6-Animation / 15-Cartoon / 126-The Walt Disney Company / 2696-Cinderella / 3092-Cinderella / \n", "0-Games / 8-Football / 21-Stadium / \n", "95-Talent show / 115-Ballet / 130-Gymnastics / 1255-Ribbon / \n", "350-Circus / 788-Juggling / 882-Pen / \n", "916-Spa / 3386-Hot tub / \n", "5-Dance / 10-Animal / 31-Disc jockey / 80-Horse / \n", "12-Food / 49-School / 1152-Plush / 1707-Microwave oven / \n", "6-Animation / 15-Cartoon / \n", "0-Games / 2-Video game / 263-Light / \n", "6-Animation / 15-Cartoon / \n", "31-Disc jockey / \n", "3-Concert / \n", "0-Games / 8-Football / 21-Stadium / 65-Kick / 73-Ball / 76-Ball / 81-Athlete / \n", "1-Vehicle / 4-Car / 449-Honda / 2900-Honda Fit / \n", "12-Food / 26-Cooking / 52-Dish / 58-Cuisine / \n", "248-Hotel / \n", "3-Concert / 7-Musician / 13-Musical ensemble / 30-Drummer / 39-Drums / \n", "1-Vehicle / 4-Car / \n", "7-Musician / 14-Guitar / 24-String instrument / 63-Acoustic guitar / 242-Paper / 3935-Papier-mâché / \n", "1-Vehicle / 12-Food / 18-Outdoor recreation / 74-Truck / 617-Maize / \n", "12-Food / 52-Dish / 58-Cuisine / 215-Eating / 771-Noodle / 1539-Ramen / 3459-Instant noodle / \n", "6-Animation / \n", "3-Concert / 7-Musician / \n", "83-Skateboarding / 108-Skateboard / \n", "56-Hair / 106-Hairstyle / \n", "3-Concert / 7-Musician / 13-Musical ensemble / 230-Brass instrument / \n", "5-Dance / 9-Music video / \n", "12-Food / 348-Fruit / 2995-Blueberry / 3020-Raspberry / \n", "0-Games / 2-Video game / 19-PC game / 33-Weapon / 43-Call of Duty / 85-Combat / 96-Soldier / 194-Call of Duty: Modern Warfare 2 / 216-Call of Duty: Modern Warfare 3 / 302-Call of Duty 4: Modern Warfare / \n", "1-Vehicle / 4-Car / 11-Motorsport / 17-Racing / 40-Road / 41-Sports car / 57-Race track / 378-Drag racing / \n", "0-Games / 6-Animation / 15-Cartoon / 2648-Larva / \n", "10-Animal / 18-Outdoor recreation / 22-Nature / 42-Fishing / 147-River / 268-Recreational fishing / 2094-Rainbow trout / \n", "12-Food / 26-Cooking / 32-Recipe / 290-Pony / 829-Bead / 859-Beadwork / \n", "31-Disc jockey / \n", "0-Games / 2-Video game / 803-DC Comics / 1311-Amazon.com; Inc. / 1515-DC Universe Online / \n", "6-Animation / \n", "25-Toy / 126-The Walt Disney Company / 354-Egg / 417-Play-Doh / 611-Kinder Surprise / \n", "36-Piano / 53-Keyboard / 104-Pianist / 107-Musical keyboard / \n", "10-Animal / 2735-Raccoon / \n", "93-Comedy / 355-Comedian / \n", "0-Games / 3235-Monopoly / \n", "13-Musical ensemble / \n", "6-Animation / 15-Cartoon / 2644-Fantage / \n", "10-Animal / 48-Pet / 71-Dog / 1839-Rottweiler / 2049-Police dog / \n", "20-Trailer / \n", "1-Vehicle / 4-Car / 11-Motorsport / \n", "256-Tennis / \n", "5-Dance / \n", "23-Mobile phone / 29-Smartphone / 37-Gadget / \n", "205-Newscaster / \n", "35-Minecraft / \n", "49-School / 835-Backpack / 2454-Notebook / \n", "0-Games / 2-Video game / 55-Video game console / 176-PlayStation 4 / 314-Xbox One / 386-Xbox / \n", "5-Dance / \n", "3-Concert / 7-Musician / \n", "1-Vehicle / 4-Car / 517-Audi / 2420-Audi TT / \n", "9-Music video / \n", "2-Video game / \n", "5-Dance / 21-Stadium / \n", "1-Vehicle / 4-Car / 11-Motorsport / 17-Racing / 57-Race track / 92-Radio-controlled model / 202-Radio-controlled car / 935-Off-road racing / 3419-Truggy / \n", "1-Vehicle / 4-Car / 234-Ford / 2288-Ford Escort / \n", "1-Vehicle / 17-Racing / 18-Outdoor recreation / 83-Skateboarding / 108-Skateboard / 555-Longboard / 667-Longboarding / \n", "2-Video game / 6-Animation / 20-Trailer / 320-Super Smash Bros. / \n", "3-Concert / \n", "25-Toy / 140-LEGO / 264-Star Wars / 852-Lego Star Wars / 3429-Yoda / \n", "46-Choir / \n", "0-Games / 781-Pachinko / \n", "0-Games / 2-Video game / 638-Gears of War / 773-Gears of War / 1425-Gears of War 2 / \n", "23-Mobile phone / 29-Smartphone / 37-Gadget / 101-Telephone / 178-Samsung Galaxy / 1379-HTC One / 2328-HTC One (M8) / \n", "812-Plastic / 1008-Kite / \n", "1-Vehicle / 4-Car / \n", "122-Weight training / \n", "122-Weight training / \n", "5-Dance / 16-Performance art / \n", "1-Vehicle / 4-Car / 1148-Hot rod / \n", "0-Games / 54-Highlight film / 60-Basketball / 76-Ball / 3291-Basketball Coach / \n", "31-Disc jockey / \n", "3-Concert / \n", "3-Concert / 5-Dance / \n", "12-Food / 1489-Coffeemaker / \n", "593-Moon / 1181-Mars / \n", "1-Vehicle / 11-Motorsport / 20-Trailer / 27-Motorcycle / 105-Motorcycling / 172-Motocross / \n", "6-Animation / 434-Ninja / \n", "0-Games / 1029-Aion: The Tower of Eternity / \n", "56-Hair / 106-Hairstyle / 405-Braid / \n", "12-Food / 113-House / 3150-Supper / \n", "31-Disc jockey / \n", "18-Outdoor recreation / 59-Winter sport / 112-Ice skating / 262-Ice / 316-Figure skating / 833-Ice skate / \n", "2-Video game / 243-Sonic the Hedgehog / 2028-Sonic World Adventure / \n", "3-Concert / 7-Musician / 30-Drummer / \n", "0-Games / 2-Video game / 19-PC game / 33-Weapon / 43-Call of Duty / 96-Soldier / 194-Call of Duty: Modern Warfare 2 / \n", "0-Games / 51-Strategy video game / 643-Hero / 1687-Heroes of Newerth / \n", "1-Vehicle / 4-Car / 11-Motorsport / 41-Sports car / 70-Driving / 245-BMW / 344-Coupé / 3549-BMW M6 / \n", "6-Animation / 15-Cartoon / 61-Art / 89-Comics / 123-Naruto / 183-Manga / 285-Sasuke Uchiha / 1556-Dōjinshi / \n", "53-Keyboard / 107-Musical keyboard / \n", "0-Games / 1-Vehicle / 2-Video game / 4-Car / 34-Action-adventure game / 47-Personal computer / 297-Grand Theft Auto IV / \n", "0-Games / 88-Machine / 535-Slot machine / 3599-The Wonderful Wizard of Oz / \n", "18-Outdoor recreation / 42-Fishing / 548-Parachuting / \n", "0-Games / 19-PC game / 332-Mercedes-Benz / 411-MapleStory / \n", "3092-Cinderella / \n", "3-Concert / 7-Musician / 30-Drummer / 39-Drums / 44-Drums / \n", "1-Vehicle / 4-Car / 11-Motorsport / 17-Racing / 40-Road / 41-Sports car / 57-Race track / 231-Supercar / 270-Sedan / 378-Drag racing / 1122-Mitsubishi Lancer Evolution / \n", "1-Vehicle / 103-Rail transport / \n", "372-Playing card / 1191-Tarot / \n", "9-Music video / \n", "376-Flute / 4605-Bansuri / \n", "3-Concert / 7-Musician / 13-Musical ensemble / 30-Drummer / 39-Drums / \n", "5-Dance / 16-Performance art / \n", "1-Vehicle / 4-Car / 270-Sedan / 344-Coupé / 1466-Pontiac / 1493-Supercharger / \n", "7-Musician / 14-Guitar / 24-String instrument / 63-Acoustic guitar / \n", "0-Games / 2-Video game / 20-Trailer / 35-Minecraft / 843-The Doctor / \n", "42-Fishing / \n", "3-Concert / \n", "0-Games / 211-RuneScape / 1035-Bank / \n", "25-Toy / 89-Comics / 435-Action figure / 2121-Marvel Legends / \n", "93-Comedy / 341-Sketch comedy / 355-Comedian / 416-Apartment / \n", "117-Boxing / 399-Kickboxing / \n", "237-Final Fantasy / 730-Final Fantasy VII / \n", "7-Musician / 14-Guitar / 100-Electric guitar / 606-String / 4002-Eight-string guitar / \n", "75-Wrestling / \n", "61-Art / 148-Painting / 303-Paint / 3928-Still life / \n", "0-Games / 60-Basketball / 168-Basketball moves / 255-Slam dunk / \n", "23-Mobile phone / 29-Smartphone / 37-Gadget / 72-iPhone / \n", "1-Vehicle / 10-Animal / 48-Pet / 71-Dog / 87-Boat / \n", "0-Games / 3-Concert / 60-Basketball / \n", "0-Games / 2-Video game / 779-Point Blank / \n", "10-Animal / 80-Horse / 438-Mare / \n", "0-Games / 2-Video game / 34-Action-adventure game / 297-Grand Theft Auto IV / \n", "1-Vehicle / 262-Ice / \n", "36-Piano / 53-Keyboard / \n", "59-Winter sport / 112-Ice skating / 316-Figure skating / \n", "36-Piano / \n", "0-Games / 411-MapleStory / \n", "0-Games / 2-Video game / 6-Animation / 15-Cartoon / \n", "0-Games / 2-Video game / 55-Video game console / 199-Wii / 243-Sonic the Hedgehog / 292-Handheld game console / 335-Nintendo 3DS / 3016-Mario & Sonic at the Olympic Games / \n", "0-Games / 8-Football / \n", "1-Vehicle / 124-Tractor / \n", "0-Games / 2-Video game / 23-Mobile phone / 29-Smartphone / 37-Gadget / 178-Samsung Galaxy / 3361-Samsung Galaxy Y / \n", "5-Dance / 66-Bollywood / 78-Wedding / \n", "5-Dance / 13-Musical ensemble / 44-Drums / 253-Carnival / \n", "120-Winter / 253-Carnival / \n", "1-Vehicle / 4-Car / 40-Road / 70-Driving / 74-Truck / 84-Snow / 120-Winter / 1145-Winter storm / \n", "0-Games / 2-Video game / 19-PC game / 34-Action-adventure game / 2205-Bully / \n", "230-Brass instrument / 296-Saxophone / 1006-Tenor saxophone / \n", "1-Vehicle / 4-Car / 25-Toy / 155-Camera / 1169-Hot Wheels / \n", "170-Home improvement / 1314-Garage / \n", "3-Concert / 5-Dance / 16-Performance art / 133-Nightclub / \n", "3-Concert / 16-Performance art / \n", "28-Fashion / 1459-Magazine / \n", "0-Games / 2-Video game / 55-Video game console / 111-PlayStation 3 / 127-Xbox 360 / 181-Xbox / \n", "0-Games / 75-Wrestling / 116-Pokémon / 186-Pokémon / 1204-Pokémon Ruby and Sapphire / 1665-Pokémon Emerald / 2236-Emerald / \n", "3-Concert / 16-Performance art / \n", "30-Drummer / 39-Drums / 44-Drums / 146-Snare drum / \n", "36-Piano / \n", "31-Disc jockey / \n", "42-Fishing / 222-Surfing / 815-Surfboard / \n", "5-Dance / \n", "2-Video game / 4550-Conker's Bad Fur Day / \n", "5-Dance / \n", "287-Human swimming / 2605-Medley swimming / \n", "6-Animation / \n", "3-Concert / 5-Dance / 16-Performance art / 95-Talent show / \n", "0-Games / 35-Minecraft / \n", "0-Games / \n", "3-Concert / 7-Musician / \n", "0-Games / 8-Football / 21-Stadium / 65-Kick / 595-Champion / \n", "1-Vehicle / 33-Weapon / 271-Tank / 1409-Missile / 1894-Cannon / \n", "6-Animation / \n", "3-Concert / 7-Musician / \n", "1-Vehicle / 11-Motorsport / 27-Motorcycle / 105-Motorcycling / 164-Beach / \n", "6-Animation / 9-Music video / 967-Text / \n", "2411-Howcast / \n", "1-Vehicle / 4-Car / 11-Motorsport / 41-Sports car / 297-Grand Theft Auto IV / 658-Grand Theft Auto: The Lost and Damned / \n", "3-Concert / \n", "0-Games / 2-Video game / 241-Fighting game / 511-Mortal Kombat / 3859-Mortal Kombat II / 4684-Kung Lao / \n", "9-Music video / \n", "23-Mobile phone / 29-Smartphone / 37-Gadget / 72-iPhone / 189-iPad / 272-iPod / 331-iPod touch / \n", "0-Games / 211-RuneScape / \n", "2-Video game / 20-Trailer / 34-Action-adventure game / 380-Assassin's Creed / 1241-Assassin's Creed IV: Black Flag / \n", "860-Christian Church / \n", "6-Animation / 15-Cartoon / 396-One Piece / \n", "10-Animal / 18-Outdoor recreation / 42-Fishing / 91-Fish / \n", "36-Piano / 53-Keyboard / \n", "5-Dance / 20-Trailer / \n", "1-Vehicle / 4-Car / 11-Motorsport / 17-Racing / 40-Road / 41-Sports car / 207-Exhaust system / 344-Coupé / 349-Chevrolet / 482-Asphalt / 952-Chevrolet Corvette / 1040-Street racing / \n", "0-Games / 2-Video game / 19-PC game / 33-Weapon / 928-Borderlands 2 / 2613-Borderlands: The Pre-Sequel! / \n", "3-Concert / 7-Musician / 14-Guitar / 39-Drums / \n", "6-Animation / 25-Toy / 89-Comics / 396-One Piece / 2525-Gashapon / \n", "0-Games / 60-Basketball / 77-Arena / 168-Basketball moves / \n", "653-Thread / 1082-Knot / 2507-Macramé / 3063-Friendship bracelet / \n", "1-Vehicle / 59-Winter sport / 84-Snow / 120-Winter / 610-Snowmobile / 1201-Ski-Doo / \n", "7-Musician / 107-Musical keyboard / 219-Accordion / \n", "28-Fashion / 120-Winter / 757-Boot / \n", "0-Games / 8-Football / 21-Stadium / 65-Kick / 73-Ball / 76-Ball / 81-Athlete / \n", "6-Animation / \n", "5-Dance / \n", "0-Games / 2-Video game / 1983-Infestation: Survivor Stories / \n", "0-Games / 2-Video game / 1118-Kirby / 1133-Kirby / 4449-King Dedede / \n", "3-Concert / \n", "3-Concert / \n", "23-Mobile phone / 29-Smartphone / 37-Gadget / 72-iPhone / 508-iPhone 5 / 2038-Credit card / \n", "0-Games / 1-Vehicle / 2-Video game / 6-Animation / 50-Aircraft / 82-Airplane / 129-Aviation / 451-Cockpit / \n", "6-Animation / 15-Cartoon / 4122-Marceline the Vampire Queen / \n", "3-Concert / 7-Musician / 13-Musical ensemble / \n", "36-Piano / 53-Keyboard / 104-Pianist / 107-Musical keyboard / \n", "1465-FOX / \n", "121-Photography / 155-Camera / \n", "0-Games / 2-Video game / 2774-Tom Clancy's Rainbow Six / \n", "12-Food / 26-Cooking / 32-Recipe / 52-Dish / 109-Cooking show / 179-Kitchen / 597-Soup / 742-Countertop / 1108-Mushroom / 2309-Low-carbohydrate diet / \n", "28-Fashion / 60-Basketball / 206-Shoe / 333-Nike; Inc. / 368-Sneakers / \n", "0-Games / 2-Video game / 34-Action-adventure game / 1125-God of War / 1743-God of War / \n", "9-Music video / 14-Guitar / 24-String instrument / \n", "102-Building / 162-Medicine / 209-University / \n", "5-Dance / \n", "1-Vehicle / 11-Motorsport / 27-Motorcycle / 105-Motorcycling / 154-Wheel / 158-Tire / 207-Exhaust system / 806-Chopper / \n", "3-Concert / 7-Musician / 978-Love song / \n", "5-Dance / \n", "5-Dance / 16-Performance art / \n", "0-Games / 2-Video game / 19-PC game / 51-Strategy video game / 603-StarCraft II: Wings of Liberty / 732-StarCraft / 2353-Heroes of the Storm / 2647-StarCraft II: Heart of the Swarm / \n", "7-Musician / 13-Musical ensemble / 24-String instrument / 38-Orchestra / 150-Violin / 450-Viola / 639-Fiddle / \n", "0-Games / 2-Video game / 6-Animation / 701-Beyblade / \n", "0-Games / 51-Strategy video game / 382-Dota 2 / 549-Defense of the Ancients / \n", "0-Games / 116-Pokémon / 186-Pokémon / 4521-Deoxys / \n", "0-Games / 2-Video game / 33-Weapon / \n", "6-Animation / 97-Drawing / 538-Illustration / 2945-Fan art / 3865-Blue Dragon / \n", "31-Disc jockey / \n", "5-Dance / 9-Music video / \n", "2-Video game / \n", "12-Food / 1346-Wood carving / \n", "12-Food / 26-Cooking / 32-Recipe / 442-Bread / \n", "20-Trailer / \n", "0-Games / 8-Football / 65-Kick / 1247-Penalty kick / \n", "5-Dance / 9-Music video / \n", "0-Games / 2-Video game / 33-Weapon / 43-Call of Duty / 194-Call of Duty: Modern Warfare 2 / \n", "359-Harry Potter (Literary Series) / 4489-Harry Potter and the Philosopher's Stone / \n", "67-Cycling / 69-Bicycle / 88-Machine / 1067-Lock / 2516-Lock picking / \n", "61-Art / 148-Painting / \n", "6-Animation / 2984-.hack / \n", "46-Choir / \n", "0-Games / 2-Video game / 20-Trailer / 111-PlayStation 3 / \n", "0-Games / 35-Minecraft / 55-Video game console / 127-Xbox 360 / 181-Xbox / \n", "6-Animation / \n", "0-Games / 2-Video game / 6-Animation / 514-Tekken / 797-Tekken / 1273-Tekken 6 / \n", "7-Musician / \n", "31-Disc jockey / \n", "0-Games / 54-Highlight film / 163-Running / \n", "42-Fishing / 275-Diving / 381-Underwater diving / 407-Scuba diving / \n", "6-Animation / 15-Cartoon / 89-Comics / \n", "6-Animation / \n", "273-News program / 3257-Hindu temple / \n", "0-Games / 8-Football / \n", "36-Piano / 53-Keyboard / 107-Musical keyboard / 265-Electronic keyboard / \n", "6-Animation / 25-Toy / \n", "0-Games / 2-Video game / 1137-Crash Bandicoot / \n", "9-Music video / \n", "364-Sitcom / \n", "3-Concert / 7-Musician / 13-Musical ensemble / 36-Piano / 53-Keyboard / 104-Pianist / 107-Musical keyboard / \n", "10-Animal / 18-Outdoor recreation / 80-Horse / 247-Stallion / 289-Dressage / 404-Livestock / 438-Mare / \n", "1-Vehicle / 4-Car / \n", "66-Bollywood / \n", "3-Concert / 13-Musical ensemble / \n", "0-Games / 2-Video game / 19-PC game / 33-Weapon / 43-Call of Duty / 47-Personal computer / 96-Soldier / 302-Call of Duty 4: Modern Warfare / \n", "10-Animal / 22-Nature / 537-Zoo / \n", "28-Fashion / 56-Hair / 106-Hairstyle / \n", "1271-Beauty salon / \n", "0-Games / 59-Winter sport / 81-Athlete / 112-Ice skating / 142-Hockey / 262-Ice / 312-Ice rink / \n", "609-Tattoo / 863-Body piercing / \n", "1-Vehicle / 4-Car / 84-Snow / 277-Sport utility vehicle / \n", "59-Winter sport / 281-Snowboarding / 4321-Mountainboarding / \n", "1-Vehicle / 11-Motorsport / 27-Motorcycle / 2857-Yamaha Raptor 700R / \n", "3-Concert / 7-Musician / 1666-Hammond organ / \n", "122-Weight training / 204-Gym / 214-Muscle / 469-Bag / 529-Squat / 532-Barbell / \n", "1-Vehicle / 4-Car / 11-Motorsport / 41-Sports car / 207-Exhaust system / 231-Supercar / 344-Coupé / 517-Audi / 1537-Audi R8 / 2949-V10 engine / \n", "6-Animation / 15-Cartoon / 412-Monster / 503-Monster High / \n", "676-Handbag / \n", "8-Football / 21-Stadium / \n", "25-Toy / 140-LEGO / 264-Star Wars / 852-Lego Star Wars / 2576-Clone trooper / \n", "8-Football / 81-Athlete / \n", "31-Disc jockey / \n", "12-Food / 1870-Chewing gum / 1943-Bakery / \n", "1-Vehicle / 4-Car / 74-Truck / 92-Radio-controlled model / 167-Four-wheel drive / 202-Radio-controlled car / \n", "23-Mobile phone / 29-Smartphone / \n", "1-Vehicle / 4-Car / 11-Motorsport / 17-Racing / 57-Race track / 651-Stock car racing / 728-Dirt track racing / \n", "0-Games / 1467-Pump It Up / 2651-Pump It Up Fiesta / 3591-Pump It Up Fiesta 2 / \n", "0-Games / 2-Video game / 19-PC game / 928-Borderlands 2 / \n", "93-Comedy / \n", "0-Games / 2-Video game / 127-Xbox 360 / 445-Guitar Hero / 965-Guitar Hero III: Legends of Rock / \n", "31-Disc jockey / \n", "8-Football / 21-Stadium / \n", "3-Concert / 16-Performance art / 93-Comedy / \n", "7-Musician / 14-Guitar / 24-String instrument / 63-Acoustic guitar / \n", "46-Choir / \n", "6-Animation / 2435-Tokyo Ghoul / \n", "36-Piano / 53-Keyboard / 104-Pianist / 107-Musical keyboard / \n", "12-Food / 32-Recipe / 617-Maize / 684-Ice cream / \n", "5-Dance / \n", "455-Unidentified flying object / 2918-Aurora / \n", "1-Vehicle / 189-iPad / 4558-Spektrum RC / \n", "6-Animation / 15-Cartoon / 61-Art / 97-Drawing / 408-Sketch / 528-Pencil / \n", "9-Music video / \n", "49-School / \n", "14-Guitar / 63-Acoustic guitar / 361-Flamenco / 2380-Flamenco guitar / \n", "14-Guitar / 24-String instrument / 100-Electric guitar / \n", "3-Concert / 7-Musician / 13-Musical ensemble / 14-Guitar / 361-Flamenco / \n", "102-Building / 113-House / 521-Metal / 2428-Barn / \n", "102-Building / 113-House / 170-Home improvement / 276-Room / 353-Wall / 1350-Plaster / \n", "86-Plant / 137-Tree / 1208-Coconut / \n", "163-Running / 674-Marathon / \n", "68-Lighting / 263-Light / \n", "0-Games / 1-Vehicle / 2-Video game / 11-Motorsport / 17-Racing / 27-Motorcycle / 57-Race track / 172-Motocross / 2930-MX vs. ATV / \n", "23-Mobile phone / 29-Smartphone / \n", "6-Animation / 15-Cartoon / 20-Trailer / 123-Naruto / 285-Sasuke Uchiha / 3232-Naruto Uzumaki / \n", "0-Games / 2-Video game / 584-Super Mario Bros. / 1170-Super Mario World / \n", "6-Animation / 15-Cartoon / 396-One Piece / \n", "5-Dance / \n", "12-Food / 26-Cooking / 32-Recipe / 52-Dish / 58-Cuisine / 454-Sauce / 620-Barbecue / 966-Sandwich / 1517-Spaghetti / \n", "23-Mobile phone / 29-Smartphone / 72-iPhone / 760-iPhone 5s / \n", "56-Hair / 106-Hairstyle / \n", "0-Games / 2-Video game / 19-PC game / 638-Gears of War / 773-Gears of War / \n", "313-Book / \n", "55-Video game console / 581-PlayStation / \n", "56-Hair / 106-Hairstyle / 1359-Hair iron / \n", "85-Combat / 117-Boxing / 399-Kickboxing / \n", "6-Animation / 28-Fashion / \n", "6-Animation / 126-The Walt Disney Company / \n", "2-Video game / 47-Personal computer / 131-Computer / \n", "0-Games / 489-Board game / 922-Tabletop game / \n", "3343-The Real Housewives / \n", "1-Vehicle / 4-Car / 74-Truck / 158-Tire / 167-Four-wheel drive / 375-Pickup truck / 3541-Nissan Navara / \n", "0-Games / 2-Video game / 19-PC game / 33-Weapon / 43-Call of Duty / 96-Soldier / 194-Call of Duty: Modern Warfare 2 / 302-Call of Duty 4: Modern Warfare / \n", "89-Comics / 1791-Stargate / \n", "3770-Toothpaste / \n", "47-Personal computer / 831-Plants vs. Zombies / \n", "756-Tibia / 1356-Paladin / \n", "3-Concert / 7-Musician / 13-Musical ensemble / 14-Guitar / 24-String instrument / 30-Drummer / 39-Drums / 44-Drums / \n", "0-Games / 467-Chipmunk / 3104-Tree squirrel / \n", "25-Toy / 83-Skateboarding / 1105-Fingerboard / \n", "1-Vehicle / 4-Car / 299-Volkswagen Passenger Cars / 526-Volkswagen / 1761-Volkswagen Jetta / \n", "347-Climbing / \n", "8-Football / 73-Ball / \n", "102-Building / 113-House / 310-The Sims / 328-Simulation video game / 564-The Sims 3 / \n", "3-Concert / 16-Performance art / 30-Drummer / 68-Lighting / \n", "1-Vehicle / 18-Outdoor recreation / 42-Fishing / 50-Aircraft / 548-Parachuting / 573-Paragliding / 765-Parachute / 778-Gliding / 1561-Paramotor / 1602-Hang gliding / \n", "1-Vehicle / \n", "1-Vehicle / 4-Car / 270-Sedan / 332-Mercedes-Benz / 891-Mercedes-AMG / 1310-Mercedes-Benz C-Class / 2317-Twin-turbo / 2365-Mercedes-Benz S-Class / \n", "9-Music video / \n", "0-Games / 1-Vehicle / 2-Video game / 4-Car / 1091-Saints Row / 2470-Hitman: Absolution / 2495-Saints Row 2 / \n", "1109-Counter-Strike Online / 3712-Labyrinth / \n", "10-Animal / 17-Racing / 80-Horse / 384-Horse racing / \n", "14-Guitar / 56-Hair / 106-Hairstyle / \n", "402-Jewellery / 859-Beadwork / 907-Necklace / \n", "25-Toy / 749-Figurine / \n", "78-Wedding / 190-Bride / \n", "0-Games / 8-Football / 54-Highlight film / 76-Ball / 767-Goalkeeper / \n", "1-Vehicle / 4-Car / 11-Motorsport / 158-Tire / \n", "1-Vehicle / 67-Cycling / 69-Bicycle / 573-Paragliding / 1561-Paramotor / \n", "452-Watch / 2168-G-Shock / \n", "0-Games / 59-Winter sport / 81-Athlete / 112-Ice skating / 142-Hockey / 312-Ice rink / \n", "0-Games / 19-PC game / 51-Strategy video game / 165-League of Legends / \n", "5-Dance / 16-Performance art / \n", "3-Concert / \n", "20-Trailer / \n", "0-Games / 2-Video game / 33-Weapon / 1382-Half-Life 2 / \n", "12-Food / 56-Hair / 106-Hairstyle / 340-Afro-textured hair / 857-Afro / \n", "3-Concert / 5-Dance / \n", "22-Nature / 86-Plant / \n", "0-Games / 2-Video game / 34-Action-adventure game / 199-Wii / 1447-Skylanders / 2321-Skylanders: Spyro's Adventure / \n", "10-Animal / 48-Pet / 71-Dog / 182-Puppy / 3046-Maltese / \n", "61-Art / 82-Airplane / 242-Paper / 370-Origami / 1933-Paper plane / \n", "3-Concert / 5-Dance / \n", "9-Music video / \n", "110-Album / \n", "45-Cosmetics / 235-Lipstick / 291-Rouge / 1012-Lip gloss / \n", "122-Weight training / 214-Muscle / 529-Squat / \n", "6-Animation / 15-Cartoon / 1599-Bone / \n", "0-Games / 1-Vehicle / 2-Video game / 4-Car / 19-PC game / 34-Action-adventure game / 111-PlayStation 3 / 118-Grand Theft Auto V / \n", "31-Disc jockey / \n", "6-Animation / 10-Animal / 15-Cartoon / 3175-Hide-and-seek / \n", "1-Vehicle / 4-Car / 41-Sports car / 650-Convertible / \n", "14-Guitar / \n", "6-Animation / 15-Cartoon / 4410-PriPara / \n", "0-Games / 2-Video game / 19-PC game / 310-The Sims / 328-Simulation video game / 1713-The Sims / \n", "42-Fishing / 164-Beach / 213-Forest / \n", "260-Prayer / \n", "56-Hair / 106-Hairstyle / \n", "1-Vehicle / 4-Car / 11-Motorsport / 27-Motorcycle / 88-Machine / 608-Wheelie / \n", "31-Disc jockey / \n", "28-Fashion / 45-Cosmetics / 56-Hair / 106-Hairstyle / 583-Wig / 607-Lace / \n", "0-Games / 756-Tibia / \n", "3-Concert / 5-Dance / \n", "0-Games / 60-Basketball / 76-Ball / \n", "1-Vehicle / 4-Car / 11-Motorsport / 17-Racing / 27-Motorcycle / 70-Driving / 212-Highway / 2892-GSX-R750 / \n", "390-Model / \n", "83-Skateboarding / 108-Skateboard / 482-Asphalt / 555-Longboard / 556-Skateboarding trick / 667-Longboarding / 1315-Kickflip / \n", "0-Games / 1-Vehicle / 2-Video game / 4-Car / 19-PC game / 34-Action-adventure game / 118-Grand Theft Auto V / 1035-Bank / \n", "1-Vehicle / 50-Aircraft / 82-Airplane / 139-Airport / 151-Landing / 157-Jet aircraft / 175-Airliner / 188-Airline / 1340-Airbus A380 / \n", "0-Games / 2-Video game / 33-Weapon / 229-Halo / 470-Halo 3 / 644-Halo: Reach / \n", "0-Games / 4554-Marvel Super Hero Squad Online / \n", "10-Animal / 22-Nature / 86-Plant / 540-Insect / 642-Leaf / 775-Bee / 784-Honey / 1121-Beehive / \n", "6-Animation / 1054-Vacuum cleaner / \n", "56-Hair / 106-Hairstyle / 583-Wig / \n", "1-Vehicle / 62-Train / 64-Transport / 103-Rail transport / 143-Railroad car / 428-Rail transport modelling / 1450-Model building / \n", "1-Vehicle / 11-Motorsport / 17-Racing / 18-Outdoor recreation / 27-Motorcycle / 105-Motorcycling / 172-Motocross / 491-Enduro / \n", "6-Animation / \n", "61-Art / 148-Painting / \n", "247-Stallion / 438-Mare / 1871-Foal / 2018-Arabian horse / \n", "5-Dance / \n", "9-Music video / \n", "3-Concert / 5-Dance / 7-Musician / 13-Musical ensemble / 38-Orchestra / 130-Gymnastics / 1835-Majorette / \n", "1-Vehicle / 4-Car / 114-Engine / 517-Audi / \n", "30-Drummer / 39-Drums / 44-Drums / 128-Cymbal / 146-Snare drum / \n", "66-Bollywood / \n", "3-Concert / 68-Lighting / \n", "8-Football / 21-Stadium / \n", "0-Games / 2-Video game / 116-Pokémon / \n", "113-House / 1731-Haunted house / \n", "9-Music video / \n", "0-Games / 2-Video game / 19-PC game / 506-Garry's Mod / 2290-Trouble in Terrorist Town / 3841-The Hidden / \n", "0-Games / 2-Video game / 127-Xbox 360 / \n", "5-Dance / 889-Swing / \n", "5-Dance / 219-Accordion / \n", "0-Games / 756-Tibia / \n", "101-Telephone / 131-Computer / \n", "9-Music video / \n", "20-Trailer / \n", "75-Wrestling / 85-Combat / \n", "0-Games / 2-Video game / 10-Animal / 80-Horse / 384-Horse racing / \n", "9-Music video / \n", "0-Games / 2-Video game / 89-Comics / 396-One Piece / \n", "2-Video game / 20-Trailer / 89-Comics / 294-Batman / 838-Batman: Arkham / 2082-Catwoman / 2813-Batmobile / 3473-Nightwing / \n", "12-Food / \n", "30-Drummer / 39-Drums / 44-Drums / 128-Cymbal / \n", "1-Vehicle / 4-Car / 11-Motorsport / 17-Racing / 41-Sports car / 57-Race track / 378-Drag racing / 449-Honda / 552-Honda Civic / \n", "1-Vehicle / 4-Car / 17-Racing / 92-Radio-controlled model / 202-Radio-controlled car / \n", "0-Games / 1-Vehicle / 2-Video game / 4-Car / 11-Motorsport / 17-Racing / 57-Race track / 413-PlayStation 2 / 3510-Ridge Racer / \n", "6-Animation / 15-Cartoon / \n", "0-Games / 2-Video game / 6-Animation / 241-Fighting game / 514-Tekken / 797-Tekken / 1273-Tekken 6 / \n", "25-Toy / 626-Puzzle / 631-Cube / 740-Rubik's Cube / \n", "18-Outdoor recreation / 22-Nature / 147-River / 213-Forest / \n", "6-Animation / 3144-Tinker Bell / \n", "55-Video game console / 127-Xbox 360 / 181-Xbox / 832-Headset / \n", "110-Album / \n", "3-Concert / 68-Lighting / \n", "45-Cosmetics / 173-Eye shadow / 193-Eye liner / 195-Mascara / 196-Eye / 218-Eyelash / 291-Rouge / \n", "0-Games / 2-Video game / 20-Trailer / 34-Action-adventure game / 111-PlayStation 3 / 127-Xbox 360 / 656-Metal Gear / 1419-GameSpot / 2698-Metal Gear Rising: Revengeance / \n", "0-Games / 60-Basketball / 168-Basketball moves / \n", "9-Music video / 14-Guitar / \n", "30-Drummer / 39-Drums / 44-Drums / 128-Cymbal / 146-Snare drum / 735-Kamen Rider Series / 1360-Magician / \n", "0-Games / 2-Video game / 33-Weapon / 43-Call of Duty / 138-Call of Duty: Black Ops II / \n", "9-Music video / \n", "3-Concert / 44-Drums / \n", "49-School / \n", "3-Concert / 16-Performance art / \n", "3-Concert / \n", "6-Animation / 15-Cartoon / 259-Fire / 434-Ninja / \n", "12-Food / 26-Cooking / \n", "1-Vehicle / 4-Car / \n", "3-Concert / 7-Musician / \n", "1-Vehicle / 4-Car / 11-Motorsport / 17-Racing / 169-Rallying / \n", "0-Games / 35-Minecraft / \n", "45-Cosmetics / \n", "0-Games / 2-Video game / 34-Action-adventure game / \n", "0-Games / 30-Drummer / 39-Drums / 44-Drums / 128-Cymbal / 146-Snare drum / \n", "0-Games / 6-Animation / 97-Drawing / 1563-Stick figure / \n", "1-Vehicle / 50-Aircraft / 129-Aviation / 388-Fighter aircraft / 2954-McDonnell Douglas F-15 Eagle / 4141-Lock On: Modern Air Combat / \n", "5-Dance / \n", "8-Football / 54-Highlight film / \n", "5-Dance / 430-Ballroom dance / \n", "1-Vehicle / 27-Motorcycle / 154-Wheel / 1124-Yamaha YZF-R1 / \n", "1-Vehicle / 4-Car / 270-Sedan / 332-Mercedes-Benz / 1310-Mercedes-Benz C-Class / 2603-Mercedes-Benz E-Class / \n", "6-Animation / 25-Toy / 140-LEGO / 434-Ninja / 865-Turtle / 884-Lego minifigure / 951-Teenage Mutant Ninja Turtles / 3667-Raphael / \n", "30-Drummer / 39-Drums / 44-Drums / 128-Cymbal / 146-Snare drum / \n", "1-Vehicle / 50-Aircraft / 82-Airplane / 129-Aviation / 139-Airport / 157-Jet aircraft / 175-Airliner / 187-Takeoff / 254-Runway / 388-Fighter aircraft / \n", "12-Food / 26-Cooking / 32-Recipe / 52-Dish / 58-Cuisine / 210-Cookware and bakeware / 274-Meat / 454-Sauce / \n", "1-Vehicle / 4-Car / 40-Road / 70-Driving / 84-Snow / \n", "123-Naruto / \n", "20-Trailer / \n", "3-Concert / 16-Performance art / \n", "12-Food / 26-Cooking / 32-Recipe / 52-Dish / 58-Cuisine / 109-Cooking show / 210-Cookware and bakeware / 239-Roasting / 274-Meat / 542-Potato / 1509-Stew / \n", "1-Vehicle / 4-Car / 349-Chevrolet / \n", "6-Animation / 2408-Lucky Star / \n", "28-Fashion / 209-University / \n", "3-Concert / \n", "0-Games / 2-Video game / 19-PC game / 928-Borderlands 2 / \n", "0-Games / 2-Video game / 43-Call of Duty / 196-Eye / 609-Tattoo / \n", "162-Medicine / 780-Fat / 1162-Dietary supplement / \n", "3-Concert / 133-Nightclub / \n", "25-Toy / 264-Star Wars / 435-Action figure / 2576-Clone trooper / \n", "0-Games / 8-Football / \n", "96-Soldier / 758-Air force / 2202-Armed forces / \n", "0-Games / 2-Video game / 33-Weapon / \n", "8-Football / \n", "86-Plant / 136-Gardening / 137-Tree / 213-Forest / 279-Garden / \n", "10-Animal / 48-Pet / 970-Monkey / 1948-Ape / 3157-Chimpanzee / \n", "3-Concert / 5-Dance / 31-Disc jockey / 133-Nightclub / \n", "7-Musician / 13-Musical ensemble / 30-Drummer / 44-Drums / 296-Saxophone / \n", "2-Video game / 739-Guild Wars / \n", "5-Dance / 16-Performance art / 115-Ballet / \n", "125-Television / \n", "93-Comedy / 125-Television / \n", "1-Vehicle / 4-Car / 34-Action-adventure game / 41-Sports car / 297-Grand Theft Auto IV / \n", "1-Vehicle / 2-Video game / 864-Speedometer / 1099-Euro Truck Simulator 2 / \n", "319-Robot / \n", "20-Trailer / \n", "2207-Kemenche / \n", "6-Animation / 35-Minecraft / \n", "5-Dance / 16-Performance art / \n", "279-Garden / 585-Lawn / \n", "45-Cosmetics / 173-Eye shadow / 193-Eye liner / 195-Mascara / 196-Eye / 218-Eyelash / 291-Rouge / \n", "12-Food / 26-Cooking / 32-Recipe / 52-Dish / 58-Cuisine / 86-Plant / 210-Cookware and bakeware / 620-Barbecue / 662-Pizza / 752-Pasta / 1108-Mushroom / 1230-Italian food / 3021-Tomato sauce / \n", "61-Art / 479-Graffiti / \n", "0-Games / 2176-Fist of the North Star / \n", "3-Concert / \n", "7-Musician / 46-Choir / \n", "3-Concert / \n", "1-Vehicle / 259-Fire / 293-Slide show / 371-Firefighter / 796-Fire engine / 1020-Emergency vehicle / 1633-Fire station / \n", "14-Guitar / 1970-Axe / \n", "25-Toy / 524-Littlest Pet Shop / 587-Clay / \n", "10-Animal / 80-Horse / \n", "20-Trailer / \n", "162-Medicine / \n", "1-Vehicle / 50-Aircraft / 82-Airplane / 129-Aviation / 157-Jet aircraft / 388-Fighter aircraft / 929-Aerobatics / \n", "0-Games / 10-Animal / \n", "0-Games / 2-Video game / 20-Trailer / 165-League of Legends / 2004-Riot Games / \n", "45-Cosmetics / 193-Eye liner / 235-Lipstick / 2716-The Hunger Games / \n", "6-Animation / \n", "1-Vehicle / 27-Motorcycle / 67-Cycling / 69-Bicycle / 114-Engine / 154-Wheel / 520-Bicycle frame / \n", "1163-Carpet / \n", "23-Mobile phone / 29-Smartphone / 37-Gadget / 72-iPhone / 272-iPod / 331-iPod touch / 675-iPhone 4S / \n", "352-Need for Speed / 1583-Toyota Corolla / 1600-Need for Speed: World / \n", "23-Mobile phone / 29-Smartphone / 345-Loudspeaker / \n", "28-Fashion / 78-Wedding / 190-Bride / 590-Wedding dress / \n", "1-Vehicle / 4-Car / 3762-PSA HDi engine / \n", "1-Vehicle / 17-Racing / 27-Motorcycle / 67-Cycling / 69-Bicycle / 570-Road bicycle racing / \n", "8-Football / \n", "6-Animation / 15-Cartoon / 61-Art / \n", "0-Games / 55-Video game console / 199-Wii / 2750-Mii / 3226-Chucky / \n", "12-Food / 26-Cooking / 32-Recipe / 52-Dish / 58-Cuisine / \n", "36-Piano / 53-Keyboard / 104-Pianist / 107-Musical keyboard / 265-Electronic keyboard / \n", "1-Vehicle / 50-Aircraft / 82-Airplane / 139-Airport / 151-Landing / 187-Takeoff / 188-Airline / 451-Cockpit / \n", "5-Dance / \n", "1-Vehicle / 67-Cycling / 69-Bicycle / 197-Trail / 213-Forest / 220-Mountain bike / 284-Mountain biking / \n", "0-Games / 379-Touhou Project / 455-Unidentified flying object / 2170-The Embodiment of Scarlet Devil / \n", "426-Resident Evil / 3514-Resident Evil: Revelations / \n", "122-Weight training / \n", "36-Piano / 53-Keyboard / 104-Pianist / 107-Musical keyboard / \n", "10-Animal / 48-Pet / 71-Dog / 182-Puppy / 824-Terrier / 2388-Bull Terrier / 3052-Staffordshire Bull Terrier / \n", "75-Wrestling / \n", "45-Cosmetics / 56-Hair / \n", "45-Cosmetics / 238-Nail / 300-Nail art / 304-Nail polish / 307-Hand / 324-Finger / 338-Manicure / 823-Artificial nails / \n", "238-Nail / 300-Nail art / 304-Nail polish / \n", "18-Outdoor recreation / 42-Fishing / 164-Beach / 222-Surfing / 233-Ocean / 672-Bodyboarding / \n", "0-Games / 73-Ball / 326-Metin2 / \n", "5-Dance / 16-Performance art / \n", "170-Home improvement / 488-Door / 1078-Installation art / 1314-Garage / 3457-Garage door / \n", "31-Disc jockey / 861-Mixing console / \n", "12-Food / 26-Cooking / 32-Recipe / 52-Dish / 58-Cuisine / 215-Eating / 301-Meal / 389-Restaurant / 589-Grilling / 1438-Street food / \n", "1-Vehicle / 4-Car / 62-Train / 286-Bus / \n", "0-Games / 2-Video game / 35-Minecraft / 243-Sonic the Hedgehog / 2327-Sonic the Hedgehog / \n", "1-Vehicle / 87-Boat / 344-Coupé / 498-Motorboat / 579-Yacht / \n", "28-Fashion / 153-Dress / \n", "117-Boxing / \n", "0-Games / 2-Video game / 21-Stadium / 35-Minecraft / 77-Arena / 79-American football / 127-Xbox 360 / 181-Xbox / \n", "9-Music video / \n", "9-Music video / 541-Black-and-white / \n", "18-Outdoor recreation / 42-Fishing / \n", "370-Origami / \n", "14-Guitar / 100-Electric guitar / \n", "6-Animation / 1527-Pretty Cure / \n", "0-Games / 2-Video game / 34-Action-adventure game / 921-Grand Theft Auto: Vice City / \n", "1-Vehicle / 18-Outdoor recreation / 92-Radio-controlled model / 149-Model aircraft / 160-Radio-controlled aircraft / 266-Helicopter / \n", "0-Games / 2-Video game / 34-Action-adventure game / 85-Combat / 380-Assassin's Creed / 1140-Assassin's Creed III / 1241-Assassin's Creed IV: Black Flag / \n", "0-Games / 8-Football / 21-Stadium / 54-Highlight film / 65-Kick / \n", "2-Video game / 47-Personal computer / 208-Wood / 1045-Computer case / \n", "14-Guitar / 24-String instrument / 63-Acoustic guitar / \n", "0-Games / 2-Video game / 19-PC game / 33-Weapon / 96-Soldier / 257-Counter-Strike / 464-First-person Shooter / \n", "205-Newscaster / \n", "0-Games / 8-Football / 21-Stadium / 81-Athlete / \n", "117-Boxing / \n", "10-Animal / 48-Pet / 166-Cat / 346-Kitten / 2924-Persian cat / \n", "0-Games / 2-Video game / 55-Video game console / 292-Handheld game console / 1446-Game Boy / \n", "0-Games / 2-Video game / 19-PC game / 33-Weapon / 85-Combat / 693-Sword / 1166-Dark Souls II / \n", "0-Games / 8-Football / \n", "3-Concert / 5-Dance / \n", "0-Games / 2-Video game / 19-PC game / 35-Minecraft / \n", "6-Animation / 325-Vampire / 4524-Rosario + Vampire / \n", "0-Games / 8-Football / 21-Stadium / 54-Highlight film / 79-American football / 81-Athlete / \n", "904-Cocktail / 1128-Glitter / 2184-Lush / \n", "3-Concert / 7-Musician / 13-Musical ensemble / 38-Orchestra / 219-Accordion / \n", "45-Cosmetics / 162-Medicine / 4597-Jaw / \n", "0-Games / 2-Video game / 33-Weapon / 506-Garry's Mod / \n", "0-Games / 51-Strategy video game / 165-League of Legends / \n", "12-Food / 3117-Match / \n", "20-Trailer / \n", "0-Games / 2-Video game / 194-Call of Duty: Modern Warfare 2 / \n", "0-Games / 7-Musician / 14-Guitar / 1762-Rock Band 3 / \n", "0-Games / 2-Video game / 43-Call of Duty / 216-Call of Duty: Modern Warfare 3 / \n", "1-Vehicle / 4-Car / 11-Motorsport / 17-Racing / 41-Sports car / 57-Race track / 70-Driving / 94-Dashcam / \n", "1-Vehicle / 67-Cycling / 69-Bicycle / 520-Bicycle frame / 645-Road bicycle / \n", "36-Piano / 53-Keyboard / \n", "3-Concert / 7-Musician / 13-Musical ensemble / 14-Guitar / 30-Drummer / 38-Orchestra / 39-Drums / 2116-Hard Rock Cafe / \n", "0-Games / 60-Basketball / 77-Arena / 168-Basketball moves / 255-Slam dunk / \n", "0-Games / 60-Basketball / 168-Basketball moves / 255-Slam dunk / \n", "1353-Bishop / \n", "1-Vehicle / 50-Aircraft / 129-Aviation / 266-Helicopter / \n", "0-Games / 8-Football / 54-Highlight film / 79-American football / \n", "467-Chipmunk / \n", "0-Games / 59-Winter sport / 81-Athlete / 112-Ice skating / 142-Hockey / 312-Ice rink / \n", "9-Music video / \n", "1-Vehicle / 4-Car / 64-Transport / 425-City / \n", "6-Animation / 15-Cartoon / \n", "5-Dance / 98-Festival / 253-Carnival / \n", "10-Animal / \n", "248-Hotel / 916-Spa / \n", "1-Vehicle / 4-Car / 11-Motorsport / 17-Racing / 40-Road / 41-Sports car / 705-Forza Motorsport / 2106-Forza Horizon 2 / \n", "1-Vehicle / 50-Aircraft / 451-Cockpit / 2724-North American P-51 Mustang / \n", "10-Animal / 71-Dog / 385-Face / \n", "0-Games / 2-Video game / 194-Call of Duty: Modern Warfare 2 / \n", "1-Vehicle / 27-Motorcycle / \n", "9-Music video / \n", "0-Games / 2302-Dying Light / \n", "0-Games / 75-Wrestling / \n", "0-Games / 2-Video game / 35-Minecraft / 55-Video game console / 111-PlayStation 3 / 127-Xbox 360 / 176-PlayStation 4 / 181-Xbox / 314-Xbox One / 317-PlayStation / \n", "425-City / \n", "355-Comedian / \n", "0-Games / 257-Counter-Strike / 826-Source Engine / \n", "30-Drummer / 39-Drums / 44-Drums / 128-Cymbal / 146-Snare drum / \n", "45-Cosmetics / 235-Lipstick / \n", "12-Food / 26-Cooking / 32-Recipe / 179-Kitchen / \n", "7-Musician / 13-Musical ensemble / 44-Drums / 122-Weight training / 203-Marching band / 820-Bagpipes / 1646-Pipe band / \n", "12-Food / 26-Cooking / 32-Recipe / 904-Cocktail / 4295-Mojito / \n", "259-Fire / 1245-Alarm device / \n", "56-Hair / \n", "28-Fashion / 1168-Neighbourhood / \n", "101-Telephone / \n", "7-Musician / 13-Musical ensemble / 46-Choir / 49-School / \n", "17-Racing / 41-Sports car / 231-Supercar / 711-Gran Turismo / 953-Gran Turismo 5 / \n", "1-Vehicle / 4-Car / 403-Toyota / \n", "6-Animation / 15-Cartoon / 123-Naruto / \n", "30-Drummer / 39-Drums / 44-Drums / 128-Cymbal / 146-Snare drum / \n", "0-Games / 8-Football / 77-Arena / \n", "0-Games / 6-Animation / 10-Animal / 15-Cartoon / 537-Zoo / 2157-Guppy / \n", "1-Vehicle / 114-Engine / 854-Diesel engine / \n", "117-Boxing / 609-Tattoo / 628-Portrait / \n", "0-Games / 8-Football / 21-Stadium / 65-Kick / 73-Ball / 76-Ball / \n", "93-Comedy / \n", "9-Music video / \n", "6-Animation / 15-Cartoon / 126-The Walt Disney Company / \n", "23-Mobile phone / 29-Smartphone / 72-iPhone / 101-Telephone / 508-iPhone 5 / 760-iPhone 5s / 1572-Husband / \n", "0-Games / 1-Vehicle / 2-Video game / 27-Motorcycle / 34-Action-adventure game / 201-Grand Theft Auto: San Andreas / \n", "12-Food / \n", "31-Disc jockey / \n", "3-Concert / 5-Dance / 16-Performance art / 95-Talent show / \n", "121-Photography / 541-Black-and-white / \n", "5-Dance / 13-Musical ensemble / \n", "9-Music video / \n", "12-Food / 389-Restaurant / \n", "23-Mobile phone / 37-Gadget / 72-iPhone / 516-Headphones / \n", "3-Concert / \n", "1-Vehicle / 4-Car / 11-Motorsport / 17-Racing / 27-Motorcycle / 105-Motorcycling / 1282-Honda CBR series / \n", "0-Games / 20-Trailer / 60-Basketball / \n", "1-Vehicle / 18-Outdoor recreation / 136-Gardening / 137-Tree / 2795-Root / \n", "110-Album / \n", "2-Video game / 121-Photography / 125-Television / \n", "10-Animal / 18-Outdoor recreation / 80-Horse / 164-Beach / 247-Stallion / \n", "9-Music video / \n", "3-Concert / 7-Musician / 16-Performance art / \n", "9-Music video / \n", "1-Vehicle / 4-Car / 11-Motorsport / 17-Racing / 41-Sports car / 118-Grand Theft Auto V / 135-Drifting / \n", "9-Music video / 153-Dress / \n", "0-Games / \n", "40-Road / 67-Cycling / 69-Bicycle / 220-Mountain bike / 520-Bicycle frame / 645-Road bicycle / \n", "3-Concert / 7-Musician / 13-Musical ensemble / 38-Orchestra / \n", "0-Games / 51-Strategy video game / 373-Clash of Clans / \n", "123-Naruto / 183-Manga / 285-Sasuke Uchiha / 1180-Itachi Uchiha / \n", "8-Football / \n", "66-Bollywood / \n", "3-Concert / 99-Christmas / 3581-Rudolph the Red-Nosed Reindeer / \n", "2-Video game / 20-Trailer / 229-Halo / 644-Halo: Reach / 2889-343 Industries / 3360-Halo: Combat Evolved Anniversary / \n", "28-Fashion / \n", "318-Church / \n", "8-Football / \n", "162-Medicine / \n", "9-Music video / \n", "0-Games / 60-Basketball / 168-Basketball moves / \n", "3-Concert / 5-Dance / 13-Musical ensemble / 1253-Ballet Dancer / \n", "117-Boxing / \n", "23-Mobile phone / 29-Smartphone / 47-Personal computer / 414-Computer monitor / \n", "0-Games / 47-Personal computer / 141-Microsoft Windows / \n", "0-Games / 2-Video game / 19-PC game / 47-Personal computer / \n", "0-Games / 2-Video game / 311-Street Fighter / 461-Street Fighter IV / 496-Super Street Fighter IV / 1988-Chun-Li / 3981-Fei Long / \n", "3-Concert / 5-Dance / 16-Performance art / \n", "0-Games / 2-Video game / 1777-Ratchet / 1824-Clank / 1859-Ratchet & Clank / \n", "2-Video game / 20-Trailer / 34-Action-adventure game / 111-PlayStation 3 / 127-Xbox 360 / 181-Xbox / 1028-Devil May Cry / 2344-DmC: Devil May Cry / \n", "9-Music video / \n", "8-Football / \n", "0-Games / 51-Strategy video game / 373-Clash of Clans / \n", "1-Vehicle / 27-Motorcycle / \n", "67-Cycling / 69-Bicycle / 221-Skatepark / 377-BMX bike / \n", "233-Ocean / 246-Furniture / 276-Room / 308-Ship / 605-Cruise ship / 2151-Carnival Cruise Lines / \n", "1-Vehicle / 20-Trailer / 27-Motorcycle / \n", "6-Animation / 15-Cartoon / 97-Drawing / 3023-Line Rider / \n", "1-Vehicle / 4-Car / 11-Motorsport / 17-Racing / 261-Tool / 378-Drag racing / \n", "8-Football / 21-Stadium / 77-Arena / \n", "12-Food / 26-Cooking / 32-Recipe / 52-Dish / 179-Kitchen / 210-Cookware and bakeware / 3008-Slow cooker / \n", "1-Vehicle / 4-Car / 11-Motorsport / 509-Classic car / 1148-Hot rod / \n", "766-President of the United States / 986-Suit / \n", "3-Concert / 68-Lighting / \n", "0-Games / 2-Video game / 33-Weapon / 474-Destiny / \n", "0-Games / 2-Video game / 19-PC game / 51-Strategy video game / 382-Dota 2 / 549-Defense of the Ancients / \n", "12-Food / 26-Cooking / 32-Recipe / 52-Dish / 58-Cuisine / 210-Cookware and bakeware / 274-Meat / 589-Grilling / 620-Barbecue / 688-Beef / 841-Butter / 862-Pork / 1085-Barbecue grill / 1115-Steak / \n", "3-Concert / 5-Dance / 16-Performance art / 46-Choir / \n", "0-Games / 89-Comics / 387-Spider-Man / \n", "661-Stretching / \n", "458-Gold / \n", "1-Vehicle / 17-Racing / 87-Boat / 579-Yacht / 917-Sail / 1309-Sailing ship / 1957-Luxury yacht / \n", "8-Football / 595-Champion / \n", "25-Toy / 140-LEGO / 3181-Lego Legends of Chima / \n", "31-Disc jockey / \n", "1-Vehicle / 4-Car / 11-Motorsport / 169-Rallying / 198-Off-road vehicle / \n", "0-Games / 8-Football / 21-Stadium / 65-Kick / 73-Ball / 77-Arena / \n", "6-Animation / 15-Cartoon / 89-Comics / 385-Face / 1164-Cartoon Network / 1696-Raven / \n", "49-School / \n", "0-Games / 8-Football / 85-Combat / \n", "10-Animal / 48-Pet / 132-Bird / \n", "0-Games / 2-Video game / 237-Final Fantasy / 1131-Final Fantasy XIII / \n", "85-Combat / 117-Boxing / 2851-Fight Night / \n", "0-Games / 2-Video game / 19-PC game / 20-Trailer / 33-Weapon / 43-Call of Duty / 96-Soldier / 134-Call of Duty: Black Ops / 138-Call of Duty: Black Ops II / 194-Call of Duty: Modern Warfare 2 / \n", "1193-Orange / \n", "1-Vehicle / 87-Boat / 147-River / \n", "10-Animal / 91-Fish / 126-The Walt Disney Company / 144-Amusement park / 258-Aquarium / 576-Walt Disney World / 716-Coral / 1463-Magic Kingdom / \n", "1-Vehicle / 4-Car / 11-Motorsport / 17-Racing / 57-Race track / 456-Go-kart / \n", "1-Vehicle / 4-Car / 11-Motorsport / 1455-Cadillac / \n", "0-Games / 1-Vehicle / 2-Video game / 4-Car / 11-Motorsport / 17-Racing / 40-Road / 41-Sports car / 352-Need for Speed / 924-Trophy / 1770-Need for Speed: Hot Pursuit / 2545-Need for Speed III: Hot Pursuit / \n", "0-Games / 2-Video game / 43-Call of Duty / 194-Call of Duty: Modern Warfare 2 / \n", "1-Vehicle / 4-Car / 74-Truck / 249-Heavy equipment / 1611-Garbage truck / \n", "1-Vehicle / 12-Food / 1699-Cart / 2371-Hot dog / \n", "1-Vehicle / 4-Car / 64-Transport / 84-Snow / 223-Weather / 490-Rain / 663-Street / 915-Town / \n", "14-Guitar / 100-Electric guitar / \n", "0-Games / 1-Vehicle / 50-Aircraft / 151-Landing / 1715-Aircraft carrier / \n", "0-Games / 2-Video game / 20-Trailer / 1028-Devil May Cry / 2344-DmC: Devil May Cry / \n", "5-Dance / \n", "415-Sewing / 997-Sewing machine / 1479-Quilt / \n", "1-Vehicle / 50-Aircraft / 139-Airport / 151-Landing / 254-Runway / 856-Airbus A320 family / \n", "476-Textile / 1184-Flag / \n", "3-Concert / \n", "209-University / \n", "28-Fashion / 337-Advertising / \n", "1-Vehicle / 27-Motorcycle / 207-Exhaust system / \n", "56-Hair / 106-Hairstyle / 583-Wig / \n", "1-Vehicle / 4-Car / \n", "1525-Wallet / 2068-Duct tape / \n", "0-Games / 59-Winter sport / 81-Athlete / 112-Ice skating / 142-Hockey / 312-Ice rink / \n", "242-Paper / 415-Sewing / 1525-Wallet / 2067-Duct / 2068-Duct tape / \n", "9-Music video / \n", "246-Furniture / 248-Hotel / 276-Room / \n", "5-Dance / 9-Music video / \n", "1-Vehicle / 4-Car / 945-MINI Cooper / 1855-Video camera / \n", "0-Games / 8-Football / 54-Highlight film / 65-Kick / 76-Ball / \n", "28-Fashion / 153-Dress / 253-Carnival / 2793-Disguise / \n", "22-Nature / 61-Art / 148-Painting / 164-Beach / \n", "819-World Wide Web / \n", "5-Dance / 9-Music video / \n", "23-Mobile phone / 29-Smartphone / 553-Sony Xperia / 3019-Sony Ericsson Xperia X10 / \n", "5-Dance / 16-Performance art / 95-Talent show / \n", "0-Games / 8-Football / \n", "10-Animal / 80-Horse / \n", "0-Games / 8-Football / 54-Highlight film / 81-Athlete / 358-Rugby football / \n", "1-Vehicle / \n", "9-Music video / \n", "1-Vehicle / 4-Car / 11-Motorsport / 114-Engine / 449-Honda / 552-Honda Civic / \n", "440-Lion / 1156-Simba / \n", "843-The Doctor / \n", "20-Trailer / 125-Television / \n", "6-Animation / 15-Cartoon / 97-Drawing / 3137-Stitch / \n", "1-Vehicle / 27-Motorcycle / 337-Advertising / \n", "3165-Turban / \n", "0-Games / 8-Football / 76-Ball / \n", "10-Animal / 48-Pet / 166-Cat / 346-Kitten / \n", "20-Trailer / 444-Family / 1014-Movieclips / 3996-Fandango / \n", "0-Games / 269-Comic book / 365-Fireworks / \n", "0-Games / 6-Animation / 77-Arena / 89-Comics / 123-Naruto / \n", "0-Games / 2-Video game / 1983-Infestation: Survivor Stories / \n", "0-Games / 8-Football / 21-Stadium / 73-Ball / \n", "0-Games / 1-Vehicle / 2-Video game / 4-Car / 34-Action-adventure game / 41-Sports car / 118-Grand Theft Auto V / 330-Money / \n", "49-School / \n", "1791-Stargate / \n", "1-Vehicle / \n", "25-Toy / 200-Doll / 412-Monster / 503-Monster High / \n", "319-Robot / \n", "0-Games / 2-Video game / 1427-TalesRunner / \n", "85-Combat / \n", "3-Concert / \n", "9-Music video / 200-Doll / \n", "849-Tales / \n", "3-Concert / 36-Piano / 2120-Tears / \n", "1-Vehicle / 4-Car / 40-Road / 62-Train / 64-Transport / 103-Rail transport / 119-Locomotive / 143-Railroad car / 2655-Vehicle horn / \n", "0-Games / 2-Video game / 199-Wii / 426-Resident Evil / \n", "75-Wrestling / 77-Arena / 777-WWE 2K / 1080-WWE 2K15 / \n", "3-Concert / \n", "0-Games / 2-Video game / 6-Animation / \n", "0-Games / 2-Video game / \n", "0-Games / 2-Video game / 33-Weapon / 192-Battlefield / 383-Battlefield 4 / 1267-Battlefield Hardline / \n", "3-Concert / 38-Orchestra / 98-Festival / \n", "67-Cycling / 69-Bicycle / 154-Wheel / \n", "260-Prayer / \n", "83-Skateboarding / 1937-Inline skates / \n", "23-Mobile phone / 29-Smartphone / 101-Telephone / 131-Computer / 141-Microsoft Windows / 1379-HTC One / 2328-HTC One (M8) / \n", "45-Cosmetics / \n", "3-Concert / 5-Dance / 16-Performance art / \n", "3-Concert / 14-Guitar / 16-Performance art / \n", "1-Vehicle / 12-Food / 156-Agriculture / 174-Farm / 931-Combine Harvester / 1158-Wheat / \n", "3-Concert / 30-Drummer / 39-Drums / 44-Drums / \n", "1892-Dormitory / \n", "1-Vehicle / 67-Cycling / 69-Bicycle / \n", "7-Musician / 13-Musical ensemble / 46-Choir / \n", "5-Dance / 1100-Hat / \n", "66-Bollywood / \n", "6-Animation / 183-Manga / 2357-Vampire Knight / \n", "9-Music video / \n", "0-Games / 159-World of Warcraft / \n", "10-Animal / 12-Food / 26-Cooking / 32-Recipe / 274-Meat / 315-Knife / \n", "9-Music video / \n", "1-Vehicle / 4-Car / 11-Motorsport / 17-Racing / 41-Sports car / 94-Dashcam / 473-Ford Mustang / 552-Honda Civic / 1040-Street racing / \n", "28-Fashion / 956-Glasses / \n", "22-Nature / \n", "8-Football / 21-Stadium / 77-Arena / \n", "9-Music video / \n", "59-Winter sport / 112-Ice skating / 142-Hockey / \n", "47-Personal computer / \n", "3-Concert / 31-Disc jockey / 133-Nightclub / \n", "0-Games / 2-Video game / 47-Personal computer / 429-Website / \n", "0-Games / 637-Five Nights at Freddy's / \n", "22-Nature / 147-River / 283-Rock / 1416-Cliff / \n", "22-Nature / 86-Plant / 376-Flute / 2532-Pan flute / \n", "3-Concert / \n", "9-Music video / \n", "10-Animal / 131-Computer / \n", "122-Weight training / 204-Gym / 214-Muscle / \n", "0-Games / 2-Video game / 23-Mobile phone / 29-Smartphone / 72-iPhone / 331-iPod touch / \n", "9-Music video / 78-Wedding / 190-Bride / \n", "3-Concert / 49-School / \n", "10-Animal / 68-Lighting / 91-Fish / 258-Aquarium / 716-Coral / 1051-Reef aquarium / \n", "323-Toddler / \n", "2-Video game / 9-Music video / 2173-Dead Space / \n", "10-Animal / 48-Pet / 71-Dog / 182-Puppy / \n", "1-Vehicle / 4-Car / 11-Motorsport / 27-Motorcycle / 40-Road / 64-Transport / 105-Motorcycling / 154-Wheel / 482-Asphalt / 608-Wheelie / \n", "1-Vehicle / 50-Aircraft / 82-Airplane / 129-Aviation / 157-Jet aircraft / \n", "12-Food / 20-Trailer / \n", "6-Animation / 15-Cartoon / 183-Manga / 762-Angel / \n", "256-Tennis / \n", "0-Games / 2-Video game / 6-Animation / 35-Minecraft / \n", "1-Vehicle / 4-Car / 11-Motorsport / 17-Racing / 169-Rallying / \n", "33-Weapon / 217-Hunting / \n", "638-Gears of War / 773-Gears of War / 1425-Gears of War 2 / \n", "0-Games / 2-Video game / 33-Weapon / 229-Halo / 470-Halo 3 / \n", "3938-Large Hadron Collider / \n", "0-Games / 6-Animation / 89-Comics / \n", "0-Games / 8-Football / \n", "3-Concert / 7-Musician / 13-Musical ensemble / 39-Drums / 763-Snake / \n", "45-Cosmetics / 890-Brush / 3655-Makeup brush / \n", "1-Vehicle / 4-Car / 11-Motorsport / 17-Racing / 57-Race track / 92-Radio-controlled model / 135-Drifting / 202-Radio-controlled car / \n", "3-Concert / \n", "7-Musician / 9-Music video / 13-Musical ensemble / 14-Guitar / 30-Drummer / 44-Drums / \n", "0-Games / 8-Football / 54-Highlight film / \n", "0-Games / 2-Video game / 33-Weapon / 96-Soldier / 834-Combat Arms / \n", "12-Food / 88-Machine / 227-Water / 938-Steel / \n", "1-Vehicle / \n", "1-Vehicle / 11-Motorsport / 17-Racing / 27-Motorcycle / 57-Race track / 172-Motocross / 769-Red Bull / \n", "1-Vehicle / 4-Car / 41-Sports car / 245-BMW / 344-Coupé / 493-Rim / 1233-BMW 3 Series / 2325-BMW 3 Series (E90) / \n", "1-Vehicle / 4-Car / 1742-Lowrider / 3040-Subaru Legacy / \n", "3-Concert / 7-Musician / \n", "205-Newscaster / \n", "1-Vehicle / 4-Car / 9-Music video / \n", "0-Games / 2-Video game / 19-PC game / 35-Minecraft / 111-PlayStation 3 / 127-Xbox 360 / 176-PlayStation 4 / 181-Xbox / 314-Xbox One / 843-The Doctor / \n", "10-Animal / 22-Nature / 42-Fishing / 91-Fish / 147-River / 814-Pond / \n", "22-Nature / 36-Piano / 150-Violin / \n", "1-Vehicle / 4-Car / 11-Motorsport / 17-Racing / 41-Sports car / 57-Race track / 378-Drag racing / \n", "0-Games / 257-Counter-Strike / 1109-Counter-Strike Online / \n", "0-Games / 2-Video game / 1187-Ragnarok Online / \n", "3-Concert / 7-Musician / 16-Performance art / \n", "0-Games / 2-Video game / 6-Animation / 15-Cartoon / 111-PlayStation 3 / 1492-Persona / \n", "7-Musician / 14-Guitar / 738-Guitar amplifier / \n", "3-Concert / \n", "47-Personal computer / 72-iPhone / 145-Tablet computer / 189-iPad / \n", "275-Diving / 287-Human swimming / \n", "12-Food / 28-Fashion / 45-Cosmetics / 56-Hair / 106-Hairstyle / 909-Updo / \n", "3-Concert / 46-Choir / 318-Church / \n", "1-Vehicle / 50-Aircraft / 82-Airplane / 139-Airport / 151-Landing / 157-Jet aircraft / 175-Airliner / 187-Takeoff / 188-Airline / 254-Runway / 472-Boeing 737 / 499-Cloud / \n", "10-Animal / 18-Outdoor recreation / 22-Nature / 217-Hunting / 706-Deer / \n", "8-Football / \n", "9-Music video / \n", "0-Games / 2-Video game / 19-PC game / 35-Minecraft / \n", "5-Dance / 60-Basketball / 191-Cheerleading / 3960-Pep squad / \n", "9-Music video / \n", "0-Games / 2-Video game / 116-Pokémon / 186-Pokémon / 690-Pokémon X and Y / \n", "14-Guitar / 24-String instrument / 63-Acoustic guitar / \n", "3-Concert / \n", "747-Retail / \n", "1-Vehicle / 50-Aircraft / 175-Airliner / 188-Airline / 1340-Airbus A380 / 2587-Air France / \n", "122-Weight training / 214-Muscle / \n", "25-Toy / 295-Chocolate / 1192-Hello Kitty / \n", "0-Games / 2-Video game / 33-Weapon / 257-Counter-Strike / 702-Counter-Strike: Source / 826-Source Engine / \n", "0-Games / 2-Video game / 111-PlayStation 3 / \n", "23-Mobile phone / 29-Smartphone / \n", "9-Music video / \n", "0-Games / 2-Video game / 34-Action-adventure game / 398-The Legend of Zelda / 2442-The Legend of Zelda: The Wind Waker / \n", "59-Winter sport / 112-Ice skating / 142-Hockey / 262-Ice / \n", "1-Vehicle / 4-Car / 74-Truck / 84-Snow / 167-Four-wheel drive / 198-Off-road vehicle / 277-Sport utility vehicle / 327-Mud / 360-Jeep / \n", "1-Vehicle / 62-Train / 840-Museum / \n", "3-Concert / 5-Dance / 10-Animal / 16-Performance art / 80-Horse / \n", "10-Animal / \n", "121-Photography / \n", "1-Vehicle / 4-Car / 6-Animation / 809-Logo / \n", "1-Vehicle / 92-Radio-controlled model / 149-Model aircraft / 160-Radio-controlled aircraft / 266-Helicopter / 392-Radio-controlled helicopter / 2867-Wasp / \n", "0-Games / 2-Video game / 1312-Dynasty Warriors / \n", "3-Concert / 7-Musician / 16-Performance art / 38-Orchestra / \n", "0-Games / 211-RuneScape / \n", "0-Games / 8-Football / 90-Sports game / 453-Pro Evolution Soccer / 1719-Pro Evolution Soccer 2015 / \n", "0-Games / 35-Minecraft / 1149-Castle / \n", "0-Games / 1680-Seafight / \n", "3-Concert / 16-Performance art / 98-Festival / \n", "1791-Stargate / \n", "1-Vehicle / 94-Dashcam / 102-Building / 416-Apartment / 915-Town / 1960-Traffic light / \n", "3-Concert / \n", "31-Disc jockey / \n", "1-Vehicle / 4-Car / 41-Sports car / 70-Driving / 234-Ford / 473-Ford Mustang / 1361-Ford GT / 1388-V6 engine / \n", "3-Concert / 16-Performance art / \n", "0-Games / 2-Video game / 4140-Quake Live / \n", "365-Fireworks / \n", "81-Athlete / 163-Running / \n", "6-Animation / \n", "28-Fashion / 205-Newscaster / 1386-ABS-CBN / \n", "3-Concert / 30-Drummer / \n", "0-Games / 1-Vehicle / 2-Video game / 4-Car / 34-Action-adventure game / 297-Grand Theft Auto IV / 658-Grand Theft Auto: The Lost and Damned / 1134-Taxicab / 1307-Police car / \n", "253-Carnival / \n", "0-Games / 2-Video game / 229-Halo / 470-Halo 3 / \n", "5-Dance / 16-Performance art / \n", "0-Games / 2-Video game / 34-Action-adventure game / 2863-Max Payne 3 / \n", "0-Games / 2-Video game / 19-PC game / 686-Diablo III / 872-Diablo / \n", "9-Music video / \n", "12-Food / 26-Cooking / 32-Recipe / 52-Dish / 58-Cuisine / 301-Meal / \n", "0-Games / 82-Airplane / 506-Garry's Mod / \n", "3-Concert / \n", "56-Hair / \n", "6-Animation / 15-Cartoon / 440-Lion / 1212-Stuffed toy / \n", "307-Hand / \n", "22-Nature / 113-House / 208-Wood / 267-Mountain / 2280-Cottage / \n", "78-Wedding / 190-Bride / 421-Gown / \n", "3-Concert / 7-Musician / 13-Musical ensemble / 14-Guitar / 16-Performance art / 24-String instrument / \n", "0-Games / 33-Weapon / 474-Destiny / \n", "1-Vehicle / 4-Car / 11-Motorsport / 261-Tool / \n", "116-Pokémon / 186-Pokémon / 690-Pokémon X and Y / \n", "5-Dance / \n", "1-Vehicle / 4-Car / 11-Motorsport / 17-Racing / 70-Driving / 207-Exhaust system / 509-Classic car / \n", "0-Games / 2-Video game / \n", "49-School / \n", "1-Vehicle / 4-Car / 234-Ford / 3338-Ford Crown Victoria / \n", "0-Games / 54-Highlight film / 81-Athlete / \n", "3-Concert / 98-Festival / 253-Carnival / \n", "0-Games / 75-Wrestling / 85-Combat / 2822-Submission wrestling / \n", "1-Vehicle / 11-Motorsport / 18-Outdoor recreation / 22-Nature / 327-Mud / 565-Mud bogging / \n", "3-Concert / 16-Performance art / \n", "0-Games / 2-Video game / 116-Pokémon / 186-Pokémon / \n", "5-Dance / 130-Gymnastics / 350-Circus / \n", "12-Food / 26-Cooking / 32-Recipe / 52-Dish / 210-Cookware and bakeware / 239-Roasting / 391-Chicken meat / 1284-Frying / 2563-Gravy / 2591-Filipino food / 2757-Fried chicken / \n", "0-Games / 8-Football / \n", "6-Animation / 123-Naruto / 285-Sasuke Uchiha / \n", "1-Vehicle / 50-Aircraft / 129-Aviation / 157-Jet aircraft / 388-Fighter aircraft / 758-Air force / 2326-Royal Air Force / \n", "0-Games / 2-Video game / 47-Personal computer / 131-Computer / 228-Laptop / 309-Computer hardware / 602-Central processing unit / \n", "1-Vehicle / 18-Outdoor recreation / 87-Boat / 579-Yacht / \n", "1-Vehicle / 4-Car / 40-Road / 158-Tire / 167-Four-wheel drive / 277-Sport utility vehicle / 403-Toyota / 1511-Toyota Land Cruiser / \n", "447-Sand / \n", "21-Stadium / 49-School / 562-Pitcher / \n", "36-Piano / 53-Keyboard / 541-Black-and-white / \n", "15-Cartoon / 1134-Taxicab / \n", "0-Games / 2-Video game / 685-Luigi / 3900-Mario & Luigi: Superstar Saga / \n", "1-Vehicle / 4-Car / 270-Sedan / 1074-Hyundai / 3316-Hyundai Genesis / \n", "0-Games / 2-Video game / 8-Football / 90-Sports game / 180-FIFA 15 / \n", "117-Boxing / \n", "42-Fishing / 510-Fishing lure / \n", "0-Games / 2-Video game / 61-Art / \n", "1-Vehicle / 6-Animation / 25-Toy / 140-LEGO / 1079-Lego City / 1302-The Lego Group / \n", "341-Sketch comedy / \n", "219-Accordion / 655-Diatonic button accordion / \n", "5-Dance / 126-The Walt Disney Company / \n", "1-Vehicle / 4-Car / 11-Motorsport / 349-Chevrolet / 406-Chevrolet / 1148-Hot rod / \n", "162-Medicine / \n", "61-Art / 148-Painting / 628-Portrait / \n", "1572-Husband / \n", "0-Games / 2-Video game / 2799-Ys / \n", "31-Disc jockey / \n", "5-Dance / 115-Ballet / \n", "85-Combat / 117-Boxing / \n", "31-Disc jockey / \n", "14-Guitar / \n", "1-Vehicle / 50-Aircraft / 187-Takeoff / 254-Runway / 266-Helicopter / 4657-Bell Boeing V-22 Osprey / \n", "6-Animation / 9-Music video / 1160-Giant panda / \n", "1-Vehicle / 666-Concrete / \n", "5-Dance / \n", "609-Tattoo / \n", "45-Cosmetics / \n", "3-Concert / 5-Dance / 16-Performance art / \n", "6-Animation / 15-Cartoon / \n", "191-Cheerleading / 776-Kindergarten / \n", "46-Choir / \n", "2-Video game / 20-Trailer / \n", "6-Animation / 15-Cartoon / \n", "36-Piano / 3117-Match / 3352-Helianthus annuus / \n", "0-Games / 2-Video game / 489-Board game / \n", "0-Games / 2-Video game / 19-PC game / 523-Cattle / 686-Diablo III / \n", "9-Music video / \n", "3-Concert / \n", "3-Concert / 7-Musician / \n", "125-Television / 355-Comedian / \n", "1-Vehicle / 4-Car / 114-Engine / 261-Tool / 585-Lawn / 741-Mower / 805-Lawn mower / \n", "504-Web page / \n", "12-Food / 215-Eating / 762-Angel / \n", "3-Concert / \n", "59-Winter sport / 112-Ice skating / 142-Hockey / \n", "208-Wood / 400-Woodturning / 972-Lathe / \n", "30-Drummer / 39-Drums / 44-Drums / 128-Cymbal / 146-Snare drum / \n", "31-Disc jockey / \n", "1-Vehicle / 4-Car / 27-Motorcycle / 105-Motorcycling / 212-Highway / \n", "6-Animation / 15-Cartoon / \n", "6-Animation / 15-Cartoon / 61-Art / 97-Drawing / 183-Manga / 408-Sketch / \n", "10-Animal / 132-Bird / 217-Hunting / \n", "1-Vehicle / 11-Motorsport / 27-Motorcycle / 105-Motorcycling / 169-Rallying / \n", "47-Personal computer / 497-Macintosh / 1922-iMac / 3974-Mac Mini / \n", "7-Musician / 13-Musical ensemble / 38-Orchestra / \n", "12-Food / 26-Cooking / 32-Recipe / 52-Dish / 58-Cuisine / 122-Weight training / 179-Kitchen / 210-Cookware and bakeware / 742-Countertop / 2027-Spinach / 2066-Omelette / 2309-Low-carbohydrate diet / 3150-Supper / \n", "1868-Preacher / \n", "121-Photography / 155-Camera / \n", "38-Orchestra / 290-Pony / 343-My Little Pony / 1319-Apple / 4042-Applejack / \n", "3-Concert / 98-Festival / \n", "3085-Charango / \n", "0-Games / 326-Metin2 / \n", "0-Games / 19-PC game / 35-Minecraft / \n", "148-Painting / 208-Wood / \n", "8-Football / 21-Stadium / \n", "0-Games / 2-Video game / 19-PC game / 51-Strategy video game / 603-StarCraft II: Wings of Liberty / 732-StarCraft / \n", "0-Games / 8-Football / 21-Stadium / 60-Basketball / \n", "12-Food / 533-Oil / \n", "31-Disc jockey / \n", "5-Dance / 31-Disc jockey / \n", "8-Football / \n", "28-Fashion / 333-Nike; Inc. / 1795-Sock / \n", "33-Weapon / 315-Knife / \n", "130-Gymnastics / \n", "0-Games / 2-Video game / 43-Call of Duty / 134-Call of Duty: Black Ops / \n", "3-Concert / \n", "98-Festival / 448-Planet / \n", "45-Cosmetics / 282-Skin / 2351-Moisturizer / 3861-Toner / \n", "18-Outdoor recreation / 59-Winter sport / 84-Snow / 120-Winter / 177-Skiing / 267-Mountain / 305-Ski / 789-Alpine skiing / \n", "1-Vehicle / 11-Motorsport / 17-Racing / 27-Motorcycle / 57-Race track / \n", "5-Dance / 115-Ballet / \n", "214-Muscle / \n", "369-Resort / \n", "12-Food / 26-Cooking / 32-Recipe / 109-Cooking show / 226-Dessert / 232-Cake / 236-Baking / 560-Sugar / 1641-Cheesecake / \n", "23-Mobile phone / 29-Smartphone / 47-Personal computer / 72-iPhone / 145-Tablet computer / 189-iPad / \n", "79-American football / 209-University / \n", "12-Food / 86-Plant / 136-Gardening / 137-Tree / \n", "0-Games / 2-Video game / 19-PC game / 33-Weapon / 85-Combat / 1166-Dark Souls II / \n", "4303-Apple Store / \n", "56-Hair / 106-Hairstyle / \n", "10-Animal / 1108-Mushroom / \n", "1-Vehicle / 11-Motorsport / 565-Mud bogging / \n", "448-Planet / 654-Sun / 1702-Telescope / 2761-Venus / \n", "1-Vehicle / 62-Train / 64-Transport / 103-Rail transport / 119-Locomotive / 143-Railroad car / 152-Track / 161-Train station / 1904-Electric locomotive / \n", "1-Vehicle / 4-Car / 23-Mobile phone / 234-Ford / 1634-Ford Fiesta / \n", "1-Vehicle / 25-Toy / 126-The Walt Disney Company / \n", "7-Musician / 9-Music video / 14-Guitar / 627-Clock / \n", "0-Games / 2-Video game / 19-PC game / 33-Weapon / 192-Battlefield / 383-Battlefield 4 / \n", "0-Games / 51-Strategy video game / 165-League of Legends / \n", "1-Vehicle / 4-Car / 11-Motorsport / 17-Racing / 41-Sports car / 57-Race track / \n", "148-Painting / \n", "10-Animal / 18-Outdoor recreation / 22-Nature / 42-Fishing / 91-Fish / 275-Diving / 288-Underwater / 381-Underwater diving / 407-Scuba diving / 716-Coral / \n", "30-Drummer / 39-Drums / 44-Drums / \n", "3-Concert / 7-Musician / 16-Performance art / 30-Drummer / 279-Garden / 4329-Busch Gardens Tampa Bay / \n", "36-Piano / 53-Keyboard / 107-Musical keyboard / \n", "983-Lexus / \n", "1-Vehicle / 27-Motorcycle / 336-Scooter / \n", "1858-Poster / \n", "260-Prayer / \n", "7-Musician / 14-Guitar / 24-String instrument / 100-Electric guitar / 1971-Epiphone / \n", "634-Trombone / \n", "0-Games / 1-Vehicle / 4-Car / 11-Motorsport / 70-Driving / 837-Roblox / \n", "3-Concert / 9-Music video / 13-Musical ensemble / \n", "222-Surfing / 1008-Kite / \n", "0-Games / 211-RuneScape / \n", "0-Games / 2-Video game / 19-PC game / 35-Minecraft / 734-DayZ / \n", "3-Concert / 7-Musician / 13-Musical ensemble / 38-Orchestra / 46-Choir / \n", "139-Airport / 159-World of Warcraft / 171-Warcraft / \n", "0-Games / 23-Mobile phone / 29-Smartphone / 72-iPhone / \n", "653-Thread / 1497-Wool / \n", "0-Games / 211-RuneScape / \n", "5-Dance / 78-Wedding / \n", "1-Vehicle / 50-Aircraft / 82-Airplane / 129-Aviation / 157-Jet aircraft / 388-Fighter aircraft / 758-Air force / 929-Aerobatics / \n", "9-Music video / \n", "9-Music video / \n", "0-Games / 2-Video game / 257-Counter-Strike / 315-Knife / \n", "3-Concert / 5-Dance / 16-Performance art / \n", "12-Food / 88-Machine / 227-Water / \n", "0-Games / 2-Video game / 19-PC game / 35-Minecraft / 506-Garry's Mod / 2290-Trouble in Terrorist Town / \n", "0-Games / 75-Wrestling / 777-WWE 2K / 977-WWE '13 / 1582-WWE 2K14 / \n", "322-Mixtape / \n", "88-Machine / 3270-Jukebox / \n", "9-Music video / 541-Black-and-white / \n", "31-Disc jockey / \n", "93-Comedy / \n", "1-Vehicle / 4-Car / 11-Motorsport / 17-Racing / 57-Race track / 111-PlayStation 3 / 456-Go-kart / 711-Gran Turismo / 769-Red Bull / 1519-Gran Turismo 6 / \n", "36-Piano / 53-Keyboard / 104-Pianist / 107-Musical keyboard / \n", "111-PlayStation 3 / 849-Tales / \n", "0-Games / 2-Video game / 25-Toy / 140-LEGO / \n", "5-Dance / 16-Performance art / \n", "1-Vehicle / 87-Boat / \n", "10-Animal / 22-Nature / 132-Bird / 137-Tree / 213-Forest / 217-Hunting / 280-Wildlife / 2333-Wild turkey / 4351-Turkey / \n", "3-Concert / 7-Musician / 13-Musical ensemble / 14-Guitar / 16-Performance art / 30-Drummer / 44-Drums / \n", "12-Food / 28-Fashion / 232-Cake / 236-Baking / \n", "20-Trailer / \n", "24-String instrument / 150-Violin / 2477-Electric violin / \n", "0-Games / 8-Football / 21-Stadium / 73-Ball / \n", "28-Fashion / 121-Photography / \n", "3-Concert / 68-Lighting / \n", "14-Guitar / 293-Slide show / \n", "1-Vehicle / 62-Train / 64-Transport / \n", "419-Earth / 593-Moon / 654-Sun / 2022-Eclipse / 2741-Solar eclipse / 3919-Lunar eclipse / \n", "102-Building / 248-Hotel / \n", "0-Games / 8-Football / \n", "1-Vehicle / 4-Car / 332-Mercedes-Benz / 3826-Mercedes-Benz A-Class / \n", "0-Games / 2-Video game / \n", "1-Vehicle / 4-Car / 64-Transport / \n", "5-Dance / \n", "36-Piano / 107-Musical keyboard / 265-Electronic keyboard / \n", "1-Vehicle / 1257-Turbine / \n", "1-Vehicle / 87-Boat / \n", "7-Musician / 13-Musical ensemble / 38-Orchestra / 150-Violin / \n", "25-Toy / 200-Doll / 269-Comic book / 290-Pony / 343-My Little Pony / \n", "12-Food / 162-Medicine / 780-Fat / \n", "616-The Bible / \n", "36-Piano / 53-Keyboard / 107-Musical keyboard / 4159-Grease / \n", "3-Concert / 7-Musician / 13-Musical ensemble / 46-Choir / \n", "1066-Aikido / \n", "44-Drums / 376-Flute / \n", "9-Music video / 30-Drummer / 39-Drums / 44-Drums / 128-Cymbal / 146-Snare drum / \n", "365-Fireworks / \n", "60-Basketball / \n", "840-Museum / \n", "1-Vehicle / 4-Car / 11-Motorsport / 17-Racing / 40-Road / 87-Boat / 169-Rallying / \n", "0-Games / 2-Video game / 791-Magic: The Gathering / \n", "1-Vehicle / 50-Aircraft / 87-Boat / 147-River / 266-Helicopter / 308-Ship / 1107-United States Navy / \n", "2013-The Twilight Saga / \n", "98-Festival / \n", "0-Games / 2-Video game / 19-PC game / 596-Monster Hunter / 1339-Monster Hunter Freedom Unite / \n", "3-Concert / 5-Dance / 16-Performance art / 115-Ballet / \n", "34-Action-adventure game / 201-Grand Theft Auto: San Andreas / \n", "1-Vehicle / 4-Car / 41-Sports car / 68-Lighting / 92-Radio-controlled model / 202-Radio-controlled car / 245-BMW / 1418-Model car / \n", "5-Dance / 6-Animation / 15-Cartoon / 2408-Lucky Star / \n", "3-Concert / 7-Musician / 46-Choir / \n", "0-Games / 2-Video game / 19-PC game / 51-Strategy video game / 165-League of Legends / 1494-Jungle / \n", "0-Games / 8-Football / \n", "1-Vehicle / 4-Car / 11-Motorsport / 17-Racing / 135-Drifting / \n", "3-Concert / \n", "144-Amusement park / \n", "75-Wrestling / \n", "6-Animation / 89-Comics / 183-Manga / \n", "1-Vehicle / 4-Car / 447-Sand / 751-Desert / 853-Dune buggy / 975-Dune / \n", "9-Music video / 14-Guitar / \n", "425-City / \n", "22-Nature / \n", "7-Musician / 14-Guitar / 100-Electric guitar / 738-Guitar amplifier / \n", "0-Games / 8-Football / 21-Stadium / 65-Kick / 73-Ball / 76-Ball / 77-Arena / 81-Athlete / \n", "294-Batman / 838-Batman: Arkham / 1881-Arkham Asylum / 2084-Batman: Arkham Origins / 2146-Bane / \n", "10-Animal / 550-Dinosaur / 1126-Tyrannosaurus / \n", "10-Animal / 18-Outdoor recreation / 132-Bird / 718-Duck / \n", "9-Music video / \n", "260-Prayer / \n", "12-Food / 26-Cooking / 32-Recipe / 52-Dish / 58-Cuisine / 454-Sauce / 752-Pasta / 1517-Spaghetti / \n", "12-Food / 26-Cooking / 539-Home appliance / 649-Oven / 943-Stove / 2490-Kitchen stove / 4586-Hotpoint / \n", "3-Concert / 7-Musician / 24-String instrument / \n", "0-Games / 2-Video game / 34-Action-adventure game / 55-Video game console / 176-PlayStation 4 / 317-PlayStation / \n", "6-Animation / 10-Animal / 15-Cartoon / 290-Pony / 343-My Little Pony / \n", "85-Combat / 117-Boxing / 399-Kickboxing / \n", "75-Wrestling / \n", "10-Animal / 17-Racing / 18-Outdoor recreation / 1117-Camcorder / 2883-Flip Video / \n", "10-Animal / 22-Nature / 132-Bird / 213-Forest / 217-Hunting / 280-Wildlife / \n", "6-Animation / 123-Naruto / 285-Sasuke Uchiha / \n", "0-Games / 2-Video game / 237-Final Fantasy / 1571-Final Fantasy XIV / 1760-Final Fantasy XIV: A Realm Reborn / \n", "42-Fishing / 222-Surfing / 672-Bodyboarding / \n", "0-Games / 586-Nintendo Entertainment System / \n", "1-Vehicle / 4-Car / 260-Prayer / \n", "325-Vampire / \n", "1-Vehicle / 4-Car / 11-Motorsport / 17-Racing / 57-Race track / \n", "1-Vehicle / 4-Car / 270-Sedan / 1074-Hyundai / \n", "1-Vehicle / 4-Car / 11-Motorsport / 17-Racing / 70-Driving / \n", "22-Nature / \n", "2-Video game / 20-Trailer / 3159-Dishonored / \n", "12-Food / 26-Cooking / 32-Recipe / 52-Dish / \n", "238-Nail / 304-Nail polish / 1424-Coupon / \n", "0-Games / 8-Football / 54-Highlight film / 79-American football / 90-Sports game / 561-Madden NFL / \n", "7-Musician / 13-Musical ensemble / 46-Choir / 99-Christmas / \n", "1-Vehicle / 4-Car / 299-Volkswagen Passenger Cars / 531-Hatchback / 1661-Volkswagen Gol / \n", "22-Nature / 121-Photography / 155-Camera / \n", "18-Outdoor recreation / 458-Gold / 1426-Treasure / \n", "59-Winter sport / 112-Ice skating / 316-Figure skating / \n", "0-Games / 2-Video game / 19-PC game / 33-Weapon / 111-PlayStation 3 / 1106-Character / 1312-Dynasty Warriors / 3721-Dynasty Warriors 7 / 3813-Dynasty Warriors: Gundam / \n", "38-Orchestra / \n", "0-Games / 2-Video game / 6-Animation / 2237-Kantai Collection / \n", "1-Vehicle / 4-Car / 270-Sedan / 2509-Mazda3 / \n", "3-Concert / 38-Orchestra / 68-Lighting / \n", "45-Cosmetics / 173-Eye shadow / 193-Eye liner / \n", "31-Disc jockey / \n", "67-Cycling / 69-Bicycle / 377-BMX bike / \n", "0-Games / 7-Musician / 14-Guitar / \n", "1-Vehicle / 103-Rail transport / 224-Rapid transit / \n", "113-House / 170-Home improvement / 246-Furniture / 416-Apartment / 580-Bedroom / \n", "6-Animation / 379-Touhou Project / \n", "93-Comedy / \n", "1-Vehicle / 25-Toy / 50-Aircraft / 92-Radio-controlled model / 149-Model aircraft / 155-Camera / 160-Radio-controlled aircraft / 266-Helicopter / 392-Radio-controlled helicopter / \n", "23-Mobile phone / 29-Smartphone / 37-Gadget / 178-Samsung Galaxy / 883-Samsung Galaxy S4 / 1015-Samsung Galaxy Note series / 1776-Samsung Galaxy Note II / \n", "30-Drummer / 39-Drums / 44-Drums / 128-Cymbal / 146-Snare drum / 1127-Drum stick / \n", "5-Dance / 46-Choir / 209-University / \n", "347-Climbing / \n", "12-Food / 26-Cooking / 313-Book / \n", "341-Sketch comedy / 364-Sitcom / \n", "3-Concert / 68-Lighting / \n", "694-Tai chi / \n", "0-Games / 2-Video game / 1290-Dragon Nest / \n", "12-Food / 28-Fashion / 476-Textile / \n", "10-Animal / 18-Outdoor recreation / 22-Nature / 42-Fishing / 91-Fish / 225-Lake / 814-Pond / \n", "1-Vehicle / 4-Car / 769-Red Bull / \n", "130-Gymnastics / 1417-Balance beam / \n", "694-Tai chi / \n", "3-Concert / \n", "8-Football / 710-Handball / \n", "3-Concert / 16-Performance art / 30-Drummer / 39-Drums / 44-Drums / \n", "14-Guitar / 606-String / 670-Banjo / \n", "10-Animal / 17-Racing / 57-Race track / 80-Horse / 384-Horse racing / \n", "1-Vehicle / 4-Car / 299-Volkswagen Passenger Cars / 526-Volkswagen / 969-Volkswagen Beetle / \n", "0-Games / 79-American football / 993-Quarterback / \n", "3-Concert / 7-Musician / 13-Musical ensemble / 16-Performance art / \n", "14-Guitar / 36-Piano / 99-Christmas / \n", "227-Water / 278-GoPro / 287-Human swimming / 288-Underwater / 329-Swimming pool / \n", "12-Food / 26-Cooking / 32-Recipe / 52-Dish / 58-Cuisine / 389-Restaurant / \n", "125-Television / \n", "89-Comics / 483-Costume / 803-DC Comics / \n", "144-Amusement park / \n", "30-Drummer / 39-Drums / 44-Drums / 128-Cymbal / 146-Snare drum / 1127-Drum stick / \n", "1-Vehicle / 4-Car / 40-Road / 449-Honda / 4099-Honda Prelude / \n", "12-Food / 26-Cooking / 32-Recipe / 1329-Blender / \n", "10-Animal / 18-Outdoor recreation / 22-Nature / 42-Fishing / 91-Fish / 814-Pond / \n", "33-Weapon / 261-Tool / 315-Knife / \n", "163-Running / \n", "12-Food / \n", "136-Gardening / 1384-Greenhouse / \n", "25-Toy / 290-Pony / 343-My Little Pony / 524-Littlest Pet Shop / \n", "6-Animation / 15-Cartoon / 1156-Simba / 3261-Princess Jasmine / 3600-The Jungle Book / 4202-Timon and Pumbaa / \n", "1-Vehicle / 74-Truck / 433-Construction / 4339-Deep foundation / \n", "748-Iron Man / 897-Avengers / 1173-Captain America / \n", "111-PlayStation 3 / 420-The Elder Scrolls V: Skyrim / 924-Trophy / \n", "163-Running / \n", "0-Games / 2-Video game / 6-Animation / 311-Street Fighter / 514-Tekken / 3634-Juri / 3740-Pandora / \n", "12-Food / 26-Cooking / 274-Meat / 589-Grilling / 620-Barbecue / 1085-Barbecue grill / \n", "8-Football / 21-Stadium / 77-Arena / \n", "1-Vehicle / 4-Car / 683-Renault / \n", "36-Piano / 53-Keyboard / 104-Pianist / 107-Musical keyboard / \n", "0-Games / 2-Video game / 345-Loudspeaker / \n", "6-Animation / 25-Toy / 200-Doll / 390-Model / 439-Barbie / \n", "75-Wrestling / \n", "427-Trumpet / \n", "1091-Saints Row / 1695-Saints Row: The Third / \n", "0-Games / 2-Video game / 43-Call of Duty / 111-PlayStation 3 / 125-Television / 134-Call of Duty: Black Ops / 2786-3D television / \n", "0-Games / 2-Video game / 19-PC game / 43-Call of Duty / 138-Call of Duty: Black Ops II / 422-Call of Duty: Ghosts / \n", "46-Choir / \n", "1-Vehicle / 4-Car / 299-Volkswagen Passenger Cars / 526-Volkswagen / 1761-Volkswagen Jetta / 3196-DVD player / \n", "318-Church / 885-Bell / \n", "36-Piano / 53-Keyboard / 107-Musical keyboard / 397-Organ / \n", "471-Wing / 3333-Cavalry / \n", "3114-Surveillance / \n", "3-Concert / 68-Lighting / \n", "23-Mobile phone / 29-Smartphone / \n", "3-Concert / \n", "12-Food / 26-Cooking / 32-Recipe / 52-Dish / 99-Christmas / 1332-Onion / \n", "49-School / \n", "59-Winter sport / 112-Ice skating / 316-Figure skating / 632-Ice dancing / \n", "1-Vehicle / 4-Car / 64-Transport / 74-Truck / \n", "28-Fashion / 45-Cosmetics / 56-Hair / 2235-Avon Products / \n", "45-Cosmetics / \n", "0-Games / 2-Video game / 8-Football / 90-Sports game / 180-FIFA 15 / 480-FIFA 13 / 848-FIFA 12 / \n", "49-School / \n", "730-Final Fantasy VII / \n", "59-Winter sport / 84-Snow / 177-Skiing / 281-Snowboarding / 305-Ski / 501-Snowboard / \n", "3-Concert / \n", "56-Hair / \n", "1-Vehicle / 4-Car / 11-Motorsport / 40-Road / 299-Volkswagen Passenger Cars / 526-Volkswagen / 969-Volkswagen Beetle / \n", "18-Outdoor recreation / 33-Weapon / 76-Ball / 3264-Golf ball / \n", "8-Football / \n", "162-Medicine / \n", "222-Surfing / 672-Bodyboarding / \n", "12-Food / 1060-Raw food / \n", "9-Music video / \n", "0-Games / 2-Video game / 6-Animation / 241-Fighting game / 486-M.U.G.E.N / 764-The King of Fighters / 1988-Chun-Li / \n", "0-Games / 116-Pokémon / 186-Pokémon / 1005-Pokémon Omega Ruby and Alpha Sapphire / \n", "6-Animation / 25-Toy / \n", "2-Video game / 6-Animation / \n", "18-Outdoor recreation / 22-Nature / 1254-Cave / \n", "0-Games / 2-Video game / 8-Football / 21-Stadium / 90-Sports game / 362-PlayStation Portable / 561-Madden NFL / \n", "7-Musician / 13-Musical ensemble / 49-School / 203-Marching band / 306-High school / 1675-Drum and bugle corps / \n", "0-Games / 792-DarkOrbit / \n", "14-Guitar / \n", "0-Games / 12-Food / 662-Pizza / \n", "593-Moon / \n", "12-Food / \n", "0-Games / 2-Video game / 55-Video game console / 413-PlayStation 2 / \n", "45-Cosmetics / 238-Nail / 300-Nail art / 304-Nail polish / 307-Hand / 324-Finger / 338-Manicure / \n", "1-Vehicle / 2616-Chevrolet S-10 / \n", "1-Vehicle / 4-Car / 277-Sport utility vehicle / 299-Volkswagen Passenger Cars / 526-Volkswagen / 982-V8 engine / 3587-Volkswagen Touareg / \n", "3-Concert / 7-Musician / 13-Musical ensemble / 46-Choir / 49-School / \n", "9-Music video / \n", "3-Concert / 13-Musical ensemble / 38-Orchestra / \n", "3-Concert / 7-Musician / 16-Performance art / 24-String instrument / \n", "6-Animation / 15-Cartoon / \n", "36-Piano / 237-Final Fantasy / 730-Final Fantasy VII / \n", "469-Bag / 653-Thread / 2329-Felt / \n", "25-Toy / 1052-Yo-Kai Watch / \n", "1-Vehicle / 4-Car / 11-Motorsport / 25-Toy / 3602-Mini-Z / \n", "677-Game controller / 1046-Remote control / \n", "0-Games / 1-Vehicle / 2-Video game / 4-Car / 434-Ninja / 2433-Trials / 3723-Trials Evolution / \n", "0-Games / 2-Video game / 47-Personal computer / 294-Batman / 838-Batman: Arkham / 1078-Installation art / 2084-Batman: Arkham Origins / \n", "9-Music video / \n", "3-Concert / 7-Musician / 13-Musical ensemble / 14-Guitar / 30-Drummer / 39-Drums / 3162-Postage stamp / \n", "121-Photography / 155-Camera / 469-Bag / 835-Backpack / \n", "322-Mixtape / \n", "1-Vehicle / 4-Car / 349-Chevrolet / 3353-Chevrolet Monte Carlo / 3371-Airbag / \n", "56-Hair / 340-Afro-textured hair / 583-Wig / 607-Lace / \n", "29-Smartphone / \n", "5-Dance / 6-Animation / \n", "7-Musician / 219-Accordion / 655-Diatonic button accordion / \n", "294-Batman / 4701-Scarecrow / \n", "1-Vehicle / 4-Car / 484-Subwoofer / \n", "0-Games / 35-Minecraft / 625-Factory / 1913-Solar panel / \n", "81-Athlete / 562-Pitcher / \n", "46-Choir / \n", "14-Guitar / 2738-Fender Jazz Bass / \n", "8-Football / 767-Goalkeeper / \n", "3-Concert / 16-Performance art / \n", "2-Video game / 116-Pokémon / 186-Pokémon / 1153-Pokémon HeartGold and SoulSilver / \n", "3-Concert / 5-Dance / 16-Performance art / 49-School / 306-High school / \n", "42-Fishing / 164-Beach / 222-Surfing / 233-Ocean / 672-Bodyboarding / \n", "5-Dance / \n", "0-Games / 60-Basketball / 168-Basketball moves / 255-Slam dunk / 1608-Injury / \n", "6-Animation / 232-Cake / 2313-Bánh / \n", "0-Games / 257-Counter-Strike / \n", "7-Musician / 14-Guitar / 24-String instrument / 100-Electric guitar / \n", "88-Machine / \n", "110-Album / \n", "15-Cartoon / 61-Art / 97-Drawing / 366-Dragon / \n", "4300-Tambourine / \n", "5-Dance / 7-Musician / 16-Performance art / 99-Christmas / \n", "6-Animation / 15-Cartoon / \n", "20-Trailer / \n", "1183-Karaoke box / \n", "3-Concert / 16-Performance art / \n", "6-Animation / 15-Cartoon / \n", "9-Music video / \n", "5-Dance / \n", "25-Toy / 200-Doll / 962-McDonald's / \n", "3-Concert / 7-Musician / 38-Orchestra / \n", "0-Games / 8-Football / 21-Stadium / 65-Kick / 73-Ball / \n", "0-Games / 2-Video game / 55-Video game console / 181-Xbox / \n", "2-Video game / 25-Toy / 126-The Walt Disney Company / 1106-Character / 2453-Disney Infinity / \n", "130-Gymnastics / 350-Circus / \n", "3-Concert / 68-Lighting / \n", "0-Games / 494-Card game / 615-Poker / \n", "85-Combat / 1239-K-1 / \n", "0-Games / 2-Video game / 37-Gadget / 47-Personal computer / 516-Headphones / 832-Headset / \n", "0-Games / 8-Football / 54-Highlight film / 79-American football / \n", "45-Cosmetics / 195-Mascara / 641-Mask / \n", "3-Concert / 7-Musician / \n", "110-Album / \n", "8-Football / \n", "1-Vehicle / 4-Car / 11-Motorsport / 88-Machine / 2623-Hose / \n", "0-Games / 49-School / \n", "0-Games / 8-Football / 21-Stadium / 65-Kick / 73-Ball / 81-Athlete / \n", "1-Vehicle / 11-Motorsport / 17-Racing / 27-Motorcycle / 57-Race track / 105-Motorcycling / 172-Motocross / \n", "0-Games / 2-Video game / 21-Stadium / 79-American football / 90-Sports game / 561-Madden NFL / \n", "3-Concert / 7-Musician / 13-Musical ensemble / \n", "28-Fashion / 45-Cosmetics / 56-Hair / 173-Eye shadow / 193-Eye liner / 195-Mascara / 196-Eye / 218-Eyelash / 235-Lipstick / 291-Rouge / 325-Vampire / 385-Face / 436-Concealer / 3655-Makeup brush / \n", "0-Games / 8-Football / \n", "12-Food / 26-Cooking / 32-Recipe / 52-Dish / 58-Cuisine / 620-Barbecue / 1186-Seafood / 2213-Shrimp / 3379-Prawn / \n", "0-Games / 2-Video game / \n", "10-Animal / 18-Outdoor recreation / 42-Fishing / 91-Fish / 268-Recreational fishing / \n", "93-Comedy / \n", "14-Guitar / 24-String instrument / 670-Banjo / \n", "1-Vehicle / 50-Aircraft / 139-Airport / 151-Landing / 187-Takeoff / 188-Airline / 254-Runway / \n", "85-Combat / 399-Kickboxing / 623-Coach / \n", "1-Vehicle / 4-Car / 270-Sedan / 299-Volkswagen Passenger Cars / 526-Volkswagen / 1761-Volkswagen Jetta / \n", "3-Concert / 7-Musician / 16-Performance art / 30-Drummer / \n", "0-Games / 2-Video game / 35-Minecraft / 1176-Pumpkin / \n", "6-Animation / 15-Cartoon / 185-Dragon Ball / 244-Goku / 418-Gohan / 1304-Cell / \n", "93-Comedy / 341-Sketch comedy / \n", "0-Games / 2-Video game / 55-Video game console / 586-Nintendo Entertainment System / \n", "3-Concert / 7-Musician / 9-Music video / \n", "222-Surfing / 233-Ocean / 672-Bodyboarding / \n", "79-American football / \n", "75-Wrestling / \n", "0-Games / 2-Video game / 6-Animation / 123-Naruto / 591-Naruto: Ultimate Ninja / \n", "3-Concert / \n", "0-Games / 35-Minecraft / \n", "3-Concert / 7-Musician / 13-Musical ensemble / \n", "7-Musician / 14-Guitar / 24-String instrument / 63-Acoustic guitar / \n", "809-Logo / 3233-Screencast / \n", "337-Advertising / 2032-Pepsi / \n", "3-Concert / 7-Musician / \n", "22-Nature / 425-City / \n", "18-Outdoor recreation / 83-Skateboarding / 108-Skateboard / \n", "15-Cartoon / 379-Touhou Project / \n", "1107-United States Navy / \n", "0-Games / 2-Video game / 33-Weapon / 1382-Half-Life 2 / 1412-Half-Life / \n", "0-Games / 19-PC game / 2043-Mabinogi / \n", "44-Drums / \n", "12-Food / 295-Chocolate / 1222-Fountain / 2256-Bearing / \n", "0-Games / 2-Video game / \n", "150-Violin / 545-Quartet (ensemble) / 1002-String quartet / 2477-Electric violin / \n", "162-Medicine / 288-Underwater / \n", "12-Food / 26-Cooking / 32-Recipe / 52-Dish / 109-Cooking show / 184-Vegetable / 210-Cookware and bakeware / 1114-Stuffing / 4053-Calabaza / \n", "0-Games / 8-Football / 90-Sports game / 180-FIFA 15 / \n", "99-Christmas / \n", "0-Games / 19-PC game / 664-CrossFire / \n", "0-Games / 2-Video game / 33-Weapon / 43-Call of Duty / 302-Call of Duty 4: Modern Warfare / \n", "8-Football / \n", "8-Football / \n", "1-Vehicle / 4-Car / 11-Motorsport / 74-Truck / 167-Four-wheel drive / 198-Off-road vehicle / 327-Mud / 565-Mud bogging / \n", "3-Concert / 7-Musician / 13-Musical ensemble / 24-String instrument / 38-Orchestra / 150-Violin / 1710-Pirates of the Caribbean / \n", "0-Games / 2-Video game / \n", "1-Vehicle / 11-Motorsport / 18-Outdoor recreation / 27-Motorcycle / \n", "0-Games / 51-Strategy video game / 165-League of Legends / \n", "21-Stadium / 54-Highlight film / \n", "1-Vehicle / 92-Radio-controlled model / 149-Model aircraft / 160-Radio-controlled aircraft / 393-Unmanned aerial vehicle / 604-Quadcopter / \n", "5-Dance / \n", "3-Concert / 16-Performance art / \n", "45-Cosmetics / 173-Eye shadow / 193-Eye liner / 195-Mascara / 196-Eye / 218-Eyelash / 235-Lipstick / 291-Rouge / \n", "98-Festival / \n", "14-Guitar / 20-Trailer / \n", "1-Vehicle / 4-Car / 11-Motorsport / 17-Racing / 57-Race track / 4612-Sepang International Circuit / \n", "12-Food / 26-Cooking / 32-Recipe / 52-Dish / 58-Cuisine / 301-Meal / 454-Sauce / 1567-Cook / \n", "0-Games / 8-Football / 21-Stadium / 54-Highlight film / 81-Athlete / \n", "35-Minecraft / 141-Microsoft Windows / \n", "3-Concert / \n", "5-Dance / \n", "0-Games / 2-Video game / 35-Minecraft / 394-Map / 2716-The Hunger Games / \n", "25-Toy / 200-Doll / 439-Barbie / \n", "3-Concert / 95-Talent show / \n", "2-Video game / 55-Video game console / 199-Wii / 443-Wii U / \n", "3-Concert / \n", "5-Dance / 20-Trailer / \n", "85-Combat / 117-Boxing / \n", "3-Concert / 7-Musician / 13-Musical ensemble / 38-Orchestra / 39-Drums / \n", "0-Games / 2-Video game / 55-Video game console / \n", "0-Games / 2-Video game / 19-PC game / 159-World of Warcraft / \n", "0-Games / 54-Highlight film / 60-Basketball / 255-Slam dunk / \n", "9-Music video / \n", "28-Fashion / 1534-H&M / \n", "7-Musician / 219-Accordion / 1994-Garmon / \n", "261-Tool / 673-Saw / 2437-String trimmer / \n", "121-Photography / \n", "3-Concert / 68-Lighting / \n", "261-Tool / \n", "9-Music video / \n", "3-Concert / 68-Lighting / \n", "0-Games / 1187-Ragnarok Online / \n", "8-Football / \n", "0-Games / 2966-Patchwork / \n", "1-Vehicle / 4-Car / 70-Driving / 94-Dashcam / 277-Sport utility vehicle / 332-Mercedes-Benz / 344-Coupé / 891-Mercedes-AMG / \n", "1-Vehicle / 4-Car / 11-Motorsport / 27-Motorcycle / \n", "1-Vehicle / 4-Car / 23-Mobile phone / 72-iPhone / \n", "4665-Day spa / \n", "3-Concert / 7-Musician / 21-Stadium / \n", "609-Tattoo / 863-Body piercing / \n", "23-Mobile phone / 29-Smartphone / 37-Gadget / 72-iPhone / 189-iPad / 272-iPod / 331-iPod touch / 554-iPhone 4 / 675-iPhone 4S / 1009-iPhone 3G / 1155-iPhone 3GS / 1504-iPad 3 / \n", "602-Central processing unit / 3391-Alesis / \n", "1-Vehicle / 10-Animal / 25-Toy / 74-Truck / 1420-Gear / 1611-Garbage truck / \n", "12-Food / 26-Cooking / 32-Recipe / 52-Dish / 58-Cuisine / 109-Cooking show / 184-Vegetable / 500-Rice / 727-Curry / 2000-Mango / \n", "9-Music video / \n", "416-Apartment / \n", "1-Vehicle / 4-Car / 245-BMW / 344-Coupé / \n", "1572-Husband / \n", "1-Vehicle / 4-Car / 11-Motorsport / 17-Racing / 27-Motorcycle / \n", "0-Games / 1167-Tanki Online / \n", "0-Games / 2-Video game / 354-Egg / 426-Resident Evil / \n", "20-Trailer / \n", "3-Concert / \n", "37-Gadget / 2009-Electromagnetic coil / \n", "6-Animation / 2769-Decal / \n", "0-Games / 10-Animal / 159-World of Warcraft / 171-Warcraft / 3301-Mop / \n", "1789-Slender: The Eight Pages / \n", "10-Animal / 80-Horse / \n", "0-Games / 60-Basketball / 168-Basketball moves / 255-Slam dunk / \n", "9-Music video / 10-Animal / 48-Pet / 71-Dog / \n", "0-Games / 2-Video game / 9-Music video / 43-Call of Duty / 138-Call of Duty: Black Ops II / \n", "1-Vehicle / 4-Car / \n", "1-Vehicle / 62-Train / 64-Transport / 103-Rail transport / 119-Locomotive / 143-Railroad car / 152-Track / 161-Train station / \n", "0-Games / 2-Video game / 23-Mobile phone / 29-Smartphone / 72-iPhone / 508-iPhone 5 / 760-iPhone 5s / \n", "10-Animal / 18-Outdoor recreation / 22-Nature / 42-Fishing / 91-Fish / 120-Winter / 147-River / 423-Coast / 825-Stream / 895-Trout / 1018-Salmon / 2094-Rainbow trout / \n", "243-Sonic the Hedgehog / \n", "3-Concert / 5-Dance / 7-Musician / \n", "6-Animation / 15-Cartoon / \n", "2337-Passport / \n", "67-Cycling / 69-Bicycle / \n", "1-Vehicle / 50-Aircraft / 82-Airplane / 92-Radio-controlled model / 149-Model aircraft / 160-Radio-controlled aircraft / 388-Fighter aircraft / \n", "45-Cosmetics / 238-Nail / 300-Nail art / 304-Nail polish / 338-Manicure / \n", "0-Games / 2-Video game / 2572-Lupin III / \n", "18-Outdoor recreation / 22-Nature / 283-Rock / 347-Climbing / 828-Rock climbing / \n", "3-Concert / 7-Musician / 13-Musical ensemble / 14-Guitar / 24-String instrument / \n", "0-Games / 712-Outer space / 3585-Space Engineers / \n", "75-Wrestling / \n", "6-Animation / 89-Comics / 269-Comic book / 803-DC Comics / 1106-Character / \n", "12-Food / 52-Dish / 162-Medicine / 215-Eating / 780-Fat / \n", "0-Games / 2-Video game / 6-Animation / 241-Fighting game / 311-Street Fighter / 461-Street Fighter IV / 496-Super Street Fighter IV / 1331-Ken Masters / \n", "1-Vehicle / 4-Car / 11-Motorsport / 154-Wheel / 158-Tire / \n", "49-School / 191-Cheerleading / \n", "3-Concert / \n", "0-Games / 1-Vehicle / 2-Video game / 4-Car / 34-Action-adventure game / 297-Grand Theft Auto IV / 658-Grand Theft Auto: The Lost and Damned / \n", "0-Games / 6-Animation / 15-Cartoon / 1118-Kirby / 1133-Kirby / 3421-Meta Knight / 4449-King Dedede / \n", "0-Games / 8-Football / 65-Kick / 76-Ball / 595-Champion / \n", "5-Dance / \n", "0-Games / 18-Outdoor recreation / 155-Camera / \n", "1-Vehicle / 259-Fire / 371-Firefighter / 796-Fire engine / \n", "0-Games / 2-Video game / 464-First-person Shooter / \n", "0-Games / 8-Football / 65-Kick / 73-Ball / \n", "28-Fashion / 45-Cosmetics / \n", "31-Disc jockey / \n", "18-Outdoor recreation / 42-Fishing / 147-River / 225-Lake / 896-Largemouth bass / \n", "122-Weight training / 204-Gym / 492-Floor / 529-Squat / 532-Barbell / 1389-Dumbbell / \n", "95-Talent show / \n", "0-Games / 2-Video game / 33-Weapon / \n", "25-Toy / 396-One Piece / 435-Action figure / \n", "1-Vehicle / 4-Car / 11-Motorsport / 17-Racing / 552-Honda Civic / \n", "0-Games / 2-Video game / 15-Cartoon / 89-Comics / 185-Dragon Ball / 244-Goku / 418-Gohan / 876-Freeza / 1244-Trunks / 2384-Dragon Ball: Raging Blast / \n", "0-Games / 1427-TalesRunner / \n", "273-News program / \n", "1-Vehicle / 4-Car / 74-Truck / 114-Engine / 527-Battery / 659-Van / \n", "1-Vehicle / 4-Car / 360-Jeep / 1205-Jeep Wrangler / \n", "0-Games / 1-Vehicle / 2-Video game / 11-Motorsport / 55-Video game console / 199-Wii / 456-Go-kart / 518-Mario Kart / 2750-Mii / \n", "0-Games / 2-Video game / 35-Minecraft / 174-Farm / 1158-Wheat / \n", "303-Paint / 319-Robot / \n", "20-Trailer / \n", "0-Games / 779-Point Blank / \n", "23-Mobile phone / 29-Smartphone / 72-iPhone / \n", "9-Music video / \n", "12-Food / 26-Cooking / 32-Recipe / 109-Cooking show / 1765-Muffin / \n", "8-Football / \n", "8-Football / 661-Stretching / \n", "0-Games / 2-Video game / 19-PC game / 43-Call of Duty / 120-Winter / 134-Call of Duty: Black Ops / 138-Call of Duty: Black Ops II / 422-Call of Duty: Ghosts / \n", "1-Vehicle / 4-Car / 41-Sports car / \n", "0-Games / 2-Video game / 34-Action-adventure game / \n", "56-Hair / 106-Hairstyle / 340-Afro-textured hair / \n", "18-Outdoor recreation / 42-Fishing / 222-Surfing / \n", "296-Saxophone / \n", "242-Paper / 313-Book / \n", "123-Naruto / \n", "337-Advertising / \n", "5-Dance / \n", "1-Vehicle / 4-Car / 40-Road / \n", "3-Concert / 7-Musician / 9-Music video / \n", "10-Animal / 91-Fish / 258-Aquarium / \n", "12-Food / 786-Juice / \n", "6-Animation / 1579-Circle / 2010-Chart / \n", "3-Concert / \n", "8-Football / \n", "0-Games / 35-Minecraft / 47-Personal computer / 131-Computer / 141-Microsoft Windows / \n", "1-Vehicle / 4-Car / 11-Motorsport / 41-Sports car / 207-Exhaust system / 790-Nissan GT-R / 1301-Nissan Skyline / 2247-Nissan Skyline GT-R / \n", "141-Microsoft Windows / \n", "3895-Curling / \n", "0-Games / 8-Football / 21-Stadium / 65-Kick / 73-Ball / 76-Ball / 81-Athlete / \n", "0-Games / 2-Video game / 241-Fighting game / 311-Street Fighter / 486-M.U.G.E.N / 764-The King of Fighters / \n", "3-Concert / 7-Musician / 39-Drums / 98-Festival / \n", "23-Mobile phone / 29-Smartphone / 37-Gadget / 125-Television / 487-Samsung Electronics / 1595-Television set / \n", "204-Gym / 4667-Gold's Gym / \n", "1-Vehicle / 62-Train / 64-Transport / 103-Rail transport / 119-Locomotive / 428-Rail transport modelling / \n", "102-Building / 170-Home improvement / 433-Construction / 1821-Office / \n", "1-Vehicle / 11-Motorsport / 18-Outdoor recreation / 22-Nature / 27-Motorcycle / 105-Motorcycling / 213-Forest / 491-Enduro / \n", "0-Games / 530-Cue sports / 568-Pool / 1334-Eight-ball / \n", "36-Piano / \n", "30-Drummer / 39-Drums / 44-Drums / 128-Cymbal / \n", "12-Food / 26-Cooking / 52-Dish / 109-Cooking show / 215-Eating / 274-Meat / 301-Meal / \n", "18-Outdoor recreation / 22-Nature / 164-Beach / 233-Ocean / 423-Coast / 1453-Balloon / \n", "79-American football / \n", "28-Fashion / 45-Cosmetics / \n", "6-Animation / 9-Music video / 1632-Villain / \n", "68-Lighting / 99-Christmas / 729-Christmas decoration / 1098-Christmas lights / \n", "0-Games / 2-Video game / 6-Animation / 586-Nintendo Entertainment System / \n", "376-Flute / 1326-Recorder / \n", "3-Concert / 5-Dance / \n", "110-Album / \n", "2-Video game / 43-Call of Duty / 431-Call of Duty: Advanced Warfare / \n", "14-Guitar / 24-String instrument / 63-Acoustic guitar / \n", "0-Games / 1-Vehicle / 2-Video game / 4-Car / 11-Motorsport / 17-Racing / 240-Arcade game / 1333-Wangan Midnight / \n", "10-Animal / 17-Racing / 80-Horse / 384-Horse racing / \n", "1-Vehicle / 4-Car / 22-Nature / 114-Engine / \n", "3-Concert / 7-Musician / \n", "6-Animation / 3162-Postage stamp / 3544-Rubber stamp / \n", "3-Concert / \n", "1-Vehicle / 4-Car / 11-Motorsport / 17-Racing / \n", "25-Toy / 139-Airport / 140-LEGO / 1079-Lego City / \n", "1-Vehicle / 11-Motorsport / 124-Tractor / 156-Agriculture / 174-Farm / 252-Tractor pulling / 459-Plough / 660-Rural area / \n", "0-Games / 2-Video game / 34-Action-adventure game / 35-Minecraft / \n", "25-Toy / 269-Comic book / \n", "28-Fashion / 153-Dress / \n", "164-Beach / 248-Hotel / 369-Resort / \n", "1-Vehicle / 74-Truck / 208-Wood / 492-Floor / 619-Manufacturing / \n", "6-Animation / 851-Digimon / \n", "1-Vehicle / 4-Car / 11-Motorsport / 17-Racing / 40-Road / 41-Sports car / 57-Race track / 231-Supercar / 344-Coupé / 349-Chevrolet / 406-Chevrolet / 761-Chevrolet Camaro / 1752-Chevrolet SS / \n", "325-Vampire / \n", "120-Winter / 657-Autumn / \n", "3-Concert / 7-Musician / 13-Musical ensemble / 46-Choir / 318-Church / \n", "48-Pet / 182-Puppy / 1278-Labrador Retriever / \n", "45-Cosmetics / 173-Eye shadow / 193-Eye liner / 195-Mascara / 218-Eyelash / \n", "10-Animal / 48-Pet / 132-Bird / \n", "150-Violin / 1747-Erhu / \n", "0-Games / 1-Vehicle / 2-Video game / 4-Car / 11-Motorsport / 17-Racing / 41-Sports car / 57-Race track / 70-Driving / 352-Need for Speed / 1056-Porsche 911 / 1600-Need for Speed: World / \n", "0-Games / 33-Weapon / 34-Action-adventure game / 118-Grand Theft Auto V / 201-Grand Theft Auto: San Andreas / 297-Grand Theft Auto IV / \n", "12-Food / 26-Cooking / 32-Recipe / 58-Cuisine / 226-Dessert / 295-Chocolate / 684-Ice cream / 1322-Biscuit / 2418-Gelatin / 2777-Mousse / 2825-Rum / 2858-Gelatin dessert / \n", "3-Concert / 5-Dance / 16-Performance art / \n", "45-Cosmetics / 238-Nail / 300-Nail art / 304-Nail polish / 307-Hand / 324-Finger / 338-Manicure / \n", "1-Vehicle / 11-Motorsport / 17-Racing / 27-Motorcycle / 57-Race track / 105-Motorcycling / \n", "7-Musician / 9-Music video / 14-Guitar / 30-Drummer / \n", "0-Games / 2-Video game / 43-Call of Duty / 134-Call of Duty: Black Ops / \n", "1-Vehicle / 18-Outdoor recreation / 22-Nature / 67-Cycling / 69-Bicycle / 220-Mountain bike / 283-Rock / 284-Mountain biking / 347-Climbing / 800-Mountain pass / \n", "1726-Harlem Shake (meme) / \n", "3-Concert / 5-Dance / 16-Performance art / 357-DVD / \n", "0-Games / 8-Football / \n", "59-Winter sport / 84-Snow / 281-Snowboarding / 501-Snowboard / 827-Pump / \n", "0-Games / 49-School / 60-Basketball / 95-Talent show / 191-Cheerleading / \n", "131-Computer / 141-Microsoft Windows / \n", "9-Music video / \n", "1701-French fries / \n", "33-Weapon / \n", "3-Concert / 5-Dance / 16-Performance art / \n", "0-Games / 8-Football / 90-Sports game / 180-FIFA 15 / \n", "61-Art / 1846-Mural / \n", "36-Piano / 53-Keyboard / 104-Pianist / 107-Musical keyboard / 265-Electronic keyboard / \n", "1-Vehicle / 4-Car / 18-Outdoor recreation / \n", "0-Games / 2-Video game / 352-Need for Speed / 875-Need for Speed: Most Wanted / \n", "2-Video game / 240-Arcade game / 488-Door / 731-Coin / \n", "0-Games / 35-Minecraft / \n", "462-Table / 695-Balloon / 1327-Minnie Mouse / \n", "1-Vehicle / 4-Car / 11-Motorsport / 17-Racing / 41-Sports car / 57-Race track / 169-Rallying / 245-BMW / 522-Helmet / \n", "661-Stretching / \n", "159-World of Warcraft / 171-Warcraft / 372-Playing card / 791-Magic: The Gathering / 1021-Booster pack / 1049-Collectible card game / \n", "0-Games / 2-Video game / 20-Trailer / 34-Action-adventure game / 89-Comics / 294-Batman / 838-Batman: Arkham / 2084-Batman: Arkham Origins / \n", "0-Games / 2-Video game / 55-Video game console / 111-PlayStation 3 / 317-PlayStation / 1482-PlayStation Network / \n", "1-Vehicle / 4-Car / 11-Motorsport / 17-Racing / 135-Drifting / \n", "23-Mobile phone / 29-Smartphone / 37-Gadget / 101-Telephone / 178-Samsung Galaxy / 2508-Samsung Galaxy Ace / \n", "0-Games / 463-Cricket / \n", "3-Concert / 5-Dance / \n", "12-Food / 215-Eating / 968-Pie / 1594-Nerd / 3613-Apple pie / \n", "861-Mixing console / 3391-Alesis / \n", "8-Football / 21-Stadium / 77-Arena / \n", "3-Concert / 7-Musician / \n", "25-Toy / 140-LEGO / \n", "12-Food / 1227-Sushi / \n", "85-Combat / 117-Boxing / \n", "1-Vehicle / 62-Train / 64-Transport / 103-Rail transport / 161-Train station / 224-Rapid transit / \n", "9-Music video / 98-Festival / \n", "1-Vehicle / 4-Car / 299-Volkswagen Passenger Cars / \n", "1-Vehicle / 4-Car / 519-Recreational vehicle / 742-Countertop / \n", "1-Vehicle / 4-Car / 27-Motorcycle / 114-Engine / 154-Wheel / 207-Exhaust system / \n", "9-Music video / \n", "14-Guitar / 24-String instrument / 63-Acoustic guitar / 1936-Bouzouki / \n", "0-Games / 596-Monster Hunter / 1339-Monster Hunter Freedom Unite / 2994-Monster Hunter Freedom / \n", "3-Concert / 1303-Fullmetal Alchemist / \n", "10-Animal / 22-Nature / 1039-Tiger / \n", "8-Football / \n", "320-Super Smash Bros. / 398-The Legend of Zelda / \n", "1-Vehicle / 4-Car / 11-Motorsport / 27-Motorcycle / 40-Road / 105-Motorcycling / 212-Highway / 608-Wheelie / \n", "10-Animal / 709-Bull / \n", "1-Vehicle / 27-Motorcycle / 105-Motorcycling / 1395-Valve / \n", "36-Piano / \n", "12-Food / 26-Cooking / 32-Recipe / 226-Dessert / 236-Baking / 295-Chocolate / 1223-Pastry / 1576-Doughnut / \n", "0-Games / 2-Video game / \n", "1-Vehicle / 4-Car / \n", "13-Musical ensemble / \n", "0-Games / 8-Football / 65-Kick / \n", "31-Disc jockey / \n", "6-Animation / 10-Animal / 15-Cartoon / \n", "22-Nature / 279-Garden / 369-Resort / \n", "1-Vehicle / 4-Car / 303-Paint / \n", "86-Plant / 136-Gardening / 279-Garden / \n", "66-Bollywood / \n", "6-Animation / 183-Manga / \n", "121-Photography / 155-Camera / 566-Digital camera / 1117-Camcorder / \n", "8-Football / 21-Stadium / 77-Arena / \n", "133-Nightclub / \n", "162-Medicine / 925-Brain / \n", "0-Games / 2-Video game / 116-Pokémon / 186-Pokémon / 4521-Deoxys / \n", "56-Hair / 106-Hairstyle / \n", "59-Winter sport / 112-Ice skating / 262-Ice / 316-Figure skating / 632-Ice dancing / 833-Ice skate / \n", "1-Vehicle / 11-Motorsport / 27-Motorcycle / 207-Exhaust system / \n", "0-Games / 51-Strategy video game / 373-Clash of Clans / \n", "1113-Toilet / 1308-Shower / 2472-Tap / \n", "3-Concert / 726-Clarinet / \n", "390-Model / \n", "3-Concert / 1231-Mandolin / \n", "162-Medicine / 196-Eye / \n", "0-Games / 2-Video game / 23-Mobile phone / 29-Smartphone / 72-iPhone / 113-House / 1234-Mouse / \n", "1-Vehicle / 1409-Missile / \n", "227-Water / \n", "2794-Marionette / \n", "9-Music video / \n", "1-Vehicle / 4-Car / 94-Dashcam / 490-Rain / \n", "20-Trailer / \n", "0-Games / 67-Cycling / 69-Bicycle / 570-Road bicycle racing / \n", "6-Animation / 15-Cartoon / 272-iPod / 1674-Ouran High School Host Club / \n", "3-Concert / 7-Musician / 16-Performance art / \n", "3-Concert / 7-Musician / 13-Musical ensemble / \n", "1-Vehicle / 4-Car / 74-Truck / 114-Engine / 349-Chevrolet / 375-Pickup truck / 406-Chevrolet / \n", "1-Vehicle / 4-Car / 27-Motorcycle / 40-Road / 94-Dashcam / 105-Motorcycling / \n", "0-Games / 256-Tennis / \n", "0-Games / 19-PC game / 35-Minecraft / \n", "56-Hair / 60-Basketball / \n", "36-Piano / 53-Keyboard / \n", "3-Concert / \n", "1-Vehicle / 4-Car / 11-Motorsport / 84-Snow / 120-Winter / 135-Drifting / 680-Subaru / 1041-Subaru Impreza / \n", "12-Food / 26-Cooking / 32-Recipe / 269-Comic book / 367-Drink / 613-Bar / 904-Cocktail / 1669-Vodka / \n", "1487-Reborn doll / \n", "28-Fashion / 206-Shoe / \n", "10-Animal / 588-Reptile / 1444-Lizard / 3060-Chameleon / \n", "18-Outdoor recreation / 42-Fishing / 222-Surfing / 2415-Big wave surfing / \n", "99-Christmas / 558-Santa Claus / \n", "5-Dance / 115-Ballet / \n", "0-Games / 2-Video game / 116-Pokémon / 186-Pokémon / 4267-GameShark / \n", "83-Skateboarding / 108-Skateboard / 221-Skatepark / 2919-Skate 2 / \n", "237-Final Fantasy / 1131-Final Fantasy XIII / 2506-Final Fantasy XIII-2 / \n", "1-Vehicle / 62-Train / 64-Transport / 103-Rail transport / 119-Locomotive / \n", "1-Vehicle / 10-Animal / 18-Outdoor recreation / 42-Fishing / 87-Boat / 91-Fish / 233-Ocean / 268-Recreational fishing / 698-Fisherman / 3763-Atlantic bluefin tuna / \n", "379-Touhou Project / \n", "2-Video game / 3515-Gas tungsten arc welding / 3929-Gas metal arc welding / \n", "0-Games / 35-Minecraft / \n", "5-Dance / 16-Performance art / 46-Choir / \n", "283-Rock / 492-Floor / \n", "7-Musician / 13-Musical ensemble / 46-Choir / 78-Wedding / 190-Bride / \n", "264-Star Wars / 714-Jedi / 717-Darth Vader / \n", "5-Dance / 16-Performance art / 130-Gymnastics / 1255-Ribbon / \n", "22-Nature / \n", "7-Musician / 36-Piano / 53-Keyboard / 104-Pianist / 107-Musical keyboard / 307-Hand / \n", "8-Football / \n", "66-Bollywood / \n", "31-Disc jockey / 1143-Marimba / \n", "1-Vehicle / 4-Car / 27-Motorcycle / 84-Snow / 105-Motorcycling / 120-Winter / \n", "0-Games / 2-Video game / 3717-Suikoden / \n", "1-Vehicle / 164-Beach / 423-Coast / \n", "88-Machine / 400-Woodturning / 972-Lathe / \n", "6-Animation / 15-Cartoon / 99-Christmas / \n", "1-Vehicle / 88-Machine / 249-Heavy equipment / 817-Bulldozer / \n", "86-Plant / 136-Gardening / 279-Garden / 1460-Herb / \n", "1-Vehicle / 136-Gardening / 249-Heavy equipment / 363-Soil / 2138-Compost / \n", "3-Concert / 7-Musician / 14-Guitar / 24-String instrument / 30-Drummer / \n", "1-Vehicle / 11-Motorsport / 17-Racing / 27-Motorcycle / 172-Motocross / \n", "3-Concert / \n", "10-Animal / 48-Pet / 71-Dog / 84-Snow / 280-Wildlife / \n", "0-Games / 2-Video game / 19-PC game / 229-Halo / 470-Halo 3 / \n", "0-Games / 1427-TalesRunner / \n", "10-Animal / 137-Tree / 542-Potato / 725-Human / \n", "12-Food / 2496-Kebab / \n", "0-Games / 2-Video game / 51-Strategy video game / 537-Zoo / 865-Turtle / 4255-Zoo Tycoon 2 / \n", "117-Boxing / \n", "46-Choir / \n", "3-Concert / 7-Musician / 24-String instrument / \n", "3-Concert / 5-Dance / 16-Performance art / 95-Talent show / 115-Ballet / 130-Gymnastics / \n", "10-Animal / 22-Nature / 280-Wildlife / 540-Insect / \n", "1-Vehicle / 4-Car / 11-Motorsport / 135-Drifting / 2739-Midnight Club: Los Angeles / \n", "3-Concert / 7-Musician / \n", "0-Games / 2-Video game / 43-Call of Duty / 216-Call of Duty: Modern Warfare 3 / \n", "3-Concert / 7-Musician / 13-Musical ensemble / 14-Guitar / \n", "0-Games / 13-Musical ensemble / 203-Marching band / \n", "9-Music video / 66-Bollywood / \n", "1-Vehicle / 4-Car / 50-Aircraft / 259-Fire / 559-Night / \n", "61-Art / 97-Drawing / 479-Graffiti / 1385-Street art / 1846-Mural / \n", "0-Games / 60-Basketball / 1399-Streetball / \n", "45-Cosmetics / 56-Hair / \n", "283-Rock / \n", "0-Games / 88-Machine / 240-Arcade game / 535-Slot machine / 3235-Monopoly / \n", "5-Dance / \n", "731-Coin / 912-Silver / \n", "6-Animation / 15-Cartoon / 505-Mega Man / 1664-Zero / \n", "0-Games / 2-Video game / 88-Machine / 535-Slot machine / \n", "8-Football / \n", "1-Vehicle / 25-Toy / 62-Train / \n", "10-Animal / 48-Pet / 71-Dog / \n", "0-Games / \n", "5-Dance / \n", "8-Football / \n", "12-Food / 26-Cooking / 210-Cookware and bakeware / 589-Grilling / 620-Barbecue / 1085-Barbecue grill / \n", "10-Animal / 22-Nature / 440-Lion / 537-Zoo / \n", "3-Concert / \n", "8-Football / \n", "72-iPhone / 272-iPod / \n", "85-Combat / \n", "6-Animation / 35-Minecraft / 99-Christmas / \n", "1-Vehicle / 27-Motorcycle / 1380-Kawasaki motorcycles / 1951-Sport bike / 2404-Kawasaki Ninja / 2494-Ninja ZX-6R / 3873-Kawasaki Ninja 300 / \n", "6-Animation / 15-Cartoon / 396-One Piece / \n", "3-Concert / 7-Musician / 9-Music video / \n", "123-Naruto / \n", "3-Concert / 7-Musician / \n", "1-Vehicle / 4-Car / 84-Snow / 120-Winter / 277-Sport utility vehicle / 680-Subaru / 1286-Subaru / 2862-Subaru Forester / \n", "59-Winter sport / 84-Snow / 177-Skiing / 281-Snowboarding / 3694-Burton Snowboards / \n", "7-Musician / 13-Musical ensemble / 14-Guitar / \n", "0-Games / 2-Video game / 474-Destiny / \n", "0-Games / 51-Strategy video game / 165-League of Legends / \n", "7-Musician / 14-Guitar / \n", "10-Animal / 48-Pet / 166-Cat / 346-Kitten / \n", "99-Christmas / 729-Christmas decoration / \n", "1-Vehicle / 64-Transport / 286-Bus / \n", "0-Games / 2-Video game / 34-Action-adventure game / 123-Naruto / \n", "5-Dance / \n", "979-Star Trek / 2981-Spock / \n", "5-Dance / 16-Performance art / 115-Ballet / 980-Solo dance / 3643-Starbound / \n", "0-Games / 2-Video game / 23-Mobile phone / 29-Smartphone / 72-iPhone / 1009-iPhone 3G / \n", "3-Concert / \n", "237-Final Fantasy / 1371-Yuna / 1706-Final Fantasy X-2 / \n", "2-Video game / 23-Mobile phone / 29-Smartphone / 37-Gadget / 101-Telephone / 178-Samsung Galaxy / 899-Samsung Galaxy S III / \n", "3-Concert / \n", "372-Playing card / 746-Card manipulation / \n", "6-Animation / 10-Animal / 15-Cartoon / 577-Penguin / \n", "1061-Marriage proposal / \n", "6-Animation / 15-Cartoon / 750-Sailor Moon / 1342-Sailor Moon / \n", "3-Concert / \n", "8-Football / \n", "242-Paper / 370-Origami / 1023-Rose / 1403-Flower bouquet / \n", "0-Games / 2-Video game / 269-Comic book / 489-Board game / 494-Card game / \n", "1-Vehicle / 59-Winter sport / 84-Snow / 120-Winter / 351-Jumping / 608-Wheelie / 610-Snowmobile / 1201-Ski-Doo / 1210-Sled / \n", "1-Vehicle / 25-Toy / 62-Train / 64-Transport / 103-Rail transport / 119-Locomotive / 507-Thomas the Tank Engine / \n", "9-Music video / \n", "2-Video game / 192-Battlefield / 1370-Battlefield 2 / \n", "110-Album / 956-Glasses / \n", "7-Musician / 13-Musical ensemble / 46-Choir / \n", "1-Vehicle / 2143-18 Wheels of Steel / \n", "83-Skateboarding / 108-Skateboard / \n", "5-Dance / \n", "1-Vehicle / 27-Motorcycle / \n", "313-Book / \n", "0-Games / 8-Football / 21-Stadium / 65-Kick / 73-Ball / 81-Athlete / \n", "55-Video game console / \n", "1-Vehicle / 4-Car / 92-Radio-controlled model / 202-Radio-controlled car / 945-MINI Cooper / \n", "14-Guitar / \n", "1-Vehicle / 4-Car / 11-Motorsport / 27-Motorcycle / 70-Driving / \n", "49-School / 860-Christian Church / \n", "8-Football / 21-Stadium / 77-Arena / \n", "85-Combat / 699-Kick / 1146-Wing Chun / \n", "0-Games / 311-Street Fighter / 461-Street Fighter IV / 496-Super Street Fighter IV / 563-Ryu / \n", "3969-Patience / \n", "0-Games / 2-Video game / 387-Spider-Man / \n", "9-Music video / \n", "1-Vehicle / 4-Car / 154-Wheel / 299-Volkswagen Passenger Cars / 493-Rim / 526-Volkswagen / 531-Hatchback / 668-Volkswagen Golf / \n", "12-Food / \n", "3-Concert / \n", "6-Animation / 10-Animal / 15-Cartoon / 61-Art / 97-Drawing / 132-Bird / 148-Painting / 408-Sketch / \n", "13-Musical ensemble / 30-Drummer / 39-Drums / 44-Drums / 49-School / 128-Cymbal / 146-Snare drum / 203-Marching band / \n", "85-Combat / 117-Boxing / \n", "1-Vehicle / 42-Fishing / 87-Boat / 88-Machine / 896-Largemouth bass / \n", "1-Vehicle / 4-Car / 41-Sports car / 154-Wheel / 3834-Porsche Panamera / \n", "3-Concert / \n", "28-Fashion / \n", "2-Video game / 25-Toy / 140-LEGO / 434-Ninja / 884-Lego minifigure / 951-Teenage Mutant Ninja Turtles / \n", "0-Games / 2-Video game / 211-RuneScape / \n", "83-Skateboarding / 108-Skateboard / \n", "1-Vehicle / 18-Outdoor recreation / 42-Fishing / 87-Boat / \n", "6-Animation / 15-Cartoon / \n", "0-Games / 1187-Ragnarok Online / \n", "0-Games / 2-Video game / 6-Animation / 89-Comics / 185-Dragon Ball / 244-Goku / \n", "1436-Draco Malfoy / \n", "9-Music video / \n", "261-Tool / 1033-Drill / 1226-Metalworking / \n", "3-Concert / \n", "630-Superman / 2051-Justice League / \n", "9-Music video / \n", "5-Dance / 164-Beach / \n", "66-Bollywood / \n", "2-Video game / 43-Call of Duty / 216-Call of Duty: Modern Warfare 3 / \n", "122-Weight training / 204-Gym / 2505-Slingshot / \n", "9-Music video / 1183-Karaoke box / \n", "3-Concert / 68-Lighting / 133-Nightclub / \n", "85-Combat / \n", "1-Vehicle / 4-Car / 245-BMW / 3320-BMW 5 Series (E60) / \n", "1-Vehicle / 4-Car / \n", "775-Bee / 4198-Khene / \n", "20-Trailer / \n", "2-Video game / 55-Video game console / 127-Xbox 360 / 181-Xbox / 386-Xbox / 700-Hard disk drive / \n", "99-Christmas / 137-Tree / 964-Christmas tree / \n", "96-Soldier / 99-Christmas / \n", "18-Outdoor recreation / 83-Skateboarding / 108-Skateboard / \n", "28-Fashion / \n", "93-Comedy / \n", "51-Strategy video game / 366-Dragon / 1718-Spore / \n", "102-Building / 248-Hotel / 574-Elevator / 1746-Otis Elevator Company / 3809-Marriott International / \n", "0-Games / 788-Juggling / \n", "20-Trailer / 625-Factory / \n", "9-Music video / 133-Nightclub / \n", "0-Games / 8-Football / 54-Highlight film / \n", "205-Newscaster / 624-Number / \n", "313-Book / \n", "0-Games / 1-Vehicle / 11-Motorsport / 76-Ball / 169-Rallying / 256-Tennis / 974-Racket / 1337-Serve / \n", "0-Games / 6-Animation / 15-Cartoon / \n", "208-Wood / 956-Glasses / \n", "75-Wrestling / 1042-Jumbotron / \n", "1-Vehicle / 4-Car / 6-Animation / \n", "23-Mobile phone / 37-Gadget / 47-Personal computer / 145-Tablet computer / 189-iPad / 1504-iPad 3 / 2253-CNET / \n", "1-Vehicle / 4-Car / 11-Motorsport / 17-Racing / 41-Sports car / 57-Race track / 135-Drifting / 3409-Nissan Altima / \n", "8-Football / 205-Newscaster / \n", "89-Comics / \n", "130-Gymnastics / \n", "9-Music video / \n", "28-Fashion / 121-Photography / \n", "1-Vehicle / 50-Aircraft / 82-Airplane / 92-Radio-controlled model / 149-Model aircraft / 160-Radio-controlled aircraft / 585-Lawn / 741-Mower / 805-Lawn mower / \n", "3-Concert / \n", "85-Combat / 117-Boxing / \n", "0-Games / 2-Video game / 19-PC game / 420-The Elder Scrolls V: Skyrim / 441-The Elder Scrolls / \n", "31-Disc jockey / \n", "6-Animation / 625-Factory / 2189-Warehouse / \n", "66-Bollywood / \n", "0-Games / 2-Video game / 43-Call of Duty / 431-Call of Duty: Advanced Warfare / 737-Joker / \n", "36-Piano / 885-Bell / \n", "49-School / 747-Retail / \n", "0-Games / 3387-Monster Retsuden Oreca Battle / \n", "12-Food / 2676-Citrus / \n", "18-Outdoor recreation / 22-Nature / 147-River / 283-Rock / 614-Hiking / 722-Waterfall / 825-Stream / \n", "14-Guitar / 24-String instrument / \n", "1-Vehicle / 4-Car / 11-Motorsport / 124-Tractor / 805-Lawn mower / \n", "1-Vehicle / 4-Car / 11-Motorsport / 17-Racing / 57-Race track / 3345-V8 Supercars / \n", "5-Dance / \n", "144-Amusement park / 321-Roller coaster / \n", "196-Eye / \n", "1-Vehicle / 4-Car / 25-Toy / 74-Truck / 140-LEGO / \n", "3471-Armlock / \n", "86-Plant / 136-Gardening / 279-Garden / \n", "3-Concert / 5-Dance / 13-Musical ensemble / 16-Performance art / \n", "0-Games / 8-Football / 21-Stadium / \n", "3-Concert / 7-Musician / 13-Musical ensemble / 16-Performance art / 38-Orchestra / \n", "0-Games / 163-Running / \n", "36-Piano / 53-Keyboard / 446-Sheet music / 592-Synthesia / \n", "23-Mobile phone / 29-Smartphone / 37-Gadget / 47-Personal computer / 145-Tablet computer / 178-Samsung Galaxy / 899-Samsung Galaxy S III / 1727-Samsung Galaxy Note 3 / \n", "117-Boxing / \n", "3-Concert / 68-Lighting / \n", "5-Dance / 16-Performance art / 95-Talent show / 430-Ballroom dance / \n", "9-Music video / \n", "2-Video game / 1251-The Witcher / 1401-The Witcher 3: Wild Hunt / \n", "3-Concert / 7-Musician / 13-Musical ensemble / 46-Choir / \n", "3-Concert / 5-Dance / 7-Musician / 16-Performance art / 361-Flamenco / \n", "3-Concert / 7-Musician / \n", "61-Art / 97-Drawing / \n", "102-Building / 113-House / 840-Museum / \n", "102-Building / 339-Architecture / \n", "0-Games / 2-Video game / 51-Strategy video game / \n", "0-Games / 2-Video game / 123-Naruto / 1410-The Elder Scrolls IV: Oblivion / \n", "10-Animal / 280-Wildlife / 440-Lion / 3986-Cougar / \n", "257-Counter-Strike / 1078-Installation art / \n", "322-Mixtape / \n", "86-Plant / 136-Gardening / 279-Garden / 1176-Pumpkin / \n", "3-Concert / 7-Musician / 24-String instrument / \n", "66-Bollywood / \n", "3-Concert / 30-Drummer / 68-Lighting / 2032-Pepsi / \n", "5-Dance / 66-Bollywood / 293-Slide show / \n", "12-Food / 86-Plant / 2343-Caridean Shrimp / \n", "3-Concert / 16-Performance art / \n", "1-Vehicle / 59-Winter sport / 112-Ice skating / 142-Hockey / \n", "0-Games / 59-Winter sport / 112-Ice skating / 142-Hockey / \n", "3-Concert / \n", "1-Vehicle / 4-Car / 11-Motorsport / 57-Race track / 70-Driving / 135-Drifting / \n", "5-Dance / 9-Music video / 31-Disc jockey / \n", "1-Vehicle / 4-Car / 11-Motorsport / \n", "6-Animation / 15-Cartoon / 2944-La Tale / \n", "0-Games / 59-Winter sport / 81-Athlete / 112-Ice skating / 142-Hockey / 312-Ice rink / \n", "2-Video game / 23-Mobile phone / 37-Gadget / 47-Personal computer / 145-Tablet computer / 189-iPad / 1926-Amazon Kindle / 2908-Kindle Fire / \n", "23-Mobile phone / 29-Smartphone / 37-Gadget / 72-iPhone / \n", "1-Vehicle / 4-Car / 129-Aviation / 695-Balloon / 1879-Hot air balloon / \n", "12-Food / 26-Cooking / 32-Recipe / 236-Baking / \n", "25-Toy / 425-City / 1850-Playmobil / \n", "122-Weight training / \n", "3-Concert / 7-Musician / 13-Musical ensemble / \n", "31-Disc jockey / \n", "102-Building / 170-Home improvement / 208-Wood / 2221-Deck / \n", "61-Art / \n", "10-Animal / 48-Pet / 71-Dog / 182-Puppy / 824-Terrier / \n", "13-Musical ensemble / 30-Drummer / 39-Drums / \n", "6-Animation / 15-Cartoon / 99-Christmas / 2047-Popeye / \n", "3-Concert / \n", "75-Wrestling / \n", "1-Vehicle / 4-Car / 11-Motorsport / 17-Racing / 57-Race track / 378-Drag racing / 4223-Funny Car / \n", "2-Video game / 37-Gadget / 55-Video game console / 111-PlayStation 3 / 176-PlayStation 4 / 292-Handheld game console / 317-PlayStation / 682-PlayStation Vita / 1482-PlayStation Network / \n", "12-Food / 32-Recipe / \n", "31-Disc jockey / \n", "0-Games / 10-Animal / 48-Pet / \n", "61-Art / 479-Graffiti / 1385-Street art / \n", "342-Windows Media Video / \n", "10-Animal / 42-Fishing / 144-Amusement park / 1075-Whale / 1135-Dolphin / 1780-Killer whale / \n", "0-Games / 2-Video game / 25-Toy / 140-LEGO / 2276-Lego Marvel Super Heroes / \n", "6-Animation / 25-Toy / \n", "3-Concert / 98-Festival / \n", "8-Football / \n", "6-Animation / 20-Trailer / 1899-E-book / \n", "18-Outdoor recreation / 213-Forest / 639-Fiddle / 1137-Crash Bandicoot / 2569-Crash Bandicoot / 3720-Crash Bandicoot 2: Cortex Strikes Back / \n", "1483-United States Air Force / \n", "0-Games / 76-Ball / 772-Bowling / 930-Ten-pin bowling / 1217-Bowling ball / \n", "0-Games / 1-Vehicle / 2-Video game / 4-Car / 11-Motorsport / 40-Road / 41-Sports car / 118-Grand Theft Auto V / 231-Supercar / 1979-Custom car / \n", "12-Food / 26-Cooking / 32-Recipe / 58-Cuisine / 109-Cooking show / 179-Kitchen / 215-Eating / 966-Sandwich / 1355-Vegetarian food / \n", "1-Vehicle / 4-Car / 70-Driving / 270-Sedan / 680-Subaru / 1286-Subaru / 3040-Subaru Legacy / \n", "988-Ibiza / \n", "192-Battlefield / 374-Battlefield 3 / \n", "6-Animation / 15-Cartoon / \n", "10-Animal / 22-Nature / 86-Plant / 540-Insect / \n", "0-Games / 60-Basketball / 81-Athlete / 168-Basketball moves / 255-Slam dunk / \n", "0-Games / 75-Wrestling / 977-WWE '13 / \n", "28-Fashion / 56-Hair / 106-Hairstyle / 583-Wig / 607-Lace / \n", "8-Football / 21-Stadium / 595-Champion / \n", "29-Smartphone / 504-Web page / \n", "5-Dance / \n", "1-Vehicle / 40-Road / 94-Dashcam / 663-Street / \n", "2-Video game / 20-Trailer / 739-Guild Wars / \n", "75-Wrestling / \n", "5-Dance / 16-Performance art / 77-Arena / 253-Carnival / \n", "9-Music video / 15-Cartoon / 290-Pony / 343-My Little Pony / \n", "1-Vehicle / 62-Train / 64-Transport / 94-Dashcam / 103-Rail transport / 119-Locomotive / 143-Railroad car / 152-Track / 691-Cargo / 821-Rail freight transport / \n", "12-Food / 26-Cooking / 58-Cuisine / 315-Knife / 1567-Cook / \n", "1-Vehicle / \n", "10-Animal / 48-Pet / 71-Dog / 182-Puppy / \n", "1-Vehicle / 4-Car / 11-Motorsport / 40-Road / 515-Dodge / \n", "0-Games / 1138-Tower of Saviors / \n", "6-Animation / \n", "9-Music video / \n", "163-Running / \n", "7-Musician / 13-Musical ensemble / 49-School / 79-American football / 203-Marching band / 250-Parade / \n", "39-Drums / 44-Drums / \n", "28-Fashion / 45-Cosmetics / 173-Eye shadow / 193-Eye liner / 195-Mascara / 196-Eye / 218-Eyelash / \n", "75-Wrestling / \n", "13-Musical ensemble / 754-Gramophone record / \n", "3-Concert / 5-Dance / 133-Nightclub / \n", "1-Vehicle / 3311-Hovercraft / \n", "18-Outdoor recreation / 22-Nature / 164-Beach / 233-Ocean / 423-Coast / 722-Waterfall / \n", "1-Vehicle / 50-Aircraft / 82-Airplane / 129-Aviation / 157-Jet aircraft / 388-Fighter aircraft / 758-Air force / 1483-United States Air Force / \n", "3-Concert / \n", "1791-Stargate / \n", "12-Food / 26-Cooking / 32-Recipe / 58-Cuisine / 109-Cooking show / 179-Kitchen / 236-Baking / 485-Dough / \n", "7-Musician / 13-Musical ensemble / 230-Brass instrument / 296-Saxophone / 427-Trumpet / 634-Trombone / \n", "1-Vehicle / 4-Car / 11-Motorsport / 17-Racing / \n", "5-Dance / \n", "287-Human swimming / \n", "10-Animal / 233-Ocean / 288-Underwater / 412-Monster / \n", "0-Games / 2-Video game / 420-The Elder Scrolls V: Skyrim / 441-The Elder Scrolls / \n", "1-Vehicle / 4-Car / \n", "60-Basketball / \n", "6-Animation / 15-Cartoon / \n", "7-Musician / 14-Guitar / 1754-Rocksmith / \n", "0-Games / 2-Video game / 19-PC game / 33-Weapon / 43-Call of Duty / 96-Soldier / 216-Call of Duty: Modern Warfare 3 / \n", "3756-Carpet cleaning / \n", "0-Games / 8-Football / 21-Stadium / 65-Kick / 76-Ball / \n", "1-Vehicle / 4-Car / 40-Road / 94-Dashcam / 212-Highway / \n", "1-Vehicle / 50-Aircraft / 82-Airplane / 129-Aviation / 388-Fighter aircraft / \n", "5-Dance / \n", "681-Leather / 1082-Knot / 2814-Keychain / \n", "0-Games / 54-Highlight film / 60-Basketball / 4466-Korfball / \n", "22-Nature / 137-Tree / 213-Forest / 444-Family / \n", "18-Outdoor recreation / 83-Skateboarding / 108-Skateboard / \n", "12-Food / 232-Cake / 485-Dough / \n", "10-Animal / 80-Horse / 247-Stallion / 289-Dressage / 342-Windows Media Video / \n", "45-Cosmetics / \n", "12-Food / 26-Cooking / \n", "191-Cheerleading / \n", "5-Dance / 16-Performance art / 25-Toy / 1094-Puppet / 1327-Minnie Mouse / \n", "7-Musician / 14-Guitar / \n", "1-Vehicle / 74-Truck / \n", "0-Games / 76-Ball / 3355-Bingo / \n", "0-Games / 6-Animation / 261-Tool / 1106-Character / \n", "1-Vehicle / 4-Car / 11-Motorsport / 17-Racing / 41-Sports car / \n", "341-Sketch comedy / \n", "78-Wedding / 190-Bride / \n", "238-Nail / 300-Nail art / 338-Manicure / \n", "3-Concert / \n", "5-Dance / 199-Wii / 4437-Just Dance 2 / \n", "67-Cycling / 69-Bicycle / 377-BMX bike / 520-Bicycle frame / \n", "7-Musician / 46-Choir / 49-School / \n", "3-Concert / 5-Dance / 16-Performance art / \n", "3-Concert / \n", "28-Fashion / 45-Cosmetics / 56-Hair / 106-Hairstyle / 173-Eye shadow / 193-Eye liner / 291-Rouge / 340-Afro-textured hair / \n", "1-Vehicle / 4-Car / 154-Wheel / 245-BMW / 3058-BMW 7 Series / \n", "6-Animation / 15-Cartoon / 183-Manga / 2408-Lucky Star / \n", "56-Hair / 106-Hairstyle / 954-Dreadlocks / \n", "86-Plant / 106-Hairstyle / \n", "1-Vehicle / 4-Car / 74-Truck / 158-Tire / \n", "0-Games / 2-Video game / 34-Action-adventure game / 380-Assassin's Creed / 1140-Assassin's Creed III / \n", "293-Slide show / \n", "23-Mobile phone / 29-Smartphone / 37-Gadget / 178-Samsung Galaxy / 487-Samsung Electronics / \n", "1-Vehicle / 4-Car / \n", "5-Dance / \n", "6-Animation / 15-Cartoon / \n", "7-Musician / 14-Guitar / 24-String instrument / \n", "25-Toy / 89-Comics / 435-Action figure / 2121-Marvel Legends / \n", "16-Performance art / \n", "1-Vehicle / 27-Motorcycle / \n", "1-Vehicle / 11-Motorsport / 124-Tractor / 156-Agriculture / 174-Farm / 252-Tractor pulling / 363-Soil / 459-Plough / \n", "253-Carnival / \n", "25-Toy / 140-LEGO / \n", "1-Vehicle / 4-Car / 40-Road / 94-Dashcam / 207-Exhaust system / 854-Diesel engine / 947-Muffler / \n", "5-Dance / 16-Performance art / 430-Ballroom dance / 889-Swing / \n", "1-Vehicle / 4-Car / 197-Trail / 1895-Werewolf / \n", "36-Piano / 53-Keyboard / 104-Pianist / 107-Musical keyboard / \n", "23-Mobile phone / 29-Smartphone / 37-Gadget / 72-iPhone / 189-iPad / 508-iPhone 5 / 760-iPhone 5s / \n", "710-Handball / \n", "1-Vehicle / 4-Car / \n", "102-Building / 248-Hotel / \n", "12-Food / 26-Cooking / 52-Dish / 58-Cuisine / 1116-Goat / \n", "5-Dance / 16-Performance art / 332-Mercedes-Benz / \n", "10-Animal / 588-Reptile / 2222-Crocodile / 3785-Crocodilia / \n", "3-Concert / 7-Musician / 13-Musical ensemble / 16-Performance art / \n", "5-Dance / 59-Winter sport / 112-Ice skating / 316-Figure skating / \n", "1-Vehicle / 4-Car / 114-Engine / 533-Oil / \n", "10-Animal / 48-Pet / 71-Dog / 824-Terrier / \n", "20-Trailer / \n", "5-Dance / \n", "153-Dress / 205-Newscaster / \n", "1-Vehicle / 4-Car / 1134-Taxicab / \n", "2-Video game / 23-Mobile phone / 29-Smartphone / 72-iPhone / \n", "1-Vehicle / 11-Motorsport / 17-Racing / 27-Motorcycle / 172-Motocross / 278-GoPro / 522-Helmet / 1323-Helmet camera / \n", "3-Concert / \n", "223-Weather / 419-Earth / 448-Planet / 593-Moon / 654-Sun / 1918-Comet / \n", "0-Games / 2-Video game / 19-PC game / 51-Strategy video game / 1187-Ragnarok Online / \n", "83-Skateboarding / 108-Skateboard / \n", "0-Games / 2-Video game / 6-Animation / 15-Cartoon / 816-Gundam / 1343-Super Robot Wars / \n", "33-Weapon / 1409-Missile / 4341-Anti-aircraft warfare / \n", "1-Vehicle / 4-Car / 11-Motorsport / 17-Racing / 40-Road / 41-Sports car / 70-Driving / 231-Supercar / 344-Coupé / 2335-Aston Martin / 4711-Aston Martin V8 Vantage (2005) / \n", "12-Food / 695-Balloon / 1972-Toy balloon / \n", "0-Games / 2-Video game / 19-PC game / 464-First-person Shooter / 928-Borderlands 2 / 3434-Gearbox Software / \n", "1-Vehicle / 4-Car / 11-Motorsport / 17-Racing / 27-Motorcycle / 40-Road / 57-Race track / 1867-Honda CB600F / 2494-Ninja ZX-6R / \n", "23-Mobile phone / 29-Smartphone / 101-Telephone / 1776-Samsung Galaxy Note II / \n", "28-Fashion / 305-Ski / \n", "75-Wrestling / 85-Combat / \n", "20-Trailer / 1014-Movieclips / \n", "0-Games / 411-MapleStory / \n", "31-Disc jockey / 567-Compact disc / \n", "0-Games / 34-Action-adventure game / \n", "3-Concert / 16-Performance art / \n", "5-Dance / 16-Performance art / 1579-Circle / \n", "12-Food / 26-Cooking / 32-Recipe / 52-Dish / 58-Cuisine / 109-Cooking show / 179-Kitchen / 454-Sauce / 589-Grilling / 620-Barbecue / 1018-Salmon / 1085-Barbecue grill / 1256-Smoking / \n", "1-Vehicle / 4-Car / 11-Motorsport / 17-Racing / 40-Road / 41-Sports car / 57-Race track / 169-Rallying / \n", "1-Vehicle / 4-Car / 64-Transport / 74-Truck / 1429-Dump truck / 4242-Mack Trucks / \n", "10-Animal / 48-Pet / 182-Puppy / 824-Terrier / \n", "0-Games / 2-Video game / 19-PC game / 271-Tank / 850-Left 4 Dead / 963-Left 4 Dead 2 / \n", "0-Games / 2-Video game / 34-Action-adventure game / 1588-Prince of Persia / 1653-Prince / 3894-Prince of Persia: The Sands of Time / \n", "0-Games / 2-Video game / 33-Weapon / 1749-Titanfall / \n", "83-Skateboarding / 1105-Fingerboard / \n", "164-Beach / 1208-Coconut / \n", "9-Music video / \n", "12-Food / 84-Snow / 262-Ice / \n", "44-Drums / 3465-Davul / \n", "10-Animal / 12-Food / 42-Fishing / 91-Fish / 893-Fish as food / 1018-Salmon / 2125-Fillet / \n", "5-Dance / \n", "0-Games / 8-Football / 21-Stadium / 65-Kick / 73-Ball / \n", "3-Concert / 7-Musician / 13-Musical ensemble / 14-Guitar / 30-Drummer / 39-Drums / 44-Drums / 133-Nightclub / \n", "444-Family / \n", "269-Comic book / \n", "13-Musical ensemble / \n", "3804-Music of Eritrea / \n", "0-Games / 2-Video game / \n", "3-Concert / 7-Musician / 30-Drummer / 39-Drums / 44-Drums / 146-Snare drum / \n", "0-Games / 6-Animation / 116-Pokémon / 186-Pokémon / \n", "0-Games / 2-Video game / 23-Mobile phone / 687-Taiko no Tatsujin / \n", "22-Nature / \n", "5-Dance / 3088-Mosquito / \n", "1-Vehicle / 4-Car / 11-Motorsport / 169-Rallying / \n", "0-Games / 2-Video game / 19-PC game / 2491-Spec Ops: The Line / \n", "0-Games / 2-Video game / 1503-Dungeon Fighter Online / \n", "12-Food / 26-Cooking / 32-Recipe / 52-Dish / 58-Cuisine / 239-Roasting / 771-Noodle / 1295-Stir frying / 3379-Prawn / \n", "1-Vehicle / 4-Car / 11-Motorsport / 17-Racing / 41-Sports car / \n", "215-Eating / 770-Banana / \n", "295-Chocolate / \n", "1-Vehicle / 4-Car / \n", "0-Games / 8-Football / 21-Stadium / 81-Athlete / \n", "110-Album / \n", "0-Games / 2-Video game / 33-Weapon / 43-Call of Duty / 216-Call of Duty: Modern Warfare 3 / \n", "18-Outdoor recreation / 59-Winter sport / 84-Snow / 120-Winter / 177-Skiing / 305-Ski / \n", "0-Games / 8-Football / 21-Stadium / 358-Rugby football / \n", "0-Games / 8-Football / 21-Stadium / 65-Kick / 73-Ball / 76-Ball / \n", "6-Animation / 15-Cartoon / \n", "6-Animation / \n", "23-Mobile phone / 29-Smartphone / 37-Gadget / 101-Telephone / \n", "59-Winter sport / 120-Winter / 177-Skiing / 267-Mountain / 278-GoPro / 347-Climbing / \n", "49-School / 444-Family / \n", "31-Disc jockey / 3458-DMC World DJ Championships / \n", "0-Games / 2-Video game / 34-Action-adventure game / \n", "20-Trailer / \n", "28-Fashion / 120-Winter / 785-Jeans / 2188-Coat / \n", "7-Musician / 14-Guitar / 24-String instrument / 63-Acoustic guitar / \n", "3-Concert / 7-Musician / 14-Guitar / 38-Orchestra / \n", "1-Vehicle / 12-Food / 156-Agriculture / \n", "3-Concert / 13-Musical ensemble / 253-Carnival / \n", "36-Piano / 53-Keyboard / \n", "1-Vehicle / 4-Car / 11-Motorsport / 27-Motorcycle / 114-Engine / 947-Muffler / \n", "0-Games / 259-Fire / 1138-Tower of Saviors / \n", "1-Vehicle / 12-Food / 124-Tractor / 156-Agriculture / 174-Farm / 931-Combine Harvester / 1158-Wheat / \n", "12-Food / 26-Cooking / 32-Recipe / 52-Dish / 58-Cuisine / 184-Vegetable / 301-Meal / 1707-Microwave oven / 2066-Omelette / \n", "5-Dance / 31-Disc jockey / \n", "10-Animal / 48-Pet / 166-Cat / 182-Puppy / 346-Kitten / \n", "9-Music video / 110-Album / \n", "78-Wedding / \n", "0-Games / 2-Video game / 85-Combat / 1166-Dark Souls II / \n", "3-Concert / 594-Music festival / \n", "0-Games / 2-Video game / 35-Minecraft / 394-Map / \n", "7-Musician / 14-Guitar / 24-String instrument / 63-Acoustic guitar / \n", "0-Games / 19-PC game / 51-Strategy video game / 1967-Sangokushi Taisen / \n", "1-Vehicle / 2-Video game / 4-Car / 11-Motorsport / 17-Racing / 41-Sports car / 57-Race track / 127-Xbox 360 / 135-Drifting / 169-Rallying / 1351-Rallycross / 3047-Colin McRae: Dirt 2 / \n", "30-Drummer / 39-Drums / 44-Drums / 128-Cymbal / 146-Snare drum / \n", "66-Bollywood / \n", "144-Amusement park / 369-Resort / 916-Spa / \n", "2-Video game / 6-Animation / 15-Cartoon / \n", "12-Food / 841-Butter / 966-Sandwich / 1472-Fruit preserves / 2197-Peanut butter / \n", "1449-Rubber band / \n", "5-Dance / 115-Ballet / \n", "359-Harry Potter (Literary Series) / \n", "3-Concert / \n", "8-Football / \n", "2-Video game / 37-Gadget / 55-Video game console / 292-Handheld game console / 317-PlayStation / 335-Nintendo 3DS / 362-PlayStation Portable / 682-PlayStation Vita / \n", "45-Cosmetics / 56-Hair / \n", "641-Mask / \n", "1-Vehicle / 4-Car / 4638-Nissan Micra / \n", "31-Disc jockey / \n", "3-Concert / 5-Dance / 7-Musician / 13-Musical ensemble / 16-Performance art / 46-Choir / 49-School / \n", "56-Hair / 106-Hairstyle / 1271-Beauty salon / \n", "7-Musician / 14-Guitar / 24-String instrument / 63-Acoustic guitar / \n", "3-Concert / 5-Dance / 7-Musician / 150-Violin / \n", "3-Concert / \n", "1-Vehicle / 4-Car / 114-Engine / 618-Fiat Automobiles / \n", "1033-Drill / 3224-European perch / \n", "10-Animal / 48-Pet / 351-Jumping / 404-Livestock / 987-Sheep / 1116-Goat / \n", "0-Games / 59-Winter sport / 112-Ice skating / 142-Hockey / \n", "1-Vehicle / 11-Motorsport / 27-Motorcycle / 67-Cycling / 69-Bicycle / 105-Motorcycling / 154-Wheel / 608-Wheelie / \n", "20-Trailer / \n", "357-DVD / 3449-Box set / \n", "12-Food / 26-Cooking / 32-Recipe / 226-Dessert / 232-Cake / 236-Baking / 475-Cheese / 1641-Cheesecake / \n", "1-Vehicle / 18-Outdoor recreation / 22-Nature / 197-Trail / 213-Forest / 267-Mountain / 614-Hiking / 1203-Backpacking / \n", "3-Concert / 7-Musician / 30-Drummer / 39-Drums / 2825-Rum / \n", "0-Games / 19-PC game / \n", "14-Guitar / 24-String instrument / 63-Acoustic guitar / 100-Electric guitar / 1259-Fender Telecaster / \n", "0-Games / 8-Football / \n", "117-Boxing / \n", "0-Games / 2-Video game / 20-Trailer / 3440-Serious Sam / \n", "0-Games / 2-Video game / 19-PC game / 51-Strategy video game / 85-Combat / 382-Dota 2 / \n", "0-Games / 2-Video game / 116-Pokémon / 186-Pokémon / 2139-Dungeon crawl / 2465-Pokémon Mystery Dungeon / \n", "6-Animation / 15-Cartoon / 877-Hatsune Miku: Project DIVA / \n", "0-Games / 256-Tennis / 544-Table tennis / \n", "0-Games / 35-Minecraft / 2035-Forge / \n", "5-Dance / 46-Choir / \n", "1-Vehicle / 4-Car / 11-Motorsport / 17-Racing / 40-Road / 41-Sports car / 207-Exhaust system / 212-Highway / 245-BMW / 344-Coupé / 874-BMW M3 / 1709-BMW 3 Series (E46) / \n", "0-Games / 8-Football / 21-Stadium / 54-Highlight film / 65-Kick / 73-Ball / \n", "0-Games / 35-Minecraft / \n", "0-Games / 49-School / 54-Highlight film / 60-Basketball / 168-Basketball moves / 306-High school / \n", "122-Weight training / 204-Gym / 214-Muscle / 529-Squat / 532-Barbell / 1584-Kettlebell / 2044-Bodyweight exercise / \n", "6-Animation / 25-Toy / \n", "36-Piano / 53-Keyboard / 104-Pianist / 107-Musical keyboard / 265-Electronic keyboard / \n", "1-Vehicle / 4-Car / 11-Motorsport / 41-Sports car / 1056-Porsche 911 / \n", "0-Games / 60-Basketball / 77-Arena / 81-Athlete / 168-Basketball moves / 255-Slam dunk / \n", "3-Concert / 7-Musician / 16-Performance art / \n", "25-Toy / 1302-The Lego Group / \n", "1-Vehicle / 4-Car / 11-Motorsport / \n", "1-Vehicle / 18-Outdoor recreation / 22-Nature / 67-Cycling / 69-Bicycle / 197-Trail / 213-Forest / 220-Mountain bike / 284-Mountain biking / \n", "0-Games / 51-Strategy video game / 165-League of Legends / 356-Star / \n", "6-Animation / \n", "3-Concert / 7-Musician / 13-Musical ensemble / \n", "22-Nature / 86-Plant / 136-Gardening / 184-Vegetable / 279-Garden / 642-Leaf / \n", "3-Concert / \n", "1-Vehicle / 22-Nature / 147-River / 223-Weather / 283-Rock / 490-Rain / \n", "7-Musician / 9-Music video / \n", "110-Album / \n", "9-Music video / 402-Jewellery / 587-Clay / 907-Necklace / 1300-Earrings / 1369-Necktie / \n", "22-Nature / \n", "8-Football / \n", "1-Vehicle / 2-Video game / 4-Car / 70-Driving / 194-Call of Duty: Modern Warfare 2 / \n", "9-Music video / \n", "1-Vehicle / 50-Aircraft / 266-Helicopter / 1409-Missile / \n", "1-Vehicle / 18-Outdoor recreation / 22-Nature / 67-Cycling / 69-Bicycle / 220-Mountain bike / 284-Mountain biking / \n", "0-Games / 2-Video game / 23-Mobile phone / 29-Smartphone / 37-Gadget / \n", "1-Vehicle / 62-Train / 64-Transport / 103-Rail transport / 119-Locomotive / 143-Railroad car / 152-Track / 161-Train station / 821-Rail freight transport / \n", "6-Animation / 15-Cartoon / \n", "3-Concert / 7-Musician / 16-Performance art / \n", "5-Dance / 16-Performance art / \n", "76-Ball / 2389-Exercise ball / \n", "44-Drums / 3465-Davul / \n", "3-Concert / 7-Musician / 13-Musical ensemble / 38-Orchestra / 230-Brass instrument / 427-Trumpet / \n", "2-Video game / 23-Mobile phone / 29-Smartphone / 37-Gadget / 72-iPhone / 101-Telephone / 508-iPhone 5 / \n", "1-Vehicle / 4-Car / 11-Motorsport / 531-Hatchback / 683-Renault / \n", "1-Vehicle / 25-Toy / 74-Truck / 88-Machine / 822-Crane / 2632-Crusher / \n", "22-Nature / 86-Plant / 136-Gardening / 184-Vegetable / 279-Garden / 363-Soil / 1863-Fertilizer / 2138-Compost / 2218-Manure / \n", "9-Music video / \n", "3-Concert / \n", "878-Composer / \n", "60-Basketball / 81-Athlete / \n", "10-Animal / 48-Pet / 71-Dog / 84-Snow / 120-Winter / 1215-Wild boar / \n", "1-Vehicle / 4-Car / 11-Motorsport / 17-Racing / 40-Road / 94-Dashcam / 790-Nissan GT-R / \n", "66-Bollywood / \n", "9-Music video / \n", "3-Concert / \n", "31-Disc jockey / \n", "1-Vehicle / 11-Motorsport / 88-Machine / 114-Engine / 124-Tractor / 252-Tractor pulling / \n", "1-Vehicle / 42-Fishing / \n", "0-Games / 2-Video game / 34-Action-adventure game / 243-Sonic the Hedgehog / 251-Sonic the Hedgehog / 2946-Sonic the Hedgehog / \n", "9-Music video / \n", "28-Fashion / 45-Cosmetics / \n", "36-Piano / 53-Keyboard / 646-Pipe organ / \n", "12-Food / 26-Cooking / 32-Recipe / 109-Cooking show / 236-Baking / 442-Bread / 485-Dough / 1693-Pudding / 2125-Fillet / 3003-Roll / \n", "49-School / 410-Teacher / 745-Writing / 845-Classroom / \n", "66-Bollywood / \n", "0-Games / 240-Arcade game / 514-Tekken / \n", "130-Gymnastics / 2322-Horizontal bar / \n", "12-Food / 26-Cooking / 32-Recipe / 52-Dish / 58-Cuisine / 210-Cookware and bakeware / 236-Baking / 239-Roasting / \n", "22-Nature / 1531-Sunrise / 1625-Sunset / \n", "1-Vehicle / 18-Outdoor recreation / 22-Nature / 137-Tree / 213-Forest / 825-Stream / \n", "3-Concert / 7-Musician / 13-Musical ensemble / 14-Guitar / 24-String instrument / \n", "28-Fashion / 206-Shoe / 785-Jeans / \n", "162-Medicine / 2311-Human tooth / \n", "99-Christmas / \n", "1-Vehicle / 4-Car / 299-Volkswagen Passenger Cars / \n", "151-Landing / \n", "20-Trailer / \n", "3-Concert / 68-Lighting / \n", "10-Animal / 48-Pet / 71-Dog / 166-Cat / 346-Kitten / \n", "707-Roof / 2729-Balcony / \n", "5-Dance / \n", "5-Dance / 16-Performance art / 350-Circus / \n", "0-Games / 411-MapleStory / \n", "7-Musician / 219-Accordion / \n", "36-Piano / 53-Keyboard / 61-Art / 97-Drawing / 148-Painting / 528-Pencil / 628-Portrait / \n", "3-Concert / \n", "1-Vehicle / 62-Train / 64-Transport / 102-Building / 144-Amusement park / 161-Train station / 224-Rapid transit / 359-Harry Potter (Literary Series) / \n", "10-Animal / 25-Toy / 524-Littlest Pet Shop / \n", "20-Trailer / 325-Vampire / \n", "0-Games / 192-Battlefield / 1370-Battlefield 2 / \n", "1-Vehicle / 87-Boat / 147-River / 481-Kayak / \n", "12-Food / \n", "1-Vehicle / 64-Transport / 74-Truck / 1046-Remote control / \n", "28-Fashion / 45-Cosmetics / \n", "5-Dance / 16-Performance art / 783-Zee TV / \n", "23-Mobile phone / 29-Smartphone / 345-Loudspeaker / 465-Microphone / \n", "0-Games / 2-Video game / 1482-PlayStation Network / 2006-BlazBlue / 3627-BlazBlue: Continuum Shift / \n", "0-Games / 35-Minecraft / \n", "0-Games / 8-Football / 90-Sports game / 180-FIFA 15 / 623-Coach / \n", "10-Animal / 48-Pet / 132-Bird / \n", "205-Newscaster / 780-Fat / \n", "3-Concert / 7-Musician / 13-Musical ensemble / 38-Orchestra / 1338-Devil / \n", "512-Printing / 525-Printer / 1063-Ink / \n", "3-Concert / 133-Nightclub / \n", "3-Concert / 7-Musician / 13-Musical ensemble / 16-Performance art / 30-Drummer / \n", "56-Hair / 106-Hairstyle / 405-Braid / 1064-French braid / 1292-Ponytail / 2456-Tail / \n", "0-Games / 2-Video game / 51-Strategy video game / 55-Video game console / 199-Wii / \n", "1-Vehicle / 4-Car / 3138-Pressure washer / \n", "15-Cartoon / 185-Dragon Ball / 244-Goku / \n", "5-Dance / 130-Gymnastics / 3830-Double Dutch / \n", "36-Piano / 53-Keyboard / 104-Pianist / 107-Musical keyboard / 265-Electronic keyboard / \n", "25-Toy / 140-LEGO / \n", "5-Dance / \n", "75-Wrestling / \n", "5-Dance / 16-Performance art / 95-Talent show / \n", "3-Concert / \n", "10-Animal / 18-Outdoor recreation / 80-Horse / 247-Stallion / 289-Dressage / 404-Livestock / 1434-Saddle / \n", "0-Games / 2-Video game / 51-Strategy video game / 373-Clash of Clans / \n", "223-Weather / \n", "56-Hair / 106-Hairstyle / \n", "5-Dance / 44-Drums / 98-Festival / \n", "31-Disc jockey / \n", "23-Mobile phone / 29-Smartphone / 72-iPhone / \n", "113-House / 329-Swimming pool / 369-Resort / \n", "1640-Library / \n", "402-Jewellery / 1010-Ring / 1300-Earrings / \n", "12-Food / 26-Cooking / 32-Recipe / 52-Dish / 58-Cuisine / 1355-Vegetarian food / \n", "162-Medicine / \n", "1-Vehicle / 4-Car / 179-Kitchen / 519-Recreational vehicle / 742-Countertop / \n", "7-Musician / 13-Musical ensemble / 30-Drummer / 39-Drums / 44-Drums / \n", "15-Cartoon / \n", "14-Guitar / 24-String instrument / 63-Acoustic guitar / \n", "59-Winter sport / 112-Ice skating / 316-Figure skating / \n", "1-Vehicle / 4-Car / 11-Motorsport / 17-Racing / 41-Sports car / \n", "0-Games / 59-Winter sport / 81-Athlete / 112-Ice skating / 142-Hockey / 312-Ice rink / \n", "3-Concert / 5-Dance / 16-Performance art / \n", "18-Outdoor recreation / 144-Amusement park / 321-Roller coaster / 334-Amusement ride / \n", "0-Games / 2-Video game / 320-Super Smash Bros. / \n", "6-Animation / 15-Cartoon / \n", "0-Games / 2-Video game / 60-Basketball / 90-Sports game / \n", "0-Games / 1638-Soulcalibur / 2633-Soulcalibur V / \n", "0-Games / 85-Combat / 4690-Shadow Fighter / \n", "0-Games / 2-Video game / 505-Mega Man / 1664-Zero / 2474-Mega Man Battle Network / \n", "31-Disc jockey / 227-Water / \n", "1-Vehicle / 4-Car / \n", "1-Vehicle / 11-Motorsport / 74-Truck / 124-Tractor / 135-Drifting / 249-Heavy equipment / \n", "0-Games / 1-Vehicle / 2-Video game / 4-Car / 11-Motorsport / 17-Racing / 41-Sports car / 57-Race track / 70-Driving / 901-Mitsubishi / 3047-Colin McRae: Dirt 2 / 3136-Mitsubishi Eclipse / 3500-Colin McRae: Dirt / \n", "0-Games / 2-Video game / 55-Video game console / 586-Nintendo Entertainment System / 910-Dragon Quest / \n", "56-Hair / 106-Hairstyle / 1886-Bob cut / \n", "9-Music video / \n", "3-Concert / 7-Musician / 13-Musical ensemble / 30-Drummer / \n", "372-Playing card / 746-Card manipulation / \n", "1-Vehicle / 4-Car / 207-Exhaust system / 299-Volkswagen Passenger Cars / 526-Volkswagen / 668-Volkswagen Golf / \n", "59-Winter sport / 84-Snow / 177-Skiing / 789-Alpine skiing / \n", "0-Games / 2-Video game / 23-Mobile phone / 29-Smartphone / 37-Gadget / 101-Telephone / 201-Grand Theft Auto: San Andreas / 3307-HTC One X / \n", "0-Games / 8-Football / 21-Stadium / 65-Kick / \n", "12-Food / 1823-Espresso machine / \n", "36-Piano / 53-Keyboard / 104-Pianist / 107-Musical keyboard / \n", "0-Games / 6-Animation / 269-Comic book / 460-Yu-Gi-Oh! Trading Card Game / \n", "9-Music video / 14-Guitar / \n", "46-Choir / \n", "0-Games / 59-Winter sport / 81-Athlete / 112-Ice skating / 142-Hockey / 262-Ice / 312-Ice rink / 1973-Hockey puck / \n", "6-Animation / 158-Tire / \n", "46-Choir / 446-Sheet music / \n", "31-Disc jockey / \n", "10-Animal / 18-Outdoor recreation / 22-Nature / 42-Fishing / 147-River / 225-Lake / 268-Recreational fishing / 825-Stream / 895-Trout / 1873-Crossbow / 3145-Spinnerbait / \n", "156-Agriculture / 1579-Circle / 3483-Crop circle / \n", "98-Festival / 209-University / \n", "1-Vehicle / 4-Car / 11-Motorsport / 17-Racing / 41-Sports car / 57-Race track / 169-Rallying / \n", "60-Basketball / 76-Ball / 168-Basketball moves / 1101-Point guard / 3291-Basketball Coach / \n", "6-Animation / 15-Cartoon / \n", "10-Animal / 17-Racing / 80-Horse / 384-Horse racing / \n", "1-Vehicle / 12-Food / 88-Machine / \n", "35-Minecraft / \n", "49-School / \n", "1944-Vatican City / \n", "3-Concert / 5-Dance / 16-Performance art / 153-Dress / \n", "14-Guitar / 63-Acoustic guitar / 100-Electric guitar / \n", "23-Mobile phone / 29-Smartphone / 37-Gadget / 121-Photography / 155-Camera / 629-Camera lens / 4431-Camera phone / \n", "43-Call of Duty / \n", "9-Music video / 920-Chair / \n", "10-Animal / 91-Fish / 2090-Siamese fighting fish / \n", "59-Winter sport / 112-Ice skating / 142-Hockey / \n", "9-Music video / \n", "657-Autumn / 1459-Magazine / 1545-Floristry / \n", "0-Games / 256-Tennis / 462-Table / 544-Table tennis / 1337-Serve / 1615-Pong / \n", "147-River / \n", "3-Concert / \n", "23-Mobile phone / 29-Smartphone / 37-Gadget / 72-iPhone / 554-iPhone 4 / \n", "10-Animal / 48-Pet / 71-Dog / 706-Deer / \n", "1-Vehicle / 11-Motorsport / 27-Motorcycle / 105-Motorcycling / \n", "3-Concert / 16-Performance art / \n", "5-Dance / 46-Choir / \n", "5-Dance / 125-Television / \n", "310-The Sims / 543-The Sims 2 / \n", "3-Concert / \n", "61-Art / 148-Painting / 538-Illustration / \n", "0-Games / 21-Stadium / 60-Basketball / 77-Arena / 168-Basketball moves / 255-Slam dunk / \n", "9-Music video / \n", "1-Vehicle / 4-Car / 205-Newscaster / \n", "0-Games / 10-Animal / 48-Pet / \n", "3-Concert / \n", "3-Concert / 133-Nightclub / \n", "0-Games / 1-Vehicle / 2-Video game / 4-Car / 17-Racing / 34-Action-adventure game / 111-PlayStation 3 / 118-Grand Theft Auto V / \n", "766-President of the United States / \n", "1-Vehicle / 4-Car / 27-Motorcycle / 40-Road / 94-Dashcam / 212-Highway / 245-BMW / 1622-BMW M5 / \n", "1-Vehicle / 4-Car / 11-Motorsport / 17-Racing / 40-Road / 57-Race track / 70-Driving / 94-Dashcam / 212-Highway / 2214-Multi-valve / 4352-Citroën Saxo / \n", "3-Concert / 2765-Mail / \n", "1-Vehicle / 4-Car / 70-Driving / 618-Fiat Automobiles / 864-Speedometer / \n", "3-Concert / 7-Musician / 16-Performance art / 30-Drummer / 39-Drums / 44-Drums / \n", "0-Games / 460-Yu-Gi-Oh! Trading Card Game / \n", "93-Comedy / 355-Comedian / \n", "6-Animation / 67-Cycling / 69-Bicycle / 126-The Walt Disney Company / \n", "28-Fashion / 880-Trousers / 986-Suit / 2188-Coat / 3781-Hoodie / 4505-GQ / \n", "3-Concert / 7-Musician / 13-Musical ensemble / 38-Orchestra / \n", "10-Animal / 48-Pet / 166-Cat / 346-Kitten / \n", "0-Games / 59-Winter sport / 81-Athlete / 112-Ice skating / 142-Hockey / \n", "6-Animation / \n", "6-Animation / 3689-Woody Woodpecker / \n", "5-Dance / 95-Talent show / \n", "23-Mobile phone / 29-Smartphone / 37-Gadget / 72-iPhone / 101-Telephone / 178-Samsung Galaxy / 512-Printing / \n", "3-Concert / 7-Musician / \n", "0-Games / 2-Video game / 33-Weapon / 176-PlayStation 4 / 948-Far Cry / 1683-Far Cry 4 / \n", "18-Outdoor recreation / 315-Knife / 1970-Axe / \n", "159-World of Warcraft / 171-Warcraft / \n", "0-Games / 1-Vehicle / 2-Video game / 4-Car / 11-Motorsport / 34-Action-adventure game / 118-Grand Theft Auto V / \n", "619-Manufacturing / 1363-Candle / \n", "5-Dance / 115-Ballet / \n", "1-Vehicle / 4-Car / 167-Four-wheel drive / 198-Off-road vehicle / 213-Forest / \n", "497-Macintosh / 512-Printing / 2711-Label / \n", "20-Trailer / \n", "67-Cycling / 69-Bicycle / 204-Gym / 284-Mountain biking / \n", "23-Mobile phone / 29-Smartphone / 37-Gadget / 101-Telephone / \n", "1-Vehicle / 4-Car / 245-BMW / 983-Lexus / 1622-BMW M5 / \n", "1-Vehicle / 4-Car / 27-Motorcycle / 40-Road / 67-Cycling / 69-Bicycle / 105-Motorcycling / \n", "0-Games / 8-Football / \n", "425-City / \n", "12-Food / 26-Cooking / 184-Vegetable / \n", "1-Vehicle / 4-Car / 10-Animal / 132-Bird / 598-Parrot / \n", "313-Book / \n", "0-Games / 2-Video game / 1170-Super Mario World / 1383-Yoshi / 2589-Yoshi's Island / \n", "3-Concert / 7-Musician / 13-Musical ensemble / 39-Drums / \n", "0-Games / 159-World of Warcraft / 171-Warcraft / \n", "1382-Half-Life 2 / 1412-Half-Life / \n", "28-Fashion / 206-Shoe / 757-Boot / \n", "191-Cheerleading / 293-Slide show / \n", "9-Music video / 78-Wedding / 190-Bride / \n", "59-Winter sport / 84-Snow / 120-Winter / 177-Skiing / 305-Ski / \n", "3-Concert / \n", "5-Dance / 115-Ballet / 430-Ballroom dance / 678-Latin dance / \n", "51-Strategy video game / 774-Galaxy / 1718-Spore / \n", "98-Festival / 253-Carnival / \n", "0-Games / 1-Vehicle / 4-Car / 17-Racing / \n", "10-Animal / \n", "28-Fashion / \n", "1-Vehicle / 11-Motorsport / 18-Outdoor recreation / 27-Motorcycle / 42-Fishing / 1025-Personal water craft / 1336-Jet Ski / \n", "6-Animation / 15-Cartoon / \n", "20-Trailer / \n", "0-Games / 2-Video game / 43-Call of Duty / 194-Call of Duty: Modern Warfare 2 / \n", "1-Vehicle / 11-Motorsport / \n", "47-Personal computer / 131-Computer / 309-Computer hardware / 602-Central processing unit / \n", "0-Games / 2-Video game / 20-Trailer / \n", "3-Concert / 7-Musician / 13-Musical ensemble / 38-Orchestra / 150-Violin / \n", "61-Art / 148-Painting / \n", "6-Animation / 15-Cartoon / 183-Manga / 708-Mermaid / \n", "121-Photography / 155-Camera / 629-Camera lens / \n", "6-Animation / 25-Toy / 319-Robot / 477-Transformers / 990-Optimus Prime / \n", "88-Machine / 512-Printing / 525-Printer / 536-Glass / 2634-Inkjet printing / \n", "164-Beach / 191-Cheerleading / \n", "18-Outdoor recreation / 144-Amusement park / 321-Roller coaster / 334-Amusement ride / \n", "8-Football / \n", "0-Games / 35-Minecraft / 1027-Terraria / \n", "3391-Alesis / \n", "1-Vehicle / 4-Car / 40-Road / 64-Transport / 70-Driving / \n", "12-Food / \n", "170-Home improvement / 246-Furniture / 276-Room / 2385-Closet / \n", "0-Games / 2-Video game / 20-Trailer / 34-Action-adventure game / 1144-Castlevania / 3768-Castlevania: Lords of Shadow / \n", "8-Football / \n", "5-Dance / 49-School / 95-Talent show / 191-Cheerleading / \n", "0-Games / 1754-Rocksmith / 2050-Rocksmith 2014 / \n", "0-Games / 19-PC game / 35-Minecraft / 2956-Furnace / \n", "61-Art / 479-Graffiti / \n", "30-Drummer / 39-Drums / 44-Drums / 128-Cymbal / \n", "24-String instrument / 2980-Sitar / \n", "8-Football / 49-School / 79-American football / 306-High school / \n", "5-Dance / \n", "6-Animation / 51-Strategy video game / 269-Comic book / 1915-Duel Masters Trading Card Game / \n", "9-Music video / \n", "0-Games / 2-Video game / 33-Weapon / 43-Call of Duty / 422-Call of Duty: Ghosts / 431-Call of Duty: Advanced Warfare / \n", "1-Vehicle / 88-Machine / \n", "31-Disc jockey / \n", "214-Muscle / \n", "5-Dance / 3240-Matryoshka doll / \n", "2-Video game / 6-Animation / \n", "6-Animation / \n", "1-Vehicle / 11-Motorsport / \n", "0-Games / 1-Vehicle / 2-Video game / 4-Car / 34-Action-adventure game / 47-Personal computer / 118-Grand Theft Auto V / \n", "0-Games / 2-Video game / 33-Weapon / 420-The Elder Scrolls V: Skyrim / 441-The Elder Scrolls / 1130-Armour / \n", "341-Sketch comedy / \n", "15-Cartoon / 123-Naruto / 285-Sasuke Uchiha / \n", "141-Microsoft Windows / 504-Web page / \n", "0-Games / 8-Football / 21-Stadium / 65-Kick / 73-Ball / 90-Sports game / \n", "0-Games / 199-Wii / 456-Go-kart / 518-Mario Kart / \n", "99-Christmas / \n", "467-Chipmunk / 846-Squirrel / \n", "1-Vehicle / 4-Car / 270-Sedan / 517-Audi / 1587-Audi Quattro / 2306-Audi A6 / \n", "31-Disc jockey / \n", "23-Mobile phone / 29-Smartphone / 72-iPhone / \n", "12-Food / 32-Recipe / 109-Cooking show / 226-Dessert / 1223-Pastry / \n", "0-Games / 2-Video game / 19-PC game / 211-RuneScape / \n", "130-Gymnastics / 1030-Trampoline / 1335-Trampolining / 2161-Playground / \n", "1-Vehicle / 4-Car / 25-Toy / 88-Machine / 319-Robot / \n", "18-Outdoor recreation / 51-Strategy video game / 144-Amusement park / 321-Roller coaster / 334-Amusement ride / 1293-RollerCoaster Tycoon 3 / \n", "36-Piano / 53-Keyboard / 107-Musical keyboard / 219-Accordion / \n", "1-Vehicle / 4-Car / 11-Motorsport / 17-Racing / 40-Road / 57-Race track / 169-Rallying / \n", "2393-Didgeridoo / \n", "8-Football / 54-Highlight film / \n", "380-Assassin's Creed / \n", "6-Animation / 849-Tales / 3519-Tales of the Abyss / \n", "162-Medicine / \n", "6-Animation / 49-School / 102-Building / 113-House / 1246-Campus / \n", "2-Video game / 35-Minecraft / 380-Assassin's Creed / 1048-Assassin's Creed / 1140-Assassin's Creed III / \n", "0-Games / 34-Action-adventure game / 201-Grand Theft Auto: San Andreas / \n", "30-Drummer / 39-Drums / 44-Drums / 128-Cymbal / 146-Snare drum / \n", "6-Animation / 15-Cartoon / \n", "0-Games / 903-Mass Effect / 1655-Mass Effect 2 / \n", "0-Games / 47-Personal computer / \n", "3-Concert / 7-Musician / 13-Musical ensemble / 16-Performance art / \n", "18-Outdoor recreation / 227-Water / 287-Human swimming / 329-Swimming pool / \n", "67-Cycling / 69-Bicycle / 570-Road bicycle racing / 645-Road bicycle / \n", "0-Games / 2-Video game / 34-Action-adventure game / 237-Final Fantasy / 1571-Final Fantasy XIV / \n", "0-Games / 2-Video game / 19-PC game / 51-Strategy video game / 165-League of Legends / 1494-Jungle / \n", "47-Personal computer / 53-Keyboard / 145-Tablet computer / 513-Computer keyboard / 1565-Logitech / \n", "1-Vehicle / 50-Aircraft / 82-Airplane / 129-Aviation / 139-Airport / 151-Landing / 157-Jet aircraft / 175-Airliner / 187-Takeoff / 188-Airline / 254-Runway / 472-Boeing 737 / 719-Boeing 747 / \n", "3-Concert / 7-Musician / 13-Musical ensemble / 38-Orchestra / \n", "0-Games / 2-Video game / 33-Weapon / 43-Call of Duty / 989-Warface / \n", "3-Concert / 7-Musician / \n", "1-Vehicle / \n", "12-Food / 26-Cooking / 32-Recipe / 52-Dish / 232-Cake / 500-Rice / 2591-Filipino food / \n", "0-Games / 1-Vehicle / 2-Video game / 4-Car / 19-PC game / 34-Action-adventure game / 111-PlayStation 3 / 413-PlayStation 2 / 921-Grand Theft Auto: Vice City / 2893-Grand Theft Auto: Vice City Stories / \n", "8-Football / 21-Stadium / \n", "1-Vehicle / 4-Car / 1742-Lowrider / \n", "1-Vehicle / 67-Cycling / 69-Bicycle / 220-Mountain bike / 284-Mountain biking / \n", "1-Vehicle / 67-Cycling / 69-Bicycle / 570-Road bicycle racing / 645-Road bicycle / \n", "1208-Coconut / \n", "10-Animal / 12-Food / 71-Dog / 232-Cake / 323-Toddler / \n", "3-Concert / \n", "3-Concert / 7-Musician / 13-Musical ensemble / 38-Orchestra / \n", "474-Destiny / \n", "6-Animation / 15-Cartoon / 185-Dragon Ball / 244-Goku / \n", "3-Concert / 7-Musician / 30-Drummer / \n", "8-Football / \n", "9-Music video / 31-Disc jockey / \n", "1-Vehicle / 2-Video game / 4-Car / 11-Motorsport / 17-Racing / 55-Video game console / 57-Race track / 111-PlayStation 3 / \n", "83-Skateboarding / 108-Skateboard / \n", "6-Animation / 137-Tree / 1738-Elmo / \n", "0-Games / 2-Video game / 55-Video game console / 413-PlayStation 2 / 2486-Scooby-Doo / \n", "3-Concert / 5-Dance / 13-Musical ensemble / 16-Performance art / \n", "3-Concert / 7-Musician / 13-Musical ensemble / \n", "6-Animation / 15-Cartoon / 1594-Nerd / \n", "0-Games / 2-Video game / 19-PC game / 33-Weapon / 96-Soldier / 638-Gears of War / 773-Gears of War / 1073-Chainsaw / 1425-Gears of War 2 / \n", "313-Book / \n", "0-Games / 2-Video game / 34-Action-adventure game / 118-Grand Theft Auto V / \n", "22-Nature / 1151-Temple / 3039-Wat / \n", "8-Football / \n", "0-Games / 1394-Alliance of Valiant Arms / \n", "3-Concert / 7-Musician / 13-Musical ensemble / 306-High school / \n", "9-Music video / \n", "1-Vehicle / 4-Car / 11-Motorsport / 17-Racing / 40-Road / 41-Sports car / 57-Race track / 378-Drag racing / 482-Asphalt / \n", "3544-Rubber stamp / \n", "3-Concert / 68-Lighting / 1712-Sweater / \n", "5-Dance / 44-Drums / 115-Ballet / \n", "1-Vehicle / 4-Car / 64-Transport / 286-Bus / \n", "45-Cosmetics / 49-School / 56-Hair / \n", "28-Fashion / 206-Shoe / 368-Sneakers / 2014-Vans / 3831-Supreme / \n", "695-Balloon / \n", "1-Vehicle / 4-Car / 114-Engine / \n", "30-Drummer / 39-Drums / 44-Drums / 128-Cymbal / 146-Snare drum / \n", "0-Games / 8-Football / 21-Stadium / 54-Highlight film / 65-Kick / 73-Ball / \n", "6-Animation / 15-Cartoon / 1492-Persona / \n", "10-Animal / 15-Cartoon / 61-Art / 71-Dog / 97-Drawing / 408-Sketch / 528-Pencil / \n", "0-Games / 2-Video game / 35-Minecraft / 394-Map / \n", "31-Disc jockey / \n", "10-Animal / 48-Pet / 71-Dog / \n", "3-Concert / 7-Musician / 16-Performance art / \n", "1-Vehicle / 4-Car / 449-Honda / 519-Recreational vehicle / \n", "1-Vehicle / 4-Car / 11-Motorsport / 17-Racing / 57-Race track / \n", "3-Concert / 16-Performance art / \n", "5-Dance / 16-Performance art / 95-Talent show / \n", "6-Animation / 15-Cartoon / 116-Pokémon / 1104-Ash Ketchum / \n", "125-Television / \n", "0-Games / 8-Football / 21-Stadium / \n", "5-Dance / 115-Ballet / \n", "640-Bottle / 1175-Soap / 4325-Detergent / \n", "0-Games / 2-Video game / 34-Action-adventure game / \n", "445-Guitar Hero / 965-Guitar Hero III: Legends of Rock / \n", "10-Animal / 20-Trailer / 550-Dinosaur / 1126-Tyrannosaurus / \n", "0-Games / 8-Football / 54-Highlight film / 73-Ball / \n", "3-Concert / 5-Dance / 16-Performance art / \n", "5-Dance / 16-Performance art / \n", "206-Shoe / \n", "18-Outdoor recreation / 42-Fishing / 67-Cycling / 69-Bicycle / \n", "778-Gliding / 1602-Hang gliding / \n", "12-Food / 26-Cooking / 32-Recipe / 52-Dish / 58-Cuisine / 109-Cooking show / 179-Kitchen / 295-Chocolate / \n", "23-Mobile phone / 29-Smartphone / 37-Gadget / 101-Telephone / 1820-Motorola Droid / \n", "31-Disc jockey / \n", "28-Fashion / \n", "13-Musical ensemble / \n", "1-Vehicle / 4-Car / \n", "8-Football / 21-Stadium / \n", "3-Concert / 7-Musician / 13-Musical ensemble / \n", "8-Football / \n", "22-Nature / \n", "122-Weight training / 214-Muscle / \n", "0-Games / 8-Football / 21-Stadium / 79-American football / \n", "3-Concert / \n", "28-Fashion / 45-Cosmetics / 56-Hair / 106-Hairstyle / 340-Afro-textured hair / 857-Afro / 909-Updo / 1320-Bangs / \n", "24-String instrument / 502-Sky / 842-Harp / \n", "14-Guitar / 24-String instrument / 63-Acoustic guitar / 3637-Steel-string acoustic guitar / \n", "14-Guitar / \n", "1-Vehicle / 4-Car / 11-Motorsport / 17-Racing / 40-Road / 41-Sports car / 482-Asphalt / \n", "12-Food / 812-Plastic / 1870-Chewing gum / \n", "6-Animation / 15-Cartoon / 290-Pony / 343-My Little Pony / \n", "1-Vehicle / 74-Truck / \n", "1-Vehicle / 11-Motorsport / 17-Racing / 27-Motorcycle / 57-Race track / 105-Motorcycling / 172-Motocross / \n", "3-Concert / 28-Fashion / 495-Runway / \n", "75-Wrestling / 85-Combat / \n", "13-Musical ensemble / 356-Star / \n", "5-Dance / \n", "6-Animation / 25-Toy / 749-Figurine / \n", "215-Eating / 534-Milk / 2080-Cup / \n", "3-Concert / \n", "0-Games / 8-Football / \n", "0-Games / 6-Animation / 995-Attack on Titan / \n", "8-Football / \n", "30-Drummer / 39-Drums / 44-Drums / 128-Cymbal / \n", "0-Games / 2-Video game / 229-Halo / 1161-Halo 2 / \n", "5-Dance / 46-Choir / 318-Church / \n", "6-Animation / 25-Toy / 323-Toddler / 3785-Crocodilia / \n", "2077-Chemical reaction / \n", "1-Vehicle / 11-Motorsport / 17-Racing / \n", "28-Fashion / 469-Bag / 676-Handbag / 2188-Coat / \n", "337-Advertising / \n", "9-Music video / \n", "223-Weather / \n", "1-Vehicle / 11-Motorsport / 88-Machine / 124-Tractor / 156-Agriculture / 174-Farm / 252-Tractor pulling / \n", "7-Musician / 13-Musical ensemble / 230-Brass instrument / 925-Brain / \n", "1-Vehicle / 4-Car / 41-Sports car / 114-Engine / 245-BMW / \n", "59-Winter sport / 112-Ice skating / 262-Ice / 316-Figure skating / 833-Ice skate / \n", "0-Games / 2-Video game / 33-Weapon / 635-Achievement / 924-Trophy / \n", "12-Food / 26-Cooking / 32-Recipe / 52-Dish / 58-Cuisine / 184-Vegetable / 572-Indian cuisine / 597-Soup / 727-Curry / 1509-Stew / 2000-Mango / \n", "1-Vehicle / 4-Car / 41-Sports car / \n", "20-Trailer / 886-James Bond / \n", "54-Highlight film / 60-Basketball / \n", "0-Games / 2-Video game / 4499-Fable II / \n", "25-Toy / \n", "3-Concert / 7-Musician / 13-Musical ensemble / 30-Drummer / 39-Drums / 44-Drums / \n", "31-Disc jockey / \n", "0-Games / 2-Video game / 8-Football / 90-Sports game / 180-FIFA 15 / \n", "3-Concert / 5-Dance / 16-Performance art / 490-Rain / \n", "1-Vehicle / 27-Motorcycle / \n", "31-Disc jockey / \n", "6-Animation / 1454-Stairs / \n", "0-Games / 8-Football / \n", "2-Video game / 6-Animation / \n", "3-Concert / 7-Musician / \n", "164-Beach / 233-Ocean / \n", "110-Album / 567-Compact disc / \n", "0-Games / 8-Football / 21-Stadium / 65-Kick / \n", "219-Accordion / \n", "2-Video game / 2277-The Lord of the Rings Online / \n", "24-String instrument / 150-Violin / 401-Cello / \n", "0-Games / 2-Video game / 241-Fighting game / 2301-Dead or Alive / 4630-Dead or Alive 5 Ultimate / \n", "1-Vehicle / 4-Car / \n", "10-Animal / 48-Pet / 132-Bird / \n", "31-Disc jockey / \n", "1-Vehicle / 62-Train / 64-Transport / 103-Rail transport / 119-Locomotive / 143-Railroad car / 152-Track / 224-Rapid transit / 794-Bridge / \n", "6-Animation / 20-Trailer / \n", "23-Mobile phone / 29-Smartphone / 37-Gadget / 72-iPhone / 101-Telephone / 675-iPhone 4S / \n", "75-Wrestling / \n", "6-Animation / 15-Cartoon / 116-Pokémon / 186-Pokémon / 1104-Ash Ketchum / \n", "75-Wrestling / \n", "1-Vehicle / 4-Car / 11-Motorsport / 17-Racing / 40-Road / 212-Highway / 299-Volkswagen Passenger Cars / 482-Asphalt / 668-Volkswagen Golf / 2286-Grand Tourer Injection / 3839-Volkswagen Golf Mk5 / \n", "6-Animation / 538-Illustration / \n", "1-Vehicle / 50-Aircraft / 82-Airplane / 102-Building / 129-Aviation / 139-Airport / 151-Landing / 175-Airliner / 187-Takeoff / 188-Airline / 254-Runway / 856-Airbus A320 family / 923-Boeing 777 / \n", "53-Keyboard / 230-Brass instrument / 3391-Alesis / \n", "257-Counter-Strike / \n", "1-Vehicle / 4-Car / 11-Motorsport / 17-Racing / 27-Motorcycle / 245-BMW / 491-Enduro / 4403-BMW F series parallel-twin / \n", "1-Vehicle / 4-Car / 41-Sports car / \n", "3-Concert / 13-Musical ensemble / 30-Drummer / 44-Drums / 253-Carnival / \n", "0-Games / 35-Minecraft / \n", "101-Telephone / 3233-Screencast / \n", "45-Cosmetics / \n", "3-Concert / \n", "13-Musical ensemble / 44-Drums / \n", "5-Dance / \n", "31-Disc jockey / 861-Mixing console / 2193-CDJ / \n", "8-Football / \n", "7-Musician / 13-Musical ensemble / 98-Festival / 203-Marching band / 250-Parade / \n", "3-Concert / \n", "1-Vehicle / 88-Machine / 114-Engine / \n", "1-Vehicle / 164-Beach / 423-Coast / 1013-Walking / \n", "141-Microsoft Windows / \n", "1-Vehicle / 27-Motorcycle / 663-Street / \n", "5-Dance / 16-Performance art / 98-Festival / 678-Latin dance / \n", "0-Games / 54-Highlight film / 60-Basketball / \n", "45-Cosmetics / 49-School / 193-Eye liner / 195-Mascara / 196-Eye / \n", "7-Musician / 14-Guitar / 38-Orchestra / \n", "59-Winter sport / 281-Snowboarding / \n", "0-Games / 2054-Diablo II / \n", "2-Video game / \n", "8-Football / \n", "1-Vehicle / \n", "0-Games / 8-Football / 21-Stadium / \n", "59-Winter sport / 177-Skiing / 1764-Ski jumping / \n", "10-Animal / 48-Pet / 166-Cat / 346-Kitten / 2211-Fur / 2924-Persian cat / \n", "14-Guitar / \n", "387-Spider-Man / \n", "0-Games / 19-PC game / 35-Minecraft / 272-iPod / \n", "1-Vehicle / 50-Aircraft / 82-Airplane / 92-Radio-controlled model / 149-Model aircraft / 160-Radio-controlled aircraft / \n", "355-Comedian / \n", "0-Games / 2-Video game / 116-Pokémon / 186-Pokémon / 1387-Pokémon Battle Revolution / \n", "12-Food / 26-Cooking / 32-Recipe / 52-Dish / 58-Cuisine / 109-Cooking show / 184-Vegetable / 239-Roasting / 500-Rice / 572-Indian cuisine / 727-Curry / 1355-Vegetarian food / 1509-Stew / 2363-Fried rice / 4236-Biryani / \n", "5-Dance / 16-Performance art / \n", "9-Music video / \n", "115-Ballet / 1253-Ballet Dancer / \n", "9-Music video / \n", "0-Games / 366-Dragon / \n", "0-Games / 2-Video game / 216-Call of Duty: Modern Warfare 3 / \n", "0-Games / 8-Football / 21-Stadium / 65-Kick / 73-Ball / 76-Ball / 81-Athlete / \n", "0-Games / 159-World of Warcraft / 171-Warcraft / \n", "7-Musician / 14-Guitar / 24-String instrument / 100-Electric guitar / \n", "60-Basketball / 255-Slam dunk / \n", "6-Animation / 15-Cartoon / \n", "21-Stadium / 85-Combat / 117-Boxing / \n", "78-Wedding / 190-Bride / 421-Gown / \n", "6-Animation / 238-Nail / 300-Nail art / 343-My Little Pony / 823-Artificial nails / \n", "3-Concert / 7-Musician / 13-Musical ensemble / 14-Guitar / \n", "3-Concert / \n", "3-Concert / 7-Musician / \n", "3-Concert / 7-Musician / \n", "49-School / \n", "10-Animal / 468-Biology / \n", "15-Cartoon / 200-Doll / 412-Monster / 503-Monster High / \n", "9-Music video / \n", "1-Vehicle / 4-Car / 114-Engine / 234-Ford / 473-Ford Mustang / \n", "7-Musician / 14-Guitar / \n", "14-Guitar / 219-Accordion / 296-Saxophone / \n", "31-Disc jockey / \n", "34-Action-adventure game / 506-Garry's Mod / \n", "12-Food / 26-Cooking / 32-Recipe / 52-Dish / 58-Cuisine / 109-Cooking show / 179-Kitchen / 184-Vegetable / 210-Cookware and bakeware / 239-Roasting / 274-Meat / 2125-Fillet / \n", "12-Food / 1060-Raw food / \n", "0-Games / 1167-Tanki Online / \n", "1-Vehicle / 4-Car / 11-Motorsport / 70-Driving / 207-Exhaust system / 2787-Saab Automobile AB / 4008-Saab / \n", "1-Vehicle / 22-Nature / 50-Aircraft / 139-Airport / 151-Landing / 425-City / \n", "1-Vehicle / 62-Train / 64-Transport / 103-Rail transport / 119-Locomotive / 143-Railroad car / \n", "93-Comedy / \n", "42-Fishing / \n", "1-Vehicle / 50-Aircraft / \n", "1-Vehicle / 11-Motorsport / 124-Tractor / 156-Agriculture / 174-Farm / 252-Tractor pulling / 459-Plough / 906-Sowing / \n", "0-Games / 2-Video game / 1288-Marvel vs. Capcom / 1499-Marvel vs. Capcom 3: Fate of Two Worlds / 1505-Ultimate Marvel vs. Capcom 3 / \n", "3-Concert / \n", "1-Vehicle / 11-Motorsport / 87-Boat / 351-Jumping / 1025-Personal water craft / 1076-Wakeboarding / 2282-Waterskiing / \n", "0-Games / 2-Video game / 17-Racing / 1137-Crash Bandicoot / \n", "1-Vehicle / 11-Motorsport / 27-Motorcycle / 522-Helmet / 1903-Motorcycle helmet / \n", "93-Comedy / 355-Comedian / \n", "3-Concert / 13-Musical ensemble / 68-Lighting / \n", "31-Disc jockey / \n", "10-Animal / 166-Cat / 280-Wildlife / 440-Lion / 537-Zoo / 1039-Tiger / 1591-Leopard / \n", "3-Concert / 5-Dance / \n", "22-Nature / 559-Night / 654-Sun / 998-Landscape / \n", "1-Vehicle / 11-Motorsport / 17-Racing / 27-Motorcycle / 105-Motorcycling / 172-Motocross / 197-Trail / 198-Off-road vehicle / 491-Enduro / \n", "8-Football / 18-Outdoor recreation / 42-Fishing / \n", "5-Dance / 98-Festival / 1289-Caporales / \n", "0-Games / 1-Vehicle / 2-Video game / 4-Car / 34-Action-adventure game / 111-PlayStation 3 / 118-Grand Theft Auto V / \n", "3-Concert / 5-Dance / 38-Orchestra / \n", "3-Concert / 5-Dance / 16-Performance art / 95-Talent show / \n", "6-Animation / \n", "0-Games / 2-Video game / 849-Tales / \n", "457-Knitting / 1804-Sewing needle / \n", "0-Games / 159-World of Warcraft / 171-Warcraft / 1356-Paladin / \n", "1-Vehicle / 4-Car / 11-Motorsport / 17-Racing / 41-Sports car / 70-Driving / \n", "10-Animal / 18-Outdoor recreation / 22-Nature / 213-Forest / 217-Hunting / 706-Deer / 1750-White-tailed deer / 2135-Bowhunting / \n", "609-Tattoo / \n", "14-Guitar / 738-Guitar amplifier / 4104-Fender Precision Bass / \n", "28-Fashion / 206-Shoe / 785-Jeans / 1811-Forever 21 / 2315-Top / \n", "1-Vehicle / 4-Car / 40-Road / 41-Sports car / 1763-Lotus Cars / \n", "0-Games / 2-Video game / 19-PC game / 43-Call of Duty / 422-Call of Duty: Ghosts / \n", "3-Concert / \n", "0-Games / 2-Video game / 60-Basketball / 991-NBA 2K15 / 2156-Rookie / \n", "7-Musician / 13-Musical ensemble / 46-Choir / 1599-Bone / \n", "3-Concert / 7-Musician / 36-Piano / 104-Pianist / \n", "2-Video game / 269-Comic book / 950-Blu-ray disc / \n", "0-Games / 2-Video game / 22-Nature / 86-Plant / 1410-The Elder Scrolls IV: Oblivion / \n", "3-Concert / 7-Musician / 13-Musical ensemble / \n", "1-Vehicle / 11-Motorsport / 351-Jumping / 794-Bridge / \n", "1-Vehicle / 74-Truck / 1070-Semi-trailer truck / \n", "31-Disc jockey / \n", "0-Games / 2387-Military tactics / \n", "22-Nature / 86-Plant / \n", "0-Games / 2-Video game / 311-Street Fighter / 461-Street Fighter IV / 496-Super Street Fighter IV / \n", "13-Musical ensemble / \n", "0-Games / 2-Video game / 1566-Doom / 3440-Serious Sam / 3527-Doom 3 / 4556-Serious Sam / \n", "222-Surfing / 1008-Kite / \n", "2-Video game / 20-Trailer / 159-World of Warcraft / 171-Warcraft / 2828-World of Warcraft: The Burning Crusade / \n", "3-Concert / 38-Orchestra / \n", "0-Games / 47-Personal computer / 2265-Subway Surfers / \n", "5-Dance / 13-Musical ensemble / 16-Performance art / \n", "3-Concert / 7-Musician / 14-Guitar / \n", "447-Sand / 751-Desert / \n", "0-Games / 59-Winter sport / 112-Ice skating / 142-Hockey / 312-Ice rink / \n", "1-Vehicle / 94-Dashcam / \n", "25-Toy / 116-Pokémon / 269-Comic book / 1021-Booster pack / 1049-Collectible card game / 1062-Pokémon Trading Card Game / \n", "1-Vehicle / 2-Video game / 4-Car / 11-Motorsport / 17-Racing / 41-Sports car / 57-Race track / 70-Driving / 169-Rallying / \n", "35-Minecraft / 43-Call of Duty / 113-House / 194-Call of Duty: Modern Warfare 2 / 394-Map / \n", "1-Vehicle / 4-Car / 40-Road / 70-Driving / 94-Dashcam / 212-Highway / 271-Tank / \n", "0-Games / 2-Video game / 55-Video game console / 199-Wii / \n", "0-Games / 2-Video game / 47-Personal computer / 176-PlayStation 4 / 181-Xbox / 1251-The Witcher / 1401-The Witcher 3: Wild Hunt / 2752-God Mode / \n", "5-Dance / \n", "123-Naruto / 285-Sasuke Uchiha / 622-Sakura Haruno / \n", "31-Disc jockey / \n", "0-Games / 2-Video game / 6-Animation / 692-Link / 1857-Sephiroth / \n", "819-World Wide Web / \n", "36-Piano / 53-Keyboard / 107-Musical keyboard / \n", "3-Concert / \n", "0-Games / 2-Video game / 15-Cartoon / 320-Super Smash Bros. / 782-Super Smash Bros. Melee / \n", "3-Concert / 16-Performance art / \n", "208-Wood / 492-Floor / 1011-Tile / 1357-Ceramic / 2362-Porcelain / \n", "1-Vehicle / 4-Car / 74-Truck / 86-Plant / 137-Tree / 167-Four-wheel drive / 349-Chevrolet / 585-Lawn / 854-Diesel engine / 1969-Shrub / \n", "9-Music video / \n", "0-Games / 8-Football / 21-Stadium / 77-Arena / \n", "0-Games / 2-Video game / 47-Personal computer / 394-Map / 497-Macintosh / 506-Garry's Mod / \n", "9-Music video / \n", "1-Vehicle / 11-Motorsport / 17-Racing / 18-Outdoor recreation / 22-Nature / 27-Motorcycle / 105-Motorcycling / 172-Motocross / 2420-Audi TT / \n", "5-Dance / 115-Ballet / \n", "9-Music video / 22-Nature / 657-Autumn / \n", "8-Football / \n", "3-Concert / 7-Musician / 13-Musical ensemble / \n", "3-Concert / 21-Stadium / 68-Lighting / \n", "0-Games / 2-Video game / 23-Mobile phone / 29-Smartphone / \n", "0-Games / 51-Strategy video game / 165-League of Legends / \n", "5-Dance / \n", "10-Animal / 18-Outdoor recreation / 80-Horse / 247-Stallion / 289-Dressage / \n", "0-Games / 8-Football / 54-Highlight film / 76-Ball / 81-Athlete / \n", "9-Music video / 15-Cartoon / 412-Monster / 503-Monster High / \n", "209-University / \n", "36-Piano / 53-Keyboard / \n", "9-Music video / 110-Album / \n", "5-Dance / 98-Festival / 250-Parade / 253-Carnival / \n", "0-Games / 14-Guitar / 211-RuneScape / 366-Dragon / \n", "3-Concert / 7-Musician / 13-Musical ensemble / 38-Orchestra / 46-Choir / \n", "365-Fireworks / \n", "0-Games / 2-Video game / 33-Weapon / 779-Point Blank / \n", "5-Dance / 6-Animation / 1523-Alice / 2201-Alice's Adventures in Wonderland / 3342-The Mad Hatter / \n", "8-Football / 4270-Old Trafford / \n", "47-Personal computer / 228-Laptop / 525-Printer / 2634-Inkjet printing / \n", "12-Food / 26-Cooking / \n", "0-Games / 1-Vehicle / 4-Car / 6-Animation / 15-Cartoon / 25-Toy / 74-Truck / 189-iPad / 433-Construction / 1429-Dump truck / \n", "99-Christmas / 558-Santa Claus / \n", "3-Concert / 30-Drummer / \n", "0-Games / 211-RuneScape / \n", "15-Cartoon / \n", "219-Accordion / \n", "1-Vehicle / 62-Train / 249-Heavy equipment / 433-Construction / 817-Bulldozer / 976-Demolition / \n", "9-Music video / \n", "1-Vehicle / 11-Motorsport / 17-Racing / 57-Race track / 67-Cycling / 69-Bicycle / 2003-Track cycling / \n", "0-Games / 2-Video game / 6-Animation / 1343-Super Robot Wars / 3128-Super Robot / \n", "9-Music video / \n", "66-Bollywood / 1606-Butterfly / \n", "46-Choir / \n", "45-Cosmetics / 173-Eye shadow / 193-Eye liner / 195-Mascara / 196-Eye / 218-Eyelash / 235-Lipstick / 1012-Lip gloss / \n", "1-Vehicle / 50-Aircraft / 82-Airplane / 129-Aviation / \n", "7-Musician / 13-Musical ensemble / 38-Orchestra / 46-Choir / 3042-The Salvation Army / \n", "390-Model / \n", "0-Games / 2-Video game / 6-Animation / 241-Fighting game / 514-Tekken / 797-Tekken / 1470-Tekken Tag Tournament 2 / 4354-Ling Xiaoyu / \n", "9-Music video / \n", "1-Vehicle / 87-Boat / \n", "5-Dance / 16-Performance art / 95-Talent show / \n", "6-Animation / \n", "3-Concert / 7-Musician / 13-Musical ensemble / \n", "78-Wedding / \n", "10-Animal / 91-Fish / 258-Aquarium / 288-Underwater / 601-Shark / \n", "1-Vehicle / 4-Car / \n", "6-Animation / 701-Beyblade / \n", "162-Medicine / \n", "439-Barbie / \n", "319-Robot / 3256-Robotic vacuum cleaner / 4127-Roomba / \n", "1-Vehicle / 153-Dress / \n", "3-Concert / 133-Nightclub / 530-Cue sports / 613-Bar / 1280-Club / \n", "18-Outdoor recreation / 59-Winter sport / 84-Snow / 120-Winter / 177-Skiing / 305-Ski / 501-Snowboard / 789-Alpine skiing / 1830-Downhill / \n", "0-Games / 6-Animation / 15-Cartoon / 89-Comics / \n", "3329-Upholstery / \n", "3-Concert / 68-Lighting / \n", "5-Dance / 115-Ballet / \n", "0-Games / 8-Football / 21-Stadium / 54-Highlight film / \n", "0-Games / 1-Vehicle / 2-Video game / 4-Car / 34-Action-adventure game / 201-Grand Theft Auto: San Andreas / 1612-Multi Theft Auto / \n", "22-Nature / \n", "1411-ESPN / \n", "3-Concert / 7-Musician / 13-Musical ensemble / 1240-Mambo / \n", "0-Games / 18-Outdoor recreation / 54-Highlight film / 81-Athlete / 163-Running / 674-Marathon / \n", "1-Vehicle / 4-Car / 11-Motorsport / 40-Road / 41-Sports car / 70-Driving / 1391-Kia / 1415-Kia Motors / 4498-Kia Optima / \n", "0-Games / 2-Video game / 1029-Aion: The Tower of Eternity / \n", "1-Vehicle / 4-Car / 41-Sports car / 154-Wheel / 245-BMW / 874-BMW M3 / \n", "689-Harmonica / \n", "0-Games / 2-Video game / 411-MapleStory / \n", "0-Games / \n", "1-Vehicle / 27-Motorcycle / 207-Exhaust system / 970-Monkey / \n", "0-Games / 2-Video game / 33-Weapon / 85-Combat / 96-Soldier / 989-Warface / \n", "310-The Sims / 1553-The Sims 4 / \n", "1-Vehicle / 22-Nature / 62-Train / 64-Transport / 103-Rail transport / 119-Locomotive / 143-Railroad car / 267-Mountain / 800-Mountain pass / \n", "0-Games / 8-Football / 54-Highlight film / 73-Ball / 90-Sports game / \n", "0-Games / 2-Video game / 35-Minecraft / \n", "21-Stadium / 77-Arena / 79-American football / \n", "114-Engine / 778-Gliding / 1602-Hang gliding / \n", "1-Vehicle / 18-Outdoor recreation / 87-Boat / 2338-Inflatable boat / \n", "5-Dance / 78-Wedding / 190-Bride / \n", "23-Mobile phone / 29-Smartphone / 37-Gadget / 101-Telephone / 178-Samsung Galaxy / \n", "330-Money / \n", "1-Vehicle / 67-Cycling / 69-Bicycle / 570-Road bicycle racing / 2003-Track cycling / \n", "14-Guitar / 24-String instrument / \n", "0-Games / 2-Video game / 19-PC game / 33-Weapon / 43-Call of Duty / 96-Soldier / 166-Cat / 216-Call of Duty: Modern Warfare 3 / 302-Call of Duty 4: Modern Warfare / \n", "5-Dance / \n", "9-Music video / 31-Disc jockey / \n", "125-Television / \n", "9-Music video / 14-Guitar / \n", "1-Vehicle / 4-Car / 901-Mitsubishi / 1413-Mitsubishi Lancer / \n", "110-Album / \n", "49-School / \n", "5-Dance / 16-Performance art / 678-Latin dance / \n", "36-Piano / 53-Keyboard / 104-Pianist / 107-Musical keyboard / 265-Electronic keyboard / \n", "22-Nature / 121-Photography / \n", "20-Trailer / 359-Harry Potter (Literary Series) / \n", "3-Concert / 5-Dance / \n", "5-Dance / 16-Performance art / \n", "12-Food / 26-Cooking / 52-Dish / 58-Cuisine / 109-Cooking show / 215-Eating / 225-Lake / 301-Meal / 771-Noodle / \n", "28-Fashion / \n", "1-Vehicle / 50-Aircraft / 82-Airplane / \n", "0-Games / 2-Video game / 357-DVD / 413-PlayStation 2 / 581-PlayStation / \n", "6-Animation / 10-Animal / 15-Cartoon / \n", "949-Rowing / \n", "6-Animation / 15-Cartoon / 61-Art / 97-Drawing / 196-Eye / 408-Sketch / 528-Pencil / \n", "20-Trailer / 630-Superman / \n", "3-Concert / 7-Musician / 14-Guitar / \n", "230-Brass instrument / 296-Saxophone / \n", "9-Music video / \n", "174-Farm / 523-Cattle / 660-Rural area / \n", "5-Dance / 98-Festival / 191-Cheerleading / \n", "6-Animation / 15-Cartoon / \n", "10-Animal / 71-Dog / 83-Skateboarding / 108-Skateboard / \n", "0-Games / 60-Basketball / 255-Slam dunk / \n", "66-Bollywood / \n", "365-Fireworks / \n", "60-Basketball / 73-Ball / 76-Ball / \n", "1-Vehicle / 27-Motorcycle / 67-Cycling / 69-Bicycle / 114-Engine / 1952-Carburetor / 3578-Motorized bicycle / \n", "6-Animation / 15-Cartoon / \n", "1-Vehicle / 4-Car / 270-Sedan / 403-Toyota / 1583-Toyota Corolla / \n", "9-Music video / 14-Guitar / \n", "98-Festival / \n", "609-Tattoo / \n", "3-Concert / 7-Musician / 14-Guitar / 30-Drummer / 39-Drums / \n", "5-Dance / \n", "3-Concert / 68-Lighting / \n", "1-Vehicle / 4-Car / 2842-Drive-through / 2852-Burger King / \n", "1-Vehicle / 11-Motorsport / 27-Motorcycle / 67-Cycling / 69-Bicycle / 105-Motorcycling / \n", "3-Concert / \n", "6-Animation / 15-Cartoon / 185-Dragon Ball / 244-Goku / 418-Gohan / \n", "0-Games / 2-Video game / 34-Action-adventure game / 111-PlayStation 3 / 118-Grand Theft Auto V / \n", "10-Animal / 12-Food / 91-Fish / \n", "1-Vehicle / 4-Car / 74-Truck / 375-Pickup truck / 406-Chevrolet / \n", "0-Games / 8-Football / 21-Stadium / 76-Ball / \n", "350-Circus / 788-Juggling / \n", "31-Disc jockey / \n", "9-Music video / \n", "5-Dance / 6-Animation / \n", "45-Cosmetics / 173-Eye shadow / 193-Eye liner / 195-Mascara / 196-Eye / 218-Eyelash / 235-Lipstick / \n", "3-Concert / 7-Musician / 13-Musical ensemble / 46-Choir / 230-Brass instrument / \n", "7-Musician / 14-Guitar / 24-String instrument / 100-Electric guitar / \n", "20-Trailer / \n", "1-Vehicle / 4-Car / 27-Motorcycle / 40-Road / 64-Transport / 575-Traffic / 1020-Emergency vehicle / \n", "293-Slide show / 444-Family / \n", "18-Outdoor recreation / 83-Skateboarding / 108-Skateboard / \n", "14-Guitar / 63-Acoustic guitar / \n", "0-Games / 59-Winter sport / 81-Athlete / 112-Ice skating / 142-Hockey / 312-Ice rink / 2730-NHL 13 / \n", "1-Vehicle / 50-Aircraft / 82-Airplane / 129-Aviation / \n", "86-Plant / 136-Gardening / 279-Garden / 807-Seed / \n", "0-Games / 2-Video game / 19-PC game / 47-Personal computer / \n", "31-Disc jockey / 861-Mixing console / 3024-Numark Industries / \n", "1-Vehicle / 4-Car / 40-Road / 70-Driving / 94-Dashcam / 212-Highway / 286-Bus / \n", "445-Guitar Hero / 2180-Frets on Fire / \n", "762-Angel / \n", "12-Food / 88-Machine / \n", "1-Vehicle / 22-Nature / 62-Train / 64-Transport / 103-Rail transport / 119-Locomotive / 152-Track / 213-Forest / \n", "1-Vehicle / 27-Motorcycle / 40-Road / 105-Motorcycling / 336-Scooter / \n", "113-House / 310-The Sims / 328-Simulation video game / 564-The Sims 3 / \n", "30-Drummer / 39-Drums / 44-Drums / 128-Cymbal / 146-Snare drum / \n", "3-Concert / 5-Dance / 16-Performance art / \n", "1-Vehicle / 4-Car / 40-Road / 70-Driving / 74-Truck / 94-Dashcam / 212-Highway / \n", "123-Naruto / \n", "6-Animation / 15-Cartoon / 72-iPhone / 101-Telephone / \n", "0-Games / 2-Video game / 8-Football / 90-Sports game / 180-FIFA 15 / \n", "3-Concert / 5-Dance / 16-Performance art / 68-Lighting / \n", "6-Animation / 15-Cartoon / 25-Toy / 140-LEGO / 359-Harry Potter (Literary Series) / \n", "3-Concert / 7-Musician / 30-Drummer / 133-Nightclub / \n", "1970-Axe / \n", "619-Manufacturing / \n", "3-Concert / 5-Dance / 16-Performance art / \n", "0-Games / 90-Sports game / 176-PlayStation 4 / 453-Pro Evolution Soccer / 1719-Pro Evolution Soccer 2015 / \n", "56-Hair / 340-Afro-textured hair / \n", "9-Music video / \n", "1-Vehicle / 23-Mobile phone / 29-Smartphone / 101-Telephone / \n", "1-Vehicle / 74-Truck / 86-Plant / 137-Tree / \n", "0-Games / 269-Comic book / 1244-Trunks / \n", "5-Dance / 49-School / 191-Cheerleading / 776-Kindergarten / \n", "18-Outdoor recreation / 83-Skateboarding / 108-Skateboard / 221-Skatepark / 278-GoPro / 522-Helmet / 1323-Helmet camera / \n", "25-Toy / 200-Doll / 412-Monster / 503-Monster High / 4332-Mother 3 / \n", "1-Vehicle / 4-Car / 74-Truck / 659-Van / 2737-GAZ / \n", "3-Concert / 7-Musician / \n", "6-Animation / 313-Book / \n", "25-Toy / 200-Doll / 483-Costume / \n", "3-Concert / \n", "3-Concert / 5-Dance / 16-Performance art / \n", "3-Concert / \n", "0-Games / 256-Tennis / \n", "261-Tool / 1226-Metalworking / \n", "3-Concert / 68-Lighting / 263-Light / \n", "9-Music video / \n", "1532-Military band / \n", "30-Drummer / 39-Drums / 44-Drums / \n", "61-Art / \n", "211-RuneScape / \n", "12-Food / 26-Cooking / 32-Recipe / 236-Baking / 239-Roasting / 274-Meat / 560-Sugar / 1596-Cinnamon / 2042-Nut / \n", "0-Games / 2-Video game / 1911-StepMania / 1990-Pac-Man / \n", "6-Animation / 15-Cartoon / \n", "93-Comedy / 125-Television / \n", "12-Food / 26-Cooking / 32-Recipe / 109-Cooking show / 215-Eating / 808-Breakfast / \n", "131-Computer / 513-Computer keyboard / \n", "1-Vehicle / 4-Car / 449-Honda / 531-Hatchback / 2900-Honda Fit / \n", "23-Mobile phone / 29-Smartphone / 125-Television / 1565-Logitech / \n", "6-Animation / 15-Cartoon / \n", "39-Drums / 44-Drums / 687-Taiko no Tatsujin / \n", "6-Animation / 15-Cartoon / 116-Pokémon / 186-Pokémon / 1005-Pokémon Omega Ruby and Alpha Sapphire / 1153-Pokémon HeartGold and SoulSilver / 4379-Team Rocket / \n", "25-Toy / 109-Cooking show / 447-Sand / \n", "23-Mobile phone / 29-Smartphone / 125-Television / \n", "10-Animal / 166-Cat / 346-Kitten / \n", "23-Mobile phone / 29-Smartphone / 37-Gadget / 101-Telephone / 141-Microsoft Windows / 932-Microsoft Lumia / \n", "12-Food / 88-Machine / 625-Factory / 812-Plastic / 2031-Molding / \n", "1-Vehicle / 286-Bus / \n", "5-Dance / \n", "8-Football / \n", "0-Games / 88-Machine / 366-Dragon / 535-Slot machine / \n", "0-Games / 1083-Role-playing game / \n", "275-Diving / 287-Human swimming / \n", "5-Dance / 16-Performance art / \n", "1-Vehicle / 4-Car / 12-Food / 25-Toy / 215-Eating / 475-Cheese / 3822-Mold / \n", "9-Music video / \n", "18-Outdoor recreation / 59-Winter sport / 112-Ice skating / 316-Figure skating / \n", "3-Concert / 7-Musician / 16-Performance art / \n", "269-Comic book / \n", "448-Planet / 1508-Spacecraft / 3413-Pluto / \n", "6-Animation / 15-Cartoon / \n", "1398-Triangle / \n", "6-Animation / 15-Cartoon / 290-Pony / 343-My Little Pony / \n", "86-Plant / 136-Gardening / 137-Tree / \n", "0-Games / 2-Video game / 596-Monster Hunter / 2260-Monster Hunter Portable 3rd / \n", "28-Fashion / 56-Hair / \n", "1902-Flyff / \n", "1612-Multi Theft Auto / \n", "3-Concert / 38-Orchestra / \n", "113-House / 279-Garden / 363-Soil / \n", "341-Sketch comedy / 364-Sitcom / \n", "42-Fishing / 275-Diving / 288-Underwater / 381-Underwater diving / 407-Scuba diving / \n", "12-Food / 227-Water / 3639-Heat pump / \n", "5-Dance / 78-Wedding / \n", "10-Animal / 18-Outdoor recreation / 80-Horse / 247-Stallion / 289-Dressage / \n", "1-Vehicle / 92-Radio-controlled model / 149-Model aircraft / 160-Radio-controlled aircraft / 266-Helicopter / 392-Radio-controlled helicopter / 601-Shark / \n", "0-Games / 59-Winter sport / 112-Ice skating / 142-Hockey / \n", "0-Games / 8-Football / 13-Musical ensemble / 21-Stadium / 77-Arena / 79-American football / 203-Marching band / \n", "3837-Kyūdō / \n", "12-Food / 26-Cooking / 32-Recipe / 367-Drink / 867-Tea / \n", "3-Concert / \n", "325-Vampire / \n", "6-Animation / 15-Cartoon / \n", "3-Concert / 16-Performance art / \n", "12-Food / 341-Sketch comedy / \n", "5-Dance / 430-Ballroom dance / 1231-Mandolin / \n", "10-Animal / 80-Horse / 247-Stallion / 289-Dressage / 2553-Colt / \n", "209-University / 319-Robot / \n", "1-Vehicle / 50-Aircraft / 266-Helicopter / 758-Air force / \n", "3-Concert / 7-Musician / 13-Musical ensemble / 14-Guitar / \n", "36-Piano / 53-Keyboard / \n", "5-Dance / 49-School / 209-University / \n", "7-Musician / 36-Piano / 46-Choir / \n", "1-Vehicle / 4-Car / 245-BMW / 3058-BMW 7 Series / \n", "1-Vehicle / 4-Car / 11-Motorsport / 27-Motorcycle / 154-Wheel / \n", "56-Hair / 367-Drink / \n", "45-Cosmetics / 56-Hair / \n", "7-Musician / 13-Musical ensemble / 203-Marching band / \n", "5-Dance / \n", "0-Games / 19-PC game / 35-Minecraft / \n", "0-Games / 2-Video game / 8-Football / 21-Stadium / 54-Highlight film / 79-American football / 90-Sports game / 561-Madden NFL / \n", "9-Music video / 31-Disc jockey / \n", "0-Games / 8-Football / 90-Sports game / 180-FIFA 15 / 480-FIFA 13 / 848-FIFA 12 / \n", "3-Concert / 7-Musician / 13-Musical ensemble / 14-Guitar / \n", "9-Music video / \n", "15-Cartoon / 61-Art / 97-Drawing / 148-Painting / 408-Sketch / 528-Pencil / 628-Portrait / \n", "0-Games / 2-Video game / 19-PC game / 43-Call of Duty / 47-Personal computer / 1749-Titanfall / \n", "3-Concert / \n", "0-Games / 2-Video game / 685-Luigi / 1170-Super Mario World / \n", "0-Games / 1-Vehicle / 11-Motorsport / 17-Racing / 27-Motorcycle / 172-Motocross / 2930-MX vs. ATV / \n", "67-Cycling / 69-Bicycle / 377-BMX bike / \n", "22-Nature / 225-Lake / 2280-Cottage / \n", "0-Games / 2-Video game / 19-PC game / 51-Strategy video game / 165-League of Legends / \n", "8-Football / 73-Ball / \n", "3-Concert / 7-Musician / 13-Musical ensemble / 46-Choir / \n", "15-Cartoon / 185-Dragon Ball / 244-Goku / \n", "10-Animal / 12-Food / 26-Cooking / 1514-Crab / \n", "0-Games / 2-Video game / 474-Destiny / \n", "10-Animal / 80-Horse / \n", "0-Games / 8-Football / 21-Stadium / 65-Kick / 76-Ball / \n", "3-Concert / 68-Lighting / 99-Christmas / 263-Light / 729-Christmas decoration / 1098-Christmas lights / \n", "0-Games / 10-Animal / 80-Horse / 1871-Foal / \n", "372-Playing card / 762-Angel / 1191-Tarot / \n", "20-Trailer / 264-Star Wars / 1791-Stargate / \n", "0-Games / 51-Strategy video game / 165-League of Legends / \n", "0-Games / 1-Vehicle / 2-Video game / 140-LEGO / 2276-Lego Marvel Super Heroes / 4016-Lego City Undercover / \n", "0-Games / 23-Mobile phone / 29-Smartphone / 72-iPhone / 4567-Cut the Rope / \n", "1-Vehicle / 4-Car / 50-Aircraft / 1554-Range Rover / 3890-Range Rover Sport / \n", "3-Concert / 7-Musician / 30-Drummer / 39-Drums / \n", "1-Vehicle / 4-Car / 349-Chevrolet / 406-Chevrolet / 659-Van / \n", "0-Games / 2-Video game / 19-PC game / 51-Strategy video game / 61-Art / 732-StarCraft / \n", "46-Choir / \n", "9-Music video / 22-Nature / 243-Sonic the Hedgehog / \n", "5-Dance / 13-Musical ensemble / \n", "10-Animal / 71-Dog / \n", "3-Concert / 7-Musician / 13-Musical ensemble / 38-Orchestra / 46-Choir / 115-Ballet / \n", "1-Vehicle / 4-Car / 41-Sports car / 67-Cycling / 69-Bicycle / \n", "6-Animation / 15-Cartoon / 61-Art / 97-Drawing / \n", "49-School / 102-Building / \n", "75-Wrestling / \n", "1-Vehicle / 4-Car / 11-Motorsport / 17-Racing / 27-Motorcycle / 57-Race track / 105-Motorcycling / \n", "49-School / \n", "5-Dance / \n", "3-Concert / 7-Musician / 13-Musical ensemble / 16-Performance art / 38-Orchestra / \n", "322-Mixtape / \n", "61-Art / 479-Graffiti / \n", "6-Animation / 15-Cartoon / 61-Art / 97-Drawing / 148-Painting / \n", "330-Money / 458-Gold / 731-Coin / 912-Silver / 2095-Cash / 4247-Bullion / \n", "3954-Camera Operator / \n", "20-Trailer / \n", "20-Trailer / \n", "1-Vehicle / 4-Car / 11-Motorsport / \n", "18-Outdoor recreation / 84-Snow / \n", "3-Concert / 7-Musician / 13-Musical ensemble / 16-Performance art / \n", "0-Games / 6-Animation / 15-Cartoon / \n", "66-Bollywood / \n", "0-Games / 2-Video game / 601-Shark / \n", "12-Food / 26-Cooking / 32-Recipe / 52-Dish / 58-Cuisine / 109-Cooking show / 210-Cookware and bakeware / \n", "3-Concert / 68-Lighting / \n", "23-Mobile phone / 29-Smartphone / 37-Gadget / 47-Personal computer / 145-Tablet computer / \n", "5-Dance / 892-Aerobics / \n", "12-Food / 52-Dish / 58-Cuisine / 215-Eating / 301-Meal / 1227-Sushi / \n", "2207-Kemenche / \n", "36-Piano / 53-Keyboard / 265-Electronic keyboard / \n", "28-Fashion / 78-Wedding / 121-Photography / 190-Bride / 421-Gown / 2269-Wedding photography / 2873-Kimono / \n", "8-Football / \n", "0-Games / 2-Video game / 19-PC game / 33-Weapon / 43-Call of Duty / 96-Soldier / \n", "3-Concert / 7-Musician / 13-Musical ensemble / 30-Drummer / \n", "0-Games / 535-Slot machine / 3952-Amusement arcade / \n", "1-Vehicle / 4-Car / 277-Sport utility vehicle / 901-Mitsubishi / 3896-Mitsubishi Outlander / \n", "0-Games / 2-Video game / 948-Far Cry / 1683-Far Cry 4 / \n", "25-Toy / 179-Kitchen / 200-Doll / 439-Barbie / \n", "643-Hero / \n", "2207-Kemenche / \n", "1060-Raw food / \n", "6-Animation / 15-Cartoon / \n", "0-Games / 2-Video game / 6-Animation / 15-Cartoon / 89-Comics / 185-Dragon Ball / 244-Goku / 869-Dragon Ball Z: Budokai Tenkaichi / 1165-Majin Buu / \n", "148-Painting / \n", "7-Musician / 13-Musical ensemble / 78-Wedding / 150-Violin / 401-Cello / 450-Viola / \n", "345-Loudspeaker / 484-Subwoofer / \n", "242-Paper / 370-Origami / 3300-Napkin / 4124-Napkin folding / \n", "9-Music video / 99-Christmas / \n", "28-Fashion / 45-Cosmetics / 56-Hair / 1811-Forever 21 / \n", "5-Dance / \n", "10-Animal / 91-Fish / 258-Aquarium / 1368-Gallon / \n", "36-Piano / 53-Keyboard / 1492-Persona / 2896-Shin Megami Tensei: Persona 3 / \n", "12-Food / 26-Cooking / 32-Recipe / 232-Cake / 236-Baking / 1081-Lemon / 1114-Stuffing / 1322-Biscuit / 2593-Sponge cake / \n", "341-Sketch comedy / \n", "8-Football / \n", "183-Manga / 4038-Waistcoat / \n", "130-Gymnastics / \n", "0-Games / 2-Video game / 19-PC game / 33-Weapon / 43-Call of Duty / 134-Call of Duty: Black Ops / 216-Call of Duty: Modern Warfare 3 / 704-Call of Duty: World at War / \n", "0-Games / 2-Video game / 19-PC game / 33-Weapon / 850-Left 4 Dead / 963-Left 4 Dead 2 / \n", "0-Games / 6-Animation / 269-Comic book / 460-Yu-Gi-Oh! Trading Card Game / 494-Card game / \n", "1-Vehicle / 4-Car / 114-Engine / 1521-Jaguar Cars / \n", "20-Trailer / \n", "1-Vehicle / 2-Video game / 20-Trailer / 124-Tractor / 174-Farm / 715-Farming Simulator / \n", "10-Animal / 25-Toy / 48-Pet / 71-Dog / 182-Puppy / \n", "5-Dance / 133-Nightclub / \n", "6-Animation / 15-Cartoon / \n", "9-Music video / \n", "222-Surfing / \n", "2840-Engraving / \n", "282-Skin / 4665-Day spa / \n", "1-Vehicle / 4-Car / 41-Sports car / 231-Supercar / 983-Lexus / 4158-Lexus LFA / \n", "6-Animation / 116-Pokémon / \n", "1-Vehicle / 62-Train / 103-Rail transport / 119-Locomotive / \n", "18-Outdoor recreation / 22-Nature / 137-Tree / 213-Forest / 217-Hunting / \n", "1-Vehicle / 27-Motorcycle / \n", "1-Vehicle / 62-Train / 64-Transport / 103-Rail transport / 161-Train station / 224-Rapid transit / 1547-Passenger / \n", "0-Games / \n", "1-Vehicle / 27-Motorcycle / 336-Scooter / \n", "6-Animation / 15-Cartoon / 3877-Ghost in the Shell / \n", "1-Vehicle / 4-Car / 11-Motorsport / 17-Racing / 41-Sports car / 135-Drifting / 154-Wheel / 158-Tire / 493-Rim / \n", "0-Games / 2-Video game / 89-Comics / 123-Naruto / 2771-Naruto: Clash of Ninja / \n", "0-Games / 2-Video game / 116-Pokémon / 186-Pokémon / 1204-Pokémon Ruby and Sapphire / \n", "425-City / \n", "1-Vehicle / 31-Disc jockey / 74-Truck / \n", "0-Games / 6-Animation / 15-Cartoon / 25-Toy / 1154-Anpanman / \n", "36-Piano / 150-Violin / \n", "18-Outdoor recreation / 83-Skateboarding / 108-Skateboard / \n", "75-Wrestling / 269-Comic book / \n", "3-Concert / \n", "0-Games / 23-Mobile phone / 29-Smartphone / 72-iPhone / 189-iPad / 272-iPod / \n", "13-Musical ensemble / \n", "298-Television advertisement / 337-Advertising / 3649-Metroid Prime / \n", "5-Dance / \n", "5-Dance / \n", "0-Games / 2-Video game / 37-Gadget / 55-Video game console / 292-Handheld game console / \n", "31-Disc jockey / \n", "2-Video game / 297-Grand Theft Auto IV / \n", "6-Animation / 9-Music video / 15-Cartoon / \n", "8-Football / \n", "0-Games / 2-Video game / 328-Simulation video game / 2088-City-building game / 2307-SimCity / 3290-SimCity / \n", "3-Concert / 16-Performance art / \n", "12-Food / 26-Cooking / 32-Recipe / 226-Dessert / 295-Chocolate / 2386-Chocolate brownie / \n", "131-Computer / 155-Camera / 278-GoPro / 633-Photograph / \n", "1-Vehicle / 17-Racing / 18-Outdoor recreation / 59-Winter sport / 84-Snow / 177-Skiing / 789-Alpine skiing / \n", "10-Animal / 166-Cat / 440-Lion / 537-Zoo / 3986-Cougar / \n", "6-Animation / 352-Need for Speed / 2126-Need for Speed: Underground / \n", "130-Gymnastics / \n", "0-Games / 1138-Tower of Saviors / \n", "9-Music video / \n", "14-Guitar / 841-Butter / \n", "0-Games / 2-Video game / 432-Kingdom Hearts / 665-Kingdom Hearts / 836-Sora / 941-Kingdom Hearts II / \n", "0-Games / 2-Video game / 686-Diablo III / \n", "0-Games / 2-Video game / \n", "1-Vehicle / 4-Car / 74-Truck / \n", "9-Music video / \n", "30-Drummer / 39-Drums / 44-Drums / 128-Cymbal / 146-Snare drum / \n", "56-Hair / 106-Hairstyle / 607-Lace / \n", "574-Elevator / \n", "3-Concert / 7-Musician / 14-Guitar / 24-String instrument / \n", "85-Combat / 130-Gymnastics / \n", "1-Vehicle / 67-Cycling / 69-Bicycle / 570-Road bicycle racing / \n", "45-Cosmetics / \n", "75-Wrestling / 85-Combat / \n", "82-Airplane / 778-Gliding / 1263-Glider / \n", "3-Concert / 16-Performance art / \n", "0-Games / 8-Football / \n", "23-Mobile phone / 29-Smartphone / 37-Gadget / 101-Telephone / 178-Samsung Galaxy / 1207-Samsung Galaxy S5 / \n", "1-Vehicle / 4-Car / 531-Hatchback / 683-Renault / \n", "5-Dance / 16-Performance art / \n", "22-Nature / 223-Weather / 490-Rain / 499-Cloud / 502-Sky / \n", "86-Plant / 136-Gardening / 261-Tool / 279-Garden / 585-Lawn / 2437-String trimmer / \n", "5-Dance / 1117-Camcorder / \n", "237-Final Fantasy / \n", "325-Vampire / \n", "9-Music video / \n", "259-Fire / 282-Skin / \n", "10-Animal / 132-Bird / 217-Hunting / 1373-Torero / \n", "536-Glass / \n", "3-Concert / \n", "0-Games / 2-Video game / 35-Minecraft / \n", "35-Minecraft / \n", "504-Web page / \n", "12-Food / 88-Machine / 533-Oil / 1769-Olive / 3027-Olive oil / \n", "3-Concert / 7-Musician / 13-Musical ensemble / 30-Drummer / \n", "1-Vehicle / \n", "3-Concert / 13-Musical ensemble / 16-Performance art / \n", "9-Music video / \n", "9-Music video / \n", "0-Games / 791-Magic: The Gathering / \n", "3-Concert / 5-Dance / 98-Festival / \n", "3-Concert / 46-Choir / \n", "36-Piano / 53-Keyboard / 104-Pianist / \n", "66-Bollywood / \n", "0-Games / 2-Video game / 229-Halo / 866-Halo 4 / \n", "394-Map / \n", "35-Minecraft / \n", "110-Album / \n", "5-Dance / \n", "2-Video game / 43-Call of Duty / 134-Call of Duty: Black Ops / 635-Achievement / 924-Trophy / \n", "44-Drums / \n", "46-Choir / \n", "3-Concert / 7-Musician / 14-Guitar / \n", "0-Games / 8-Football / 21-Stadium / 77-Arena / \n", "385-Face / 2062-Kidney / \n", "3-Concert / 68-Lighting / \n", "88-Machine / 539-Home appliance / 733-Washing / 818-Washing machine / 1142-Laundry / 1175-Soap / 1456-Clothes dryer / 4325-Detergent / \n", "0-Games / 2-Video game / 55-Video game console / 581-PlayStation / \n", "9-Music video / 14-Guitar / 63-Acoustic guitar / 465-Microphone / \n", "16-Performance art / 350-Circus / 648-Clown / \n", "0-Games / 1112-Candy Crush Saga / \n", "2222-Crocodile / \n", "0-Games / 1-Vehicle / 2-Video game / 19-PC game / 33-Weapon / 96-Soldier / 192-Battlefield / 383-Battlefield 4 / \n", "10-Animal / 48-Pet / 71-Dog / 182-Puppy / 1278-Labrador Retriever / 1377-Golden Retriever / \n", "0-Games / 75-Wrestling / 85-Combat / \n", "0-Games / 2-Video game / 6-Animation / 126-The Walt Disney Company / 1077-Lightning McQueen / 3484-Cars 2 / \n", "376-Flute / \n", "3889-Periodic Table / \n", "3-Concert / 16-Performance art / \n", "364-Sitcom / \n", "6-Animation / 15-Cartoon / 61-Art / 97-Drawing / 343-My Little Pony / \n", "10-Animal / 48-Pet / 71-Dog / 182-Puppy / \n", "452-Watch / \n", "33-Weapon / 521-Metal / \n", "6-Animation / 15-Cartoon / 801-Inuyasha / \n", "3-Concert / 7-Musician / 13-Musical ensemble / 966-Sandwich / 1843-Ham / \n", "0-Games / 2-Video game / 33-Weapon / 43-Call of Duty / 138-Call of Duty: Black Ops II / 216-Call of Duty: Modern Warfare 3 / \n", "72-iPhone / 131-Computer / 504-Web page / \n", "222-Surfing / 815-Surfboard / \n", "20-Trailer / 43-Call of Duty / 422-Call of Duty: Ghosts / \n", "10-Animal / 17-Racing / 80-Horse / 384-Horse racing / \n", "1-Vehicle / 4-Car / 11-Motorsport / 659-Van / 4453-Kei car / \n", "6-Animation / 15-Cartoon / 4221-Saiyuki / \n", "0-Games / 2-Video game / 6-Animation / \n", "1-Vehicle / 4-Car / 11-Motorsport / 17-Racing / 169-Rallying / \n", "75-Wrestling / 85-Combat / \n", "1-Vehicle / 50-Aircraft / 82-Airplane / 129-Aviation / 157-Jet aircraft / 388-Fighter aircraft / 758-Air force / 1483-United States Air Force / \n", "1-Vehicle / 18-Outdoor recreation / 50-Aircraft / 92-Radio-controlled model / 149-Model aircraft / 160-Radio-controlled aircraft / 266-Helicopter / 392-Radio-controlled helicopter / 1126-Tyrannosaurus / \n", "10-Animal / 523-Cattle / 534-Milk / \n", "3-Concert / \n", "3-Concert / \n", "5-Dance / \n", "7-Musician / 14-Guitar / 24-String instrument / 100-Electric guitar / \n", "0-Games / 2-Video game / 23-Mobile phone / 29-Smartphone / 72-iPhone / 189-iPad / \n", "6-Animation / \n", "2-Video game / 121-Photography / 155-Camera / 276-Room / \n", "66-Bollywood / 78-Wedding / 466-Gift / \n", "68-Lighting / 286-Bus / \n", "163-Running / 769-Red Bull / \n", "31-Disc jockey / \n", "1-Vehicle / 4-Car / 1074-Hyundai / \n", "122-Weight training / 529-Squat / \n", "1-Vehicle / 4-Car / 11-Motorsport / 17-Racing / \n", "3-Concert / 7-Musician / 13-Musical ensemble / 38-Orchestra / 46-Choir / 1019-Christ / \n", "1-Vehicle / 4-Car / 493-Rim / \n", "9-Music video / 66-Bollywood / \n", "25-Toy / 412-Monster / 713-Egg / 1047-Easter egg / \n", "1151-Temple / 1353-Bishop / \n", "31-Disc jockey / \n", "0-Games / 599-Dofus / \n", "6-Animation / 15-Cartoon / \n", "122-Weight training / \n", "341-Sketch comedy / \n", "0-Games / 2-Video game / 33-Weapon / 2436-Zeus / \n", "1-Vehicle / 62-Train / 119-Locomotive / 2369-RailWorks / \n", "121-Photography / 293-Slide show / \n", "55-Video game console / 127-Xbox 360 / 181-Xbox / 386-Xbox / \n", "0-Games / 8-Football / 21-Stadium / 65-Kick / 76-Ball / \n", "1-Vehicle / 4-Car / 158-Tire / 403-Toyota / 659-Van / 1624-Minivan / \n", "7-Musician / 2641-Radha / \n", "448-Planet / 4349-Neptune / \n", "3-Concert / 7-Musician / 13-Musical ensemble / 44-Drums / \n", "23-Mobile phone / 29-Smartphone / 37-Gadget / 2163-Rechargeable battery / \n", "23-Mobile phone / 29-Smartphone / 178-Samsung Galaxy / 1776-Samsung Galaxy Note II / \n", "83-Skateboarding / 108-Skateboard / 221-Skatepark / \n", "3-Concert / 7-Musician / 13-Musical ensemble / 30-Drummer / \n", "3-Concert / 5-Dance / 16-Performance art / \n", "60-Basketball / \n", "3-Concert / 7-Musician / \n", "1-Vehicle / 18-Outdoor recreation / 42-Fishing / 87-Boat / 225-Lake / 268-Recreational fishing / 896-Largemouth bass / \n", "10-Animal / 80-Horse / 247-Stallion / 404-Livestock / 438-Mare / 2972-Andalusian horse / \n", "0-Games / 6-Animation / 15-Cartoon / 25-Toy / 82-Airplane / 125-Television / 323-Toddler / \n", "47-Personal computer / 145-Tablet computer / 1802-Jubeat / \n", "12-Food / 26-Cooking / 32-Recipe / 52-Dish / 184-Vegetable / 236-Baking / 454-Sauce / 589-Grilling / 1186-Seafood / \n", "8-Football / \n", "5-Dance / 49-School / \n", "36-Piano / 53-Keyboard / 265-Electronic keyboard / \n", "3-Concert / \n", "10-Animal / 25-Toy / 166-Cat / 346-Kitten / 1234-Mouse / \n", "3-Concert / 7-Musician / 30-Drummer / \n", "25-Toy / \n", "20-Trailer / 2075-Titan / \n", "5-Dance / 66-Bollywood / \n", "7-Musician / 9-Music video / \n", "1-Vehicle / 4-Car / 86-Plant / 124-Tractor / 156-Agriculture / 174-Farm / 249-Heavy equipment / 332-Mercedes-Benz / 459-Plough / 660-Rural area / \n", "9-Music video / \n", "0-Games / 8-Football / 54-Highlight film / \n", "0-Games / 1-Vehicle / 4-Car / 34-Action-adventure game / 201-Grand Theft Auto: San Andreas / 2402-Grand Theft Auto / \n", "20-Trailer / \n", "593-Moon / \n", "9-Music video / \n", "110-Album / \n", "12-Food / 2371-Hot dog / \n", "0-Games / 505-Mega Man / 1664-Zero / 3035-Mega Man Zero / \n", "1862-Corel / \n", "36-Piano / 53-Keyboard / 246-Furniture / 457-Knitting / 546-Stitch / 1479-Quilt / 2966-Patchwork / \n", "47-Personal computer / 145-Tablet computer / \n", "0-Games / 51-Strategy video game / 165-League of Legends / \n", "1151-Temple / \n", "6-Animation / 15-Cartoon / \n", "42-Fishing / 1836-Fly tying / \n", "1-Vehicle / 4-Car / 64-Transport / 74-Truck / 158-Tire / 249-Heavy equipment / 1772-Forklift / \n", "61-Art / 97-Drawing / 148-Painting / 303-Paint / 408-Sketch / 528-Pencil / \n", "47-Personal computer / 145-Tablet computer / 155-Camera / 802-Google Nexus / 2264-Nexus 7 / \n", "2060-Roulette / \n", "9-Music video / 147-River / \n", "3412-Daredevil / \n", "10-Animal / 48-Pet / 71-Dog / 182-Puppy / 2059-Doberman Pinscher / \n", "3-Concert / 68-Lighting / \n", "20-Trailer / 66-Bollywood / \n", "110-Album / \n", "0-Games / 2-Video game / 8-Football / 90-Sports game / 453-Pro Evolution Soccer / 3526-Pro Evolution Soccer 2009 / \n", "55-Video game console / 127-Xbox 360 / 181-Xbox / 386-Xbox / \n", "12-Food / \n", "117-Boxing / \n", "1-Vehicle / 4-Car / 50-Aircraft / 82-Airplane / 157-Jet aircraft / 231-Supercar / 388-Fighter aircraft / 569-Lamborghini / 2510-McDonnell Douglas F/A-18 Hornet / \n", "1-Vehicle / 4-Car / 11-Motorsport / 74-Truck / 349-Chevrolet / 406-Chevrolet / 4251-Chevrolet K5 Blazer / \n", "0-Games / 51-Strategy video game / \n", "67-Cycling / 69-Bicycle / \n", "56-Hair / \n", "6-Animation / 15-Cartoon / 1563-Stick figure / \n", "333-Nike; Inc. / \n", "0-Games / 2-Video game / 8-Football / 90-Sports game / 480-FIFA 13 / \n", "3-Concert / 68-Lighting / \n", "1-Vehicle / 4-Car / 40-Road / 70-Driving / 94-Dashcam / \n", "7-Musician / 13-Musical ensemble / 21-Stadium / 54-Highlight film / 79-American football / 203-Marching band / 1675-Drum and bugle corps / \n", "31-Disc jockey / \n", "164-Beach / 222-Surfing / \n", "0-Games / 2-Video game / 33-Weapon / 257-Counter-Strike / \n", "1-Vehicle / 11-Motorsport / 17-Racing / 27-Motorcycle / 57-Race track / 105-Motorcycling / 172-Motocross / \n", "1016-Glove / \n", "67-Cycling / 69-Bicycle / 570-Road bicycle racing / \n", "1-Vehicle / 4-Car / 527-Battery / 574-Elevator / 1454-Stairs / \n", "0-Games / 1-Vehicle / 2-Video game / 4-Car / 34-Action-adventure game / 201-Grand Theft Auto: San Andreas / \n", "5-Dance / \n", "0-Games / 2-Video game / 19-PC game / 33-Weapon / 229-Halo / 644-Halo: Reach / \n", "7-Musician / 24-String instrument / 38-Orchestra / 150-Violin / 401-Cello / 450-Viola / 545-Quartet (ensemble) / \n", "1-Vehicle / 62-Train / 64-Transport / 103-Rail transport / 119-Locomotive / 143-Railroad car / 152-Track / \n", "25-Toy / 242-Paper / 1094-Puppet / \n", "6-Animation / 123-Naruto / 285-Sasuke Uchiha / 622-Sakura Haruno / \n", "0-Games / 2-Video game / 2143-18 Wheels of Steel / \n", "110-Album / \n", "1198-Birth / \n", "0-Games / 2-Video game / 19-PC game / 33-Weapon / 257-Counter-Strike / \n", "3-Concert / 5-Dance / 49-School / \n", "5-Dance / 95-Talent show / \n", "13-Musical ensemble / \n", "6-Animation / 15-Cartoon / 538-Illustration / \n", "97-Drawing / 882-Pen / \n", "3-Concert / 7-Musician / 13-Musical ensemble / 150-Violin / 639-Fiddle / \n", "18-Outdoor recreation / 83-Skateboarding / 108-Skateboard / 221-Skatepark / \n", "78-Wedding / 121-Photography / 190-Bride / 421-Gown / 2269-Wedding photography / \n", "12-Food / 26-Cooking / 1408-Snack / \n", "12-Food / 26-Cooking / 32-Recipe / 58-Cuisine / 437-Cream / 684-Ice cream / \n", "12-Food / 26-Cooking / 32-Recipe / 52-Dish / 184-Vegetable / 210-Cookware and bakeware / 239-Roasting / 542-Potato / 597-Soup / 2405-Zucchini / 3113-Multicooker / \n", "1-Vehicle / 4-Car / 25-Toy / 92-Radio-controlled model / 140-LEGO / 202-Radio-controlled car / 319-Robot / \n", "12-Food / 1227-Sushi / \n", "102-Building / 113-House / 170-Home improvement / 433-Construction / 707-Roof / \n", "0-Games / 2-Video game / 1580-Crysis 2 / \n", "8-Football / \n", "0-Games / 6-Animation / 116-Pokémon / 186-Pokémon / 690-Pokémon X and Y / \n", "5-Dance / \n", "0-Games / 2-Video game / 34-Action-adventure game / 2519-Grand Theft Auto: Liberty City Stories / \n", "31-Disc jockey / \n", "1-Vehicle / 4-Car / 11-Motorsport / 17-Racing / 84-Snow / 169-Rallying / \n", "78-Wedding / \n", "0-Games / 2-Video game / 34-Action-adventure game / \n", "28-Fashion / 305-Ski / \n", "9-Music video / \n", "766-President of the United States / \n", "2-Video game / 34-Action-adventure game / 201-Grand Theft Auto: San Andreas / \n", "66-Bollywood / \n", "0-Games / 2-Video game / 35-Minecraft / \n", "5-Dance / \n", "6-Animation / 10-Animal / 15-Cartoon / 290-Pony / 343-My Little Pony / 1164-Cartoon Network / \n", "3-Concert / 7-Musician / 13-Musical ensemble / 38-Orchestra / 46-Choir / \n", "124-Tractor / 156-Agriculture / 174-Farm / 741-Mower / 1037-Grass / \n", "78-Wedding / 121-Photography / 293-Slide show / 633-Photograph / \n", "6-Animation / 15-Cartoon / 2624-Lelouch Lamperouge / \n", "106-Hairstyle / 1359-Hair iron / 3178-Clothes iron / \n", "75-Wrestling / 1080-WWE 2K15 / \n", "10-Animal / 80-Horse / 438-Mare / \n", "33-Weapon / 315-Knife / 434-Ninja / \n", "47-Personal computer / 131-Computer / 309-Computer hardware / 602-Central processing unit / \n", "45-Cosmetics / \n", "9-Music video / \n", "8-Football / 595-Champion / \n", "89-Comics / 1469-X-Men / \n", "122-Weight training / 204-Gym / 661-Stretching / \n", "2-Video game / 23-Mobile phone / 29-Smartphone / 37-Gadget / 101-Telephone / 155-Camera / 178-Samsung Galaxy / 487-Samsung Electronics / 883-Samsung Galaxy S4 / \n", "6-Animation / 15-Cartoon / 20-Trailer / 298-Television advertisement / \n", "6-Animation / 15-Cartoon / 679-Mickey Mouse / 1234-Mouse / 1445-Donald Duck / \n", "36-Piano / 53-Keyboard / 104-Pianist / 107-Musical keyboard / 265-Electronic keyboard / \n", "1-Vehicle / 4-Car / 11-Motorsport / 169-Rallying / 299-Volkswagen Passenger Cars / \n", "512-Printing / 1757-Screen printing / \n", "1-Vehicle / 18-Outdoor recreation / 147-River / 1018-Salmon / 1195-Rafting / \n", "3369-Diego de la Vega / \n", "10-Animal / \n", "36-Piano / 88-Machine / 457-Knitting / 476-Textile / 1100-Hat / \n", "22-Nature / 223-Weather / \n", "12-Food / 26-Cooking / 32-Recipe / 52-Dish / 210-Cookware and bakeware / \n", "7-Musician / 14-Guitar / 24-String instrument / 100-Electric guitar / 1103-Gibson Les Paul / \n", "10-Animal / 25-Toy / 71-Dog / 524-Littlest Pet Shop / \n", "5-Dance / 1726-Harlem Shake (meme) / \n", "12-Food / 26-Cooking / 32-Recipe / 52-Dish / 58-Cuisine / 109-Cooking show / 184-Vegetable / \n", "0-Games / 2-Video game / 43-Call of Duty / 118-Grand Theft Auto V / 134-Call of Duty: Black Ops / 138-Call of Duty: Black Ops II / \n", "23-Mobile phone / 29-Smartphone / 37-Gadget / 72-iPhone / 101-Telephone / 554-iPhone 4 / 1366-LG Optimus series / \n", "4089-Maracas / \n", "45-Cosmetics / 121-Photography / 155-Camera / \n", "15-Cartoon / 185-Dragon Ball / 244-Goku / 418-Gohan / \n", "0-Games / 2-Video game / 6-Animation / 51-Strategy video game / 382-Dota 2 / 549-Defense of the Ancients / \n", "25-Toy / 417-Play-Doh / 587-Clay / 1185-Plasticine / \n", "5-Dance / \n", "22-Nature / 694-Tai chi / \n", "0-Games / 1-Vehicle / 2-Video game / 4-Car / 34-Action-adventure game / 47-Personal computer / 297-Grand Theft Auto IV / 658-Grand Theft Auto: The Lost and Damned / 1923-Niko Bellic / \n", "3-Concert / \n", "0-Games / 2-Video game / 55-Video game console / 320-Super Smash Bros. / 4449-King Dedede / \n", "1-Vehicle / 4-Car / 484-Subwoofer / \n", "9-Music video / 147-River / \n", "13-Musical ensemble / \n", "3-Concert / 7-Musician / 13-Musical ensemble / 30-Drummer / 39-Drums / 44-Drums / 3066-Timbales / \n", "0-Games / 2-Video game / 739-Guild Wars / 839-Guild Wars 2 / \n", "12-Food / 226-Dessert / 232-Cake / 578-Cookie / 759-Icing / 3641-Royal icing / \n", "0-Games / 2-Video game / 19-PC game / 43-Call of Duty / 134-Call of Duty: Black Ops / 713-Egg / 1047-Easter egg / 1069-Easter egg / \n", "0-Games / 2-Video game / 6-Animation / 2450-Inazuma Eleven / 3426-Inazuma Eleven GO / \n", "0-Games / 8-Football / 21-Stadium / 73-Ball / 90-Sports game / \n", "3-Concert / 16-Performance art / 68-Lighting / \n", "1-Vehicle / 124-Tractor / 136-Gardening / 156-Agriculture / 363-Soil / 459-Plough / 2079-Two-wheel tractor / \n", "24-String instrument / 642-Leaf / 657-Autumn / 842-Harp / \n", "33-Weapon / 217-Hunting / 2396-Coyote / \n", "3-Concert / 7-Musician / 14-Guitar / 30-Drummer / \n", "7-Musician / 14-Guitar / 24-String instrument / 63-Acoustic guitar / 490-Rain / 878-Composer / \n", "1-Vehicle / 4-Car / 1237-Baby carriage / 1932-Child safety seat / \n", "88-Machine / 261-Tool / 400-Woodturning / 521-Metal / 972-Lathe / \n", "5-Dance / 16-Performance art / \n", "121-Photography / 155-Camera / 278-GoPro / \n", "1-Vehicle / 4-Car / 114-Engine / 124-Tractor / \n", "3-Concert / 5-Dance / 16-Performance art / 191-Cheerleading / \n", "61-Art / 479-Graffiti / \n", "122-Weight training / \n", "14-Guitar / \n", "14-Guitar / 63-Acoustic guitar / \n", "3-Concert / 7-Musician / 13-Musical ensemble / 38-Orchestra / 230-Brass instrument / 427-Trumpet / \n", "0-Games / 2-Video game / 34-Action-adventure game / 1125-God of War / 1432-Kratos / 1743-God of War / 2103-God of War III / \n", "1-Vehicle / 4-Car / 11-Motorsport / 17-Racing / 27-Motorcycle / 57-Race track / 105-Motorcycling / \n", "130-Gymnastics / 1255-Ribbon / \n", "0-Games / 2-Video game / 257-Counter-Strike / 702-Counter-Strike: Source / 826-Source Engine / \n", "1-Vehicle / 4-Car / 11-Motorsport / 17-Racing / 57-Race track / 651-Stock car racing / \n", "9-Music video / 14-Guitar / 23-Mobile phone / 29-Smartphone / \n", "10-Animal / 48-Pet / 71-Dog / 182-Puppy / 2626-Shih Tzu / \n", "414-Computer monitor / \n", "447-Sand / 981-Sculpture / 3282-Sand art and play / \n", "20-Trailer / \n", "0-Games / 1071-Hearthstone: Heroes of Warcraft / \n", "2-Video game / 47-Personal computer / 513-Computer keyboard / \n", "1-Vehicle / 4-Car / 11-Motorsport / 17-Racing / 41-Sports car / 57-Race track / 70-Driving / 451-Cockpit / 3353-Chevrolet Monte Carlo / \n", "1-Vehicle / 62-Train / 64-Transport / 103-Rail transport / 119-Locomotive / 143-Railroad car / 152-Track / 428-Rail transport modelling / \n", "8-Football / 54-Highlight film / \n", "3-Concert / 7-Musician / \n", "0-Games / 2-Video game / 19-PC game / 35-Minecraft / 315-Knife / \n", "3-Concert / 7-Musician / 13-Musical ensemble / \n", "1-Vehicle / 94-Dashcam / \n", "0-Games / 2-Video game / 19-PC game / 951-Teenage Mutant Ninja Turtles / \n", "183-Manga / 2065-Tsubasa: Reservoir Chronicle / \n", "278-GoPro / \n", "46-Choir / \n", "1-Vehicle / 4-Car / 61-Art / \n", "154-Wheel / \n", "2-Video game / 47-Personal computer / 228-Laptop / 2481-Cooler Master / \n", "6-Animation / 15-Cartoon / 1866-Katara / \n", "0-Games / 60-Basketball / 168-Basketball moves / \n", "1-Vehicle / 4-Car / 11-Motorsport / 17-Racing / 378-Drag racing / \n", "6-Animation / 701-Beyblade / 3868-Beyblade: Shogun Steel / \n", "0-Games / 2-Video game / 116-Pokémon / 186-Pokémon / 376-Flute / \n", "30-Drummer / 39-Drums / 44-Drums / \n", "5-Dance / 46-Choir / 49-School / \n", "12-Food / \n", "5-Dance / \n", "14-Guitar / 95-Talent show / \n", "3-Concert / 7-Musician / \n", "3-Concert / 7-Musician / 13-Musical ensemble / 14-Guitar / 24-String instrument / 30-Drummer / 39-Drums / \n", "1-Vehicle / 4-Car / 11-Motorsport / 34-Action-adventure game / 201-Grand Theft Auto: San Andreas / \n", "0-Games / 2-Video game / 33-Weapon / 43-Call of Duty / 302-Call of Duty 4: Modern Warfare / \n", "10-Animal / 42-Fishing / 91-Fish / 893-Fish as food / 1186-Seafood / 2125-Fillet / \n", "75-Wrestling / \n", "3-Concert / 7-Musician / \n", "1-Vehicle / 4-Car / 27-Motorcycle / 114-Engine / \n", "12-Food / 26-Cooking / 32-Recipe / 52-Dish / 58-Cuisine / 109-Cooking show / 184-Vegetable / 500-Rice / 572-Indian cuisine / 3598-Lime / \n", "61-Art / 148-Painting / 936-Watercolor paint / \n", "23-Mobile phone / 37-Gadget / 125-Television / 1565-Logitech / \n", "6-Animation / \n", "3-Concert / 7-Musician / 13-Musical ensemble / 38-Orchestra / 46-Choir / 2241-Plectrum / \n", "28-Fashion / 45-Cosmetics / 173-Eye shadow / 193-Eye liner / \n", "18-Outdoor recreation / 83-Skateboarding / 108-Skateboard / 556-Skateboarding trick / 1315-Kickflip / \n", "22-Nature / 376-Flute / \n", "3-Concert / 7-Musician / 13-Musical ensemble / 46-Choir / 99-Christmas / \n", "83-Skateboarding / 703-Roller skating / 1477-Aggressive inline skating / 1937-Inline skates / 2747-Inline skating / \n", "7-Musician / 13-Musical ensemble / 38-Orchestra / \n", "162-Medicine / \n", "0-Games / 2-Video game / 35-Minecraft / 55-Video game console / 111-PlayStation 3 / 127-Xbox 360 / 176-PlayStation 4 / 181-Xbox / \n", "9-Music video / \n", "1-Vehicle / 4-Car / 345-Loudspeaker / 484-Subwoofer / 618-Fiat Automobiles / 2865-Fiat Punto / \n", "445-Guitar Hero / \n", "0-Games / 2-Video game / 578-Cookie / 1170-Super Mario World / \n", "10-Animal / 12-Food / 26-Cooking / 32-Recipe / 71-Dog / 232-Cake / \n", "1-Vehicle / 27-Motorcycle / 67-Cycling / 69-Bicycle / 520-Bicycle frame / 625-Factory / \n", "3-Concert / 7-Musician / 13-Musical ensemble / 38-Orchestra / 46-Choir / 230-Brass instrument / 634-Trombone / \n", "3175-Hide-and-seek / \n", "1-Vehicle / 4-Car / 11-Motorsport / 17-Racing / 198-Off-road vehicle / 751-Desert / 975-Dune / \n", "3556-Charlie Brown / \n", "1088-Infomercial / \n", "10-Animal / 18-Outdoor recreation / 22-Nature / 275-Diving / 288-Underwater / 381-Underwater diving / 407-Scuba diving / \n", "9-Music video / 78-Wedding / 190-Bride / \n", "7-Musician / \n", "4235-American Kenpo / \n", "0-Games / 2-Video game / 47-Personal computer / 51-Strategy video game / \n", "6-Animation / 33-Weapon / \n", "7-Musician / 38-Orchestra / 150-Violin / \n", "61-Art / 98-Festival / 479-Graffiti / \n", "0-Games / 43-Call of Duty / 1078-Installation art / 2434-Call of Duty / \n", "162-Medicine / 830-Bed / \n", "0-Games / 34-Action-adventure game / 201-Grand Theft Auto: San Andreas / \n", "2-Video game / \n", "98-Festival / \n", "0-Games / 2-Video game / 89-Comics / 185-Dragon Ball / 244-Goku / \n", "0-Games / 2-Video game / 34-Action-adventure game / 201-Grand Theft Auto: San Andreas / \n", "15-Cartoon / 61-Art / 148-Painting / \n", "20-Trailer / \n", "296-Saxophone / 3199-Baritone saxophone / \n", "47-Personal computer / 1125-God of War / \n", "0-Games / 35-Minecraft / 102-Building / \n", "86-Plant / 136-Gardening / 1403-Flower bouquet / \n", "9-Music video / \n", "0-Games / 2-Video game / 55-Video game console / 320-Super Smash Bros. / 443-Wii U / 697-Super Smash Bros. for Nintendo 3DS and Wii U / \n", "3-Concert / 7-Musician / 13-Musical ensemble / 16-Performance art / 38-Orchestra / \n", "45-Cosmetics / 282-Skin / \n", "0-Games / 462-Table / 489-Board game / 922-Tabletop game / \n", "0-Games / 2-Video game / 55-Video game console / 578-Cookie / \n", "1-Vehicle / 4-Car / 41-Sports car / 650-Convertible / 3979-Maserati GranTurismo / \n", "8-Football / \n", "0-Games / 2-Video game / 19-PC game / 35-Minecraft / \n", "1-Vehicle / 4-Car / \n", "5-Dance / 95-Talent show / \n", "238-Nail / 300-Nail art / 304-Nail polish / 324-Finger / 823-Artificial nails / \n", "9-Music video / \n", "0-Games / 2-Video game / 19-PC game / 111-PlayStation 3 / \n", "3-Concert / \n", "1532-Military band / \n", "30-Drummer / 39-Drums / 44-Drums / 128-Cymbal / 146-Snare drum / \n", "84-Snow / 120-Winter / 137-Tree / 578-Cookie / 795-Cupcake / \n", "1-Vehicle / 39-Drums / 477-Transformers / 2835-Fallen / \n", "3-Concert / 7-Musician / \n", "1-Vehicle / 17-Racing / 67-Cycling / 69-Bicycle / 570-Road bicycle racing / \n", "93-Comedy / 355-Comedian / \n", "3-Concert / 7-Musician / 13-Musical ensemble / 46-Choir / 3223-Reindeer / \n", "47-Personal computer / 228-Laptop / 309-Computer hardware / 513-Computer keyboard / \n", "3-Concert / \n", "419-Earth / 448-Planet / 712-Outer space / 1686-Orbit / \n", "7-Musician / 9-Music video / 13-Musical ensemble / 14-Guitar / \n", "3-Concert / \n", "3837-Kyūdō / \n", "0-Games / 8-Football / 21-Stadium / 54-Highlight film / 79-American football / 81-Athlete / \n", "5-Dance / 16-Performance art / 78-Wedding / 190-Bride / \n", "1-Vehicle / 4-Car / \n", "14-Guitar / 24-String instrument / 63-Acoustic guitar / \n", "5-Dance / 9-Music video / 66-Bollywood / \n", "3-Concert / \n", "3-Concert / 5-Dance / 16-Performance art / \n", "1-Vehicle / 92-Radio-controlled model / 149-Model aircraft / 160-Radio-controlled aircraft / 393-Unmanned aerial vehicle / 1311-Amazon.com; Inc. / \n", "0-Games / 2-Video game / 33-Weapon / 43-Call of Duty / 431-Call of Duty: Advanced Warfare / \n", "325-Vampire / \n", "31-Disc jockey / \n", "3-Concert / 68-Lighting / \n", "162-Medicine / 468-Biology / 1596-Cinnamon / \n", "25-Toy / \n", "20-Trailer / \n", "30-Drummer / 39-Drums / 44-Drums / 128-Cymbal / 146-Snare drum / 465-Microphone / \n", "3-Concert / 16-Performance art / \n", "1-Vehicle / 4-Car / \n", "3-Concert / \n", "396-One Piece / \n", "1-Vehicle / 4-Car / 74-Truck / 158-Tire / 349-Chevrolet / 375-Pickup truck / 406-Chevrolet / 1518-Chevrolet Silverado / \n", "3-Concert / 7-Musician / 13-Musical ensemble / 38-Orchestra / 46-Choir / 318-Church / \n", "0-Games / 2-Video game / 35-Minecraft / 807-Seed / 810-Village / \n", "0-Games / 2-Video game / 55-Video game console / 199-Wii / 1220-New Super Mario Bros / \n", "22-Nature / \n", "0-Games / 2-Video game / 6-Animation / 269-Comic book / 460-Yu-Gi-Oh! Trading Card Game / \n", "7-Musician / 14-Guitar / 24-String instrument / 100-Electric guitar / 116-Pokémon / 186-Pokémon / 1254-Cave / \n", "3-Concert / 7-Musician / 30-Drummer / \n", "1737-Dance Dance Revolution / \n", "0-Games / 8-Football / 76-Ball / \n", "325-Vampire / \n", "3-Concert / 7-Musician / \n", "6-Animation / 15-Cartoon / 185-Dragon Ball / 244-Goku / 418-Gohan / 511-Mortal Kombat / 1304-Cell / \n", "369-Resort / \n", "1253-Ballet Dancer / \n", "3-Concert / 7-Musician / 13-Musical ensemble / 30-Drummer / 39-Drums / 230-Brass instrument / \n", "1-Vehicle / 4-Car / 11-Motorsport / 17-Racing / 40-Road / 57-Race track / 70-Driving / 94-Dashcam / \n", "12-Food / 720-Wine / \n", "3579-Mona Lisa / \n", "3-Concert / 7-Musician / 24-String instrument / \n", "1-Vehicle / 4-Car / 234-Ford / 1634-Ford Fiesta / \n", "12-Food / 1576-Doughnut / \n", "38-Orchestra / \n", "0-Games / 6-Animation / \n", "10-Animal / 86-Plant / 91-Fish / 136-Gardening / 258-Aquarium / \n", "6-Animation / 15-Cartoon / 116-Pokémon / 1104-Ash Ketchum / \n", "1-Vehicle / 4-Car / 11-Motorsport / 40-Road / 70-Driving / 94-Dashcam / 515-Dodge / 647-Muscle car / \n", "223-Weather / 419-Earth / 448-Planet / \n", "49-School / 99-Christmas / 908-Video card / \n", "0-Games / 2-Video game / 6-Animation / 15-Cartoon / 185-Dragon Ball / 199-Wii / 244-Goku / 518-Mario Kart / \n", "0-Games / 8-Football / 21-Stadium / 65-Kick / 76-Ball / \n", "3-Concert / 7-Musician / 14-Guitar / 30-Drummer / \n", "66-Bollywood / \n", "5-Dance / 46-Choir / 49-School / \n", "3-Concert / 7-Musician / 16-Performance art / 38-Orchestra / \n", "31-Disc jockey / \n", "0-Games / 2-Video game / 19-PC game / 34-Action-adventure game / 380-Assassin's Creed / \n", "768-Diamond / 2507-Macramé / 3063-Friendship bracelet / \n", "3-Concert / \n", "75-Wrestling / 1042-Jumbotron / \n", "0-Games / 2-Video game / 79-American football / 90-Sports game / 330-Money / 561-Madden NFL / 1813-Madden NFL 13 / 2916-Madden NFL 12 / \n", "365-Fireworks / \n", "60-Basketball / \n", "0-Games / 8-Football / 20-Trailer / 21-Stadium / 77-Arena / \n", "1-Vehicle / 4-Car / 148-Painting / \n", "12-Food / 539-Home appliance / 818-Washing machine / 1142-Laundry / 2838-Cotton / \n", "5-Dance / 9-Music video / \n", "1-Vehicle / 4-Car / 11-Motorsport / 17-Racing / 27-Motorcycle / 105-Motorcycling / 172-Motocross / \n", "1-Vehicle / 4-Car / 11-Motorsport / 17-Racing / 27-Motorcycle / 57-Race track / 1951-Sport bike / 2133-Road racing / \n", "3-Concert / \n", "0-Games / 328-Simulation video game / \n", "308-Ship / 605-Cruise ship / 1402-Suite / \n", "10-Animal / 48-Pet / 132-Bird / 342-Windows Media Video / 364-Sitcom / 598-Parrot / \n", "9-Music video / \n", "83-Skateboarding / 108-Skateboard / 1105-Fingerboard / \n", "28-Fashion / 45-Cosmetics / 238-Nail / \n", "3-Concert / \n", "1-Vehicle / 11-Motorsport / 27-Motorcycle / 67-Cycling / 69-Bicycle / 105-Motorcycling / 565-Mud bogging / \n", "1-Vehicle / 4-Car / 11-Motorsport / 17-Racing / 40-Road / 41-Sports car / 70-Driving / 94-Dashcam / 212-Highway / 332-Mercedes-Benz / \n", "68-Lighting / 113-House / 170-Home improvement / 353-Wall / 879-Ceiling / 3130-Loudspeaker enclosure / \n", "3-Concert / 7-Musician / 39-Drums / \n", "0-Games / 8-Football / 21-Stadium / 54-Highlight film / 79-American football / 81-Athlete / \n", "0-Games / 51-Strategy video game / 269-Comic book / 791-Magic: The Gathering / \n", "25-Toy / 524-Littlest Pet Shop / \n", "0-Games / 8-Football / 21-Stadium / 65-Kick / 1247-Penalty kick / \n", "0-Games / 2-Video game / 19-PC game / 33-Weapon / 43-Call of Duty / 216-Call of Duty: Modern Warfare 3 / \n", "522-Helmet / 1352-Airbrush / 4152-Aerography / \n", "61-Art / 148-Painting / 2538-Painter / \n", "83-Skateboarding / 108-Skateboard / 555-Longboard / 667-Longboarding / \n", "6-Animation / 9-Music video / 15-Cartoon / 538-Illustration / \n", "0-Games / 8-Football / 90-Sports game / 180-FIFA 15 / \n", "12-Food / 26-Cooking / \n", "12-Food / 26-Cooking / 32-Recipe / 52-Dish / 226-Dessert / 295-Chocolate / 1569-Steamed rice / \n", "3-Concert / 7-Musician / 13-Musical ensemble / 16-Performance art / \n", "3-Concert / \n", "3-Concert / 8-Football / 21-Stadium / \n", "10-Animal / 80-Horse / 247-Stallion / 289-Dressage / 290-Pony / \n", "347-Climbing / \n", "1-Vehicle / 4-Car / \n", "1-Vehicle / 50-Aircraft / 64-Transport / 82-Airplane / 139-Airport / 157-Jet aircraft / 175-Airliner / 188-Airline / 254-Runway / 719-Boeing 747 / 923-Boeing 777 / 1529-Jet engine / 2174-Delta Air Lines / \n", "1-Vehicle / 50-Aircraft / 82-Airplane / 92-Radio-controlled model / 147-River / 149-Model aircraft / 160-Radio-controlled aircraft / 225-Lake / \n", "5-Dance / \n", "28-Fashion / 56-Hair / 106-Hairstyle / 1576-Doughnut / 2663-Chignon / \n", "661-Stretching / 920-Chair / \n", "5-Dance / 49-School / \n", "0-Games / 8-Football / 21-Stadium / 54-Highlight film / 65-Kick / 73-Ball / \n", "1-Vehicle / 4-Car / 11-Motorsport / 17-Racing / 57-Race track / 728-Dirt track racing / \n", "10-Animal / 80-Horse / 290-Pony / 404-Livestock / \n", "1-Vehicle / 18-Outdoor recreation / 22-Nature / 67-Cycling / 69-Bicycle / 220-Mountain bike / 284-Mountain biking / \n", "229-Halo / 866-Halo 4 / \n", "3-Concert / 7-Musician / 13-Musical ensemble / 38-Orchestra / 46-Choir / 150-Violin / 726-Clarinet / \n", "538-Illustration / \n", "0-Games / 1-Vehicle / 2-Video game / 4-Car / 11-Motorsport / 17-Racing / \n", "28-Fashion / 45-Cosmetics / 99-Christmas / \n", "1-Vehicle / 4-Car / 154-Wheel / 158-Tire / \n", "0-Games / 19-PC game / 1071-Hearthstone: Heroes of Warcraft / \n", "209-University / \n", "45-Cosmetics / 173-Eye shadow / \n", "3-Concert / 7-Musician / 14-Guitar / \n", "122-Weight training / 529-Squat / 4483-Burpee / \n", "18-Outdoor recreation / 59-Winter sport / 267-Mountain / \n", "1-Vehicle / 4-Car / 11-Motorsport / 70-Driving / 94-Dashcam / 135-Drifting / 683-Renault / \n", "248-Hotel / \n", "30-Drummer / 39-Drums / 44-Drums / 128-Cymbal / 146-Snare drum / \n", "5-Dance / \n", "1-Vehicle / 4-Car / 114-Engine / 3934-Radiator / \n", "9-Music video / 911-Studio / \n", "0-Games / 2-Video game / 19-PC game / 686-Diablo III / 872-Diablo / \n", "1-Vehicle / 22-Nature / 124-Tractor / 136-Gardening / 156-Agriculture / 174-Farm / 459-Plough / 542-Potato / 868-Tram / \n", "1-Vehicle / 4-Car / 11-Motorsport / 17-Racing / 27-Motorcycle / 172-Motocross / \n", "242-Paper / 370-Origami / 466-Gift / \n", "18-Outdoor recreation / 23-Mobile phone / 29-Smartphone / 72-iPhone / \n", "2-Video game / 2815-F-Zero / \n", "1-Vehicle / 25-Toy / 62-Train / 1279-Plarail / \n", "3-Concert / \n", "12-Food / 1739-Fireplace / \n", "1-Vehicle / 10-Animal / 523-Cattle / \n", "3-Concert / 7-Musician / 13-Musical ensemble / 14-Guitar / 24-String instrument / \n", "2-Video game / 23-Mobile phone / 29-Smartphone / 37-Gadget / 55-Video game console / 178-Samsung Galaxy / 581-PlayStation / 3361-Samsung Galaxy Y / \n", "9-Music video / \n", "0-Games / 2-Video game / 145-Tablet computer / 178-Samsung Galaxy / \n", "1-Vehicle / 4-Car / 114-Engine / 259-Fire / \n", "10-Animal / 48-Pet / 723-Pig / \n", "0-Games / 2-Video game / 55-Video game console / 292-Handheld game console / 320-Super Smash Bros. / 782-Super Smash Bros. Melee / \n", "3-Concert / 5-Dance / \n", "0-Games / 2-Video game / 55-Video game console / 2348-Chrono Trigger / \n", "0-Games / 2-Video game / 685-Luigi / 3900-Mario & Luigi: Superstar Saga / \n", "1955-Letter / \n", "0-Games / 137-Tree / 174-Farm / 211-RuneScape / 4192-Oak / \n", "0-Games / 2-Video game / 834-Combat Arms / \n", "3-Concert / \n", "6-Animation / 15-Cartoon / 243-Sonic the Hedgehog / 251-Sonic the Hedgehog / \n", "1-Vehicle / 62-Train / 64-Transport / 103-Rail transport / 119-Locomotive / 143-Railroad car / 152-Track / 161-Train station / 224-Rapid transit / 691-Cargo / 821-Rail freight transport / 1904-Electric locomotive / \n", "624-Number / \n", "89-Comics / 2368-Marvel Universe / \n", "0-Games / 372-Playing card / 494-Card game / \n", "3-Concert / 5-Dance / 133-Nightclub / \n", "10-Animal / 48-Pet / 537-Zoo / 571-Bear / 1160-Giant panda / \n", "59-Winter sport / 112-Ice skating / 142-Hockey / \n", "0-Games / 2-Video game / 19-PC game / 20-Trailer / 1232-Perfect World / 4467-Jade Dynasty / \n", "0-Games / 326-Metin2 / \n", "28-Fashion / 45-Cosmetics / 56-Hair / 173-Eye shadow / 193-Eye liner / 195-Mascara / 196-Eye / 218-Eyelash / 235-Lipstick / 291-Rouge / 436-Concealer / \n", "75-Wrestling / \n", "31-Disc jockey / \n", "0-Games / 8-Football / 76-Ball / \n", "110-Album / \n", "10-Animal / 80-Horse / 247-Stallion / \n", "3-Concert / \n", "9-Music video / \n", "0-Games / 2-Video game / 19-PC game / 34-Action-adventure game / 118-Grand Theft Auto V / 127-Xbox 360 / 176-PlayStation 4 / 330-Money / \n", "6-Animation / 10-Animal / 15-Cartoon / 126-The Walt Disney Company / 2089-Winnie-the-Pooh / \n", "0-Games / 47-Personal computer / 201-Grand Theft Auto: San Andreas / \n", "22-Nature / 144-Amusement park / 147-River / \n", "12-Food / \n", "6-Animation / 15-Cartoon / 1563-Stick figure / \n", "3-Concert / 7-Musician / 16-Performance art / \n", "10-Animal / 12-Food / 26-Cooking / 52-Dish / \n", "1-Vehicle / 4-Car / 70-Driving / 94-Dashcam / 2214-Multi-valve / 2989-Opel Vectra / \n", "10-Animal / 17-Racing / 80-Horse / 247-Stallion / 384-Horse racing / \n", "159-World of Warcraft / \n", "45-Cosmetics / 173-Eye shadow / 195-Mascara / 196-Eye / 291-Rouge / 436-Concealer / \n", "1-Vehicle / 87-Boat / 147-River / 225-Lake / 1358-Canyon / \n", "468-Biology / \n", "10-Animal / 588-Reptile / 1444-Lizard / 3060-Chameleon / \n", "1-Vehicle / 50-Aircraft / 82-Airplane / 129-Aviation / 151-Landing / 157-Jet aircraft / 187-Takeoff / 388-Fighter aircraft / 758-Air force / \n", "0-Games / 2-Video game / 34-Action-adventure game / 201-Grand Theft Auto: San Andreas / 2402-Grand Theft Auto / 3226-Chucky / \n", "49-School / 410-Teacher / 466-Gift / 2914-Wreath / 4129-Crayon / \n", "0-Games / 2-Video game / 199-Wii / 443-Wii U / \n", "122-Weight training / 529-Squat / 661-Stretching / \n", "3-Concert / 559-Night / \n", "0-Games / 8-Football / 21-Stadium / \n", "0-Games / 51-Strategy video game / 165-League of Legends / \n", "20-Trailer / \n", "93-Comedy / \n", "23-Mobile phone / 29-Smartphone / 72-iPhone / \n", "5-Dance / 16-Performance art / \n", "1-Vehicle / 27-Motorcycle / 154-Wheel / 158-Tire / 336-Scooter / 1448-Vespa / \n", "1-Vehicle / 27-Motorcycle / 3004-Softail / \n", "1-Vehicle / 62-Train / 64-Transport / 103-Rail transport / 119-Locomotive / 143-Railroad car / 152-Track / 161-Train station / \n", "1-Vehicle / 4-Car / \n", "14-Guitar / \n", "5-Dance / \n", "5-Dance / \n", "22-Nature / \n", "3-Concert / \n", "0-Games / 2-Video game / 240-Arcade game / 1990-Pac-Man / \n", "66-Bollywood / \n", "9-Music video / \n", "93-Comedy / \n", "0-Games / 2-Video game / 4-Car / 34-Action-adventure game / 118-Grand Theft Auto V / \n", "10-Animal / 48-Pet / 71-Dog / 132-Bird / 182-Puppy / 598-Parrot / 2626-Shih Tzu / \n", "93-Comedy / \n", "1-Vehicle / 4-Car / 332-Mercedes-Benz / 891-Mercedes-AMG / 3915-Mercedes-Benz CLS-Class / \n", "0-Games / 2-Video game / 50-Aircraft / 55-Video game console / 2101-Ace Combat / \n", "8-Football / \n", "2-Video game / 23-Mobile phone / 29-Smartphone / 72-iPhone / 272-iPod / 331-iPod touch / \n", "3-Concert / 7-Musician / 13-Musical ensemble / 38-Orchestra / 46-Choir / 203-Marching band / \n", "1-Vehicle / 50-Aircraft / 82-Airplane / 129-Aviation / 139-Airport / 157-Jet aircraft / 175-Airliner / 188-Airline / 254-Runway / 472-Boeing 737 / \n", "30-Drummer / 39-Drums / 44-Drums / 128-Cymbal / 146-Snare drum / \n", "162-Medicine / \n", "0-Games / 2-Video game / 19-PC game / 35-Minecraft / 51-Strategy video game / 1120-Tower defense / \n", "5-Dance / 16-Performance art / 115-Ballet / \n", "12-Food / \n", "3-Concert / 7-Musician / 16-Performance art / \n", "10-Animal / 48-Pet / 132-Bird / \n", "5-Dance / 16-Performance art / 115-Ballet / \n", "0-Games / 2-Video game / 19-PC game / 33-Weapon / 43-Call of Duty / 96-Soldier / 134-Call of Duty: Black Ops / 138-Call of Duty: Black Ops II / \n", "6-Animation / 15-Cartoon / 20-Trailer / \n", "10-Animal / 71-Dog / \n", "1-Vehicle / 4-Car / 94-Dashcam / 155-Camera / 559-Night / \n", "809-Logo / \n", "23-Mobile phone / 29-Smartphone / 3577-Walkman / \n", "3-Concert / \n", "0-Games / 2-Video game / 90-Sports game / 180-FIFA 15 / 561-Madden NFL / \n", "1143-Marimba / 1734-Xylophone / 2878-Vibraphone / \n", "6-Animation / 15-Cartoon / 1741-Pineapple / \n", "1266-Death Note / \n", "79-American football / \n", "3-Concert / 5-Dance / \n", "369-Resort / \n", "30-Drummer / 39-Drums / 44-Drums / 128-Cymbal / 146-Snare drum / \n", "16-Performance art / 95-Talent show / \n", "9-Music video / \n", "3-Concert / 7-Musician / 30-Drummer / \n", "1-Vehicle / 4-Car / 40-Road / 41-Sports car / 70-Driving / 212-Highway / 245-BMW / 344-Coupé / 650-Convertible / 1324-Roadster / 3373-BMW Z4 / \n", "1-Vehicle / 10-Animal / 18-Outdoor recreation / 42-Fishing / 87-Boat / 91-Fish / \n", "1-Vehicle / 50-Aircraft / 82-Airplane / 1263-Glider / \n", "36-Piano / 53-Keyboard / 265-Electronic keyboard / 2773-iPod nano / \n", "0-Games / 2-Video game / 6-Animation / 15-Cartoon / 89-Comics / 185-Dragon Ball / 244-Goku / 418-Gohan / 869-Dragon Ball Z: Budokai Tenkaichi / 876-Freeza / \n", "3-Concert / \n", "260-Prayer / \n", "1-Vehicle / 4-Car / 11-Motorsport / 17-Racing / \n", "9-Music video / 322-Mixtape / \n", "7-Musician / 14-Guitar / 24-String instrument / \n", "0-Games / 2-Video game / 6-Animation / 816-Gundam / 3813-Dynasty Warriors: Gundam / \n", "84-Snow / 120-Winter / \n", "14-Guitar / 24-String instrument / 100-Electric guitar / \n", "0-Games / 2-Video game / 441-The Elder Scrolls / 1410-The Elder Scrolls IV: Oblivion / \n", "0-Games / 75-Wrestling / \n", "1-Vehicle / 4-Car / 47-Personal computer / 1996-Intel Core i7 / \n", "1-Vehicle / 11-Motorsport / 27-Motorcycle / 105-Motorcycling / 207-Exhaust system / 2892-GSX-R750 / \n", "101-Telephone / 178-Samsung Galaxy / 899-Samsung Galaxy S III / \n", "9-Music video / 541-Black-and-white / \n", "8-Football / \n", "3355-Bingo / \n", "67-Cycling / 69-Bicycle / \n", "120-Winter / 657-Autumn / \n", "1-Vehicle / 4-Car / 406-Chevrolet / \n", "6-Animation / 15-Cartoon / \n", "322-Mixtape / \n", "13-Musical ensemble / 820-Bagpipes / 1646-Pipe band / \n", "8-Football / \n", "5-Dance / \n", "5-Dance / \n", "3-Concert / 5-Dance / 16-Performance art / \n", "5-Dance / 67-Cycling / 1997-Unicycle / \n", "3-Concert / 7-Musician / 13-Musical ensemble / 545-Quartet (ensemble) / \n", "9-Music video / \n", "6-Animation / 15-Cartoon / \n", "0-Games / 398-The Legend of Zelda / 1228-The Legend of Zelda: Ocarina of Time / \n", "1-Vehicle / 4-Car / 519-Recreational vehicle / \n", "0-Games / 2-Video game / 2071-Indiana Jones / \n", "163-Running / \n", "3-Concert / 7-Musician / 14-Guitar / 30-Drummer / \n", "31-Disc jockey / \n", "0-Games / 2-Video game / 51-Strategy video game / 831-Plants vs. Zombies / 1120-Tower defense / \n", "12-Food / 232-Cake / 1641-Cheesecake / \n", "3-Concert / \n", "7-Musician / 13-Musical ensemble / 14-Guitar / 30-Drummer / 419-Earth / 448-Planet / 593-Moon / 654-Sun / 712-Outer space / \n", "0-Games / 2-Video game / 35-Minecraft / 394-Map / \n", "0-Games / 1680-Seafight / \n", "23-Mobile phone / 29-Smartphone / 37-Gadget / 101-Telephone / 178-Samsung Galaxy / 899-Samsung Galaxy S III / 1207-Samsung Galaxy S5 / 3648-Samsung Galaxy S III Mini / \n", "5-Dance / 49-School / \n", "3-Concert / 7-Musician / 13-Musical ensemble / \n", "0-Games / 8-Football / 21-Stadium / 73-Ball / \n", "2-Video game / \n", "45-Cosmetics / 282-Skin / \n", "9-Music video / \n", "2122-Compact Cassette / \n", "0-Games / 2-Video game / 19-PC game / 43-Call of Duty / 134-Call of Duty: Black Ops / 138-Call of Duty: Black Ops II / \n", "5-Dance / 95-Talent show / 115-Ballet / \n", "25-Toy / 140-LEGO / \n", "9-Music video / \n", "3-Concert / 5-Dance / \n", "0-Games / 2-Video game / 19-PC game / 192-Battlefield / 383-Battlefield 4 / \n", "6-Animation / 15-Cartoon / 20-Trailer / \n", "3-Concert / 7-Musician / \n", "303-Paint / 890-Brush / \n", "330-Money / \n", "12-Food / 26-Cooking / 389-Restaurant / \n", "141-Microsoft Windows / \n", "3-Concert / 16-Performance art / \n", "272-iPod / 497-Macintosh / \n", "10-Animal / 80-Horse / 247-Stallion / \n", "0-Games / 2-Video game / 34-Action-adventure game / 434-Ninja / 682-PlayStation Vita / 2192-Ninja Gaiden / \n", "117-Boxing / \n", "72-iPhone / 272-iPod / 331-iPod touch / \n", "222-Surfing / \n", "28-Fashion / 1082-Knot / 1369-Necktie / \n", "7-Musician / 14-Guitar / 24-String instrument / 63-Acoustic guitar / 329-Swimming pool / 639-Fiddle / 670-Banjo / 1231-Mandolin / \n", "12-Food / \n", "118-Grand Theft Auto V / 615-Poker / \n", "76-Ball / 204-Gym / \n", "12-Food / 585-Lawn / 2437-String trimmer / \n", "1113-Toilet / \n", "0-Games / 2-Video game / 34-Action-adventure game / 201-Grand Theft Auto: San Andreas / \n", "3-Concert / \n", "7-Musician / 9-Music video / \n", "1449-Rubber band / \n", "0-Games / 2-Video game / \n", "0-Games / 1-Vehicle / 2-Video game / 4-Car / 34-Action-adventure game / 297-Grand Theft Auto IV / 658-Grand Theft Auto: The Lost and Damned / 1923-Niko Bellic / \n", "1-Vehicle / 4-Car / 11-Motorsport / 40-Road / 41-Sports car / 70-Driving / 94-Dashcam / 212-Highway / 1040-Street racing / 1635-Ferrari 458 / \n", "1023-Rose / \n", "10-Animal / 80-Horse / 438-Mare / \n", "1-Vehicle / 4-Car / 11-Motorsport / 252-Tractor pulling / \n", "14-Guitar / \n", "2-Video game / 6-Animation / 229-Halo / 866-Halo 4 / 2889-343 Industries / \n", "46-Choir / 260-Prayer / 318-Church / 860-Christian Church / \n", "1-Vehicle / 50-Aircraft / 82-Airplane / 129-Aviation / 187-Takeoff / \n", "61-Art / 97-Drawing / 148-Painting / 1508-Spacecraft / \n", "215-Eating / 2411-Howcast / \n", "3-Concert / 5-Dance / \n", "0-Games / 2-Video game / 55-Video game console / 335-Nintendo 3DS / 1108-Mushroom / 1220-New Super Mario Bros / 3943-New Super Mario Bros. 2 / \n", "0-Games / 2-Video game / 240-Arcade game / 708-Mermaid / 1876-Beatmania IIDX / 2063-Beatmania / \n", "22-Nature / \n", "1-Vehicle / 4-Car / 11-Motorsport / \n", "1-Vehicle / 124-Tractor / 156-Agriculture / 174-Farm / 249-Heavy equipment / 363-Soil / 459-Plough / 660-Rural area / \n", "59-Winter sport / 112-Ice skating / 142-Hockey / 234-Ford / \n", "85-Combat / 117-Boxing / \n", "95-Talent show / \n", "0-Games / 35-Minecraft / \n", "3-Concert / 14-Guitar / 98-Festival / 594-Music festival / \n", "0-Games / 1-Vehicle / 2-Video game / 413-PlayStation 2 / 1715-Aircraft carrier / \n", "12-Food / 136-Gardening / 174-Farm / 1108-Mushroom / \n", "8-Football / \n", "3-Concert / \n", "3-Concert / 38-Orchestra / \n", "5-Dance / 3092-Cinderella / \n", "5-Dance / 60-Basketball / \n", "3-Concert / 5-Dance / 16-Performance art / \n", "28-Fashion / 696-T-shirt / 1534-H&M / \n", "6-Animation / 15-Cartoon / \n", "6-Animation / 15-Cartoon / \n", "75-Wrestling / \n", "9-Music video / \n", "1-Vehicle / 4-Car / 11-Motorsport / 40-Road / 41-Sports car / 70-Driving / 94-Dashcam / 245-BMW / \n", "410-Teacher / \n", "3-Concert / 16-Performance art / \n", "5-Dance / 28-Fashion / 121-Photography / \n", "458-Gold / \n", "3-Concert / \n", "9-Music video / \n", "162-Medicine / 1357-Ceramic / \n", "1-Vehicle / 4-Car / 234-Ford / 1395-Valve / \n", "2397-Cabbage / \n", "89-Comics / 294-Batman / 387-Spider-Man / 630-Superman / 2082-Catwoman / \n", "5-Dance / \n", "3-Concert / 14-Guitar / \n", "36-Piano / 53-Keyboard / 104-Pianist / 107-Musical keyboard / 265-Electronic keyboard / \n", "5-Dance / 16-Performance art / 115-Ballet / \n", "5-Dance / \n", "61-Art / 97-Drawing / \n", "9-Music video / \n", "7-Musician / 13-Musical ensemble / 14-Guitar / 24-String instrument / 46-Choir / \n", "2-Video game / 43-Call of Duty / 302-Call of Duty 4: Modern Warfare / \n", "1-Vehicle / 4-Car / \n", "1-Vehicle / 12-Food / 1793-Taco / \n", "9-Music video / \n", "1-Vehicle / 4-Car / 11-Motorsport / 135-Drifting / \n", "3-Concert / 68-Lighting / \n", "12-Food / 22-Nature / 86-Plant / 227-Water / \n", "0-Games / 395-Camping / \n", "159-World of Warcraft / 171-Warcraft / \n", "12-Food / 26-Cooking / 32-Recipe / 52-Dish / 58-Cuisine / 162-Medicine / 215-Eating / \n", "0-Games / \n", "139-Airport / 1061-Marriage proposal / \n", "6-Animation / 15-Cartoon / \n", "0-Games / 2-Video game / 19-PC game / 127-Xbox 360 / \n", "12-Food / 26-Cooking / 32-Recipe / 52-Dish / 58-Cuisine / 236-Baking / 274-Meat / 485-Dough / \n", "6-Animation / \n", "47-Personal computer / 131-Computer / \n", "164-Beach / 369-Resort / \n", "0-Games / 2-Video game / \n", "3-Concert / 7-Musician / 13-Musical ensemble / 38-Orchestra / \n", "0-Games / 8-Football / \n", "5-Dance / \n", "6-Animation / 15-Cartoon / 325-Vampire / \n", "12-Food / 26-Cooking / 32-Recipe / 52-Dish / 58-Cuisine / 109-Cooking show / 210-Cookware and bakeware / 239-Roasting / 274-Meat / 2650-Frying pan / \n", "23-Mobile phone / 29-Smartphone / 37-Gadget / 101-Telephone / 141-Microsoft Windows / 932-Microsoft Lumia / \n", "1-Vehicle / 4-Car / 11-Motorsport / 17-Racing / 57-Race track / 378-Drag racing / 1916-Buick / \n", "0-Games / 2-Video game / 19-PC game / 596-Monster Hunter / \n", "10-Animal / 48-Pet / 71-Dog / 4389-Basset Hound / \n", "6-Animation / 15-Cartoon / \n", "5-Dance / 16-Performance art / \n", "2-Video game / 34-Action-adventure game / 1396-Lara Croft / \n", "18-Outdoor recreation / 22-Nature / 275-Diving / 407-Scuba diving / 986-Suit / 4185-Wetsuit / \n", "28-Fashion / 115-Ballet / 2211-Fur / \n", "1-Vehicle / 4-Car / 114-Engine / 3187-Spark plug / \n", "79-American football / \n", "164-Beach / 2275-Beach volleyball / \n", "5-Dance / \n", "762-Angel / \n", "1551-The Phantom of the Opera / 3904-Miss Saigon / 4443-Jekyll & Hyde / \n", "117-Boxing / \n", "1-Vehicle / 4-Car / 154-Wheel / 403-Toyota / 2104-Toyota Camry / \n", "36-Piano / 53-Keyboard / \n", "28-Fashion / 45-Cosmetics / 1459-Magazine / \n", "6-Animation / 20-Trailer / \n", "125-Television / \n", "322-Mixtape / \n", "0-Games / 1-Vehicle / 2-Video game / 4-Car / 11-Motorsport / 17-Racing / 27-Motorcycle / 40-Road / 70-Driving / 201-Grand Theft Auto: San Andreas / 212-Highway / \n", "3-Concert / 16-Performance art / \n", "12-Food / 26-Cooking / 52-Dish / 58-Cuisine / 389-Restaurant / \n", "0-Games / 1802-Jubeat / \n", "5-Dance / 16-Performance art / 95-Talent show / 115-Ballet / 350-Circus / 788-Juggling / \n", "3-Concert / 7-Musician / 30-Drummer / \n", "3-Concert / 31-Disc jockey / 68-Lighting / 133-Nightclub / \n", "6-Animation / 61-Art / 97-Drawing / 479-Graffiti / \n", "616-The Bible / \n", "1-Vehicle / 4-Car / 245-BMW / 1590-BMW 3 Series (E30) / \n", "273-News program / \n", "0-Games / 8-Football / 410-Teacher / \n", "7-Musician / 14-Guitar / 24-String instrument / \n", "6-Animation / 123-Naruto / 183-Manga / 285-Sasuke Uchiha / 622-Sakura Haruno / \n", "0-Games / 2-Video game / 34-Action-adventure game / \n", "1672-LittleBigPlanet / \n", "3-Concert / 5-Dance / 31-Disc jockey / 133-Nightclub / \n", "23-Mobile phone / 29-Smartphone / 72-iPhone / 189-iPad / 272-iPod / 331-iPod touch / \n", "22-Nature / \n", "1-Vehicle / 84-Snow / 92-Radio-controlled model / 149-Model aircraft / 160-Radio-controlled aircraft / 392-Radio-controlled helicopter / 393-Unmanned aerial vehicle / 604-Quadcopter / \n", "0-Games / 1-Vehicle / 2-Video game / 34-Action-adventure game / 118-Grand Theft Auto V / \n", "0-Games / 1-Vehicle / 2-Video game / 4-Car / 34-Action-adventure game / 40-Road / 41-Sports car / 118-Grand Theft Auto V / \n", "8-Football / \n", "141-Microsoft Windows / 406-Chevrolet / \n", "10-Animal / 48-Pet / 71-Dog / 182-Puppy / \n", "835-Backpack / \n", "0-Games / 4004-Tug of war / \n", "102-Building / 113-House / 170-Home improvement / 276-Room / 580-Bedroom / 652-Bathroom / 879-Ceiling / 888-Living room / \n", "0-Games / \n", "6-Animation / 15-Cartoon / \n", "125-Television / \n", "101-Telephone / \n", "2-Video game / 23-Mobile phone / 29-Smartphone / 37-Gadget / 101-Telephone / 1097-Touchscreen / \n", "0-Games / 60-Basketball / \n", "0-Games / 2-Video game / 159-World of Warcraft / 171-Warcraft / 1213-World of Warcraft: Wrath of the Lich King / \n", "5-Dance / \n", "44-Drums / 2324-Hang / \n", "3-Concert / 295-Chocolate / \n", "273-News program / \n", "1-Vehicle / 11-Motorsport / 18-Outdoor recreation / 27-Motorcycle / 105-Motorcycling / 197-Trail / \n", "28-Fashion / 153-Dress / \n", "12-Food / 26-Cooking / 32-Recipe / 442-Bread / 1943-Bakery / \n", "319-Robot / 492-Floor / \n", "28-Fashion / 56-Hair / 106-Hairstyle / 909-Updo / \n", "242-Paper / 370-Origami / 466-Gift / \n", "2092-Coal / \n", "93-Comedy / \n", "0-Games / 1-Vehicle / 2-Video game / 4-Car / 34-Action-adventure game / 266-Helicopter / 362-PlayStation Portable / 921-Grand Theft Auto: Vice City / 2893-Grand Theft Auto: Vice City Stories / \n", "0-Games / 2-Video game / 4311-Mega Man Star Force / \n", "1-Vehicle / 18-Outdoor recreation / 87-Boat / 225-Lake / 481-Kayak / \n", "311-Street Fighter / 461-Street Fighter IV / \n", "0-Games / 2-Video game / 257-Counter-Strike / \n", "5-Dance / 9-Music video / \n", "7-Musician / 8-Football / 13-Musical ensemble / 21-Stadium / 79-American football / 203-Marching band / 209-University / \n", "0-Games / 2-Video game / 34-Action-adventure game / 111-PlayStation 3 / 118-Grand Theft Auto V / \n", "36-Piano / 53-Keyboard / 104-Pianist / \n", "6-Animation / 15-Cartoon / \n", "16-Performance art / 115-Ballet / 361-Flamenco / \n", "9-Music video / \n", "219-Accordion / \n", "3-Concert / \n", "0-Games / 2-Video game / 116-Pokémon / 186-Pokémon / \n", "2627-Little Red Riding Hood / \n", "9-Music video / \n", "1-Vehicle / 4-Car / 27-Motorcycle / 67-Cycling / 69-Bicycle / \n", "28-Fashion / 45-Cosmetics / \n", "0-Games / 8-Football / 79-American football / \n", "1-Vehicle / 4-Car / 11-Motorsport / 17-Racing / 41-Sports car / 57-Race track / 711-Gran Turismo / 953-Gran Turismo 5 / \n", "10-Animal / 15-Cartoon / 116-Pokémon / 596-Monster Hunter / \n", "31-Disc jockey / 861-Mixing console / \n", "60-Basketball / 256-Tennis / \n", "0-Games / 59-Winter sport / 112-Ice skating / 316-Figure skating / 351-Jumping / \n", "10-Animal / 22-Nature / 147-River / \n", "3-Concert / \n", "22-Nature / 267-Mountain / 283-Rock / 800-Mountain pass / \n", "6-Animation / 25-Toy / 126-The Walt Disney Company / \n", "97-Drawing / 880-Trousers / 1627-Shorts / \n", "20-Trailer / \n", "14-Guitar / 36-Piano / \n", "371-Firefighter / \n", "0-Games / 59-Winter sport / 81-Athlete / 112-Ice skating / 142-Hockey / 312-Ice rink / \n", "5-Dance / \n", "30-Drummer / 39-Drums / 44-Drums / 128-Cymbal / 146-Snare drum / 1127-Drum stick / \n", "83-Skateboarding / 108-Skateboard / 221-Skatepark / \n", "3-Concert / 7-Musician / 13-Musical ensemble / 38-Orchestra / \n", "0-Games / 2-Video game / 116-Pokémon / \n", "12-Food / 452-Watch / \n", "9-Music video / \n", "18-Outdoor recreation / 22-Nature / 287-Human swimming / 329-Swimming pool / \n", "5-Dance / 38-Orchestra / \n", "5-Dance / \n", "8-Football / \n", "1-Vehicle / 4-Car / 332-Mercedes-Benz / 1119-Sensor / 2603-Mercedes-Benz E-Class / \n", "0-Games / 116-Pokémon / 186-Pokémon / \n", "9-Music video / \n", "5-Dance / 16-Performance art / \n", "208-Wood / 261-Tool / 400-Woodturning / 1088-Infomercial / \n", "0-Games / 2-Video game / 43-Call of Duty / 138-Call of Duty: Black Ops II / \n", "10-Animal / 48-Pet / 71-Dog / 182-Puppy / \n", "5-Dance / \n", "8-Football / \n", "1-Vehicle / 4-Car / 11-Motorsport / 88-Machine / 114-Engine / 459-Plough / \n", "0-Games / 2-Video game / \n", "47-Personal computer / 228-Laptop / 525-Printer / 696-T-shirt / 1536-Woven fabric / \n", "0-Games / 59-Winter sport / 81-Athlete / 112-Ice skating / 142-Hockey / 312-Ice rink / \n", "36-Piano / 446-Sheet music / \n", "1-Vehicle / 11-Motorsport / 114-Engine / 124-Tractor / 252-Tractor pulling / \n", "3247-Dome / \n", "0-Games / 8-Football / 90-Sports game / 163-Running / 180-FIFA 15 / \n", "31-Disc jockey / \n", "9-Music video / 14-Guitar / 24-String instrument / \n", "237-Final Fantasy / 730-Final Fantasy VII / \n", "1-Vehicle / 50-Aircraft / 82-Airplane / 139-Airport / 151-Landing / 157-Jet aircraft / 175-Airliner / 187-Takeoff / 188-Airline / 856-Airbus A320 family / 4148-Pre-flight safety demonstration / \n", "8-Football / \n", "0-Games / 51-Strategy video game / 165-League of Legends / \n", "6-Animation / 15-Cartoon / 61-Art / 97-Drawing / 616-The Bible / 1075-Whale / \n", "0-Games / 2-Video game / 19-PC game / 33-Weapon / 96-Soldier / 257-Counter-Strike / \n", "0-Games / 2-Video game / 229-Halo / \n", "3-Concert / 8-Football / 21-Stadium / 60-Basketball / \n", "337-Advertising / 798-Shampoo / \n", "39-Drums / 44-Drums / 2030-Djembe / \n", "0-Games / 5-Dance / 2123-Just Dance / \n", "3-Concert / \n", "28-Fashion / 793-Shirt / 1100-Hat / \n", "28-Fashion / 56-Hair / 1271-Beauty salon / \n", "5-Dance / 16-Performance art / \n", "6-Animation / 15-Cartoon / 2231-Toontown Online / \n", "0-Games / 8-Football / 90-Sports game / 180-FIFA 15 / \n", "5-Dance / 892-Aerobics / \n", "0-Games / 8-Football / 205-Newscaster / 699-Kick / \n", "3-Concert / 7-Musician / 13-Musical ensemble / 14-Guitar / 24-String instrument / \n", "45-Cosmetics / 238-Nail / 873-Acrylic paint / \n", "9-Music video / \n", "12-Food / 26-Cooking / 32-Recipe / 52-Dish / 58-Cuisine / 162-Medicine / 808-Breakfast / \n", "0-Games / 21-Stadium / 54-Highlight film / 60-Basketball / 77-Arena / 81-Athlete / 168-Basketball moves / 255-Slam dunk / \n", "0-Games / 2-Video game / 19-PC game / 51-Strategy video game / 1570-Command & Conquer / 4526-Command & Conquer: Generals / \n", "56-Hair / 106-Hairstyle / 1320-Bangs / \n", "2401-Wolfenstein / \n", "0-Games / 6-Animation / 3367-Bart Simpson / \n", "18-Outdoor recreation / 22-Nature / \n", "12-Food / 2992-Energy drink / \n", "66-Bollywood / \n", "0-Games / 2-Video game / 55-Video game console / 127-Xbox 360 / 181-Xbox / 199-Wii / 229-Halo / 443-Wii U / 866-Halo 4 / 963-Left 4 Dead 2 / 1140-Assassin's Creed III / \n", "1-Vehicle / 18-Outdoor recreation / 87-Boat / 949-Rowing / \n", "25-Toy / 1017-Yo-yo / \n", "0-Games / 2-Video game / 1144-Castlevania / \n", "18-Outdoor recreation / 22-Nature / 147-River / 481-Kayak / 1195-Rafting / \n", "1-Vehicle / 18-Outdoor recreation / 42-Fishing / 87-Boat / 1135-Dolphin / \n", "3-Concert / 13-Musical ensemble / 16-Performance art / 38-Orchestra / \n", "0-Games / 1-Vehicle / 2-Video game / 4-Car / 17-Racing / 34-Action-adventure game / 201-Grand Theft Auto: San Andreas / 352-Need for Speed / \n", "6-Animation / 89-Comics / 750-Sailor Moon / \n", "5-Dance / \n", "1-Vehicle / 4-Car / 94-Dashcam / 1134-Taxicab / \n", "1-Vehicle / 87-Boat / 147-River / 222-Surfing / 481-Kayak / \n", "3-Concert / 7-Musician / 13-Musical ensemble / 24-String instrument / 36-Piano / 53-Keyboard / 150-Violin / 450-Viola / 1805-Carousel / \n", "223-Weather / 394-Map / 419-Earth / 448-Planet / 593-Moon / 654-Sun / 1145-Winter storm / 1918-Comet / 3446-Solar flare / \n", "1-Vehicle / 22-Nature / 62-Train / 64-Transport / 103-Rail transport / 119-Locomotive / 428-Rail transport modelling / \n", "10-Animal / 71-Dog / \n", "3-Concert / \n", "49-School / 102-Building / \n", "10-Animal / 71-Dog / 80-Horse / 2438-French Bulldog / \n", "25-Toy / 126-The Walt Disney Company / 417-Play-Doh / \n", "6-Animation / 432-Kingdom Hearts / 2801-Characters of Kingdom Hearts / 3299-Kingdom Hearts: Chain of Memories / \n", "9-Music video / \n", "66-Bollywood / \n", "1-Vehicle / 259-Fire / 371-Firefighter / \n", "34-Action-adventure game / 641-Mask / \n", "0-Games / 1-Vehicle / 2-Video game / 4-Car / 34-Action-adventure game / 297-Grand Theft Auto IV / 658-Grand Theft Auto: The Lost and Damned / \n", "1-Vehicle / 5-Dance / 11-Motorsport / 124-Tractor / 252-Tractor pulling / \n", "1-Vehicle / 62-Train / \n", "3-Concert / 7-Musician / 16-Performance art / \n", "9-Music video / \n", "3-Concert / \n", "1-Vehicle / 4-Car / 18-Outdoor recreation / 92-Radio-controlled model / 167-Four-wheel drive / 202-Radio-controlled car / 926-Traxxas / \n", "1-Vehicle / 4-Car / \n", "3-Concert / 5-Dance / 10-Animal / 71-Dog / \n", "1-Vehicle / 50-Aircraft / 82-Airplane / 139-Airport / 151-Landing / 157-Jet aircraft / 175-Airliner / 187-Takeoff / 188-Airline / 223-Weather / 254-Runway / \n", "330-Money / 458-Gold / 521-Metal / 731-Coin / 912-Silver / 1426-Treasure / \n", "20-Trailer / \n", "78-Wedding / 153-Dress / 190-Bride / \n", "75-Wrestling / \n", "0-Games / 2-Video game / 19-PC game / 43-Call of Duty / 134-Call of Duty: Black Ops / 138-Call of Duty: Black Ops II / \n", "242-Paper / \n", "9-Music video / \n", "3-Concert / 16-Performance art / 68-Lighting / 98-Festival / 594-Music festival / \n", "8-Football / \n", "4185-Wetsuit / \n", "479-Graffiti / \n", "1-Vehicle / 395-Camping / 914-Tent / \n", "1-Vehicle / 4-Car / 25-Toy / 74-Truck / 92-Radio-controlled model / 202-Radio-controlled car / \n", "10-Animal / 48-Pet / 71-Dog / 1377-Golden Retriever / \n", "3-Concert / 7-Musician / 16-Performance art / 594-Music festival / \n", "10-Animal / 12-Food / 713-Egg / \n", "6-Animation / 25-Toy / 140-LEGO / 2312-SpongeBob SquarePants / \n", "18-Outdoor recreation / 83-Skateboarding / 108-Skateboard / 221-Skatepark / 556-Skateboarding trick / \n", "1-Vehicle / 67-Cycling / 69-Bicycle / 520-Bicycle frame / 645-Road bicycle / \n", "6-Animation / 1523-Alice / 2201-Alice's Adventures in Wonderland / 3342-The Mad Hatter / \n", "0-Games / 1-Vehicle / 2-Video game / 4-Car / 11-Motorsport / 17-Racing / 41-Sports car / 352-Need for Speed / 3252-Burnout Paradise / \n", "1-Vehicle / 4-Car / 114-Engine / 299-Volkswagen Passenger Cars / \n", "0-Games / 2-Video game / 33-Weapon / 2379-Tom Clancy's Ghost Recon / 3423-Tom Clancy's Ghost Recon: Future Soldier / \n", "46-Choir / \n", "1-Vehicle / 18-Outdoor recreation / 42-Fishing / 87-Boat / 268-Recreational fishing / 709-Bull / \n", "28-Fashion / 350-Circus / \n", "447-Sand / 751-Desert / 2410-Camel / \n", "0-Games / 8-Football / 21-Stadium / \n", "1-Vehicle / 11-Motorsport / 87-Boat / 2056-Battleship / \n", "6-Animation / 61-Art / 99-Christmas / 447-Sand / 3282-Sand art and play / 3550-Sand animation / \n", "110-Album / 4427-Humbucker / \n", "0-Games / 60-Basketball / 168-Basketball moves / \n", "118-Grand Theft Auto V / \n", "0-Games / 2-Video game / 19-PC game / 33-Weapon / 96-Soldier / 192-Battlefield / 383-Battlefield 4 / \n", "125-Television / 205-Newscaster / \n", "3-Concert / 5-Dance / 16-Performance art / \n", "0-Games / 59-Winter sport / 81-Athlete / 112-Ice skating / 142-Hockey / \n", "22-Nature / 42-Fishing / 164-Beach / 222-Surfing / 233-Ocean / 423-Coast / 672-Bodyboarding / \n", "7-Musician / 14-Guitar / 24-String instrument / 100-Electric guitar / \n", "3-Concert / 7-Musician / 16-Performance art / 30-Drummer / 39-Drums / 44-Drums / \n", "23-Mobile phone / 29-Smartphone / 72-iPhone / 256-Tennis / 1626-Forehand / \n", "0-Games / 21-Stadium / 79-American football / 81-Athlete / \n", "25-Toy / 140-LEGO / 264-Star Wars / 627-Clock / 749-Figurine / 1245-Alarm device / 3429-Yoda / 3551-Alarm clock / \n", "5-Dance / 115-Ballet / \n", "3-Concert / 7-Musician / 13-Musical ensemble / 30-Drummer / \n", "3-Concert / 16-Performance art / \n", "1-Vehicle / 4-Car / 270-Sedan / \n", "37-Gadget / 47-Personal computer / 145-Tablet computer / \n", "42-Fishing / 510-Fishing lure / 902-Fishing bait / \n", "1-Vehicle / 4-Car / \n", "14-Guitar / 1972-Toy balloon / \n", "6-Animation / 356-Star / 411-MapleStory / \n", "230-Brass instrument / 427-Trumpet / 4673-Flugelhorn / \n", "1-Vehicle / 11-Motorsport / 17-Racing / 27-Motorcycle / 172-Motocross / 197-Trail / \n", "42-Fishing / 548-Parachuting / 573-Paragliding / 765-Parachute / \n", "205-Newscaster / 371-Firefighter / \n", "5-Dance / \n", "1788-Fencing / \n", "6-Animation / 995-Attack on Titan / \n", "93-Comedy / \n", "3-Concert / \n", "9-Music video / \n", "1089-Wire / 1616-Electrical wiring / 1864-Switch / \n", "9-Music video / 99-Christmas / \n", "3-Concert / 357-DVD / \n", "694-Tai chi / \n", "10-Animal / 18-Outdoor recreation / 280-Wildlife / 2396-Coyote / \n", "0-Games / 2-Video game / 19-PC game / 43-Call of Duty / 216-Call of Duty: Modern Warfare 3 / \n", "61-Art / 97-Drawing / 148-Painting / 242-Paper / 303-Paint / \n", "86-Plant / 136-Gardening / \n", "0-Games / 2-Video game / 1170-Super Mario World / \n", "7-Musician / 13-Musical ensemble / 22-Nature / 46-Choir / \n", "1-Vehicle / 11-Motorsport / 27-Motorcycle / 158-Tire / 3548-Yamaha YZ125 / \n", "113-House / \n", "1-Vehicle / 4-Car / 11-Motorsport / 17-Racing / 41-Sports car / 114-Engine / 2286-Grand Tourer Injection / 2826-Suzuki Swift / \n", "3-Concert / \n", "20-Trailer / 60-Basketball / \n", "75-Wrestling / 85-Combat / \n", "6-Animation / 15-Cartoon / \n", "93-Comedy / \n", "1-Vehicle / 4-Car / 11-Motorsport / 17-Racing / 27-Motorcycle / 57-Race track / \n", "0-Games / 2-Video game / 19-PC game / 33-Weapon / 96-Soldier / 114-Engine / 257-Counter-Strike / 266-Helicopter / \n", "3-Concert / \n", "93-Comedy / \n", "38-Orchestra / 356-Star / \n", "7-Musician / 14-Guitar / 24-String instrument / 63-Acoustic guitar / 361-Flamenco / 3369-Diego de la Vega / \n", "2019-Dental braces / \n", "395-Camping / \n", "0-Games / 49-School / 54-Highlight film / 60-Basketball / 168-Basketball moves / \n", "6-Animation / 15-Cartoon / 185-Dragon Ball / 244-Goku / 418-Gohan / \n", "1-Vehicle / 4-Car / 11-Motorsport / 17-Racing / 378-Drag racing / \n", "18-Outdoor recreation / 42-Fishing / 84-Snow / 120-Winter / 262-Ice / 287-Human swimming / \n", "66-Bollywood / 125-Television / \n", "20-Trailer / 286-Bus / \n", "6-Animation / 102-Building / 113-House / 416-Apartment / \n", "1-Vehicle / 50-Aircraft / 82-Airplane / 129-Aviation / 139-Airport / 175-Airliner / 188-Airline / \n", "6-Animation / 15-Cartoon / 547-Bleach / \n", "9-Music video / \n", "6-Animation / 15-Cartoon / 67-Cycling / 69-Bicycle / 126-The Walt Disney Company / 4323-Cricket / \n", "9-Music video / \n", "0-Games / 2-Video game / 328-Simulation video game / 1407-Animal Crossing / 1819-Animal Crossing: New Leaf / \n", "1-Vehicle / 67-Cycling / 69-Bicycle / \n", "1-Vehicle / 4-Car / 1950-The CW / \n", "0-Games / 47-Personal computer / 1169-Hot Wheels / \n", "1-Vehicle / 62-Train / \n", "0-Games / 2-Video game / 6-Animation / 15-Cartoon / 20-Trailer / 2398-Personal armor / \n", "10-Animal / 18-Outdoor recreation / 22-Nature / 42-Fishing / 91-Fish / \n", "22-Nature / 86-Plant / 1023-Rose / \n", "0-Games / 2-Video game / 35-Minecraft / \n", "298-Television advertisement / 2032-Pepsi / \n", "6-Animation / 15-Cartoon / 126-The Walt Disney Company / \n", "1-Vehicle / 4-Car / 49-School / 64-Transport / 74-Truck / 286-Bus / 2199-School bus / \n", "898-Pilates / \n", "0-Games / 2-Video game / 19-PC game / 35-Minecraft / 412-Monster / \n", "1-Vehicle / 4-Car / 11-Motorsport / 41-Sports car / 790-Nissan GT-R / 1699-Cart / \n", "0-Games / 2-Video game / 4392-The Evil Within / \n", "1-Vehicle / 4-Car / 11-Motorsport / 17-Racing / 40-Road / 41-Sports car / 57-Race track / 135-Drifting / 245-BMW / 344-Coupé / \n", "599-Dofus / \n", "8-Football / \n", "20-Trailer / \n", "12-Food / 26-Cooking / 32-Recipe / \n", "93-Comedy / \n", "1-Vehicle / 4-Car / 25-Toy / 344-Coupé / \n", "6-Animation / 15-Cartoon / \n", "10-Animal / 280-Wildlife / 425-City / \n", "2-Video game / 387-Spider-Man / \n", "0-Games / 458-Gold / 1787-Happy Farm / \n", "0-Games / 2-Video game / 43-Call of Duty / 47-Personal computer / 194-Call of Duty: Modern Warfare 2 / 302-Call of Duty 4: Modern Warfare / 431-Call of Duty: Advanced Warfare / 464-First-person Shooter / 602-Central processing unit / \n", "597-Soup / \n", "47-Personal computer / 228-Laptop / 512-Printing / 525-Printer / \n", "2009-Electromagnetic coil / 3806-Tesla coil / \n", "0-Games / 89-Comics / 781-Pachinko / \n", "2484-Electronic circuit / 3285-Boombox / \n", "5-Dance / 16-Performance art / 115-Ballet / \n", "130-Gymnastics / \n", "18-Outdoor recreation / 83-Skateboarding / 108-Skateboard / 221-Skatepark / \n", "287-Human swimming / 329-Swimming pool / \n", "2-Video game / 237-Final Fantasy / \n", "163-Running / \n", "7-Musician / 36-Piano / 53-Keyboard / 104-Pianist / \n", "7-Musician / 14-Guitar / 24-String instrument / 63-Acoustic guitar / \n", "13-Musical ensemble / 49-School / 98-Festival / 203-Marching band / 250-Parade / \n", "45-Cosmetics / 436-Concealer / \n", "93-Comedy / 101-Telephone / \n", "5-Dance / 8-Football / \n", "23-Mobile phone / 29-Smartphone / 37-Gadget / 47-Personal computer / 145-Tablet computer / \n", "164-Beach / 447-Sand / 573-Paragliding / 975-Dune / \n", "99-Christmas / 2649-Nativity scene / \n", "6-Animation / 61-Art / 89-Comics / 183-Manga / 547-Bleach / \n", "42-Fishing / \n", "14-Guitar / 219-Accordion / 655-Diatonic button accordion / \n", "191-Cheerleading / \n", "5-Dance / 2198-Irish dance / \n", "18-Outdoor recreation / 83-Skateboarding / 108-Skateboard / 221-Skatepark / \n", "0-Games / 6-Animation / 1563-Stick figure / \n", "616-The Bible / \n", "7-Musician / 14-Guitar / 24-String instrument / 63-Acoustic guitar / \n", "25-Toy / 524-Littlest Pet Shop / 749-Figurine / \n", "5-Dance / \n", "25-Toy / 89-Comics / 435-Action figure / 749-Figurine / 2121-Marvel Legends / 2735-Raccoon / \n", "0-Games / 734-DayZ / 736-ARMA / 1216-DayZ / \n", "45-Cosmetics / 173-Eye shadow / 193-Eye liner / 195-Mascara / 196-Eye / 218-Eyelash / 235-Lipstick / 291-Rouge / 436-Concealer / \n", "1-Vehicle / 4-Car / 11-Motorsport / 41-Sports car / 245-BMW / 270-Sedan / 874-BMW M3 / 1709-BMW 3 Series (E46) / \n", "1-Vehicle / 4-Car / 11-Motorsport / 70-Driving / 94-Dashcam / \n", "0-Games / 2-Video game / 20-Trailer / 1991-Rappelz / \n", "144-Amusement park / 334-Amusement ride / \n", "0-Games / 2-Video game / 35-Minecraft / \n", "97-Drawing / \n", "3-Concert / \n", "9-Music video / \n", "3-Concert / \n", "98-Festival / 250-Parade / 253-Carnival / \n", "260-Prayer / \n", "3-Concert / 7-Musician / 13-Musical ensemble / 46-Choir / \n", "0-Games / 5-Dance / \n", "14-Guitar / 24-String instrument / 100-Electric guitar / \n", "1054-Vacuum cleaner / \n", "20-Trailer / 264-Star Wars / \n", "31-Disc jockey / 861-Mixing console / 3458-DMC World DJ Championships / \n", "42-Fishing / \n", "28-Fashion / 153-Dress / \n", "925-Brain / \n", "28-Fashion / \n", "5-Dance / 250-Parade / \n", "14-Guitar / \n", "9-Music video / \n", "7-Musician / 14-Guitar / 24-String instrument / \n", "0-Games / 85-Combat / 116-Pokémon / 186-Pokémon / \n", "250-Parade / 253-Carnival / \n", "3-Concert / 7-Musician / 30-Drummer / 38-Orchestra / 39-Drums / 44-Drums / \n", "2-Video game / 34-Action-adventure game / 55-Video game console / 111-PlayStation 3 / 176-PlayStation 4 / 317-PlayStation / 950-Blu-ray disc / 1985-The Last of Us / \n", "1-Vehicle / 4-Car / 28-Fashion / 785-Jeans / 1534-H&M / \n", "0-Games / 19-PC game / 35-Minecraft / \n", "1-Vehicle / 156-Agriculture / 174-Farm / 715-Farming Simulator / \n", "3-Concert / 7-Musician / 13-Musical ensemble / 46-Choir / \n", "1-Vehicle / 4-Car / 945-MINI Cooper / \n", "3-Concert / \n", "162-Medicine / 3971-Dentures / \n", "28-Fashion / \n", "20-Trailer / \n", "1-Vehicle / 67-Cycling / 69-Bicycle / 377-BMX bike / \n", "14-Guitar / 44-Drums / 1825-Cajón / \n", "14-Guitar / 24-String instrument / 63-Acoustic guitar / \n", "61-Art / 479-Graffiti / \n", "6-Animation / 15-Cartoon / \n", "5-Dance / \n", "9-Music video / \n", "5-Dance / 769-Red Bull / \n", "5-Dance / \n", "10-Animal / 48-Pet / 166-Cat / \n", "162-Medicine / \n", "9-Music video / \n", "9-Music video / \n", "5-Dance / \n", "3-Concert / 7-Musician / 14-Guitar / 30-Drummer / \n", "1-Vehicle / 18-Outdoor recreation / 87-Boat / \n", "12-Food / 25-Toy / 215-Eating / \n", "3-Concert / \n", "0-Games / 411-MapleStory / \n", "22-Nature / \n", "0-Games / 1-Vehicle / 4-Car / 11-Motorsport / 17-Racing / 40-Road / 41-Sports car / 57-Race track / 3504-Honda NSX / 4302-GTR 2 – FIA GT Racing Game / \n", "162-Medicine / 196-Eye / 625-Factory / \n", "28-Fashion / 153-Dress / \n", "463-Cricket / 1495-Twenty20 / \n", "47-Personal computer / 131-Computer / 145-Tablet computer / 228-Laptop / 2615-HP Pavilion / \n", "25-Toy / 354-Egg / 439-Barbie / 611-Kinder Surprise / 1047-Easter egg / \n", "6-Animation / 126-The Walt Disney Company / \n", "20-Trailer / 222-Surfing / \n", "102-Building / 318-Church / 885-Bell / \n", "0-Games / 49-School / 54-Highlight film / 60-Basketball / 81-Athlete / 168-Basketball moves / 255-Slam dunk / 306-High school / 1101-Point guard / \n", "10-Animal / 22-Nature / 91-Fish / 258-Aquarium / 865-Turtle / \n", "102-Building / 113-House / 170-Home improvement / 276-Room / 888-Living room / \n", "30-Drummer / 39-Drums / 44-Drums / 128-Cymbal / 146-Snare drum / \n", "6-Animation / 102-Building / 113-House / \n", "13-Musical ensemble / \n", "1-Vehicle / 11-Motorsport / 17-Racing / 27-Motorcycle / 40-Road / \n", "5-Dance / 16-Performance art / 95-Talent show / \n", "35-Minecraft / 1149-Castle / \n", "0-Games / 1837-Realm of the Mad God / \n", "3-Concert / \n", "8-Football / 21-Stadium / 77-Arena / \n", "6-Animation / \n", "0-Games / 8-Football / 21-Stadium / 54-Highlight film / 65-Kick / 76-Ball / \n", "260-Prayer / \n", "5-Dance / 16-Performance art / 78-Wedding / 190-Bride / \n", "15-Cartoon / 89-Comics / \n", "5-Dance / 16-Performance art / 95-Talent show / \n", "5-Dance / 410-Teacher / \n", "14-Guitar / 24-String instrument / 63-Acoustic guitar / \n", "125-Television / \n", "1-Vehicle / 27-Motorcycle / 516-Headphones / 522-Helmet / 832-Headset / \n", "163-Running / \n", "1-Vehicle / 4-Car / 9-Music video / 60-Basketball / 255-Slam dunk / \n", "12-Food / 26-Cooking / 32-Recipe / 88-Machine / 539-Home appliance / 1489-Coffeemaker / 1729-Latte / 1823-Espresso machine / 2057-Cappuccino / \n", "46-Choir / \n", "12-Food / \n", "1-Vehicle / 40-Road / 74-Truck / 2226-Bus driver / \n", "8-Football / \n", "35-Minecraft / 61-Art / 2945-Fan art / \n", "46-Choir / 318-Church / 870-Chapel / \n", "0-Games / 2-Video game / 51-Strategy video game / 1989-Oculus Rift / \n", "121-Photography / 155-Camera / 566-Digital camera / 1855-Video camera / \n", "0-Games / 8-Football / 21-Stadium / \n", "10-Animal / 80-Horse / \n", "96-Soldier / 794-Bridge / \n", "5-Dance / 16-Performance art / \n", "0-Games / 10-Animal / 166-Cat / 1193-Orange / \n", "5-Dance / 16-Performance art / 39-Drums / \n", "0-Games / 1496-LittleBigPlanet 2 / 1672-LittleBigPlanet / \n", "12-Food / 26-Cooking / 32-Recipe / 52-Dish / 58-Cuisine / 109-Cooking show / 179-Kitchen / 301-Meal / \n", "1608-Injury / \n", "1-Vehicle / 4-Car / 11-Motorsport / 41-Sports car / 231-Supercar / 1324-Roadster / 1486-Bugatti Veyron / 1875-Bugatti / \n", "2914-Wreath / \n", "5-Dance / 93-Comedy / \n", "45-Cosmetics / 173-Eye shadow / 193-Eye liner / 195-Mascara / 282-Skin / 291-Rouge / 436-Concealer / \n", "42-Fishing / 84-Snow / 120-Winter / 262-Ice / \n", "47-Personal computer / 145-Tablet computer / 847-Angry Birds / \n", "6-Animation / 708-Mermaid / \n", "9-Music video / \n", "1-Vehicle / 4-Car / 11-Motorsport / 114-Engine / 332-Mercedes-Benz / \n", "1623-Shaolin Kung Fu / \n", "10-Animal / 22-Nature / 91-Fish / 275-Diving / 288-Underwater / 381-Underwater diving / 407-Scuba diving / 716-Coral / \n", "1-Vehicle / 74-Truck / 92-Radio-controlled model / \n", "23-Mobile phone / 29-Smartphone / 125-Television / 487-Samsung Electronics / 1595-Television set / \n", "162-Medicine / 214-Muscle / \n", "0-Games / 2-Video game / 505-Mega Man / 581-PlayStation / \n", "43-Call of Duty / 216-Call of Duty: Modern Warfare 3 / \n", "15-Cartoon / 61-Art / 97-Drawing / 424-Chicken / \n", "59-Winter sport / 84-Snow / 120-Winter / 177-Skiing / 281-Snowboarding / \n", "14-Guitar / 24-String instrument / 63-Acoustic guitar / \n", "0-Games / 2-Video game / 34-Action-adventure game / 434-Ninja / 2192-Ninja Gaiden / \n", "10-Animal / 14-Guitar / 48-Pet / 132-Bird / 1147-Hamster / \n", "0-Games / 54-Highlight film / 60-Basketball / 77-Arena / \n", "75-Wrestling / \n", "0-Games / 2-Video game / 33-Weapon / 43-Call of Duty / 431-Call of Duty: Advanced Warfare / \n", "8-Football / \n", "8-Football / \n", "1-Vehicle / 4-Car / 124-Tractor / 542-Potato / 2079-Two-wheel tractor / \n", "5-Dance / \n", "1-Vehicle / 62-Train / 119-Locomotive / 143-Railroad car / 161-Train station / \n", "61-Art / 97-Drawing / 3544-Rubber stamp / \n", "28-Fashion / \n", "153-Dress / \n", "3-Concert / 7-Musician / \n", "12-Food / 88-Machine / 625-Factory / \n", "61-Art / 82-Airplane / 242-Paper / 313-Book / 370-Origami / 1933-Paper plane / \n", "9-Music video / \n", "12-Food / 26-Cooking / 179-Kitchen / \n", "512-Printing / 525-Printer / 1063-Ink / \n", "5-Dance / 95-Talent show / \n", "0-Games / 8-Football / \n", "3-Concert / 16-Performance art / 31-Disc jockey / \n", "0-Games / 8-Football / 21-Stadium / 54-Highlight film / 73-Ball / 79-American football / 81-Athlete / \n", "93-Comedy / \n", "1106-Character / \n", "20-Trailer / 89-Comics / 2069-Flash / \n", "9-Music video / \n", "28-Fashion / 1237-Baby carriage / \n", "22-Nature / 86-Plant / 136-Gardening / 807-Seed / 1460-Herb / \n", "23-Mobile phone / 29-Smartphone / 37-Gadget / 72-iPhone / 513-Computer keyboard / \n", "56-Hair / 583-Wig / 733-Washing / \n", "3-Concert / 5-Dance / 16-Performance art / \n", "25-Toy / 435-Action figure / 735-Kamen Rider Series / 2525-Gashapon / \n", "99-Christmas / \n", "0-Games / 2-Video game / 19-PC game / 35-Minecraft / \n", "121-Photography / 541-Black-and-white / \n", "37-Gadget / 47-Personal computer / 131-Computer / 145-Tablet computer / 189-iPad / \n", "7-Musician / 9-Music video / 14-Guitar / \n", "6-Animation / 9-Music video / 15-Cartoon / 97-Drawing / \n", "12-Food / 944-Hamburger / \n", "0-Games / 2-Video game / 34-Action-adventure game / 35-Minecraft / \n", "79-American football / 465-Microphone / \n", "1-Vehicle / 11-Motorsport / 17-Racing / 21-Stadium / 27-Motorcycle / 105-Motorcycling / \n", "2649-Nativity scene / \n", "1-Vehicle / 11-Motorsport / 17-Racing / 27-Motorcycle / 172-Motocross / 447-Sand / \n", "13-Musical ensemble / \n", "75-Wrestling / 85-Combat / \n", "84-Snow / 85-Combat / 120-Winter / \n", "0-Games / 2-Video game / 35-Minecraft / \n", "42-Fishing / 287-Human swimming / 329-Swimming pool / 892-Aerobics / \n", "1-Vehicle / 62-Train / 64-Transport / 103-Rail transport / 119-Locomotive / 143-Railroad car / 152-Track / 161-Train station / 224-Rapid transit / \n", "1-Vehicle / 50-Aircraft / 129-Aviation / 736-ARMA / \n", "9-Music video / \n", "5-Dance / 387-Spider-Man / \n", "0-Games / 2-Video game / 34-Action-adventure game / 1985-The Last of Us / \n", "18-Outdoor recreation / 59-Winter sport / 177-Skiing / 789-Alpine skiing / \n", "9-Music video / 995-Attack on Titan / \n", "3-Concert / 1038-Hijab / \n", "1-Vehicle / 4-Car / 11-Motorsport / 17-Racing / 27-Motorcycle / 57-Race track / 105-Motorcycling / 378-Drag racing / \n", "2-Video game / 127-Xbox 360 / 704-Call of Duty: World at War / \n", "1-Vehicle / 4-Car / 11-Motorsport / 17-Racing / 27-Motorcycle / 57-Race track / \n", "0-Games / 2-Video game / 19-PC game / 1566-Doom / \n", "372-Playing card / 746-Card manipulation / \n", "10-Animal / 42-Fishing / 91-Fish / 275-Diving / 381-Underwater diving / \n", "4004-Tug of war / \n", "12-Food / 2040-Korean food / \n", "0-Games / 2-Video game / 6-Animation / 15-Cartoon / 199-Wii / \n", "8-Football / 21-Stadium / 77-Arena / \n", "7-Musician / 24-String instrument / 150-Violin / 842-Harp / \n", "14-Guitar / 24-String instrument / \n", "0-Games / 51-Strategy video game / 174-Farm / 373-Clash of Clans / \n", "10-Animal / 12-Food / 22-Nature / 280-Wildlife / \n", "42-Fishing / \n", "372-Playing card / 746-Card manipulation / \n", "3-Concert / 5-Dance / 16-Performance art / \n", "3-Concert / 31-Disc jockey / \n", "66-Bollywood / \n", "1-Vehicle / 4-Car / 18-Outdoor recreation / 22-Nature / 167-Four-wheel drive / 197-Trail / 198-Off-road vehicle / 360-Jeep / 1205-Jeep Wrangler / \n", "110-Album / 567-Compact disc / \n", "5-Dance / 16-Performance art / 46-Choir / \n", "10-Animal / 540-Insect / 775-Bee / 784-Honey / 1121-Beehive / \n", "47-Personal computer / 141-Microsoft Windows / \n", "3-Concert / \n", "86-Plant / 136-Gardening / 156-Agriculture / 174-Farm / 906-Sowing / 1678-Foam / \n", "6-Animation / 15-Cartoon / \n", "341-Sketch comedy / \n", "0-Games / 2-Video game / 127-Xbox 360 / 511-Mortal Kombat / 1756-Predator / 2011-Alien / 2837-Killer Instinct / 3255-Aliens vs. Predator / 4046-Killer Instinct / \n", "5-Dance / \n", "0-Games / 2-Video game / 19-PC game / 33-Weapon / 257-Counter-Strike / 702-Counter-Strike: Source / \n", "0-Games / 2-Video game / 1630-Warframe / \n", "3-Concert / 31-Disc jockey / 133-Nightclub / \n", "1802-Jubeat / \n", "3-Concert / 5-Dance / 16-Performance art / \n", "5-Dance / 115-Ballet / \n", "2-Video game / 20-Trailer / 111-PlayStation 3 / 1928-BioShock / 2645-BioShock Infinite / \n", "18-Outdoor recreation / 83-Skateboarding / 108-Skateboard / 221-Skatepark / \n", "1-Vehicle / 4-Car / 11-Motorsport / 27-Motorcycle / 40-Road / 105-Motorcycling / 864-Speedometer / \n", "1-Vehicle / 4-Car / 11-Motorsport / 17-Racing / 57-Race track / 1400-rFactor / \n", "10-Animal / 18-Outdoor recreation / 22-Nature / 147-River / 197-Trail / 1694-Harness racing / \n", "1-Vehicle / 4-Car / 11-Motorsport / 17-Racing / 41-Sports car / 57-Race track / 3679-Super GT / \n", "3-Concert / 7-Musician / 13-Musical ensemble / 38-Orchestra / \n", "0-Games / 2-Video game / 51-Strategy video game / 165-League of Legends / \n", "0-Games / 8-Football / 21-Stadium / 54-Highlight film / 65-Kick / 76-Ball / 79-American football / \n", "814-Pond / \n", "0-Games / 2-Video game / 2267-Geometry Dash / \n", "5-Dance / 8-Football / \n", "9-Music video / \n", "12-Food / 26-Cooking / 58-Cuisine / 122-Weight training / 179-Kitchen / \n", "1-Vehicle / 10-Animal / 18-Outdoor recreation / 22-Nature / 42-Fishing / 87-Boat / 91-Fish / 268-Recreational fishing / 510-Fishing lure / \n", "0-Games / 35-Minecraft / \n", "3-Concert / 13-Musical ensemble / 16-Performance art / 68-Lighting / \n", "31-Disc jockey / \n", "12-Food / 26-Cooking / 32-Recipe / 210-Cookware and bakeware / 391-Chicken meat / 542-Potato / 943-Stove / \n", "25-Toy / 113-House / 200-Doll / 246-Furniture / 1842-Dollhouse / \n", "0-Games / 2-Video game / 19-PC game / 33-Weapon / 43-Call of Duty / 134-Call of Duty: Black Ops / 138-Call of Duty: Black Ops II / \n", "10-Animal / 136-Gardening / 1384-Greenhouse / \n", "25-Toy / 200-Doll / 232-Cake / 323-Toddler / 1192-Hello Kitty / \n", "5-Dance / 16-Performance art / \n", "3-Concert / \n", "10-Animal / 18-Outdoor recreation / 42-Fishing / 91-Fish / 275-Diving / 288-Underwater / 381-Underwater diving / 407-Scuba diving / 3910-Mackerel / \n", "8-Football / 298-Television advertisement / \n", "3-Concert / 5-Dance / 8-Football / 769-Red Bull / \n", "12-Food / 26-Cooking / 32-Recipe / 52-Dish / 58-Cuisine / 109-Cooking show / 179-Kitchen / 557-Salad / 771-Noodle / \n", "1-Vehicle / 4-Car / 27-Motorcycle / 207-Exhaust system / \n", "1-Vehicle / \n", "1-Vehicle / 27-Motorcycle / \n", "1-Vehicle / 62-Train / 103-Rail transport / 152-Track / 161-Train station / \n", "12-Food / 26-Cooking / 389-Restaurant / \n", "0-Games / 35-Minecraft / \n", "2-Video game / 23-Mobile phone / 29-Smartphone / 37-Gadget / 72-iPhone / 155-Camera / \n", "6-Animation / 15-Cartoon / 1744-Black Butler / \n", "5-Dance / 16-Performance art / 95-Talent show / \n", "0-Games / 2-Video game / 6-Animation / 37-Gadget / 55-Video game console / 292-Handheld game console / 335-Nintendo 3DS / 995-Attack on Titan / \n", "0-Games / 2-Video game / 19-PC game / 229-Halo / 470-Halo 3 / 644-Halo: Reach / 1161-Halo 2 / 2002-Halo 3: ODST / \n", "1-Vehicle / 12-Food / 215-Eating / 2190-Bowl / \n", "43-Call of Duty / 138-Call of Duty: Black Ops II / \n", "0-Games / 159-World of Warcraft / 171-Warcraft / \n", "3-Concert / 16-Performance art / \n", "75-Wrestling / 777-WWE 2K / 977-WWE '13 / 1582-WWE 2K14 / \n", "20-Trailer / \n", "9-Music video / \n", "6-Animation / \n", "117-Boxing / \n", "385-Face / \n", "28-Fashion / 49-School / \n", "61-Art / 148-Painting / \n", "1-Vehicle / 11-Motorsport / 17-Racing / 27-Motorcycle / 57-Race track / 105-Motorcycling / 172-Motocross / 198-Off-road vehicle / 669-Extreme sport / 728-Dirt track racing / \n", "3-Concert / \n", "0-Games / 10-Animal / \n", "61-Art / \n", "1-Vehicle / 50-Aircraft / 82-Airplane / 129-Aviation / 139-Airport / 157-Jet aircraft / 175-Airliner / 188-Airline / 254-Runway / 856-Airbus A320 family / \n", "2-Video game / 9-Music video / 43-Call of Duty / 134-Call of Duty: Black Ops / 229-Halo / 1580-Crysis 2 / \n", "0-Games / 326-Metin2 / \n", "3-Concert / 7-Musician / 13-Musical ensemble / 24-String instrument / 38-Orchestra / 150-Violin / \n", "0-Games / 59-Winter sport / 112-Ice skating / 142-Hockey / \n", "8-Football / \n", "18-Outdoor recreation / 22-Nature / 347-Climbing / 3318-Abseiling / \n", "1-Vehicle / 4-Car / 41-Sports car / 70-Driving / 299-Volkswagen Passenger Cars / 668-Volkswagen Golf / \n", "5-Dance / \n", "6-Animation / \n", "293-Slide show / \n", "0-Games / 269-Comic book / 489-Board game / 922-Tabletop game / 2833-Go / \n", "3-Concert / 7-Musician / 1684-Hillsong Church / \n", "0-Games / 362-PlayStation Portable / \n", "2-Video game / 37-Gadget / 55-Video game console / 292-Handheld game console / 335-Nintendo 3DS / \n", "1-Vehicle / 4-Car / 11-Motorsport / 17-Racing / 41-Sports car / 57-Race track / 651-Stock car racing / \n", "9-Music video / \n", "3-Concert / 7-Musician / 16-Performance art / 30-Drummer / \n", "8-Football / 595-Champion / \n", "66-Bollywood / \n", "6-Animation / 10-Animal / 15-Cartoon / 71-Dog / \n", "117-Boxing / \n", "9-Music video / 98-Festival / \n", "25-Toy / 200-Doll / \n", "38-Orchestra / 446-Sheet music / \n", "3-Concert / \n", "3-Concert / 7-Musician / 4209-Scarlet / \n", "325-Vampire / \n", "10-Animal / 71-Dog / \n", "3-Concert / 7-Musician / \n", "0-Games / 2-Video game / 19-PC game / 192-Battlefield / 374-Battlefield 3 / \n", "61-Art / 97-Drawing / 148-Painting / \n", "0-Games / 2-Video game / 19-PC game / 35-Minecraft / 216-Call of Duty: Modern Warfare 3 / \n", "3-Concert / 7-Musician / 13-Musical ensemble / \n", "0-Games / 19-PC game / 211-RuneScape / 510-Fishing lure / \n", "0-Games / 8-Football / 49-School / \n", "49-School / \n", "211-RuneScape / \n", "10-Animal / 48-Pet / 71-Dog / 182-Puppy / 824-Terrier / \n", "0-Games / 51-Strategy video game / 373-Clash of Clans / \n", "125-Television / \n", "162-Medicine / \n", "88-Machine / 208-Wood / 261-Tool / 882-Pen / 2428-Barn / \n", "3-Concert / 7-Musician / 14-Guitar / \n", "1-Vehicle / 4-Car / 27-Motorcycle / 114-Engine / 336-Scooter / 2308-Belt drive / \n", "9-Music video / \n", "1-Vehicle / 88-Machine / 114-Engine / 533-Oil / \n", "0-Games / 49-School / 54-Highlight film / 60-Basketball / 168-Basketball moves / 1101-Point guard / \n", "9-Music video / \n", "5-Dance / \n", "1-Vehicle / 12-Food / 124-Tractor / 156-Agriculture / 174-Farm / 542-Potato / \n", "0-Games / 1-Vehicle / 4-Car / 41-Sports car / 2400-Asphalt / \n", "45-Cosmetics / 193-Eye liner / 195-Mascara / 196-Eye / 218-Eyelash / \n", "2-Video game / 6-Animation / 20-Trailer / 1928-BioShock / 2645-BioShock Infinite / \n", "5-Dance / 115-Ballet / 1253-Ballet Dancer / \n", "313-Book / \n", "5-Dance / 31-Disc jockey / \n", "93-Comedy / \n", "3-Concert / 7-Musician / 16-Performance art / \n", "12-Food / 26-Cooking / 32-Recipe / 52-Dish / 58-Cuisine / 109-Cooking show / 184-Vegetable / 475-Cheese / 557-Salad / \n", "18-Outdoor recreation / 22-Nature / 86-Plant / 136-Gardening / 137-Tree / 213-Forest / \n", "233-Ocean / \n", "1-Vehicle / 11-Motorsport / 124-Tractor / 252-Tractor pulling / \n", "3-Concert / 7-Musician / 13-Musical ensemble / 30-Drummer / 39-Drums / 44-Drums / \n", "9-Music video / \n", "3-Concert / 7-Musician / 13-Musical ensemble / 14-Guitar / 30-Drummer / 44-Drums / \n", "546-Stitch / 4246-Crochet hook / \n", "5-Dance / \n", "10-Animal / 48-Pet / 537-Zoo / 970-Monkey / \n", "9-Music video / 78-Wedding / 121-Photography / 190-Bride / \n", "0-Games / 2-Video game / 6-Animation / 15-Cartoon / 89-Comics / \n", "8-Football / \n", "9-Music video / \n", "36-Piano / 53-Keyboard / 446-Sheet music / \n", "0-Games / 2-Video game / \n", "7-Musician / 13-Musical ensemble / 38-Orchestra / 46-Choir / 870-Chapel / \n", "354-Egg / 1924-Symbol / 2227-Multimedios Televisión / \n", "130-Gymnastics / 1030-Trampoline / 1335-Trampolining / \n", "0-Games / 2277-The Lord of the Rings Online / \n", "43-Call of Duty / 229-Halo / 470-Halo 3 / \n", "0-Games / 6-Animation / 15-Cartoon / 837-Roblox / \n", "1-Vehicle / 4-Car / 11-Motorsport / 40-Road / 70-Driving / 94-Dashcam / 449-Honda / 552-Honda Civic / 1493-Supercharger / 3697-Honda K engine / \n", "2-Video game / 20-Trailer / 34-Action-adventure game / 35-Minecraft / 380-Assassin's Creed / \n", "10-Animal / 80-Horse / \n", "6-Animation / \n", "0-Games / 6-Animation / 8-Football / 90-Sports game / \n", "3-Concert / 31-Disc jockey / 133-Nightclub / 988-Ibiza / \n", "12-Food / 26-Cooking / 1227-Sushi / \n", "1-Vehicle / 12-Food / 156-Agriculture / 617-Maize / \n", "14-Guitar / 24-String instrument / 100-Electric guitar / 3014-Gibson SG / \n", "3-Concert / 7-Musician / 13-Musical ensemble / 38-Orchestra / 46-Choir / 371-Firefighter / \n", "0-Games / 1-Vehicle / 2-Video game / 4-Car / 34-Action-adventure game / 118-Grand Theft Auto V / 277-Sport utility vehicle / \n", "3-Concert / 5-Dance / 49-School / \n", "1-Vehicle / 4-Car / 11-Motorsport / 509-Classic car / 2558-Plymouth / \n", "36-Piano / \n", "4007-The House of the Dead / \n", "0-Games / 257-Counter-Strike / \n", "3-Concert / \n", "1-Vehicle / 4-Car / 11-Motorsport / 70-Driving / 207-Exhaust system / 234-Ford / 344-Coupé / 509-Classic car / 1148-Hot rod / \n", "1-Vehicle / 67-Cycling / 69-Bicycle / 691-Cargo / \n", "102-Building / 113-House / 276-Room / 462-Table / 888-Living room / \n", "66-Bollywood / 905-Flash Video / \n", "0-Games / 1-Vehicle / 2-Video game / 4-Car / 11-Motorsport / 17-Racing / 41-Sports car / 57-Race track / 352-Need for Speed / 1770-Need for Speed: Hot Pursuit / 2545-Need for Speed III: Hot Pursuit / \n", "1-Vehicle / 4-Car / 11-Motorsport / 17-Racing / 169-Rallying / \n", "9-Music video / \n", "0-Games / 59-Winter sport / 112-Ice skating / 142-Hockey / \n", "13-Musical ensemble / 44-Drums / 2444-Crêpe / 3221-Pizzica / \n", "141-Microsoft Windows / \n", "0-Games / 8-Football / 21-Stadium / 54-Highlight film / 65-Kick / 73-Ball / \n", "1-Vehicle / 4-Car / 11-Motorsport / 158-Tire / 1148-Hot rod / 2917-Rat rod / \n", "3-Concert / \n", "10-Animal / 80-Horse / 247-Stallion / 289-Dressage / 438-Mare / \n", "3-Concert / 1386-ABS-CBN / \n", "18-Outdoor recreation / 42-Fishing / 267-Mountain / 800-Mountain pass / \n", "3-Concert / \n", "10-Animal / 20-Trailer / 71-Dog / \n", "3-Concert / 7-Musician / \n", "6-Animation / \n", "1-Vehicle / 11-Motorsport / 27-Motorcycle / 40-Road / 105-Motorcycling / \n", "0-Games / 2-Video game / 1288-Marvel vs. Capcom / 1499-Marvel vs. Capcom 3: Fate of Two Worlds / 3081-Marvel vs. Capcom 2: New Age of Heroes / \n", "28-Fashion / 206-Shoe / 368-Sneakers / 2257-Athletic shoe / \n", "7-Musician / 14-Guitar / \n", "2-Video game / 432-Kingdom Hearts / 1908-Kingdom Hearts Birth by Sleep / \n", "28-Fashion / \n", "0-Games / 2-Video game / 505-Mega Man / \n", "6-Animation / 126-The Walt Disney Company / 996-Princess / \n", "3-Concert / 7-Musician / 13-Musical ensemble / \n", "0-Games / 2-Video game / 20-Trailer / \n", "6-Animation / 15-Cartoon / \n", "0-Games / 687-Taiko no Tatsujin / \n", "93-Comedy / \n", "13-Musical ensemble / \n", "28-Fashion / \n", "8-Football / \n", "1-Vehicle / 4-Car / \n", "15-Cartoon / 185-Dragon Ball / 244-Goku / 418-Gohan / 1304-Cell / \n", "7-Musician / 14-Guitar / 24-String instrument / 100-Electric guitar / 1103-Gibson Les Paul / \n", "2-Video game / 23-Mobile phone / 29-Smartphone / 72-iPhone / 189-iPad / 272-iPod / \n", "3-Concert / 68-Lighting / 263-Light / \n", "3-Concert / 7-Musician / 14-Guitar / \n", "75-Wrestling / \n", "1-Vehicle / 4-Car / 11-Motorsport / 17-Racing / 40-Road / 41-Sports car / 57-Race track / 64-Transport / 74-Truck / 378-Drag racing / 482-Asphalt / \n", "1-Vehicle / 4-Car / 11-Motorsport / 17-Racing / 41-Sports car / 57-Race track / 135-Drifting / 378-Drag racing / \n", "6-Animation / 15-Cartoon / \n", "31-Disc jockey / \n", "12-Food / 26-Cooking / 32-Recipe / 52-Dish / \n", "75-Wrestling / \n", "6-Animation / 15-Cartoon / \n", "0-Games / 8-Football / 54-Highlight film / 79-American football / \n", "1-Vehicle / 4-Car / 40-Road / 70-Driving / 84-Snow / 94-Dashcam / 120-Winter / 158-Tire / \n", "4511-Watchmen / \n", "56-Hair / \n", "3-Concert / 16-Performance art / 68-Lighting / \n", "66-Bollywood / \n", "3-Concert / 7-Musician / 14-Guitar / \n", "5-Dance / \n", "15-Cartoon / 921-Grand Theft Auto: Vice City / \n", "5-Dance / 16-Performance art / 115-Ballet / 577-Penguin / \n", "6-Animation / 185-Dragon Ball / 244-Goku / \n", "14-Guitar / 606-String / 1971-Epiphone / \n", "20-Trailer / \n", "3-Concert / 5-Dance / 16-Performance art / 31-Disc jockey / 133-Nightclub / \n", "1697-Astrological sign / \n", "5-Dance / 191-Cheerleading / 776-Kindergarten / \n", "0-Games / 21-Stadium / 49-School / 81-Athlete / 562-Pitcher / 946-Baseball park / \n", "6-Animation / 15-Cartoon / \n", "117-Boxing / 595-Champion / \n", "20-Trailer / 541-Black-and-white / \n", "1-Vehicle / 11-Motorsport / 27-Motorcycle / 105-Motorcycling / \n", "208-Wood / 400-Woodturning / \n", "0-Games / \n", "110-Album / \n", "0-Games / 2-Video game / 535-Slot machine / \n", "3-Concert / \n", "20-Trailer / 1014-Movieclips / \n", "5-Dance / 68-Lighting / 102-Building / 227-Water / 263-Light / 804-Shopping mall / 1222-Fountain / \n", "2297-Elf / \n", "6-Animation / 99-Christmas / 357-DVD / \n", "0-Games / 2-Video game / 257-Counter-Strike / \n", "9-Music video / \n", "3-Concert / 7-Musician / 13-Musical ensemble / 46-Choir / \n", "325-Vampire / \n", "6-Animation / 15-Cartoon / \n", "0-Games / 8-Football / 21-Stadium / 65-Kick / 73-Ball / 76-Ball / 81-Athlete / \n", "38-Orchestra / \n", "7-Musician / 9-Music video / 14-Guitar / \n", "36-Piano / 53-Keyboard / 107-Musical keyboard / 265-Electronic keyboard / \n", "1-Vehicle / 4-Car / \n", "1-Vehicle / 78-Wedding / \n", "8-Football / \n", "61-Art / 148-Painting / 936-Watercolor paint / \n", "7-Musician / 13-Musical ensemble / 14-Guitar / \n", "14-Guitar / 24-String instrument / 100-Electric guitar / 4165-Washburn Guitars / \n", "693-Sword / \n", "12-Food / 170-Home improvement / 652-Bathroom / 1592-Plumbing / 2153-Sink / \n", "1262-NBA 2K14 / \n", "7-Musician / 150-Violin / \n", "121-Photography / 512-Printing / 633-Photograph / \n", "148-Painting / \n", "8-Football / 54-Highlight film / \n", "10-Animal / 166-Cat / 196-Eye / 283-Rock / \n", "23-Mobile phone / 29-Smartphone / 513-Computer keyboard / \n", "1-Vehicle / 4-Car / 11-Motorsport / 17-Racing / 57-Race track / \n", "1-Vehicle / 4-Car / 27-Motorcycle / 336-Scooter / 1036-Moped / \n", "0-Games / 2-Video game / 243-Sonic the Hedgehog / 251-Sonic the Hedgehog / \n", "3-Concert / 5-Dance / 31-Disc jockey / 133-Nightclub / \n", "6-Animation / \n", "1-Vehicle / 11-Motorsport / 27-Motorcycle / 608-Wheelie / 927-Supermoto / \n", "31-Disc jockey / \n", "0-Games / 2-Video game / 19-PC game / 51-Strategy video game / 603-StarCraft II: Wings of Liberty / 732-StarCraft / \n", "36-Piano / 53-Keyboard / 104-Pianist / 107-Musical keyboard / 131-Computer / 265-Electronic keyboard / \n", "14-Guitar / 192-Battlefield / 1370-Battlefield 2 / \n", "3-Concert / 68-Lighting / 263-Light / \n", "8-Football / \n", "9-Music video / 66-Bollywood / \n", "1-Vehicle / 4-Car / \n", "28-Fashion / 153-Dress / \n", "8-Football / \n", "89-Comics / \n", "3-Concert / 7-Musician / 30-Drummer / 133-Nightclub / 1280-Club / \n", "23-Mobile phone / 29-Smartphone / 37-Gadget / 101-Telephone / 553-Sony Xperia / \n", "3804-Music of Eritrea / \n", "5-Dance / 16-Performance art / \n", "0-Games / 8-Football / \n", "7-Musician / 14-Guitar / 24-String instrument / \n", "10-Animal / 48-Pet / 71-Dog / 182-Puppy / \n", "130-Gymnastics / 154-Wheel / \n", "1-Vehicle / 4-Car / 11-Motorsport / 17-Racing / 27-Motorcycle / 40-Road / \n", "23-Mobile phone / 29-Smartphone / 465-Microphone / \n", "0-Games / 2-Video game / 55-Video game console / \n", "9-Music video / \n", "10-Animal / 865-Turtle / \n", "36-Piano / \n", "9-Music video / \n", "22-Nature / \n", "1-Vehicle / 27-Motorcycle / 99-Christmas / \n", "66-Bollywood / \n", "3497-Hair dryer / \n", "9-Music video / \n", "0-Games / 8-Football / \n", "7-Musician / 14-Guitar / 24-String instrument / 63-Acoustic guitar / 670-Banjo / \n", "93-Comedy / 355-Comedian / \n", "0-Games / 2-Video game / 6-Animation / 15-Cartoon / 199-Wii / \n", "6-Animation / 25-Toy / 200-Doll / \n", "1-Vehicle / 62-Train / 103-Rail transport / 152-Track / 428-Rail transport modelling / \n", "3-Concert / \n", "23-Mobile phone / 29-Smartphone / 37-Gadget / 72-iPhone / 101-Telephone / 272-iPod / \n", "12-Food / 295-Chocolate / 985-Strawberry / 1403-Flower bouquet / \n", "6-Animation / 643-Hero / \n", "0-Games / 2-Video game / 19-PC game / 33-Weapon / 2726-Armored Core / \n", "1-Vehicle / 4-Car / \n", "7-Musician / 14-Guitar / 63-Acoustic guitar / \n", "1184-Flag / \n", "0-Games / 8-Football / 21-Stadium / \n", "3-Concert / \n", "591-Naruto: Ultimate Ninja / \n", "0-Games / 256-Tennis / \n", "10-Animal / 48-Pet / 132-Bird / 1799-Domestic canary / \n", "0-Games / 2-Video game / 19-PC game / 725-Human / 739-Guild Wars / 839-Guild Wars 2 / \n", "125-Television / \n", "440-Lion / 1039-Tiger / \n", "0-Games / 2-Video game / 19-PC game / 33-Weapon / 43-Call of Duty / 96-Soldier / 302-Call of Duty 4: Modern Warfare / 422-Call of Duty: Ghosts / \n", "9-Music video / \n", "10-Animal / 601-Shark / \n", "1-Vehicle / 11-Motorsport / 27-Motorcycle / 105-Motorcycling / 172-Motocross / 197-Trail / 491-Enduro / \n", "1172-VHS / \n", "0-Games / 2-Video game / 34-Action-adventure game / 380-Assassin's Creed / 1522-Assassin's Creed II / \n", "3-Concert / 13-Musical ensemble / 30-Drummer / 39-Drums / 44-Drums / \n", "10-Animal / 48-Pet / 71-Dog / 695-Balloon / 1030-Trampoline / \n", "13-Musical ensemble / \n", "78-Wedding / \n", "1-Vehicle / 50-Aircraft / 266-Helicopter / 308-Ship / 605-Cruise ship / 2151-Carnival Cruise Lines / \n", "3-Concert / 7-Musician / 16-Performance art / \n", "0-Games / 8-Football / 76-Ball / \n", "3-Concert / 5-Dance / 16-Performance art / \n", "0-Games / 8-Football / 73-Ball / \n", "93-Comedy / 355-Comedian / \n", "5-Dance / 1004-Samurai / \n", "1-Vehicle / 4-Car / 70-Driving / \n", "5-Dance / 49-School / 95-Talent show / 191-Cheerleading / \n", "0-Games / 2-Video game / 55-Video game console / 586-Nintendo Entertainment System / \n", "0-Games / 2-Video game / 426-Resident Evil / \n", "0-Games / 2-Video game / 19-PC game / 51-Strategy video game / 382-Dota 2 / 549-Defense of the Ancients / \n", "1-Vehicle / 11-Motorsport / 17-Racing / 27-Motorcycle / \n", "5-Dance / \n", "0-Games / 2-Video game / 34-Action-adventure game / 656-Metal Gear / 2698-Metal Gear Rising: Revengeance / \n", "390-Model / 1605-Second Life / \n", "31-Disc jockey / 322-Mixtape / \n", "3-Concert / 7-Musician / 13-Musical ensemble / 14-Guitar / 30-Drummer / 39-Drums / 44-Drums / \n", "6-Animation / 15-Cartoon / \n", "7-Musician / 14-Guitar / 24-String instrument / 63-Acoustic guitar / \n", "1-Vehicle / 4-Car / 1952-Carburetor / \n", "0-Games / 159-World of Warcraft / 171-Warcraft / \n", "0-Games / 8-Football / 21-Stadium / 65-Kick / 81-Athlete / \n", "0-Games / 2-Video game / 20-Trailer / \n", "66-Bollywood / \n", "3-Concert / 7-Musician / 30-Drummer / \n", "75-Wrestling / 85-Combat / 117-Boxing / \n", "1-Vehicle / 4-Car / 11-Motorsport / 17-Racing / 40-Road / 41-Sports car / 234-Ford / 473-Ford Mustang / \n", "516-Headphones / \n", "31-Disc jockey / \n", "466-Gift / \n", "209-University / \n", "5-Dance / \n", "10-Animal / 91-Fish / 288-Underwater / 601-Shark / \n", "379-Touhou Project / \n", "0-Games / 2-Video game / 1841-Path of Exile / \n", "1-Vehicle / 27-Motorcycle / 94-Dashcam / 336-Scooter / \n", "0-Games / 8-Football / 21-Stadium / 54-Highlight film / 65-Kick / 76-Ball / \n", "1-Vehicle / 4-Car / 40-Road / 70-Driving / 94-Dashcam / 212-Highway / \n", "0-Games / 211-RuneScape / 1035-Bank / \n", "23-Mobile phone / 29-Smartphone / 37-Gadget / 101-Telephone / 832-Headset / \n", "394-Map / 420-The Elder Scrolls V: Skyrim / \n", "12-Food / 26-Cooking / 589-Grilling / 620-Barbecue / 1085-Barbecue grill / 1256-Smoking / 2395-Ribs / \n", "0-Games / 1343-Super Robot Wars / \n", "3-Concert / 68-Lighting / \n", "3-Concert / 4180-Alpha / \n", "444-Family / 2725-Phoenix / \n", "46-Choir / \n", "45-Cosmetics / \n", "28-Fashion / 153-Dress / 390-Model / \n", "3-Concert / 7-Musician / 13-Musical ensemble / 14-Guitar / 2393-Didgeridoo / \n", "1035-Bank / 2095-Cash / \n", "3-Concert / \n", "61-Art / 148-Painting / 303-Paint / 873-Acrylic paint / 936-Watercolor paint / 1362-Canvas / \n", "6-Animation / 10-Animal / 15-Cartoon / \n", "1-Vehicle / 4-Car / 11-Motorsport / 70-Driving / 207-Exhaust system / 680-Subaru / 1041-Subaru Impreza / 1897-Subaru Impreza WRX / \n", "122-Weight training / \n", "9-Music video / \n", "9-Music video / 30-Drummer / \n", "3-Concert / 7-Musician / \n", "1-Vehicle / 4-Car / 40-Road / 207-Exhaust system / 299-Volkswagen Passenger Cars / 526-Volkswagen / 531-Hatchback / 668-Volkswagen Golf / 2286-Grand Tourer Injection / 3464-Volkswagen Golf Mk6 / \n", "1-Vehicle / 67-Cycling / 69-Bicycle / 220-Mountain bike / 1420-Gear / \n", "310-The Sims / 328-Simulation video game / 543-The Sims 2 / \n", "14-Guitar / 100-Electric guitar / \n", "6-Animation / 9-Music video / \n", "59-Winter sport / 112-Ice skating / 142-Hockey / 262-Ice / 312-Ice rink / \n", "1625-Sunset / \n", "0-Games / 54-Highlight film / 60-Basketball / 168-Basketball moves / \n", "18-Outdoor recreation / 83-Skateboarding / 108-Skateboard / 221-Skatepark / 556-Skateboarding trick / \n", "5-Dance / 31-Disc jockey / \n", "925-Brain / \n", "25-Toy / 140-LEGO / 884-Lego minifigure / 1302-The Lego Group / 1478-Lego Ninjago / \n", "0-Games / 2-Video game / 55-Video game console / 199-Wii / 456-Go-kart / 518-Mario Kart / \n", "6-Animation / 15-Cartoon / 4082-Extra / \n", "1-Vehicle / 4-Car / 270-Sedan / 403-Toyota / 1583-Toyota Corolla / \n", "121-Photography / 293-Slide show / \n", "22-Nature / 61-Art / 97-Drawing / 148-Painting / 303-Paint / 936-Watercolor paint / 3030-Landscape painting / \n", "7-Musician / 14-Guitar / 24-String instrument / \n", "1-Vehicle / 4-Car / 11-Motorsport / 17-Racing / 57-Race track / 1351-Rallycross / \n", "25-Toy / 179-Kitchen / \n", "0-Games / 8-Football / 21-Stadium / 65-Kick / 73-Ball / 76-Ball / \n", "45-Cosmetics / 282-Skin / 385-Face / 533-Oil / 2351-Moisturizer / \n", "3-Concert / 5-Dance / 16-Performance art / \n", "6-Animation / 372-Playing card / 460-Yu-Gi-Oh! Trading Card Game / \n", "5-Dance / 16-Performance art / 911-Studio / \n", "1-Vehicle / 4-Car / 11-Motorsport / 17-Racing / 41-Sports car / 57-Race track / 2907-Suzuka Circuit / \n", "6-Animation / 15-Cartoon / 616-The Bible / \n", "0-Games / 34-Action-adventure game / 201-Grand Theft Auto: San Andreas / \n", "0-Games / 25-Toy / 78-Wedding / 200-Doll / 439-Barbie / 1441-Mattel / \n", "248-Hotel / \n", "3-Concert / \n", "0-Games / 1-Vehicle / 25-Toy / 140-LEGO / 1079-Lego City / 4006-Lego Creator / \n", "587-Clay / 2089-Winnie-the-Pooh / \n", "1-Vehicle / 62-Train / 103-Rail transport / 119-Locomotive / \n", "0-Games / 2-Video game / 19-PC game / 43-Call of Duty / 431-Call of Duty: Advanced Warfare / \n", "0-Games / 2-Video game / 311-Street Fighter / \n", "3-Concert / 1149-Castle / \n", "242-Paper / \n", "0-Games / 2-Video game / 474-Destiny / \n", "353-Wall / \n", "5-Dance / \n", "3-Concert / 9-Music video / \n", "0-Games / 2-Video game / 6-Animation / 961-Sega Genesis / \n", "3-Concert / \n", "3-Concert / 7-Musician / 13-Musical ensemble / \n", "67-Cycling / 69-Bicycle / \n", "12-Food / 56-Hair / 533-Oil / \n", "23-Mobile phone / 29-Smartphone / 72-iPhone / \n", "1-Vehicle / 4-Car / 403-Toyota / 2512-Toyota Prius / \n", "1-Vehicle / 12-Food / 124-Tractor / 156-Agriculture / 174-Farm / 249-Heavy equipment / 931-Combine Harvester / 1158-Wheat / 3260-Barley / \n", "1572-Husband / \n", "9-Music video / 310-The Sims / 543-The Sims 2 / \n", "1-Vehicle / 4-Car / 11-Motorsport / 17-Racing / 40-Road / 41-Sports car / 135-Drifting / 207-Exhaust system / 344-Coupé / 482-Asphalt / \n", "12-Food / 86-Plant / \n", "23-Mobile phone / 29-Smartphone / 37-Gadget / 72-iPhone / 272-iPod / 331-iPod touch / \n", "5-Dance / 66-Bollywood / \n", "2-Video game / 199-Wii / 335-Nintendo 3DS / \n", "0-Games / 2-Video game / 37-Gadget / 55-Video game console / 199-Wii / 292-Handheld game console / 335-Nintendo 3DS / \n", "233-Ocean / 1003-Tuba / \n", "0-Games / 159-World of Warcraft / 171-Warcraft / \n", "23-Mobile phone / 29-Smartphone / 205-Newscaster / \n", "1-Vehicle / 4-Car / 11-Motorsport / 17-Racing / 169-Rallying / 683-Renault / \n", "0-Games / 2-Video game / 19-PC game / 35-Minecraft / 229-Halo / \n", "110-Album / \n", "13-Musical ensemble / 2202-Armed forces / \n", "1-Vehicle / 4-Car / 11-Motorsport / 17-Racing / 1307-Police car / \n", "7-Musician / 36-Piano / 53-Keyboard / 104-Pianist / 107-Musical keyboard / \n", "93-Comedy / \n", "694-Tai chi / \n", "45-Cosmetics / 238-Nail / 300-Nail art / 304-Nail polish / 307-Hand / 324-Finger / 338-Manicure / 823-Artificial nails / \n", "14-Guitar / 24-String instrument / \n", "5-Dance / \n", "137-Tree / 208-Wood / 673-Saw / 1073-Chainsaw / 1294-Logging / \n", "3-Concert / 7-Musician / 13-Musical ensemble / 219-Accordion / \n", "23-Mobile phone / 29-Smartphone / 47-Personal computer / 228-Laptop / 497-Macintosh / 513-Computer keyboard / 900-MacBook / 1086-MacBook Pro / 1657-MacBook Air / \n", "0-Games / 2-Video game / 240-Arcade game / 241-Fighting game / 311-Street Fighter / 461-Street Fighter IV / 496-Super Street Fighter IV / 563-Ryu / \n", "468-Biology / 1110-Cell / \n", "0-Games / 59-Winter sport / 112-Ice skating / 142-Hockey / 312-Ice rink / \n", "3-Concert / 13-Musical ensemble / \n", "31-Disc jockey / \n", "0-Games / 8-Football / \n", "512-Printing / \n", "3-Concert / \n", "110-Album / \n", "5-Dance / \n", "8-Football / 21-Stadium / 77-Arena / \n", "12-Food / \n", "0-Games / 2-Video game / 35-Minecraft / \n", "12-Food / 26-Cooking / 32-Recipe / 52-Dish / 58-Cuisine / 109-Cooking show / 184-Vegetable / 210-Cookware and bakeware / 3003-Roll / \n", "14-Guitar / \n", "22-Nature / \n", "3-Concert / 7-Musician / 30-Drummer / 39-Drums / 44-Drums / \n", "762-Angel / \n", "0-Games / 2-Video game / 19-PC game / 43-Call of Duty / 96-Soldier / 194-Call of Duty: Modern Warfare 2 / 302-Call of Duty 4: Modern Warfare / \n", "126-The Walt Disney Company / \n", "33-Weapon / 261-Tool / 315-Knife / \n", "18-Outdoor recreation / 83-Skateboarding / 108-Skateboard / 221-Skatepark / \n", "14-Guitar / 1754-Rocksmith / \n", "12-Food / \n", "1-Vehicle / 12-Food / 86-Plant / 124-Tractor / 174-Farm / 459-Plough / 660-Rural area / 906-Sowing / 1129-Hay / 2039-Soybean / \n", "12-Food / 752-Pasta / 2362-Porcelain / \n", "125-Television / \n", "0-Games / 6-Animation / 15-Cartoon / 116-Pokémon / 2597-Sprite / \n", "6-Animation / 15-Cartoon / \n", "0-Games / 25-Toy / 298-Television advertisement / \n", "102-Building / 318-Church / 870-Chapel / 885-Bell / \n", "5-Dance / 16-Performance art / \n", "3-Concert / 8-Football / \n", "1-Vehicle / 11-Motorsport / 156-Agriculture / 252-Tractor pulling / \n", "3-Concert / 7-Musician / 13-Musical ensemble / 38-Orchestra / 46-Choir / 99-Christmas / \n", "1-Vehicle / 2-Video game / 4-Car / 11-Motorsport / 17-Racing / 41-Sports car / 111-PlayStation 3 / 135-Drifting / 953-Gran Turismo 5 / \n", "5-Dance / 16-Performance art / \n", "1-Vehicle / 11-Motorsport / 17-Racing / 27-Motorcycle / 57-Race track / 105-Motorcycling / 669-Extreme sport / \n", "36-Piano / 53-Keyboard / 107-Musical keyboard / 397-Organ / 1666-Hammond organ / \n", "0-Games / 1-Vehicle / 11-Motorsport / 21-Stadium / \n", "1-Vehicle / 18-Outdoor recreation / 42-Fishing / 87-Boat / 466-Gift / 601-Shark / \n", "7-Musician / 14-Guitar / 24-String instrument / 63-Acoustic guitar / 100-Electric guitar / \n", "28-Fashion / 155-Camera / \n", "3-Concert / 38-Orchestra / \n", "0-Games / 2-Video game / 19-PC game / 43-Call of Duty / 422-Call of Duty: Ghosts / \n", "7-Musician / 13-Musical ensemble / 49-School / 79-American football / 203-Marching band / 306-High school / \n", "12-Food / 807-Seed / 4566-Chia / \n", "25-Toy / 428-Rail transport modelling / 507-Thomas the Tank Engine / \n", "78-Wedding / 153-Dress / 190-Bride / 590-Wedding dress / \n", "0-Games / 2-Video game / 19-PC game / 33-Weapon / 96-Soldier / \n", "1650-The Idolmaster / 1755-The Idolmaster / \n", "7-Musician / 14-Guitar / 24-String instrument / 100-Electric guitar / \n", "483-Costume / \n", "0-Games / 60-Basketball / \n", "2-Video game / 111-PlayStation 3 / 237-Final Fantasy / 1131-Final Fantasy XIII / \n", "60-Basketball / 113-House / 255-Slam dunk / \n", "0-Games / 1-Vehicle / 2-Video game / 4-Car / 11-Motorsport / 17-Racing / 111-PlayStation 3 / 4507-MotorStorm / \n", "113-House / 433-Construction / \n", "0-Games / 2-Video game / 47-Personal computer / 127-Xbox 360 / 181-Xbox / 420-The Elder Scrolls V: Skyrim / \n", "1-Vehicle / 4-Car / 673-Saw / 1073-Chainsaw / 1294-Logging / 1414-Gasoline / 2209-Firewood / 3073-Log splitter / \n", "3-Concert / \n", "0-Games / 2-Video game / 116-Pokémon / 186-Pokémon / \n", "5-Dance / 16-Performance art / 46-Choir / 209-University / \n", "0-Games / 8-Football / 73-Ball / \n", "1-Vehicle / 4-Car / 27-Motorcycle / 40-Road / \n", "3-Concert / \n", "92-Radio-controlled model / 149-Model aircraft / 160-Radio-controlled aircraft / 393-Unmanned aerial vehicle / 3198-Parrot AR.Drone / \n", "5-Dance / 6-Animation / 1738-Elmo / \n", "0-Games / 2-Video game / 19-PC game / 241-Fighting game / 3149-Melty Blood / \n", "425-City / 794-Bridge / 4473-Sydney Opera House / \n", "10-Animal / 22-Nature / 588-Reptile / 865-Turtle / 2631-Tortoise / \n", "141-Microsoft Windows / 1938-Rust / \n", "3-Concert / 7-Musician / 13-Musical ensemble / 14-Guitar / \n", "12-Food / \n", "1-Vehicle / 18-Outdoor recreation / 42-Fishing / 87-Boat / 579-Yacht / \n", "3-Concert / 7-Musician / 13-Musical ensemble / 38-Orchestra / \n", "0-Games / 8-Football / 54-Highlight film / 79-American football / \n", "2704-Priston Tale / \n", "12-Food / 26-Cooking / 620-Barbecue / 2395-Ribs / \n", "5-Dance / 20-Trailer / \n", "1-Vehicle / 18-Outdoor recreation / 67-Cycling / 69-Bicycle / 197-Trail / 213-Forest / 220-Mountain bike / 284-Mountain biking / 520-Bicycle frame / \n", "1-Vehicle / 62-Train / 64-Transport / 103-Rail transport / 119-Locomotive / 143-Railroad car / 152-Track / 161-Train station / 224-Rapid transit / 821-Rail freight transport / \n", "9-Music video / \n", "18-Outdoor recreation / 283-Rock / 458-Gold / 751-Desert / \n", "8-Football / \n", "85-Combat / \n", "163-Running / \n", "1-Vehicle / 11-Motorsport / 50-Aircraft / 82-Airplane / 129-Aviation / 157-Jet aircraft / \n", "1-Vehicle / 4-Car / 11-Motorsport / 17-Racing / \n", "5-Dance / \n", "3-Concert / 7-Musician / 13-Musical ensemble / 24-String instrument / 38-Orchestra / 150-Violin / 450-Viola / \n", "5-Dance / 95-Talent show / \n", "9-Music video / 130-Gymnastics / \n", "14-Guitar / 1533-Lead guitar / \n", "1-Vehicle / 18-Outdoor recreation / 137-Tree / 1294-Logging / \n", "6-Animation / 15-Cartoon / 97-Drawing / 183-Manga / \n", "39-Drums / 44-Drums / \n", "6-Animation / 15-Cartoon / 123-Naruto / 285-Sasuke Uchiha / 1180-Itachi Uchiha / \n", "18-Outdoor recreation / 83-Skateboarding / 108-Skateboard / 555-Longboard / 667-Longboarding / \n", "86-Plant / 136-Gardening / 184-Vegetable / 279-Garden / 642-Leaf / 1460-Herb / \n", "6-Animation / 15-Cartoon / 1445-Donald Duck / \n", "3-Concert / 7-Musician / 13-Musical ensemble / 30-Drummer / 46-Choir / \n", "0-Games / 6-Animation / 35-Minecraft / \n", "0-Games / 159-World of Warcraft / 171-Warcraft / \n", "3-Concert / 7-Musician / 13-Musical ensemble / 38-Orchestra / 46-Choir / \n", "0-Games / 8-Football / 21-Stadium / \n", "0-Games / 1-Vehicle / 2-Video game / 4-Car / 34-Action-adventure game / 41-Sports car / 70-Driving / 231-Supercar / 297-Grand Theft Auto IV / \n", "66-Bollywood / \n", "12-Food / 26-Cooking / 32-Recipe / 52-Dish / 58-Cuisine / 475-Cheese / 542-Potato / \n", "3-Concert / 7-Musician / \n", "10-Animal / 48-Pet / \n", "372-Playing card / 731-Coin / 746-Card manipulation / \n", "5-Dance / 9-Music video / \n", "3-Concert / 1044-Coca-Cola / 1224-Cola / \n", "1-Vehicle / 4-Car / 1461-Electric car / 2304-Electric vehicle / \n", "9-Music video / 164-Beach / \n", "3-Concert / 7-Musician / 14-Guitar / 24-String instrument / \n", "36-Piano / 219-Accordion / \n", "126-The Walt Disney Company / 144-Amusement park / 250-Parade / \n", "18-Outdoor recreation / 83-Skateboarding / 108-Skateboard / 221-Skatepark / 482-Asphalt / 556-Skateboarding trick / 667-Longboarding / \n", "6-Animation / 75-Wrestling / \n", "6-Animation / \n", "0-Games / 8-Football / \n", "18-Outdoor recreation / 83-Skateboarding / 108-Skateboard / \n", "7-Musician / 13-Musical ensemble / 14-Guitar / \n", "6-Animation / 15-Cartoon / 89-Comics / 123-Naruto / 183-Manga / 285-Sasuke Uchiha / \n", "6-Animation / 10-Animal / \n", "0-Games / 445-Guitar Hero / \n", "60-Basketball / \n", "3-Concert / 7-Musician / 13-Musical ensemble / 38-Orchestra / 427-Trumpet / \n", "0-Games / 60-Basketball / 168-Basketball moves / \n", "12-Food / \n", "3-Concert / 7-Musician / \n", "3-Concert / \n", "3-Concert / 5-Dance / 16-Performance art / \n", "162-Medicine / 282-Skin / \n", "14-Guitar / 24-String instrument / 100-Electric guitar / \n", "82-Airplane / 92-Radio-controlled model / 149-Model aircraft / 160-Radio-controlled aircraft / \n", "5-Dance / \n", "0-Games / 8-Football / 21-Stadium / \n", "3-Concert / 7-Musician / 30-Drummer / 39-Drums / 44-Drums / \n", "6-Animation / 15-Cartoon / 116-Pokémon / \n", "1-Vehicle / 4-Car / 11-Motorsport / \n", "10-Animal / 18-Outdoor recreation / 80-Horse / 247-Stallion / 289-Dressage / 438-Mare / \n", "31-Disc jockey / \n", "4226-Neo / \n", "1-Vehicle / 11-Motorsport / 17-Racing / 27-Motorcycle / 172-Motocross / 491-Enduro / \n", "8-Football / \n", "14-Guitar / \n", "6-Animation / 97-Drawing / 725-Human / \n", "83-Skateboarding / 108-Skateboard / \n", "12-Food / 22-Nature / 26-Cooking / 86-Plant / 136-Gardening / 389-Restaurant / \n", "1-Vehicle / 4-Car / 154-Wheel / 158-Tire / 344-Coupé / 449-Honda / 552-Honda Civic / \n", "0-Games / 8-Football / 358-Rugby football / \n", "1-Vehicle / 18-Outdoor recreation / 42-Fishing / 87-Boat / 147-River / \n", "6-Animation / 9-Music video / \n", "12-Food / \n", "67-Cycling / 69-Bicycle / 197-Trail / 220-Mountain bike / 278-GoPro / 284-Mountain biking / \n", "12-Food / \n", "162-Medicine / 214-Muscle / \n", "20-Trailer / 89-Comics / 1173-Captain America / 4345-Winter Soldier / \n", "209-University / \n", "1-Vehicle / 4-Car / 245-BMW / 270-Sedan / 2305-BMW 5 Series / 4393-BMW 5 Series (F10) / \n", "1-Vehicle / 4-Car / 11-Motorsport / 70-Driving / 94-Dashcam / 114-Engine / 270-Sedan / 344-Coupé / 449-Honda / 552-Honda Civic / \n", "2-Video game / 23-Mobile phone / 29-Smartphone / 37-Gadget / 101-Telephone / 178-Samsung Galaxy / \n", "0-Games / 90-Sports game / 453-Pro Evolution Soccer / \n", "0-Games / 2-Video game / 8-Football / 21-Stadium / 90-Sports game / 180-FIFA 15 / \n", "260-Prayer / 616-The Bible / 971-Alphabet / \n", "3-Concert / 7-Musician / \n", "93-Comedy / 1567-Cook / \n", "28-Fashion / 45-Cosmetics / 238-Nail / 304-Nail polish / \n", "9-Music video / \n", "6-Animation / \n", "23-Mobile phone / 29-Smartphone / 37-Gadget / 47-Personal computer / 72-iPhone / 145-Tablet computer / 178-Samsung Galaxy / 1015-Samsung Galaxy Note series / 1727-Samsung Galaxy Note 3 / \n", "66-Bollywood / 78-Wedding / \n", "23-Mobile phone / 29-Smartphone / 37-Gadget / \n", "9-Music video / 183-Manga / \n", "1-Vehicle / 4-Car / 2160-Wagon / \n", "36-Piano / 53-Keyboard / 104-Pianist / 107-Musical keyboard / \n", "6-Animation / 15-Cartoon / 477-Transformers / 990-Optimus Prime / 1900-Autobot / 3398-Transformers: Generation 1 / \n", "448-Planet / 2270-The New York Times / 2643-Asteroid / \n", "3-Concert / 7-Musician / 14-Guitar / \n", "0-Games / 2-Video game / 37-Gadget / 55-Video game console / 116-Pokémon / 186-Pokémon / 292-Handheld game console / 335-Nintendo 3DS / 3191-Mew / 4521-Deoxys / \n", "12-Food / 215-Eating / \n", "78-Wedding / 593-Moon / \n", "30-Drummer / 39-Drums / 44-Drums / 128-Cymbal / 146-Snare drum / \n", "0-Games / 2-Video game / 43-Call of Duty / 134-Call of Duty: Black Ops / 138-Call of Duty: Black Ops II / \n", "0-Games / 2-Video game / 6-Animation / 15-Cartoon / \n", "23-Mobile phone / 29-Smartphone / \n", "10-Animal / 80-Horse / 247-Stallion / \n", "9-Music video / \n", "8-Football / 205-Newscaster / \n", "151-Landing / 164-Beach / 447-Sand / 548-Parachuting / 1061-Marriage proposal / \n", "59-Winter sport / 84-Snow / 120-Winter / 177-Skiing / 267-Mountain / 347-Climbing / \n", "7-Musician / 13-Musical ensemble / 38-Orchestra / \n", "10-Animal / 80-Horse / 290-Pony / 438-Mare / \n", "0-Games / 2-Video game / 257-Counter-Strike / \n", "31-Disc jockey / \n", "14-Guitar / \n", "142-Hockey / \n", "0-Games / 2-Video game / 51-Strategy video game / 165-League of Legends / \n", "0-Games / 2-Video game / 4409-Infinity Blade / \n", "14-Guitar / \n", "6-Animation / 15-Cartoon / 123-Naruto / \n", "66-Bollywood / \n", "30-Drummer / 39-Drums / 44-Drums / 128-Cymbal / 146-Snare drum / \n", "99-Christmas / \n", "0-Games / 43-Call of Duty / 138-Call of Duty: Black Ops II / \n", "110-Album / \n", "6-Animation / 15-Cartoon / 3853-Akane Tendo / \n", "14-Guitar / 24-String instrument / 100-Electric guitar / \n", "96-Soldier / 229-Halo / \n", "0-Games / 411-MapleStory / \n", "14-Guitar / 1231-Mandolin / \n", "0-Games / 8-Football / 21-Stadium / 65-Kick / 73-Ball / \n", "121-Photography / \n", "0-Games / 1-Vehicle / 2-Video game / 19-PC game / 286-Bus / \n", "8-Football / 21-Stadium / \n", "0-Games / 8-Football / \n", "30-Drummer / 39-Drums / 44-Drums / 128-Cymbal / 146-Snare drum / 2073-Sabian / \n", "772-Bowling / \n", "5-Dance / 9-Music video / \n", "31-Disc jockey / \n", "419-Earth / \n", "1-Vehicle / 4-Car / 74-Truck / 154-Wheel / 158-Tire / 167-Four-wheel drive / 349-Chevrolet / 375-Pickup truck / 1806-GMC / \n", "0-Games / 772-Bowling / 930-Ten-pin bowling / \n", "1-Vehicle / 92-Radio-controlled model / 94-Dashcam / 149-Model aircraft / 160-Radio-controlled aircraft / 393-Unmanned aerial vehicle / \n", "810-Village / \n", "5-Dance / 244-Goku / \n", "23-Mobile phone / 29-Smartphone / 37-Gadget / 101-Telephone / \n", "1-Vehicle / 4-Car / 40-Road / 64-Transport / 74-Truck / 94-Dashcam / 482-Asphalt / 1070-Semi-trailer truck / 3220-Dekotora / \n", "28-Fashion / 45-Cosmetics / 173-Eye shadow / 193-Eye liner / 195-Mascara / 291-Rouge / 890-Brush / 3655-Makeup brush / \n", "61-Art / \n", "1-Vehicle / 67-Cycling / 69-Bicycle / 83-Skateboarding / 377-BMX bike / \n", "5-Dance / 10-Animal / 80-Horse / 247-Stallion / 438-Mare / \n", "31-Disc jockey / \n", "5-Dance / 13-Musical ensemble / \n", "28-Fashion / 483-Costume / \n", "36-Piano / 53-Keyboard / 61-Art / 97-Drawing / \n", "9-Music video / \n", "1-Vehicle / 4-Car / 11-Motorsport / 169-Rallying / \n", "6-Animation / \n", "117-Boxing / 2008-Floyd Mayweather; Jr. vs. Manny Pacquiao / \n", "1-Vehicle / 4-Car / \n", "1-Vehicle / 4-Car / 11-Motorsport / 17-Racing / 169-Rallying / 198-Off-road vehicle / \n", "0-Games / 2-Video game / 511-Mortal Kombat / \n", "319-Robot / \n", "3-Concert / \n", "609-Tattoo / \n", "20-Trailer / \n", "0-Games / 2-Video game / 735-Kamen Rider Series / \n", "25-Toy / 200-Doll / \n", "1-Vehicle / 4-Car / 74-Truck / 154-Wheel / \n", "8-Football / 31-Disc jockey / 76-Ball / \n", "9-Music video / 541-Black-and-white / \n", "6-Animation / 2767-Captain Hook / \n", "1-Vehicle / 4-Car / 332-Mercedes-Benz / \n", "18-Outdoor recreation / 83-Skateboarding / 108-Skateboard / 221-Skatepark / \n", "93-Comedy / 355-Comedian / \n", "85-Combat / 1832-Hand-to-hand combat / \n", "1-Vehicle / 87-Boat / 308-Ship / 579-Yacht / 996-Princess / \n", "3804-Music of Eritrea / \n", "9-Music video / \n", "1-Vehicle / 18-Outdoor recreation / 67-Cycling / 69-Bicycle / 197-Trail / 220-Mountain bike / 283-Rock / 284-Mountain biking / \n", "12-Food / 45-Cosmetics / 56-Hair / 282-Skin / 340-Afro-textured hair / \n", "6-Animation / 15-Cartoon / \n", "3-Concert / 68-Lighting / \n", "3-Concert / 5-Dance / 16-Performance art / \n", "3-Concert / \n", "5-Dance / 28-Fashion / 66-Bollywood / 402-Jewellery / \n", "3580-Handbell / \n", "6-Animation / 15-Cartoon / \n", "0-Games / 19-PC game / 1837-Realm of the Mad God / \n", "0-Games / 23-Mobile phone / 29-Smartphone / 72-iPhone / 272-iPod / 331-iPod touch / \n", "144-Amusement park / 321-Roller coaster / 334-Amusement ride / 3485-PortAventura / \n", "5-Dance / \n", "6-Animation / 15-Cartoon / 432-Kingdom Hearts / 665-Kingdom Hearts / 836-Sora / 3299-Kingdom Hearts: Chain of Memories / \n", "0-Games / 51-Strategy video game / 366-Dragon / 373-Clash of Clans / \n", "3-Concert / 5-Dance / 16-Performance art / \n", "6-Animation / 15-Cartoon / 432-Kingdom Hearts / 665-Kingdom Hearts / 836-Sora / \n", "92-Radio-controlled model / 149-Model aircraft / 160-Radio-controlled aircraft / 393-Unmanned aerial vehicle / 1311-Amazon.com; Inc. / \n", "0-Games / 2-Video game / 72-iPhone / \n", "14-Guitar / 63-Acoustic guitar / 670-Banjo / 2241-Plectrum / \n", "754-Gramophone record / \n", "1-Vehicle / 4-Car / 114-Engine / \n", "0-Games / 2-Video game / 60-Basketball / 90-Sports game / 1262-NBA 2K14 / \n", "3-Concert / 98-Festival / \n", "3-Concert / 5-Dance / 253-Carnival / \n", "1-Vehicle / 4-Car / 11-Motorsport / 40-Road / 41-Sports car / 64-Transport / 70-Driving / 94-Dashcam / 212-Highway / 790-Nissan GT-R / \n", "12-Food / 88-Machine / 640-Bottle / \n", "20-Trailer / \n", "5-Dance / \n", "0-Games / 2-Video game / 55-Video game console / 240-Arcade game / 243-Sonic the Hedgehog / 251-Sonic the Hedgehog / 961-Sega Genesis / \n", "0-Games / 8-Football / 21-Stadium / 65-Kick / 90-Sports game / \n", "5-Dance / 115-Ballet / \n", "0-Games / 35-Minecraft / 174-Farm / 1699-Cart / \n", "125-Television / \n", "6-Animation / \n", "3-Concert / 7-Musician / 13-Musical ensemble / 30-Drummer / 38-Orchestra / 39-Drums / \n", "8-Football / \n", "261-Tool / 315-Knife / \n", "14-Guitar / 63-Acoustic guitar / \n", "102-Building / \n", "6-Animation / 9-Music video / \n", "12-Food / 26-Cooking / 32-Recipe / 52-Dish / 109-Cooking show / 179-Kitchen / 226-Dessert / 232-Cake / 759-Icing / 1088-Infomercial / 1365-Fondant icing / \n", "60-Basketball / 1399-Streetball / \n", "2-Video game / \n", "45-Cosmetics / \n", "113-House / 1088-Infomercial / 3301-Mop / \n", "23-Mobile phone / 29-Smartphone / 37-Gadget / \n", "0-Games / 2-Video game / 961-Sega Genesis / \n", "0-Games / 1-Vehicle / 2-Video game / 4-Car / 19-PC game / 34-Action-adventure game / 111-PlayStation 3 / 118-Grand Theft Auto V / \n", "0-Games / 51-Strategy video game / 1687-Heroes of Newerth / \n", "3-Concert / \n", "1-Vehicle / 67-Cycling / 69-Bicycle / 3611-Folding bicycle / \n", "3-Concert / 7-Musician / 30-Drummer / \n", "0-Games / 8-Football / 54-Highlight film / 79-American football / 90-Sports game / 561-Madden NFL / 1813-Madden NFL 13 / 2916-Madden NFL 12 / \n", "56-Hair / \n", "7-Musician / 9-Music video / 14-Guitar / 412-Monster / \n", "0-Games / 8-Football / 21-Stadium / 65-Kick / \n", "3-Concert / \n", "1-Vehicle / \n", "9-Music video / \n", "6-Animation / 15-Cartoon / \n", "8-Football / 79-American football / 993-Quarterback / \n", "3-Concert / 31-Disc jockey / 133-Nightclub / \n", "0-Games / 2-Video game / 51-Strategy video game / 185-Dragon Ball / 1989-Oculus Rift / \n", "1-Vehicle / 18-Outdoor recreation / 67-Cycling / 69-Bicycle / 197-Trail / 220-Mountain bike / 283-Rock / 284-Mountain biking / \n", "609-Tattoo / \n", "0-Games / 2-Video game / 1083-Role-playing game / 3351-Paper Mario: The Thousand-Year Door / 4269-Mario role-playing games / \n", "30-Drummer / 39-Drums / 44-Drums / 128-Cymbal / 146-Snare drum / \n", "0-Games / 10-Animal / 25-Toy / 323-Toddler / 550-Dinosaur / \n", "93-Comedy / 364-Sitcom / 648-Clown / \n", "1-Vehicle / 50-Aircraft / 266-Helicopter / 267-Mountain / 1058-Volcano / \n", "15-Cartoon / 126-The Walt Disney Company / \n", "47-Personal computer / 228-Laptop / 414-Computer monitor / \n", "28-Fashion / 368-Sneakers / \n", "99-Christmas / 370-Origami / 466-Gift / \n", "0-Games / 8-Football / 21-Stadium / 65-Kick / \n", "1-Vehicle / 4-Car / 121-Photography / 155-Camera / 721-Digital SLR / \n", "1-Vehicle / 11-Motorsport / 27-Motorcycle / 608-Wheelie / 4133-Yamaha DT125 / \n", "325-Vampire / \n", "5-Dance / 115-Ballet / \n", "0-Games / 2-Video game / 33-Weapon / 43-Call of Duty / 96-Soldier / 134-Call of Duty: Black Ops / \n", "47-Personal computer / 145-Tablet computer / 242-Paper / \n", "23-Mobile phone / 29-Smartphone / 37-Gadget / 101-Telephone / 1097-Touchscreen / \n", "1202-Academy Awards / \n", "6-Animation / 15-Cartoon / 183-Manga / \n", "248-Hotel / \n", "3-Concert / \n", "36-Piano / 53-Keyboard / 104-Pianist / 107-Musical keyboard / \n", "0-Games / 2-Video game / 685-Luigi / 887-King Koopa / 4464-Mario & Luigi: Bowser's Inside Story / \n", "0-Games / 2-Video game / 34-Action-adventure game / 3901-Star Fox 64 / \n", "0-Games / 2-Video game / 9-Music video / 19-PC game / 51-Strategy video game / 165-League of Legends / \n", "5-Dance / 66-Bollywood / \n", "1-Vehicle / 4-Car / 11-Motorsport / 17-Racing / 41-Sports car / 135-Drifting / 1810-Live for Speed / \n", "5-Dance / 115-Ballet / \n", "1-Vehicle / 4-Car / 163-Running / 234-Ford / 1297-Ford Focus / \n", "10-Animal / 18-Outdoor recreation / 22-Nature / 132-Bird / 223-Weather / \n", "10-Animal / 48-Pet / 71-Dog / 182-Puppy / 1278-Labrador Retriever / 1377-Golden Retriever / 4166-Rescue dog / \n", "9-Music video / 78-Wedding / 190-Bride / \n", "4366-Onigiri / \n", "8-Football / \n", "3-Concert / 5-Dance / \n", "14-Guitar / 59-Winter sport / 63-Acoustic guitar / 84-Snow / 177-Skiing / 281-Snowboarding / 305-Ski / 501-Snowboard / 1764-Ski jumping / \n", "260-Prayer / \n", "5-Dance / 43-Call of Duty / 102-Building / 302-Call of Duty 4: Modern Warfare / \n", "309-Computer hardware / 908-Video card / \n", "42-Fishing / 227-Water / 287-Human swimming / 329-Swimming pool / \n", "1-Vehicle / 11-Motorsport / 17-Racing / 27-Motorcycle / 57-Race track / 105-Motorcycling / \n", "9-Music video / \n", "616-The Bible / \n", "67-Cycling / 69-Bicycle / 377-BMX bike / \n", "9-Music video / \n", "9-Music video / \n", "45-Cosmetics / 282-Skin / 385-Face / \n", "3-Concert / \n", "1-Vehicle / 192-Battlefield / \n", "2115-Radar / \n", "1-Vehicle / 18-Outdoor recreation / 22-Nature / 67-Cycling / 69-Bicycle / 86-Plant / 197-Trail / 213-Forest / 220-Mountain bike / 284-Mountain biking / \n", "3-Concert / 7-Musician / 30-Drummer / 39-Drums / \n", "1-Vehicle / 74-Truck / 394-Map / \n", "9-Music video / \n", "1-Vehicle / 4-Car / 11-Motorsport / 17-Racing / 57-Race track / 522-Helmet / 1323-Helmet camera / 1400-rFactor / \n", "1078-Installation art / \n", "3-Concert / 68-Lighting / \n", "28-Fashion / \n", "8-Football / \n", "10-Animal / 588-Reptile / \n", "1-Vehicle / 4-Car / 11-Motorsport / 40-Road / 74-Truck / 121-Photography / 135-Drifting / \n", "84-Snow / 223-Weather / 1145-Winter storm / \n", "0-Games / 2-Video game / 19-PC game / 33-Weapon / 34-Action-adventure game / 2792-Darksiders / \n", "1-Vehicle / 4-Car / 11-Motorsport / 17-Racing / 27-Motorcycle / 40-Road / 57-Race track / \n", "85-Combat / 937-Black belt / \n", "28-Fashion / 56-Hair / 106-Hairstyle / 1038-Hijab / \n", "36-Piano / 53-Keyboard / \n", "1-Vehicle / 4-Car / \n", "5-Dance / \n", "9-Music video / \n", "754-Gramophone record / \n", "584-Super Mario Bros. / \n", "85-Combat / \n", "6-Animation / 15-Cartoon / 2110-D.Gray-man / \n", "1-Vehicle / 88-Machine / 261-Tool / 673-Saw / \n", "0-Games / 2-Video game / 260-Prayer / \n", "12-Food / 26-Cooking / 32-Recipe / 52-Dish / 58-Cuisine / 1218-Japanese Cuisine / 1227-Sushi / \n", "7-Musician / 14-Guitar / 24-String instrument / 100-Electric guitar / \n", "3-Concert / 7-Musician / 13-Musical ensemble / 16-Performance art / \n", "448-Planet / 455-Unidentified flying object / 593-Moon / 1480-Astronaut / \n", "66-Bollywood / \n", "0-Games / 54-Highlight film / 60-Basketball / 168-Basketball moves / 322-Mixtape / 1101-Point guard / 3652-3x3 / \n", "0-Games / 2-Video game / 19-PC game / 33-Weapon / 43-Call of Duty / 96-Soldier / 704-Call of Duty: World at War / \n", "5-Dance / 1443-Dance studio / \n", "0-Games / 8-Football / 21-Stadium / 65-Kick / 1247-Penalty kick / \n", "0-Games / 2-Video game / 35-Minecraft / \n", "6-Animation / \n", "23-Mobile phone / 29-Smartphone / 37-Gadget / 72-iPhone / 101-Telephone / \n", "35-Minecraft / 131-Computer / \n", "3-Concert / \n", "5-Dance / 98-Festival / 250-Parade / 253-Carnival / \n", "1-Vehicle / 139-Airport / \n", "31-Disc jockey / \n", "7-Musician / 13-Musical ensemble / 46-Choir / \n", "24-String instrument / 150-Violin / \n", "511-Mortal Kombat / 1679-Scorpion / 1929-Sub-Zero / \n", "14-Guitar / 24-String instrument / 63-Acoustic guitar / \n", "75-Wrestling / 1042-Jumbotron / \n", "0-Games / 2-Video game / 6-Animation / \n", "12-Food / 26-Cooking / 32-Recipe / 109-Cooking show / 226-Dessert / 232-Cake / 236-Baking / 354-Egg / 649-Oven / 1322-Biscuit / \n", "61-Art / 366-Dragon / \n", "2-Video game / 55-Video game console / 111-PlayStation 3 / 176-PlayStation 4 / \n", "9-Music video / \n", "5-Dance / \n", "22-Nature / \n", "0-Games / 8-Football / 21-Stadium / 65-Kick / 90-Sports game / \n", "0-Games / 39-Drums / 44-Drums / 240-Arcade game / 687-Taiko no Tatsujin / \n", "23-Mobile phone / 29-Smartphone / 72-iPhone / 272-iPod / 331-iPod touch / \n", "110-Album / \n", "3-Concert / \n", "36-Piano / 46-Choir / \n", "1-Vehicle / 4-Car / 11-Motorsport / 154-Wheel / 234-Ford / 342-Windows Media Video / 473-Ford Mustang / 647-Muscle car / \n", "0-Games / 141-Microsoft Windows / \n", "9-Music video / 2242-Pillow / \n", "849-Tales / 2195-Tales of Symphonia / \n", "14-Guitar / 24-String instrument / \n", "18-Outdoor recreation / 22-Nature / 283-Rock / \n", "3-Concert / 13-Musical ensemble / \n", "56-Hair / \n", "28-Fashion / 1712-Sweater / \n", "3-Concert / 7-Musician / 13-Musical ensemble / 38-Orchestra / \n", "22-Nature / 86-Plant / 136-Gardening / 1500-Cherry blossom / \n", "10-Animal / 13-Musical ensemble / 38-Orchestra / \n", "12-Food / 200-Doll / 439-Barbie / 1707-Microwave oven / \n", "3-Concert / 5-Dance / 7-Musician / 13-Musical ensemble / 16-Performance art / \n", "0-Games / 19-PC game / 1071-Hearthstone: Heroes of Warcraft / \n", "9-Music video / \n", "7-Musician / 14-Guitar / 24-String instrument / 63-Acoustic guitar / \n", "3-Concert / 7-Musician / 14-Guitar / \n", "113-House / 164-Beach / 170-Home improvement / 369-Resort / \n", "8-Football / 54-Highlight film / \n", "1-Vehicle / 10-Animal / 18-Outdoor recreation / 217-Hunting / 718-Duck / \n", "61-Art / 97-Drawing / 408-Sketch / 1432-Kratos / \n", "5-Dance / 16-Performance art / 115-Ballet / \n", "0-Games / 599-Dofus / \n", "3-Concert / 38-Orchestra / \n", "7-Musician / 14-Guitar / \n", "1367-Golf club / \n", "61-Art / 97-Drawing / \n", "83-Skateboarding / 108-Skateboard / 221-Skatepark / \n", "8-Football / \n", "31-Disc jockey / \n", "9-Music video / \n", "213-Forest / 395-Camping / \n", "0-Games / 2-Video game / 19-PC game / 2277-The Lord of the Rings Online / \n", "0-Games / 2-Video game / 6-Animation / 20-Trailer / 682-PlayStation Vita / 2628-Sly Cooper / \n", "0-Games / 6-Animation / 89-Comics / 269-Comic book / 460-Yu-Gi-Oh! Trading Card Game / \n", "1-Vehicle / 4-Car / \n", "0-Games / 8-Football / 21-Stadium / 54-Highlight film / 65-Kick / \n", "23-Mobile phone / 29-Smartphone / 3850-Weighing scale / \n", "3-Concert / 7-Musician / 13-Musical ensemble / 30-Drummer / 39-Drums / \n", "93-Comedy / 1088-Infomercial / \n", "6-Animation / 1866-Katara / \n", "20-Trailer / \n", "0-Games / 6-Animation / 1959-Phantasy Star / \n", "1-Vehicle / 87-Boat / 2705-Winch / \n", "0-Games / 2-Video game / \n", "1-Vehicle / 4-Car / 11-Motorsport / 17-Racing / 40-Road / 57-Race track / 456-Go-kart / \n", "0-Games / 1-Vehicle / 2-Video game / 4-Car / 11-Motorsport / 17-Racing / 1333-Wangan Midnight / \n", "2658-Mirror's Edge / \n", "0-Games / 2-Video game / 55-Video game console / 189-iPad / 387-Spider-Man / 3561-The Amazing Spider-Man / \n", "0-Games / 2-Video game / 19-PC game / 1938-Rust / \n", "102-Building / 339-Architecture / 416-Apartment / 939-Condominium / \n", "130-Gymnastics / \n", "78-Wedding / 121-Photography / 190-Bride / 293-Slide show / 590-Wedding dress / \n", "0-Games / 34-Action-adventure game / 35-Minecraft / \n", "6-Animation / 20-Trailer / \n", "1-Vehicle / 4-Car / 11-Motorsport / 17-Racing / 57-Race track / 651-Stock car racing / 728-Dirt track racing / \n", "0-Games / 60-Basketball / 81-Athlete / 168-Basketball moves / \n", "1-Vehicle / 10-Animal / 18-Outdoor recreation / 42-Fishing / 87-Boat / 91-Fish / 268-Recreational fishing / 698-Fisherman / \n", "3-Concert / \n", "1-Vehicle / 4229-Cable television / \n", "7-Musician / 14-Guitar / 24-String instrument / 63-Acoustic guitar / \n", "3-Concert / 7-Musician / 39-Drums / 44-Drums / \n", "0-Games / 1-Vehicle / 2-Video game / 4-Car / 17-Racing / \n", "0-Games / 51-Strategy video game / 165-League of Legends / \n", "66-Bollywood / \n", "1-Vehicle / 4-Car / 751-Desert / 998-Landscape / 1511-Toyota Land Cruiser / \n", "14-Guitar / 24-String instrument / 100-Electric guitar / \n", "67-Cycling / 69-Bicycle / 3665-Giant Bicycles / \n", "3-Concert / 5-Dance / 31-Disc jockey / 68-Lighting / 133-Nightclub / \n", "6-Animation / 10-Animal / 15-Cartoon / 636-Rabbit / 3090-Bugs Bunny / \n", "5-Dance / 16-Performance art / \n", "10-Animal / 18-Outdoor recreation / 91-Fish / 275-Diving / 287-Human swimming / 288-Underwater / 381-Underwater diving / \n", "539-Home appliance / 818-Washing machine / 863-Body piercing / 1456-Clothes dryer / \n", "9-Music video / \n", "23-Mobile phone / 29-Smartphone / 72-iPhone / \n", "1-Vehicle / 11-Motorsport / 17-Racing / 27-Motorcycle / 40-Road / 57-Race track / 105-Motorcycling / \n", "15-Cartoon / 123-Naruto / \n", "12-Food / 26-Cooking / 32-Recipe / 52-Dish / 58-Cuisine / 109-Cooking show / 1319-Apple / \n", "8-Football / 21-Stadium / 77-Arena / \n", "164-Beach / 222-Surfing / 815-Surfboard / \n", "9-Music video / \n", "5-Dance / \n", "1465-FOX / \n", "150-Violin / 401-Cello / 450-Viola / \n", "99-Christmas / \n", "1-Vehicle / 4-Car / 11-Motorsport / 41-Sports car / 207-Exhaust system / 234-Ford / 344-Coupé / 473-Ford Mustang / 1493-Supercharger / 1877-Shelby Mustang / \n", "1-Vehicle / 259-Fire / 371-Firefighter / \n", "8-Football / \n", "25-Toy / 200-Doll / 1842-Dollhouse / \n", "0-Games / 256-Tennis / 544-Table tennis / \n", "9-Music video / \n", "126-The Walt Disney Company / 144-Amusement park / 444-Family / 576-Walt Disney World / \n", "9-Music video / \n", "1-Vehicle / 1614-Honda Accord / \n", "23-Mobile phone / 29-Smartphone / 37-Gadget / 101-Telephone / 932-Microsoft Lumia / \n", "20-Trailer / \n", "3-Concert / 7-Musician / 13-Musical ensemble / 30-Drummer / \n", "45-Cosmetics / 111-PlayStation 3 / \n", "9-Music video / 78-Wedding / \n", "1-Vehicle / 18-Outdoor recreation / 42-Fishing / 87-Boat / 92-Radio-controlled model / 147-River / 225-Lake / 233-Ocean / 1490-Radio-controlled boat / \n", "3-Concert / 7-Musician / \n", "1533-Lead guitar / \n", "3-Concert / 7-Musician / 13-Musical ensemble / 30-Drummer / 39-Drums / 44-Drums / \n", "0-Games / 6-Animation / 460-Yu-Gi-Oh! Trading Card Game / \n", "6-Animation / 15-Cartoon / 801-Inuyasha / \n", "6-Animation / 10-Animal / 15-Cartoon / 80-Horse / 290-Pony / 343-My Little Pony / \n", "68-Lighting / 263-Light / 490-Rain / \n", "0-Games / 2-Video game / 19-PC game / 33-Weapon / 43-Call of Duty / 96-Soldier / 138-Call of Duty: Black Ops II / 302-Call of Duty 4: Modern Warfare / 431-Call of Duty: Advanced Warfare / 1311-Amazon.com; Inc. / \n", "20-Trailer / \n", "5-Dance / 49-School / 191-Cheerleading / \n", "3-Concert / \n", "9-Music video / \n", "3-Concert / 5-Dance / 31-Disc jockey / \n", "75-Wrestling / \n", "18-Outdoor recreation / 163-Running / 674-Marathon / \n", "9-Music video / \n", "0-Games / 60-Basketball / 168-Basketball moves / 255-Slam dunk / \n", "1-Vehicle / 4-Car / 158-Tire / \n", "14-Guitar / 24-String instrument / 100-Electric guitar / \n", "3-Concert / \n", "9-Music video / \n", "10-Animal / 384-Horse racing / \n", "56-Hair / \n", "1026-Foreign Exchange Market / \n", "1-Vehicle / 4-Car / \n", "12-Food / 26-Cooking / 32-Recipe / 367-Drink / 613-Bar / 904-Cocktail / 1065-Bartender / 1176-Pumpkin / \n", "0-Games / 19-PC game / 35-Minecraft / \n", "5-Dance / 78-Wedding / 190-Bride / 3693-Groomsman / \n", "6-Animation / 4624-Little Battlers Experience / \n", "25-Toy / 200-Doll / 412-Monster / 503-Monster High / \n", "7-Musician / 30-Drummer / 39-Drums / 44-Drums / 128-Cymbal / \n", "23-Mobile phone / 29-Smartphone / 1032-BlackBerry / \n", "3-Concert / 5-Dance / 16-Performance art / 133-Nightclub / \n", "22-Nature / \n", "0-Games / 1-Vehicle / 2-Video game / 4-Car / 34-Action-adventure game / 201-Grand Theft Auto: San Andreas / 921-Grand Theft Auto: Vice City / \n", "23-Mobile phone / 29-Smartphone / 37-Gadget / 72-iPhone / 178-Samsung Galaxy / 554-iPhone 4 / \n", "1-Vehicle / 266-Helicopter / \n", "25-Toy / 735-Kamen Rider Series / \n", "121-Photography / \n", "1-Vehicle / 4-Car / 41-Sports car / 234-Ford / 1297-Ford Focus / \n", "9-Music video / 388-Fighter aircraft / \n", "10-Animal / 18-Outdoor recreation / 80-Horse / 247-Stallion / 404-Livestock / 438-Mare / 2553-Colt / \n", "246-Furniture / \n", "59-Winter sport / 84-Snow / 120-Winter / 177-Skiing / 305-Ski / 789-Alpine skiing / \n", "1-Vehicle / 4-Car / 27-Motorcycle / 1124-Yamaha YZF-R1 / \n", "615-Poker / \n", "5-Dance / 10-Animal / 78-Wedding / 80-Horse / 98-Festival / 2583-Limousine / \n", "9-Music video / 13-Musical ensemble / \n", "6-Animation / 75-Wrestling / 1844-WWE action figures / \n", "3-Concert / 93-Comedy / \n", "5-Dance / 440-Lion / \n", "99-Christmas / 2649-Nativity scene / \n", "1-Vehicle / 271-Tank / \n", "364-Sitcom / \n", "3-Concert / 5-Dance / \n", "0-Games / 2-Video game / 19-PC game / 33-Weapon / 85-Combat / 96-Soldier / 111-PlayStation 3 / 257-Counter-Strike / \n", "5-Dance / 16-Performance art / \n", "25-Toy / \n", "31-Disc jockey / \n", "0-Games / 9-Music video / 35-Minecraft / \n", "3-Concert / 38-Orchestra / \n", "8-Football / 21-Stadium / 54-Highlight film / \n", "66-Bollywood / \n", "20-Trailer / \n", "170-Home improvement / 416-Apartment / \n", "88-Machine / 812-Plastic / \n", "1-Vehicle / 4-Car / 11-Motorsport / 17-Racing / 57-Race track / 378-Drag racing / 473-Ford Mustang / 647-Muscle car / \n", "489-Board game / 799-Chess / \n", "23-Mobile phone / 29-Smartphone / 37-Gadget / 516-Headphones / 832-Headset / \n", "0-Games / 2-Video game / 19-PC game / 43-Call of Duty / 138-Call of Duty: Black Ops II / 422-Call of Duty: Ghosts / \n", "205-Newscaster / 223-Weather / \n", "0-Games / 2-Video game / 243-Sonic the Hedgehog / 251-Sonic the Hedgehog / 1959-Phantasy Star / \n", "1-Vehicle / 4-Car / 11-Motorsport / 17-Racing / 40-Road / 41-Sports car / 57-Race track / 70-Driving / 231-Supercar / 711-Gran Turismo / 953-Gran Turismo 5 / 1635-Ferrari 458 / \n", "18-Outdoor recreation / 22-Nature / 147-River / 1195-Rafting / \n", "3-Concert / 46-Choir / \n", "1-Vehicle / 4-Car / 27-Motorcycle / 67-Cycling / 69-Bicycle / \n", "3-Concert / \n", "3-Concert / 16-Performance art / \n", "5-Dance / \n", "12-Food / 26-Cooking / 179-Kitchen / \n", "0-Games / 2-Video game / 125-Television / \n", "23-Mobile phone / 29-Smartphone / 37-Gadget / 72-iPhone / 101-Telephone / \n", "47-Personal computer / 228-Laptop / 465-Microphone / 513-Computer keyboard / \n", "1082-Knot / 1369-Necktie / \n", "9-Music video / \n", "1-Vehicle / 4-Car / 229-Halo / 522-Helmet / \n", "1-Vehicle / 64-Transport / 286-Bus / \n", "1-Vehicle / 11-Motorsport / 17-Racing / 57-Race track / 59-Winter sport / 84-Snow / 120-Winter / 610-Snowmobile / \n", "8-Football / \n", "0-Games / 2-Video game / 34-Action-adventure game / 1116-Goat / 3147-Goat Simulator / \n", "9-Music video / 22-Nature / 213-Forest / \n", "36-Piano / \n", "9-Music video / \n", "3-Concert / 5-Dance / 98-Festival / 126-The Walt Disney Company / 144-Amusement park / 250-Parade / 576-Walt Disney World / 679-Mickey Mouse / \n", "9-Music video / \n", "209-University / \n", "8-Football / 623-Coach / \n", "6-Animation / \n", "3-Concert / 7-Musician / 13-Musical ensemble / 46-Choir / \n", "0-Games / 35-Minecraft / 425-City / \n", "67-Cycling / 69-Bicycle / 83-Skateboarding / 108-Skateboard / 221-Skatepark / 336-Scooter / 377-BMX bike / \n", "1-Vehicle / 68-Lighting / 263-Light / 744-Window / 1020-Emergency vehicle / 3070-Emergency vehicle lighting / \n", "25-Toy / 626-Puzzle / 631-Cube / 740-Rubik's Cube / \n", "0-Games / 1-Vehicle / 2-Video game / 11-Motorsport / 17-Racing / 27-Motorcycle / 57-Race track / 172-Motocross / 2930-MX vs. ATV / \n", "5-Dance / \n", "30-Drummer / 39-Drums / 44-Drums / 128-Cymbal / 146-Snare drum / \n", "28-Fashion / 66-Bollywood / \n", "1726-Harlem Shake (meme) / \n", "102-Building / \n", "1-Vehicle / \n", "3-Concert / \n", "12-Food / 720-Wine / \n", "10-Animal / 91-Fish / \n", "842-Harp / \n", "56-Hair / 106-Hairstyle / 1089-Wire / 2697-Zebra / \n", "12-Food / 26-Cooking / 32-Recipe / 454-Sauce / 671-Tomato / 752-Pasta / \n", "707-Roof / \n", "915-Town / \n", "3-Concert / 5-Dance / 14-Guitar / 16-Performance art / 361-Flamenco / \n", "12-Food / 232-Cake / \n", "0-Games / 21-Stadium / 75-Wrestling / 77-Arena / \n", "61-Art / 479-Graffiti / 911-Studio / \n", "3-Concert / 7-Musician / 125-Television / \n", "0-Games / 2-Video game / 35-Minecraft / \n", "3-Concert / 16-Performance art / \n", "0-Games / 2-Video game / 34-Action-adventure game / \n", "2277-The Lord of the Rings Online / 2759-Lute / \n", "31-Disc jockey / \n", "25-Toy / 140-LEGO / 290-Pony / 343-My Little Pony / 524-Littlest Pet Shop / 1192-Hello Kitty / 3899-Lego Minecraft / \n", "2-Video game / 47-Personal computer / 1028-Devil May Cry / 1893-Devil May Cry 4 / \n", "3-Concert / 5-Dance / 95-Talent show / \n", "10-Animal / \n", "1016-Glove / \n", "1-Vehicle / 62-Train / 64-Transport / \n", "9-Music video / \n", "23-Mobile phone / 86-Plant / 155-Camera / 1117-Camcorder / \n", "0-Games / 1138-Tower of Saviors / \n", "1-Vehicle / 4-Car / 11-Motorsport / 17-Racing / 41-Sports car / 378-Drag racing / 935-Off-road racing / 1699-Cart / 4003-Baja 1000 / \n", "20-Trailer / \n", "130-Gymnastics / 1451-Ribbon / \n", "5-Dance / 95-Talent show / 130-Gymnastics / 1835-Majorette / \n", "0-Games / 8-Football / 65-Kick / 76-Ball / \n", "6-Animation / 15-Cartoon / 116-Pokémon / 4379-Team Rocket / \n", "0-Games / 2-Video game / 19-PC game / 127-Xbox 360 / 1196-Dragon Age / 2179-Dragon Age: Origins / \n", "10-Animal / 80-Horse / 438-Mare / \n", "22-Nature / 36-Piano / 53-Keyboard / 1500-Cherry blossom / \n", "75-Wrestling / 85-Combat / \n", "3789-Circular saw / \n", "9-Music video / \n", "222-Surfing / 329-Swimming pool / \n", "45-Cosmetics / 173-Eye shadow / 193-Eye liner / 195-Mascara / 196-Eye / 218-Eyelash / 235-Lipstick / 291-Rouge / 436-Concealer / \n", "205-Newscaster / 913-President / \n", "1-Vehicle / 4-Car / 3762-PSA HDi engine / \n", "350-Circus / 788-Juggling / 1100-Hat / \n", "67-Cycling / 69-Bicycle / 220-Mountain bike / 284-Mountain biking / \n", "3-Concert / 7-Musician / 13-Musical ensemble / \n", "56-Hair / 106-Hairstyle / 340-Afro-textured hair / 1269-Human hair color / \n", "6-Animation / 1809-Tokyo Mew Mew / \n", "248-Hotel / \n", "5-Dance / 9-Music video / 133-Nightclub / \n", "0-Games / 2-Video game / 34-Action-adventure game / 127-Xbox 360 / 243-Sonic the Hedgehog / 251-Sonic the Hedgehog / 2028-Sonic World Adventure / \n", "0-Games / 2-Video game / \n", "83-Skateboarding / 108-Skateboard / \n", "20-Trailer / \n", "6-Animation / 15-Cartoon / 396-One Piece / \n", "0-Games / 2-Video game / 55-Video game console / 199-Wii / 3325-Super Paper Mario / 4193-Pit / \n", "16-Performance art / 31-Disc jockey / \n", "208-Wood / 943-Stove / 2707-Wood-burning stove / \n", "253-Carnival / \n", "66-Bollywood / \n", "1-Vehicle / 25-Toy / 62-Train / 119-Locomotive / 140-LEGO / 1079-Lego City / \n", "3461-Noah's Ark / \n", "9-Music video / \n", "14-Guitar / \n", "12-Food / 86-Plant / 136-Gardening / 671-Tomato / \n", "1-Vehicle / 4-Car / 1414-Gasoline / 2616-Chevrolet S-10 / \n", "117-Boxing / \n", "0-Games / 489-Board game / 922-Tabletop game / \n", "45-Cosmetics / 193-Eye liner / 195-Mascara / 196-Eye / 218-Eyelash / 235-Lipstick / 1012-Lip gloss / \n", "361-Flamenco / \n", "9-Music video / \n", "93-Comedy / \n", "36-Piano / \n", "0-Games / 489-Board game / \n", "209-University / 1246-Campus / \n", "3-Concert / 38-Orchestra / \n", "28-Fashion / \n", "0-Games / 2-Video game / 19-PC game / 51-Strategy video game / \n", "10-Animal / 48-Pet / 1147-Hamster / \n", "0-Games / 2-Video game / 43-Call of Duty / 134-Call of Duty: Black Ops / 194-Call of Duty: Modern Warfare 2 / \n", "295-Chocolate / \n", "0-Games / 2-Video game / 51-Strategy video game / 382-Dota 2 / 549-Defense of the Ancients / \n", "9-Music video / \n", "162-Medicine / \n", "8-Football / 21-Stadium / \n", "45-Cosmetics / 173-Eye shadow / 193-Eye liner / 195-Mascara / 196-Eye / 218-Eyelash / 235-Lipstick / 291-Rouge / \n", "1289-Caporales / \n", "7-Musician / 9-Music video / 14-Guitar / \n", "9-Music video / \n", "23-Mobile phone / 29-Smartphone / 1820-Motorola Droid / \n", "208-Wood / 492-Floor / 1179-Brick / 1869-Cement / \n", "1-Vehicle / 50-Aircraft / 129-Aviation / 139-Airport / 151-Landing / 175-Airliner / 187-Takeoff / 188-Airline / 254-Runway / 923-Boeing 777 / \n", "59-Winter sport / 112-Ice skating / 262-Ice / 312-Ice rink / 316-Figure skating / 632-Ice dancing / 833-Ice skate / \n", "1-Vehicle / \n", "12-Food / 612-Beer / \n", "93-Comedy / 341-Sketch comedy / 2764-Mad / \n", "0-Games / 60-Basketball / 168-Basketball moves / 255-Slam dunk / 1101-Point guard / \n", "6-Animation / 15-Cartoon / \n", "0-Games / 59-Winter sport / 112-Ice skating / 142-Hockey / 312-Ice rink / 342-Windows Media Video / \n", "31-Disc jockey / \n", "5-Dance / \n", "8-Football / 21-Stadium / \n", "0-Games / 488-Door / \n", "3-Concert / 7-Musician / 13-Musical ensemble / 24-String instrument / 38-Orchestra / 1081-Lemon / \n", "1-Vehicle / 4-Car / 11-Motorsport / 17-Racing / 982-V8 engine / 1466-Pontiac / \n", "6-Animation / 310-The Sims / 328-Simulation video game / 564-The Sims 3 / \n", "49-School / 410-Teacher / 845-Classroom / \n", "10-Animal / 80-Horse / 247-Stallion / 289-Dressage / 351-Jumping / \n", "10-Animal / 48-Pet / 71-Dog / 182-Puppy / 2283-Beagle / \n", "0-Games / 2-Video game / 199-Wii / 518-Mario Kart / 887-King Koopa / \n", "507-Thomas the Tank Engine / \n", "1-Vehicle / 62-Train / 64-Transport / 103-Rail transport / 119-Locomotive / 143-Railroad car / 152-Track / 161-Train station / 224-Rapid transit / \n", "0-Games / 1-Vehicle / 2-Video game / 4-Car / 34-Action-adventure game / 40-Road / 297-Grand Theft Auto IV / 658-Grand Theft Auto: The Lost and Damned / \n", "10-Animal / 48-Pet / 166-Cat / \n", "0-Games / 8-Football / 21-Stadium / 54-Highlight film / 73-Ball / 79-American football / 81-Athlete / 358-Rugby football / \n", "31-Disc jockey / \n", "1-Vehicle / 4-Car / 11-Motorsport / 17-Racing / \n", "0-Games / 8-Football / 21-Stadium / 65-Kick / 73-Ball / \n", "341-Sketch comedy / 364-Sitcom / \n", "6-Animation / 15-Cartoon / 1946-Total Drama Island / \n", "1-Vehicle / 11-Motorsport / 17-Racing / 27-Motorcycle / 105-Motorcycling / 172-Motocross / 197-Trail / 491-Enduro / \n", "1-Vehicle / 4-Car / 11-Motorsport / 41-Sports car / 449-Honda / 2223-Honda S2000 / \n", "9-Music video / \n", "1-Vehicle / 124-Tractor / 156-Agriculture / 174-Farm / 715-Farming Simulator / \n", "3-Concert / 7-Musician / \n", "0-Games / 394-Map / 910-Dragon Quest / 1426-Treasure / \n", "1-Vehicle / 4-Car / 345-Loudspeaker / 484-Subwoofer / \n", "49-School / 2191-Text messaging / \n", "3-Concert / 7-Musician / 13-Musical ensemble / 14-Guitar / 38-Orchestra / 44-Drums / \n", "230-Brass instrument / 296-Saxophone / 1006-Tenor saxophone / \n", "1-Vehicle / 50-Aircraft / 92-Radio-controlled model / 149-Model aircraft / 160-Radio-controlled aircraft / 266-Helicopter / 392-Radio-controlled helicopter / \n", "7-Musician / 14-Guitar / 100-Electric guitar / \n", "3-Concert / 68-Lighting / \n", "1-Vehicle / 67-Cycling / 69-Bicycle / 220-Mountain bike / 284-Mountain biking / 520-Bicycle frame / \n", "10-Animal / 91-Fish / 258-Aquarium / \n", "7-Musician / 9-Music video / 30-Drummer / \n", "8-Football / 21-Stadium / 77-Arena / \n", "125-Television / 1594-Nerd / \n", "10-Animal / 18-Outdoor recreation / 22-Nature / 42-Fishing / 91-Fish / 288-Underwater / 481-Kayak / \n", "0-Games / 2-Video game / 90-Sports game / 180-FIFA 15 / \n", "28-Fashion / 56-Hair / 106-Hairstyle / 340-Afro-textured hair / 583-Wig / 857-Afro / \n", "20-Trailer / 294-Batman / 2146-Bane / \n", "36-Piano / 53-Keyboard / \n", "23-Mobile phone / 29-Smartphone / 37-Gadget / 101-Telephone / \n", "0-Games / 2-Video game / 55-Video game console / \n", "86-Plant / \n", "0-Games / 2-Video game / 19-PC game / 51-Strategy video game / 165-League of Legends / 1494-Jungle / \n", "12-Food / 26-Cooking / 32-Recipe / 52-Dish / 58-Cuisine / 109-Cooking show / 210-Cookware and bakeware / 239-Roasting / 688-Beef / 1332-Onion / \n", "25-Toy / 99-Christmas / 323-Toddler / 466-Gift / \n", "1088-Infomercial / \n", "6-Animation / 25-Toy / 435-Action figure / 477-Transformers / 4445-Transformers: Beast Wars / \n", "0-Games / 2-Video game / 872-Diablo / 2054-Diablo II / 3509-Diablo II: Lord of Destruction / \n", "0-Games / 2-Video game / 257-Counter-Strike / 702-Counter-Strike: Source / \n", "88-Machine / 666-Concrete / 1678-Foam / \n", "1-Vehicle / 4-Car / 74-Truck / \n", "3-Concert / \n", "5-Dance / \n", "1-Vehicle / 4-Car / \n", "12-Food / 32-Recipe / 226-Dessert / 232-Cake / 236-Baking / 295-Chocolate / 437-Cream / 560-Sugar / 578-Cookie / 600-Flour / 759-Icing / 795-Cupcake / 1102-Cake decorating / 1365-Fondant icing / 3087-Sugar paste / \n", "0-Games / 34-Action-adventure game / 201-Grand Theft Auto: San Andreas / \n", "18-Outdoor recreation / 573-Paragliding / 778-Gliding / 1602-Hang gliding / \n", "8-Football / \n", "0-Games / 35-Minecraft / \n", "45-Cosmetics / 173-Eye shadow / 193-Eye liner / 196-Eye / 291-Rouge / \n", "5-Dance / 16-Performance art / \n", "0-Games / 51-Strategy video game / 373-Clash of Clans / \n", "6-Animation / 15-Cartoon / 185-Dragon Ball / 244-Goku / 4618-Android 17 / \n", "191-Cheerleading / \n", "14-Guitar / 24-String instrument / 63-Acoustic guitar / \n", "0-Games / 6-Animation / 15-Cartoon / \n", "3-Concert / \n", "6-Animation / 15-Cartoon / 183-Manga / \n", "372-Playing card / 746-Card manipulation / \n", "3-Concert / \n", "9-Music video / \n", "0-Games / 2-Video game / 89-Comics / 185-Dragon Ball / \n", "5-Dance / 79-American football / \n", "0-Games / 2-Video game / 19-PC game / 33-Weapon / 264-Star Wars / \n", "318-Church / \n", "248-Hotel / 369-Resort / \n", "28-Fashion / 45-Cosmetics / 469-Bag / \n", "708-Mermaid / 981-Sculpture / 2456-Tail / \n", "6-Animation / 15-Cartoon / 20-Trailer / \n", "6-Animation / 2544-Megami Tensei / 3774-Shin Megami Tensei / \n", "3-Concert / 5-Dance / 49-School / 95-Talent show / 191-Cheerleading / 595-Champion / \n", "1-Vehicle / 4-Car / 299-Volkswagen Passenger Cars / 4067-Škoda Fabia / \n", "85-Combat / 117-Boxing / \n", "121-Photography / 633-Photograph / \n", "35-Minecraft / \n", "1-Vehicle / 11-Motorsport / 27-Motorcycle / 40-Road / 105-Motorcycling / 172-Motocross / 608-Wheelie / 927-Supermoto / 2975-BMW Motorrad / \n", "0-Games / 51-Strategy video game / 165-League of Legends / \n", "1-Vehicle / 4-Car / 40-Road / 70-Driving / 74-Truck / 94-Dashcam / 212-Highway / \n", "372-Playing card / 746-Card manipulation / \n", "12-Food / 26-Cooking / 32-Recipe / 109-Cooking show / 179-Kitchen / 442-Bread / 475-Cheese / 542-Potato / 3003-Roll / \n", "1-Vehicle / 9-Music video / \n", "6-Animation / 15-Cartoon / 185-Dragon Ball / 244-Goku / \n", "7-Musician / 78-Wedding / \n", "31-Disc jockey / \n", "12-Food / 26-Cooking / 52-Dish / 58-Cuisine / 215-Eating / 301-Meal / 389-Restaurant / 1949-Octopus / \n", "5-Dance / \n", "6-Animation / 15-Cartoon / 320-Super Smash Bros. / \n", "5-Dance / 20-Trailer / \n", "96-Soldier / 736-ARMA / \n", "0-Games / 2-Video game / 19-PC game / 96-Soldier / 192-Battlefield / 374-Battlefield 3 / \n", "1-Vehicle / 4-Car / \n", "261-Tool / \n", "0-Games / 6-Animation / 15-Cartoon / 547-Bleach / 3562-Hollow / 3744-Tōshirō Hitsugaya / 4642-Kenpachi Zaraki / \n", "1-Vehicle / 11-Motorsport / 27-Motorcycle / \n", "8-Football / 21-Stadium / \n", "6-Animation / 113-House / 246-Furniture / 416-Apartment / \n", "1-Vehicle / 4-Car / 234-Ford / 2514-Seat / \n", "171-Warcraft / \n", "9-Music video / \n", "6-Animation / 15-Cartoon / 624-Number / \n", "93-Comedy / \n", "0-Games / 2-Video game / 159-World of Warcraft / 1356-Paladin / \n", "8-Football / \n", "3-Concert / 68-Lighting / \n", "1-Vehicle / 2-Video game / 4-Car / 70-Driving / 111-PlayStation 3 / 127-Xbox 360 / 352-Need for Speed / 2820-Need For Speed: ProStreet / \n", "1510-Liquid / \n", "5-Dance / 16-Performance art / \n", "323-Toddler / 783-Zee TV / \n", "376-Flute / \n", "12-Food / 26-Cooking / 32-Recipe / 52-Dish / 58-Cuisine / 109-Cooking show / 210-Cookware and bakeware / 239-Roasting / 274-Meat / \n", "3-Concert / \n", "9-Music video / \n", "1-Vehicle / 4-Car / \n", "0-Games / 2-Video game / 19-PC game / 34-Action-adventure game / 656-Metal Gear / 3284-Metal Gear Solid 2: Sons of Liberty / \n", "0-Games / 2-Video game / 237-Final Fantasy / \n", "28-Fashion / \n", "3-Concert / \n", "3-Concert / \n", "8-Football / \n", "0-Games / 35-Minecraft / \n", "78-Wedding / 1427-TalesRunner / \n", "7-Musician / 30-Drummer / 39-Drums / 44-Drums / 584-Super Mario Bros. / \n", "10-Animal / 25-Toy / 550-Dinosaur / 1126-Tyrannosaurus / \n", "5-Dance / 49-School / 95-Talent show / 191-Cheerleading / \n", "6-Animation / 547-Bleach / \n", "3-Concert / 7-Musician / \n", "30-Drummer / 39-Drums / 44-Drums / 128-Cymbal / 146-Snare drum / \n", "1-Vehicle / 4-Car / 11-Motorsport / 167-Four-wheel drive / 197-Trail / 198-Off-road vehicle / 327-Mud / 360-Jeep / \n", "1-Vehicle / 4-Car / 299-Volkswagen Passenger Cars / 526-Volkswagen / 969-Volkswagen Beetle / \n", "3063-Friendship bracelet / \n", "30-Drummer / 39-Drums / 44-Drums / 128-Cymbal / 146-Snare drum / 1603-Fox / \n", "78-Wedding / 1061-Marriage proposal / \n", "75-Wrestling / \n", "14-Guitar / 24-String instrument / 670-Banjo / \n", "0-Games / 90-Sports game / 199-Wii / 3171-Wii Sports / \n", "4235-American Kenpo / \n", "162-Medicine / \n", "1-Vehicle / 4-Car / 11-Motorsport / 17-Racing / 41-Sports car / 57-Race track / 378-Drag racing / \n", "36-Piano / 53-Keyboard / \n", "10-Animal / 166-Cat / \n", "1-Vehicle / 4-Car / 11-Motorsport / \n", "61-Art / 102-Building / \n", "12-Food / 469-Bag / 500-Rice / \n", "6-Animation / 15-Cartoon / \n", "18-Outdoor recreation / 59-Winter sport / 177-Skiing / 305-Ski / \n", "0-Games / 1-Vehicle / 4-Car / 34-Action-adventure game / 297-Grand Theft Auto IV / 1661-Volkswagen Gol / \n", "8-Football / \n", "0-Games / 2-Video game / 19-PC game / 111-PlayStation 3 / 127-Xbox 360 / 1007-Portal / 1274-Portal 2 / \n", "12-Food / 26-Cooking / 32-Recipe / 52-Dish / 58-Cuisine / 215-Eating / 301-Meal / 662-Pizza / 1176-Pumpkin / \n", "3-Concert / 7-Musician / 14-Guitar / \n", "988-Ibiza / \n", "5-Dance / \n", "12-Food / 26-Cooking / 32-Recipe / 58-Cuisine / 109-Cooking show / 210-Cookware and bakeware / 274-Meat / 2313-Bánh / \n", "419-Earth / 712-Outer space / 1480-Astronaut / \n", "3-Concert / 5-Dance / 16-Performance art / 68-Lighting / \n", "3-Concert / 68-Lighting / \n", "75-Wrestling / 85-Combat / \n", "18-Outdoor recreation / 22-Nature / 217-Hunting / 1215-Wild boar / \n", "110-Album / \n", "38-Orchestra / \n", "141-Microsoft Windows / \n", "31-Disc jockey / \n", "17-Racing / 769-Red Bull / \n", "5-Dance / 130-Gymnastics / \n", "22-Nature / 260-Prayer / \n", "3-Concert / 7-Musician / 13-Musical ensemble / 36-Piano / 38-Orchestra / 104-Pianist / \n", "0-Games / 411-MapleStory / \n", "1-Vehicle / 18-Outdoor recreation / 67-Cycling / 69-Bicycle / 220-Mountain bike / 284-Mountain biking / 570-Road bicycle racing / \n", "25-Toy / 200-Doll / \n", "0-Games / 2-Video game / 89-Comics / 387-Spider-Man / 748-Iron Man / 855-Hulk / 897-Avengers / 1173-Captain America / 1982-Venom / 2172-Deadpool / 4401-Iron Man's armor / 4554-Marvel Super Hero Squad Online / \n", "1-Vehicle / 4-Car / 11-Motorsport / 17-Racing / 41-Sports car / 352-Need for Speed / 647-Muscle car / 761-Chevrolet Camaro / 2772-Shift 2: Unleashed / \n", "88-Machine / 319-Robot / 372-Playing card / \n", "0-Games / 8-Football / 90-Sports game / 180-FIFA 15 / \n", "83-Skateboarding / 108-Skateboard / 221-Skatepark / \n", "7-Musician / 14-Guitar / 24-String instrument / 63-Acoustic guitar / \n", "3-Concert / 7-Musician / 46-Choir / \n", "8-Football / \n", "0-Games / 51-Strategy video game / 2990-Command & Conquer: Red Alert / 3430-Command & Conquer: Red Alert 3 / \n", "2-Video game / 20-Trailer / 34-Action-adventure game / 1144-Castlevania / 3768-Castlevania: Lords of Shadow / \n", "2-Video game / 20-Trailer / 55-Video game console / 199-Wii / 320-Super Smash Bros. / 335-Nintendo 3DS / 443-Wii U / 697-Super Smash Bros. for Nintendo 3DS and Wii U / 2750-Mii / \n", "3-Concert / \n", "5-Dance / 9-Music video / \n", "0-Games / 25-Toy / 455-Unidentified flying object / 596-Monster Hunter / 1668-Claw crane / \n", "8-Football / \n", "0-Games / 2-Video game / 19-PC game / 159-World of Warcraft / 171-Warcraft / \n", "1442-Tabla / \n", "59-Winter sport / 84-Snow / 177-Skiing / 278-GoPro / 281-Snowboarding / \n", "46-Choir / \n", "1-Vehicle / 4-Car / 11-Motorsport / 17-Racing / 25-Toy / 92-Radio-controlled model / 135-Drifting / 202-Radio-controlled car / 1414-Gasoline / 2753-SEAT León / \n", "31-Disc jockey / \n", "1-Vehicle / 4-Car / 40-Road / 94-Dashcam / 800-Mountain pass / \n", "0-Games / 1680-Seafight / \n", "0-Games / 51-Strategy video game / 165-League of Legends / \n", "1-Vehicle / 50-Aircraft / 149-Model aircraft / 471-Wing / \n", "12-Food / 640-Bottle / 720-Wine / \n", "3-Concert / 7-Musician / 14-Guitar / 24-String instrument / \n", "162-Medicine / \n", "1-Vehicle / 4-Car / 124-Tractor / 249-Heavy equipment / 363-Soil / 817-Bulldozer / \n", "59-Winter sport / 112-Ice skating / 115-Ballet / 262-Ice / 312-Ice rink / 316-Figure skating / 632-Ice dancing / 833-Ice skate / \n", "6-Animation / 25-Toy / 200-Doll / 439-Barbie / \n", "0-Games / 2-Video game / 25-Toy / 140-LEGO / 264-Star Wars / 714-Jedi / 717-Darth Vader / 852-Lego Star Wars / 1393-Lightsaber / 4117-Leia Organa / \n", "12-Food / 26-Cooking / 32-Recipe / 58-Cuisine / 295-Chocolate / 1765-Muffin / 2699-Chocolate chip / \n", "36-Piano / 53-Keyboard / 104-Pianist / 107-Musical keyboard / \n", "0-Games / 2-Video game / 47-Personal computer / 131-Computer / 586-Nintendo Entertainment System / \n", "9-Music video / \n", "12-Food / 26-Cooking / 32-Recipe / 52-Dish / 58-Cuisine / 391-Chicken meat / 424-Chicken / 1509-Stew / \n", "0-Games / 8-Football / 54-Highlight film / 358-Rugby football / \n", "60-Basketball / 255-Slam dunk / \n", "0-Games / 8-Football / 54-Highlight film / \n", "5-Dance / 2198-Irish dance / \n", "7-Musician / \n", "9-Music video / \n", "3-Concert / 7-Musician / 14-Guitar / \n", "8-Football / \n", "125-Television / \n", "1-Vehicle / \n", "28-Fashion / 206-Shoe / 1800-Football boot / \n", "6-Animation / 10-Animal / 2648-Larva / \n", "6-Animation / 1934-Hunter × Hunter / \n", "5-Dance / 49-School / \n", "20-Trailer / \n", "7-Musician / 14-Guitar / 24-String instrument / 150-Violin / 639-Fiddle / \n", "3-Concert / 7-Musician / 16-Performance art / 30-Drummer / \n", "0-Games / 35-Minecraft / \n", "3-Concert / 7-Musician / 13-Musical ensemble / 14-Guitar / 24-String instrument / \n", "0-Games / 51-Strategy video game / \n", "33-Weapon / 362-PlayStation Portable / \n", "5-Dance / 115-Ballet / 889-Swing / \n", "0-Games / 21-Stadium / 1182-Softball / \n", "1-Vehicle / 4-Car / 11-Motorsport / 17-Racing / 40-Road / 169-Rallying / \n", "75-Wrestling / \n", "6-Animation / \n", "1-Vehicle / 62-Train / 64-Transport / 103-Rail transport / 119-Locomotive / 143-Railroad car / 152-Track / 161-Train station / \n", "12-Food / 86-Plant / 136-Gardening / 156-Agriculture / 542-Potato / \n", "237-Final Fantasy / 3069-Final Fantasy V / \n", "12-Food / 26-Cooking / 32-Recipe / \n", "10-Animal / 18-Outdoor recreation / 22-Nature / 42-Fishing / 91-Fish / 275-Diving / 287-Human swimming / 288-Underwater / 407-Scuba diving / \n", "28-Fashion / 153-Dress / \n", "7-Musician / 38-Orchestra / 230-Brass instrument / \n", "0-Games / \n", "5-Dance / \n", "0-Games / 8-Football / 21-Stadium / 65-Kick / 73-Ball / 81-Athlete / \n", "0-Games / 2-Video game / 25-Toy / 140-LEGO / 264-Star Wars / 852-Lego Star Wars / 4356-R2-D2 / \n", "14-Guitar / 269-Comic book / \n", "125-Television / \n", "1-Vehicle / 4-Car / 40-Road / 70-Driving / 94-Dashcam / 663-Street / 1285-Parking / \n", "2-Video game / 23-Mobile phone / 29-Smartphone / 37-Gadget / 101-Telephone / 178-Samsung Galaxy / 899-Samsung Galaxy S III / 1773-Samsung Galaxy S series / \n", "5-Dance / \n", "1-Vehicle / 286-Bus / \n", "3-Concert / 5-Dance / 16-Performance art / 68-Lighting / \n", "9-Music video / \n", "93-Comedy / \n", "6-Animation / 2237-Kantai Collection / \n", "0-Games / 8-Football / 90-Sports game / 180-FIFA 15 / \n", "28-Fashion / 59-Winter sport / 112-Ice skating / 142-Hockey / \n", "0-Games / 5-Dance / 49-School / 79-American football / 191-Cheerleading / \n", "1-Vehicle / 4-Car / \n", "10-Animal / 1177-Spider / \n", "0-Games / 8-Football / 21-Stadium / 73-Ball / 77-Arena / \n", "1-Vehicle / 4-Car / 509-Classic car / 1455-Cadillac / \n", "358-Rugby football / \n", "23-Mobile phone / 29-Smartphone / 125-Television / \n", "0-Games / 4162-Tales of Pirates / \n", "83-Skateboarding / 108-Skateboard / 433-Construction / 2866-Half-pipe / \n", "0-Games / 2-Video game / 55-Video game console / 199-Wii / 456-Go-kart / 518-Mario Kart / \n", "93-Comedy / \n", "0-Games / 8-Football / 21-Stadium / 73-Ball / 76-Ball / \n", "5-Dance / \n", "1-Vehicle / 11-Motorsport / 27-Motorcycle / 105-Motorcycling / 114-Engine / 1036-Moped / \n", "25-Toy / 262-Ice / 323-Toddler / 417-Play-Doh / 437-Cream / 684-Ice cream / \n", "89-Comics / 123-Naruto / \n", "88-Machine / 400-Woodturning / 1043-Router / \n", "0-Games / 2-Video game / 19-PC game / 35-Minecraft / \n", "12-Food / \n", "7-Musician / 13-Musical ensemble / 44-Drums / 46-Choir / 820-Bagpipes / 1646-Pipe band / \n", "0-Games / 33-Weapon / 834-Combat Arms / \n", "83-Skateboarding / 108-Skateboard / 555-Longboard / 667-Longboarding / \n", "511-Mortal Kombat / \n", "18-Outdoor recreation / 59-Winter sport / 84-Snow / 120-Winter / 177-Skiing / 281-Snowboarding / 305-Ski / 2921-Snowkiting / \n", "3-Concert / 7-Musician / \n", "10-Animal / 48-Pet / 71-Dog / 2283-Beagle / \n", "7-Musician / 886-James Bond / \n", "9-Music video / \n", "1-Vehicle / 4-Car / 864-Speedometer / 2016-Volvo Cars / \n", "1-Vehicle / 4-Car / 74-Truck / 2701-Tow truck / \n", "364-Sitcom / \n", "9-Music video / \n", "0-Games / 60-Basketball / \n", "0-Games / 2-Video game / 14-Guitar / 445-Guitar Hero / 724-Rock Band / 1703-Rock Band / \n", "0-Games / 372-Playing card / 494-Card game / 746-Card manipulation / \n", "0-Games / 8-Football / 21-Stadium / 65-Kick / 73-Ball / \n", "3-Concert / 7-Musician / 13-Musical ensemble / 44-Drums / 689-Harmonica / 2922-Richter-tuned harmonica / \n", "1-Vehicle / 10-Animal / 18-Outdoor recreation / 42-Fishing / 147-River / \n", "15-Cartoon / 648-Clown / 3696-Ronald McDonald / \n", "10-Animal / 22-Nature / 132-Bird / \n", "3816-Indie film / \n", "9-Music video / \n", "31-Disc jockey / \n", "99-Christmas / 558-Santa Claus / \n", "102-Building / 1733-Ant / \n", "3-Concert / 7-Musician / 13-Musical ensemble / 14-Guitar / 30-Drummer / 44-Drums / \n", "0-Games / 2-Video game / 34-Action-adventure game / 176-PlayStation 4 / 294-Batman / 838-Batman: Arkham / 1881-Arkham Asylum / \n", "13-Musical ensemble / 203-Marching band / 250-Parade / \n", "31-Disc jockey / \n", "2-Video game / 31-Disc jockey / 68-Lighting / \n", "0-Games / 2-Video game / 90-Sports game / \n", "28-Fashion / \n", "0-Games / 75-Wrestling / 85-Combat / 90-Sports game / 3764-EA Sports UFC / \n", "22-Nature / \n", "12-Food / 273-News program / \n", "3603-Quartz / \n", "14-Guitar / 24-String instrument / 100-Electric guitar / \n", "1119-Sensor / \n", "3-Concert / 16-Performance art / \n", "8-Football / 79-American football / \n", "286-Bus / \n", "12-Food / 32-Recipe / 162-Medicine / 786-Juice / \n", "7-Musician / 30-Drummer / 39-Drums / 44-Drums / 128-Cymbal / 146-Snare drum / \n", "7-Musician / 14-Guitar / \n", "0-Games / 8-Football / \n", "3-Concert / \n", "75-Wrestling / \n", "0-Games / 1211-Transformice / \n", "0-Games / 8-Football / 54-Highlight film / 65-Kick / 73-Ball / 90-Sports game / 180-FIFA 15 / \n", "6-Animation / 15-Cartoon / 61-Art / \n", "2-Video game / 6-Animation / 20-Trailer / 1523-Alice / \n", "0-Games / 2-Video game / 43-Call of Duty / 96-Soldier / 138-Call of Duty: Black Ops II / \n", "296-Saxophone / \n", "9-Music video / \n", "121-Photography / \n", "6-Animation / 89-Comics / 803-DC Comics / \n", "6-Animation / 10-Animal / 15-Cartoon / 126-The Walt Disney Company / 440-Lion / 1156-Simba / 4202-Timon and Pumbaa / \n", "0-Games / 8-Football / 21-Stadium / 65-Kick / \n", "7-Musician / 9-Music video / 14-Guitar / \n", "28-Fashion / 469-Bag / 676-Handbag / 1163-Carpet / \n", "5-Dance / 115-Ballet / \n", "3-Concert / 7-Musician / 13-Musical ensemble / 38-Orchestra / 133-Nightclub / \n", "59-Winter sport / 112-Ice skating / 142-Hockey / \n", "260-Prayer / \n", "3-Concert / 7-Musician / 13-Musical ensemble / \n", "1-Vehicle / 4-Car / 74-Truck / 158-Tire / 167-Four-wheel drive / 375-Pickup truck / 3541-Nissan Navara / \n", "1-Vehicle / 10-Animal / 124-Tractor / 156-Agriculture / \n", "66-Bollywood / \n", "5-Dance / 110-Album / \n", "89-Comics / 803-DC Comics / \n", "6-Animation / 25-Toy / 524-Littlest Pet Shop / \n", "0-Games / 2-Video game / 35-Minecraft / \n", "45-Cosmetics / 162-Medicine / \n", "0-Games / 6-Animation / 9-Music video / 35-Minecraft / \n", "0-Games / 2-Video game / 6-Animation / 116-Pokémon / 186-Pokémon / 320-Super Smash Bros. / \n", "28-Fashion / 206-Shoe / 757-Boot / \n", "28-Fashion / 956-Glasses / 1481-Sunglasses / \n", "6-Animation / 15-Cartoon / \n", "6-Animation / 15-Cartoon / \n", "9-Music video / \n", "209-University / 293-Slide show / \n", "7-Musician / 14-Guitar / 63-Acoustic guitar / \n", "23-Mobile phone / 29-Smartphone / 37-Gadget / 527-Battery / 1159-Battery charger / \n", "1-Vehicle / 11-Motorsport / 154-Wheel / 158-Tire / 935-Off-road racing / \n", "6-Animation / 61-Art / 97-Drawing / 479-Graffiti / 1385-Street art / \n", "31-Disc jockey / \n", "0-Games / 257-Counter-Strike / 351-Jumping / \n", "1-Vehicle / 8-Football / \n", "256-Tennis / 974-Racket / \n", "1-Vehicle / 4-Car / 11-Motorsport / 17-Racing / 27-Motorcycle / 57-Race track / \n", "23-Mobile phone / 29-Smartphone / 37-Gadget / 72-iPhone / 342-Windows Media Video / \n", "20-Trailer / 743-Edward Cullen / \n", "3-Concert / 16-Performance art / \n", "1-Vehicle / 50-Aircraft / 129-Aviation / 223-Weather / 502-Sky / 929-Aerobatics / \n", "15-Cartoon / 243-Sonic the Hedgehog / 251-Sonic the Hedgehog / \n", "1-Vehicle / 2-Video game / 4-Car / 47-Personal computer / 228-Laptop / 525-Printer / \n", "141-Microsoft Windows / \n", "12-Food / 227-Water / 640-Bottle / \n", "8-Football / \n", "10-Animal / 48-Pet / 71-Dog / 280-Wildlife / 657-Autumn / \n", "10-Animal / 18-Outdoor recreation / 42-Fishing / \n", "5-Dance / 16-Performance art / \n", "0-Games / 2-Video game / 19-PC game / 51-Strategy video game / 165-League of Legends / 643-Hero / 2004-Riot Games / \n", "0-Games / 60-Basketball / 168-Basketball moves / \n", "1-Vehicle / 64-Transport / 286-Bus / 2226-Bus driver / \n", "1-Vehicle / 92-Radio-controlled model / 149-Model aircraft / 160-Radio-controlled aircraft / 266-Helicopter / 392-Radio-controlled helicopter / \n", "61-Art / \n", "0-Games / 34-Action-adventure game / 201-Grand Theft Auto: San Andreas / \n", "49-School / \n", "1-Vehicle / 25-Toy / 62-Train / 507-Thomas the Tank Engine / 1685-Toy train / \n", "30-Drummer / 39-Drums / 44-Drums / 146-Snare drum / 1825-Cajón / 3066-Timbales / \n", "3-Concert / 7-Musician / 13-Musical ensemble / 16-Performance art / \n", "117-Boxing / \n", "5-Dance / 49-School / \n", "6-Animation / 15-Cartoon / \n", "1-Vehicle / 10-Animal / 523-Cattle / 2359-Calf / \n", "6-Animation / 15-Cartoon / 61-Art / 97-Drawing / 148-Painting / 183-Manga / \n", "12-Food / 26-Cooking / 215-Eating / 295-Chocolate / 582-Candy / 985-Strawberry / \n", "25-Toy / 61-Art / 417-Play-Doh / 587-Clay / \n", "3-Concert / 7-Musician / \n", "5-Dance / 49-School / 95-Talent show / 130-Gymnastics / 191-Cheerleading / \n", "15-Cartoon / 61-Art / 97-Drawing / \n", "3-Concert / \n", "3-Concert / 7-Musician / 13-Musical ensemble / 38-Orchestra / 39-Drums / \n", "51-Strategy video game / 174-Farm / 373-Clash of Clans / \n", "20-Trailer / \n", "10-Animal / 80-Horse / \n", "0-Games / 185-Dragon Ball / \n", "205-Newscaster / 223-Weather / \n", "0-Games / 2-Video game / 577-Penguin / \n", "1958-Monastery / \n", "345-Loudspeaker / \n", "51-Strategy video game / 366-Dragon / 373-Clash of Clans / \n", "47-Personal computer / 131-Computer / 228-Laptop / 513-Computer keyboard / \n", "28-Fashion / 415-Sewing / 2873-Kimono / \n", "12-Food / 25-Toy / 126-The Walt Disney Company / 354-Egg / 387-Spider-Man / 611-Kinder Surprise / 713-Egg / \n", "3-Concert / \n", "30-Drummer / 39-Drums / 44-Drums / 128-Cymbal / 146-Snare drum / \n", "0-Games / 2-Video game / 127-Xbox 360 / 1007-Portal / 1274-Portal 2 / \n", "0-Games / 599-Dofus / \n", "25-Toy / 3357-Monica's Gang / \n", "66-Bollywood / \n", "8-Football / \n", "783-Zee TV / \n", "0-Games / 2-Video game / \n", "1-Vehicle / 67-Cycling / 69-Bicycle / 213-Forest / 220-Mountain bike / 284-Mountain biking / \n", "6-Animation / 10-Animal / 15-Cartoon / 126-The Walt Disney Company / \n", "7-Musician / 14-Guitar / 100-Electric guitar / 1747-Erhu / \n", "59-Winter sport / 112-Ice skating / 142-Hockey / 623-Coach / \n", "79-American football / \n", "6-Animation / 4708-Visual novel / \n", "0-Games / 6-Animation / 99-Christmas / 679-Mickey Mouse / \n", "0-Games / 59-Winter sport / 77-Arena / 112-Ice skating / 142-Hockey / 312-Ice rink / \n", "155-Camera / 278-GoPro / \n", "15-Cartoon / 538-Illustration / 4122-Marceline the Vampire Queen / \n", "5-Dance / 7-Musician / 13-Musical ensemble / 38-Orchestra / \n", "1-Vehicle / 27-Motorcycle / \n", "137-Tree / 158-Tire / 315-Knife / 1468-Piston / \n", "8-Football / 21-Stadium / 77-Arena / \n", "0-Games / 535-Slot machine / 2301-Dead or Alive / \n", "8-Football / \n", "3770-Toothpaste / \n", "551-British Broadcasting Corporation / \n", "3-Concert / 5-Dance / \n", "6-Animation / 61-Art / 97-Drawing / 183-Manga / 528-Pencil / \n", "3-Concert / \n", "12-Food / 389-Restaurant / \n", "23-Mobile phone / 29-Smartphone / 37-Gadget / 101-Telephone / \n", "0-Games / 2-Video game / 6-Animation / 10-Animal / 132-Bird / 847-Angry Birds / 3319-Angry Birds / \n", "23-Mobile phone / 29-Smartphone / 72-iPhone / \n", "13-Musical ensemble / 2570-Constellation / \n", "298-Television advertisement / \n", "47-Personal computer / 145-Tablet computer / \n", "0-Games / 2-Video game / 34-Action-adventure game / 47-Personal computer / 201-Grand Theft Auto: San Andreas / \n", "0-Games / 2-Video game / 55-Video game console / \n", "1-Vehicle / 4-Car / 11-Motorsport / 17-Racing / 27-Motorcycle / 57-Race track / 378-Drag racing / \n", "7-Musician / 14-Guitar / 24-String instrument / 639-Fiddle / 670-Banjo / \n", "47-Personal computer / 228-Laptop / \n", "1-Vehicle / 50-Aircraft / 82-Airplane / 92-Radio-controlled model / 129-Aviation / 149-Model aircraft / 157-Jet aircraft / 160-Radio-controlled aircraft / 388-Fighter aircraft / 2510-McDonnell Douglas F/A-18 Hornet / \n", "275-Diving / 287-Human swimming / \n", "12-Food / 274-Meat / \n", "1311-Amazon.com; Inc. / \n", "827-Pump / \n", "3-Concert / 5-Dance / 16-Performance art / 133-Nightclub / \n", "0-Games / 19-PC game / 35-Minecraft / \n", "6-Animation / 12-Food / 536-Glass / \n", "45-Cosmetics / 56-Hair / 78-Wedding / 218-Eyelash / \n", "467-Chipmunk / \n", "204-Gym / 532-Barbell / 1584-Kettlebell / 4441-Functional training / \n", "1-Vehicle / 18-Outdoor recreation / 144-Amusement park / 321-Roller coaster / 334-Amusement ride / \n", "0-Games / 1-Vehicle / 2-Video game / 4-Car / 11-Motorsport / 17-Racing / 40-Road / 41-Sports car / 57-Race track / 352-Need for Speed / 875-Need for Speed: Most Wanted / 1577-Need for Speed: Most Wanted / \n", "10-Animal / 48-Pet / 132-Bird / 598-Parrot / \n", "0-Games / 51-Strategy video game / 373-Clash of Clans / \n", "6-Animation / 15-Cartoon / 61-Art / 97-Drawing / 183-Manga / 185-Dragon Ball / 244-Goku / \n", "1264-Head / \n", "0-Games / 1290-Dragon Nest / \n", "36-Piano / \n", "10-Animal / 1108-Mushroom / \n", "5-Dance / 16-Performance art / \n", "0-Games / 21-Stadium / 54-Highlight film / 81-Athlete / 562-Pitcher / 946-Baseball park / \n", "3-Concert / 7-Musician / 13-Musical ensemble / 30-Drummer / \n", "1035-Bank / \n", "12-Food / 56-Hair / \n", "0-Games / 2-Video game / 33-Weapon / 43-Call of Duty / 216-Call of Duty: Modern Warfare 3 / \n", "3-Concert / 7-Musician / 13-Musical ensemble / 38-Orchestra / 46-Choir / \n", "9-Music video / 15-Cartoon / \n", "60-Basketball / 2547-Staples Center / \n", "18-Outdoor recreation / 83-Skateboarding / 108-Skateboard / 221-Skatepark / \n", "1-Vehicle / 22-Nature / 42-Fishing / 87-Boat / 92-Radio-controlled model / 1490-Radio-controlled boat / 3108-Fishing vessel / \n", "2573-Domestic pigeon / \n", "1424-Coupon / \n", "5-Dance / 9-Music video / \n", "0-Games / 2-Video game / 35-Minecraft / 174-Farm / \n", "6-Animation / 10-Animal / \n", "0-Games / 1-Vehicle / 4-Car / 17-Racing / 2400-Asphalt / \n", "36-Piano / 53-Keyboard / 104-Pianist / 107-Musical keyboard / \n", "0-Games / 2-Video game / 22-Nature / 734-DayZ / 736-ARMA / 1034-ARMA 2 / 1314-Garage / \n", "3-Concert / 7-Musician / 13-Musical ensemble / 16-Performance art / 38-Orchestra / \n", "444-Family / \n", "31-Disc jockey / \n", "581-PlayStation / 973-Nintendo 64 / \n", "5-Dance / 7-Musician / 13-Musical ensemble / 203-Marching band / 2134-Teddy bear / \n", "12-Food / 88-Machine / \n", "1-Vehicle / 50-Aircraft / 139-Airport / 157-Jet aircraft / 451-Cockpit / 1705-Boeing 767 / \n", "45-Cosmetics / 282-Skin / 385-Face / \n", "915-Town / \n", "36-Piano / 53-Keyboard / 104-Pianist / 107-Musical keyboard / \n", "5-Dance / 16-Performance art / \n", "3-Concert / 133-Nightclub / \n", "9-Music video / \n", "33-Weapon / 273-News program / \n", "61-Art / 457-Knitting / 546-Stitch / 653-Thread / \n", "7-Musician / 9-Music video / 13-Musical ensemble / 14-Guitar / \n", "115-Ballet / \n", "31-Disc jockey / \n", "1-Vehicle / 27-Motorcycle / 271-Tank / \n", "6-Animation / 10-Animal / 323-Toddler / \n", "3-Concert / 16-Performance art / \n", "18-Outdoor recreation / 163-Running / 674-Marathon / \n", "0-Games / 2-Video game / 638-Gears of War / \n", "6-Animation / 15-Cartoon / 2047-Popeye / \n", "12-Food / 18-Outdoor recreation / 58-Cuisine / 239-Roasting / 770-Banana / \n", "141-Microsoft Windows / 512-Printing / 525-Printer / \n", "0-Games / 8-Football / 21-Stadium / 65-Kick / 73-Ball / 77-Arena / 81-Athlete / \n", "0-Games / 60-Basketball / 168-Basketball moves / 255-Slam dunk / \n", "286-Bus / 387-Spider-Man / \n", "248-Hotel / \n", "3-Concert / 68-Lighting / \n", "427-Trumpet / 499-Cloud / 502-Sky / \n", "293-Slide show / \n", "3-Concert / 7-Musician / 13-Musical ensemble / 14-Guitar / 24-String instrument / 39-Drums / 689-Harmonica / \n", "14-Guitar / 63-Acoustic guitar / \n", "1-Vehicle / 4-Car / 40-Road / 167-Four-wheel drive / 198-Off-road vehicle / 286-Bus / 299-Volkswagen Passenger Cars / 327-Mud / \n", "5-Dance / 95-Talent show / 191-Cheerleading / \n", "20-Trailer / 886-James Bond / \n", "0-Games / 81-Athlete / 562-Pitcher / 946-Baseball park / 2150-Baseball bat / \n", "0-Games / 8-Football / 54-Highlight film / \n", "12-Food / 26-Cooking / 32-Recipe / 58-Cuisine / 424-Chicken / 557-Salad / \n", "1-Vehicle / 4-Car / 11-Motorsport / 17-Racing / 27-Motorcycle / 40-Road / \n", "14-Guitar / 100-Electric guitar / \n", "13-Musical ensemble / \n", "0-Games / 2-Video game / 55-Video game console / 127-Xbox 360 / 181-Xbox / 1027-Terraria / \n", "694-Tai chi / \n", "5-Dance / 16-Performance art / 708-Mermaid / \n", "0-Games / 19-PC game / 51-Strategy video game / 165-League of Legends / \n", "3-Concert / 7-Musician / 545-Quartet (ensemble) / \n", "309-Computer hardware / \n", "61-Art / \n", "23-Mobile phone / 29-Smartphone / 178-Samsung Galaxy / 1727-Samsung Galaxy Note 3 / \n", "0-Games / 8-Football / 54-Highlight film / 76-Ball / 358-Rugby football / \n", "22-Nature / 86-Plant / 136-Gardening / 279-Garden / \n", "3-Concert / 21-Stadium / \n", "0-Games / 2-Video game / 229-Halo / 518-Mario Kart / 644-Halo: Reach / 3225-Princess Daisy / \n", "10-Animal / 280-Wildlife / \n", "0-Games / 49-School / 54-Highlight film / 60-Basketball / 168-Basketball moves / \n", "0-Games / 6-Animation / \n", "8-Football / \n", "842-Harp / \n", "215-Eating / \n", "0-Games / 307-Hand / 311-Street Fighter / \n", "141-Microsoft Windows / \n", "1533-Lead guitar / \n", "7-Musician / 14-Guitar / 24-String instrument / 63-Acoustic guitar / \n", "45-Cosmetics / \n", "1-Vehicle / 4-Car / 11-Motorsport / 17-Racing / 40-Road / 41-Sports car / 57-Race track / \n", "0-Games / 60-Basketball / 77-Arena / 168-Basketball moves / \n", "1738-Elmo / 3668-Ernie / \n", "248-Hotel / \n", "10-Animal / 48-Pet / 71-Dog / 182-Puppy / 1839-Rottweiler / \n", "28-Fashion / 1038-Hijab / 1096-Scarf / \n", "0-Games / 2-Video game / 19-PC game / 33-Weapon / 85-Combat / 96-Soldier / 192-Battlefield / 383-Battlefield 4 / \n", "1-Vehicle / 4-Car / 74-Truck / 249-Heavy equipment / 817-Bulldozer / \n", "9-Music video / \n", "6-Animation / 183-Manga / 541-Black-and-white / \n", "0-Games / 10-Animal / \n", "9-Music video / \n", "8-Football / 358-Rugby football / \n", "45-Cosmetics / 415-Sewing / 676-Handbag / 1941-Pattern / \n", "3-Concert / 5-Dance / \n", "0-Games / 6-Animation / 10-Animal / 15-Cartoon / \n", "0-Games / 2-Video game / 19-PC game / 928-Borderlands 2 / \n", "14-Guitar / \n", "1-Vehicle / 4-Car / 11-Motorsport / 84-Snow / 120-Winter / 135-Drifting / 349-Chevrolet / 1145-Winter storm / \n", "10-Animal / 91-Fish / 215-Eating / 258-Aquarium / \n", "6-Animation / 701-Beyblade / \n", "1-Vehicle / 4-Car / 11-Motorsport / 1661-Volkswagen Gol / \n", "9-Music video / \n", "3-Concert / 16-Performance art / \n", "1-Vehicle / 74-Truck / 158-Tire / 3431-Volvo FH / \n", "9-Music video / \n", "0-Games / 411-MapleStory / 1178-Quest / \n", "1-Vehicle / 11-Motorsport / 17-Racing / 27-Motorcycle / 172-Motocross / \n", "5-Dance / \n", "0-Games / 1-Vehicle / 2-Video game / 11-Motorsport / 199-Wii / 443-Wii U / 518-Mario Kart / 1831-Mario Kart 8 / \n", "0-Games / 2-Video game / \n", "0-Games / 8-Football / 21-Stadium / 54-Highlight film / 65-Kick / 73-Ball / 76-Ball / \n", "1-Vehicle / \n", "5-Dance / 6-Animation / \n", "1-Vehicle / 4-Car / 25-Toy / 962-McDonald's / 1169-Hot Wheels / 1418-Model car / \n", "3-Concert / 7-Musician / 16-Performance art / \n", "5-Dance / 16-Performance art / 95-Talent show / 355-Comedian / \n", "9-Music video / \n", "49-School / \n", "0-Games / 1-Vehicle / 2-Video game / 4-Car / 34-Action-adventure game / 201-Grand Theft Auto: San Andreas / 858-Carl Johnson / \n", "1-Vehicle / 4-Car / 27-Motorcycle / 94-Dashcam / 4048-Yamaha MT-09 / \n", "9-Music video / 133-Nightclub / \n", "5-Dance / 14-Guitar / 95-Talent show / \n", "1-Vehicle / 11-Motorsport / 17-Racing / 27-Motorcycle / 105-Motorcycling / 172-Motocross / 197-Trail / 491-Enduro / 608-Wheelie / \n", "298-Television advertisement / 337-Advertising / 1035-Bank / \n", "0-Games / 51-Strategy video game / 285-Sasuke Uchiha / 382-Dota 2 / 549-Defense of the Ancients / \n", "37-Gadget / 47-Personal computer / 145-Tablet computer / \n", "395-Camping / \n", "0-Games / 2-Video game / 317-PlayStation / 413-PlayStation 2 / 426-Resident Evil / 581-PlayStation / \n", "31-Disc jockey / \n", "20-Trailer / \n", "3-Concert / \n", "0-Games / 8-Football / 54-Highlight film / 81-Athlete / \n", "126-The Walt Disney Company / 144-Amusement park / \n", "242-Paper / 370-Origami / 577-Penguin / \n", "3-Concert / 153-Dress / 356-Star / \n", "93-Comedy / \n", "0-Games / 116-Pokémon / 186-Pokémon / 1204-Pokémon Ruby and Sapphire / 1387-Pokémon Battle Revolution / 1665-Pokémon Emerald / \n", "9-Music video / \n", "3-Concert / \n", "66-Bollywood / \n", "22-Nature / 164-Beach / 233-Ocean / 423-Coast / \n", "49-School / 661-Stretching / 898-Pilates / 1055-Back / \n", "45-Cosmetics / \n", "1-Vehicle / 4-Car / 11-Motorsport / 17-Racing / 57-Race track / 787-Monster truck / 3920-Maximum Destruction / \n", "3428-Costume jewelry / \n", "20-Trailer / \n", "20-Trailer / 263-Light / 419-Earth / \n", "1016-Glove / 3757-Baseball glove / 4241-Wilson Sporting Goods / \n", "5-Dance / \n", "25-Toy / 140-LEGO / 359-Harry Potter (Literary Series) / 884-Lego minifigure / \n", "45-Cosmetics / 193-Eye liner / 195-Mascara / 218-Eyelash / 235-Lipstick / 291-Rouge / 436-Concealer / 1012-Lip gloss / \n", "0-Games / 8-Football / 21-Stadium / 76-Ball / \n", "3-Concert / 5-Dance / 68-Lighting / \n", "0-Games / 2-Video game / 51-Strategy video game / 792-DarkOrbit / \n", "9-Music video / \n", "3-Concert / 16-Performance art / \n", "93-Comedy / \n", "8-Football / \n", "0-Games / 2-Video game / 90-Sports game / 480-FIFA 13 / \n", "3-Concert / 7-Musician / 16-Performance art / \n", "3-Concert / \n", "3377-Wii Remote / \n", "5-Dance / 16-Performance art / \n", "0-Games / 8-Football / 21-Stadium / 65-Kick / 73-Ball / 76-Ball / 81-Athlete / \n", "5-Dance / 16-Performance art / 95-Talent show / 130-Gymnastics / 351-Jumping / 1157-Rope / 3830-Double Dutch / \n", "412-Monster / \n", "606-String / 1326-Recorder / \n", "61-Art / 1488-Mosaic / 3266-Ancient Rome / \n", "1-Vehicle / 4-Car / 11-Motorsport / 17-Racing / 41-Sports car / 57-Race track / 135-Drifting / 169-Rallying / \n", "3-Concert / 7-Musician / 13-Musical ensemble / 38-Orchestra / 46-Choir / 1747-Erhu / \n", "3-Concert / 7-Musician / 13-Musical ensemble / 46-Choir / 209-University / \n", "31-Disc jockey / \n", "3-Concert / 7-Musician / 13-Musical ensemble / 30-Drummer / 36-Piano / 39-Drums / 53-Keyboard / 104-Pianist / 107-Musical keyboard / \n", "9-Music video / 322-Mixtape / 2461-DatPiff / \n", "20-Trailer / \n", "8-Football / \n", "10-Animal / 80-Horse / 247-Stallion / 290-Pony / 438-Mare / \n", "0-Games / 2-Video game / 55-Video game console / 396-One Piece / \n", "1-Vehicle / 87-Boat / \n", "66-Bollywood / 93-Comedy / \n", "25-Toy / 200-Doll / 290-Pony / 343-My Little Pony / 417-Play-Doh / 439-Barbie / 524-Littlest Pet Shop / \n", "0-Games / 2-Video game / 35-Minecraft / \n", "5-Dance / 16-Performance art / 59-Winter sport / 112-Ice skating / 115-Ballet / 316-Figure skating / \n", "28-Fashion / 121-Photography / 153-Dress / 390-Model / \n", "22-Nature / 67-Cycling / 69-Bicycle / 197-Trail / 220-Mountain bike / 284-Mountain biking / \n", "47-Personal computer / 131-Computer / 414-Computer monitor / 429-Website / \n", "5-Dance / 16-Performance art / 115-Ballet / \n", "10-Animal / \n", "0-Games / 2-Video game / 19-PC game / 686-Diablo III / 872-Diablo / \n", "0-Games / 8-Football / 54-Highlight film / 81-Athlete / \n", "12-Food / \n", "5-Dance / 78-Wedding / 190-Bride / 421-Gown / \n", "1-Vehicle / 4-Car / 11-Motorsport / 41-Sports car / 70-Driving / 344-Coupé / 531-Hatchback / 683-Renault / 2455-Renault Mégane / \n", "72-iPhone / \n", "3-Concert / 68-Lighting / \n", "2608-Strap / \n", "0-Games / 2-Video game / 35-Minecraft / \n", "5-Dance / 44-Drums / \n", "1-Vehicle / 11-Motorsport / 17-Racing / 57-Race track / 59-Winter sport / 84-Snow / 120-Winter / 262-Ice / 610-Snowmobile / \n", "0-Games / 19-PC game / 211-RuneScape / \n", "18-Outdoor recreation / 96-Soldier / \n", "3-Concert / 7-Musician / 13-Musical ensemble / 38-Orchestra / \n", "415-Sewing / 653-Thread / 997-Sewing machine / \n", "3-Concert / 7-Musician / 13-Musical ensemble / 38-Orchestra / \n", "3-Concert / 7-Musician / 13-Musical ensemble / 14-Guitar / 30-Drummer / 39-Drums / 44-Drums / 100-Electric guitar / \n", "12-Food / 2129-Vending machine / \n", "1-Vehicle / 88-Machine / 249-Heavy equipment / 512-Printing / 817-Bulldozer / 934-Flooring / \n", "30-Drummer / 39-Drums / 44-Drums / 128-Cymbal / 146-Snare drum / \n", "1-Vehicle / 4-Car / 41-Sports car / 1324-Roadster / \n", "12-Food / 1363-Candle / \n", "78-Wedding / \n", "9-Music video / \n", "10-Animal / 540-Insect / 775-Bee / 784-Honey / 1121-Beehive / 2858-Gelatin dessert / \n", "3-Concert / 16-Performance art / 30-Drummer / 68-Lighting / \n", "1459-Magazine / \n", "0-Games / 2-Video game / 43-Call of Duty / 134-Call of Duty: Black Ops / \n", "0-Games / 2-Video game / 19-PC game / 33-Weapon / 1394-Alliance of Valiant Arms / \n", "12-Food / 454-Sauce / 1022-Chili pepper / 3317-Salsa / \n", "3-Concert / 5-Dance / 16-Performance art / \n", "0-Games / \n", "136-Gardening / 329-Swimming pool / \n", "0-Games / 2-Video game / 34-Action-adventure game / 118-Grand Theft Auto V / \n", "8-Football / 21-Stadium / \n", "6-Animation / 15-Cartoon / \n", "3-Concert / 7-Musician / 13-Musical ensemble / 30-Drummer / 39-Drums / \n", "0-Games / 39-Drums / 724-Rock Band / \n", "20-Trailer / \n", "1-Vehicle / 4-Car / 11-Motorsport / 70-Driving / \n", "0-Games / 54-Highlight film / 60-Basketball / 76-Ball / 168-Basketball moves / \n", "18-Outdoor recreation / 83-Skateboarding / 108-Skateboard / 221-Skatepark / \n", "5-Dance / 430-Ballroom dance / \n", "0-Games / 2-Video game / 19-PC game / 51-Strategy video game / 165-League of Legends / \n", "3-Concert / 7-Musician / 13-Musical ensemble / 842-Harp / \n", "3889-Periodic Table / \n", "5-Dance / 7-Musician / 78-Wedding / \n", "0-Games / 6-Animation / 25-Toy / 140-LEGO / \n", "9-Music video / 14-Guitar / \n", "310-The Sims / 543-The Sims 2 / \n", "0-Games / 2-Video game / 20-Trailer / 23-Mobile phone / 29-Smartphone / 72-iPhone / 189-iPad / \n", "99-Christmas / \n", "46-Choir / 49-School / \n", "20-Trailer / \n", "31-Disc jockey / \n", "31-Disc jockey / \n", "122-Weight training / 204-Gym / 1250-Biceps / \n", "3-Concert / \n", "31-Disc jockey / \n", "1-Vehicle / 4-Car / 11-Motorsport / 169-Rallying / \n", "0-Games / 6-Animation / \n", "0-Games / 2-Video game / 47-Personal computer / \n", "93-Comedy / 341-Sketch comedy / \n", "0-Games / 116-Pokémon / \n", "5-Dance / \n", "45-Cosmetics / 56-Hair / 173-Eye shadow / 235-Lipstick / \n", "89-Comics / \n", "3535-Boutique / \n", "36-Piano / 53-Keyboard / 104-Pianist / 107-Musical keyboard / 265-Electronic keyboard / \n", "18-Outdoor recreation / 22-Nature / 42-Fishing / 147-River / 722-Waterfall / \n", "6-Animation / 10-Animal / 15-Cartoon / 126-The Walt Disney Company / 571-Bear / 784-Honey / 2089-Winnie-the-Pooh / \n", "551-British Broadcasting Corporation / \n", "250-Parade / 253-Carnival / \n", "5-Dance / 13-Musical ensemble / 16-Performance art / 98-Festival / \n", "75-Wrestling / \n", "14-Guitar / 24-String instrument / 100-Electric guitar / \n", "762-Angel / \n", "504-Web page / 819-World Wide Web / 3233-Screencast / \n", "18-Outdoor recreation / 83-Skateboarding / 108-Skateboard / \n", "1-Vehicle / 4-Car / 70-Driving / 2253-CNET / \n", "14-Guitar / 100-Electric guitar / \n", "6-Animation / \n", "0-Games / 2-Video game / 34-Action-adventure game / \n", "0-Games / 2-Video game / 19-PC game / 1188-Uncharted / \n", "125-Television / 504-Web page / 3166-Satellite television / \n", "36-Piano / 53-Keyboard / \n", "3-Concert / \n" ] } ], "source": [ "i = 0\n", "for item in validate_labels.iterrows():\n", " labels = item[1][\"labels\"]\n", " for l in sorted(map(int, labels.split())):\n", " print \"%d-%s\" % (l, label[l]), \"/\",\n", " print \"\"\n", " i += 1\n", " if i >= 10000:\n", " break" ] } ], "metadata": { "kernelspec": { "display_name": "Python 2", "language": "python", "name": "python2" }, "language_info": { "codemirror_mode": { "name": "ipython", "version": 2 }, "file_extension": ".py", "mimetype": "text/x-python", "name": "python", "nbconvert_exporter": "python", "pygments_lexer": "ipython2", "version": "2.7.12" } }, "nbformat": 4, "nbformat_minor": 2 } ================================================ FILE: eda/vertical.tsv ================================================ 0 0 1 1 2 0 3 2 4 1 5 2 6 2 7 2 8 3 9 2 10 4 11 3 12 5 13 2 14 2 15 2 16 2 17 3 18 6 19 0 20 2 21 2 22 7 23 8 24 2 25 9 26 5 27 1 28 10 29 8 30 2 31 2 32 5 33 11 34 0 35 0 36 2 37 12 38 2 39 2 40 13 41 1 42 6 43 0 44 2 45 10 46 2 47 12 48 4 49 14 50 1 51 0 52 5 53 12 54 2 55 12 56 10 57 0 58 5 59 3 60 3 61 2 62 13 63 2 64 13 65 3 66 2 67 3 68 15 69 1 70 1 71 4 72 12 73 3 74 1 75 3 76 3 77 2 78 6 79 3 80 4 81 3 82 1 83 3 84 16 85 3 86 7 87 1 88 13 89 2 90 0 91 13 92 6 93 2 94 12 95 2 96 11 97 2 98 2 99 6 100 2 101 8 102 2 103 13 104 2 105 1 106 10 107 2 108 3 109 2 110 2 111 12 112 3 113 15 114 1 115 2 116 2 117 3 118 0 119 13 120 16 121 2 122 10 123 2 124 13 125 2 126 2 127 12 128 2 129 13 130 3 131 12 132 4 133 2 134 0 135 3 136 15 137 13 138 0 139 17 140 9 141 12 142 3 143 13 144 17 145 12 146 2 147 18 148 2 149 13 150 2 151 13 152 13 153 9 154 1 155 12 156 13 157 13 158 1 159 0 160 6 161 13 162 19 163 3 164 17 165 0 166 4 167 1 168 3 169 3 170 15 171 0 172 3 173 10 174 13 175 13 176 12 177 3 178 12 179 15 180 0 181 12 182 4 183 2 184 5 185 2 186 0 187 13 188 17 189 12 190 6 191 3 192 0 193 10 194 0 195 10 196 7 197 13 198 1 199 12 200 9 201 0 202 6 203 2 204 10 205 16 206 9 207 1 208 13 209 14 210 15 211 0 212 18 213 13 214 7 215 5 216 0 217 6 218 10 219 2 220 1 221 3 222 3 223 16 224 13 225 17 226 5 227 13 228 12 229 0 230 2 231 1 232 5 233 7 234 1 235 10 236 5 237 0 238 10 239 5 240 0 241 0 242 13 243 0 244 2 245 1 246 15 247 4 248 17 249 13 250 2 251 2 252 13 253 6 254 13 255 3 256 3 257 0 258 4 259 7 260 20 261 13 262 7 263 7 264 2 265 2 266 1 267 7 268 6 269 2 270 1 271 11 272 12 273 2 274 5 275 6 276 21 277 1 278 12 279 15 280 4 281 3 282 7 283 7 284 1 285 2 286 18 287 3 288 7 289 6 290 4 291 10 292 12 293 14 294 2 295 5 296 2 297 0 298 2 299 1 300 10 301 5 302 0 303 2 304 10 305 3 306 14 307 7 308 1 309 12 310 0 311 0 312 3 313 22 314 12 315 6 316 3 317 12 318 2 319 7 320 0 321 17 322 2 323 20 324 7 325 18 326 0 327 7 328 0 329 15 330 23 331 12 332 1 333 9 334 17 335 12 336 1 337 13 338 10 339 2 340 10 341 2 342 12 343 9 344 1 345 12 346 4 347 3 348 5 349 1 350 2 351 3 352 0 353 13 354 5 355 2 356 7 357 2 358 3 359 22 360 1 361 2 362 12 363 7 364 2 365 2 366 20 367 5 368 9 369 17 370 6 371 11 372 0 373 0 374 0 375 1 376 2 377 1 378 3 379 0 380 0 381 6 382 0 383 0 384 0 385 7 386 12 387 2 388 11 389 5 390 10 391 5 392 6 393 13 394 18 395 6 396 2 397 2 398 12 399 3 400 13 401 2 402 9 403 1 404 13 405 10 406 1 407 6 408 2 409 0 410 14 411 0 412 18 413 12 414 12 415 6 416 21 417 20 418 2 419 7 420 0 421 9 422 0 423 18 424 13 425 18 426 0 427 2 428 6 429 8 430 2 431 0 432 0 433 13 434 2 435 9 436 10 437 5 438 4 439 9 440 4 441 0 442 5 443 12 444 20 445 2 446 2 447 13 448 7 449 1 450 2 451 13 452 9 453 0 454 5 455 2 456 3 457 6 458 7 459 13 460 0 461 0 462 15 463 3 464 0 465 2 466 9 467 4 468 7 469 9 470 0 471 13 472 13 473 1 474 0 475 5 476 13 477 2 478 6 479 2 480 0 481 1 482 13 483 9 484 12 485 5 486 0 487 12 488 15 489 0 490 16 491 3 492 13 493 1 494 0 495 10 496 0 497 12 498 1 499 7 500 5 501 3 502 7 503 9 504 8 505 0 506 0 507 2 508 12 509 1 510 6 511 0 512 13 513 12 514 0 515 1 516 12 517 1 518 0 519 1 520 1 521 13 522 1 523 13 524 9 525 12 526 1 527 12 528 2 529 10 530 0 531 1 532 10 533 13 534 5 535 0 536 13 537 17 538 2 539 15 540 7 541 2 542 5 543 0 544 0 545 2 546 6 547 2 548 3 549 0 550 7 551 16 552 1 553 8 554 12 555 3 556 3 557 5 558 6 559 18 560 5 561 0 562 3 563 0 564 0 565 3 566 12 567 12 568 0 569 1 570 3 571 4 572 5 573 1 574 13 575 1 576 17 577 4 578 5 579 1 580 15 581 12 582 5 583 10 584 0 585 15 586 12 587 7 588 4 589 5 590 9 591 2 592 2 593 7 594 2 595 3 596 0 597 5 598 4 599 0 600 5 601 4 602 12 603 0 604 1 605 1 606 2 607 13 608 3 609 10 610 1 611 5 612 5 613 2 614 6 615 0 616 22 617 13 618 1 619 13 620 5 621 12 622 2 623 3 624 22 625 13 626 0 627 15 628 2 629 12 630 2 631 7 632 3 633 2 634 2 635 0 636 4 637 0 638 12 639 2 640 5 641 9 642 7 643 18 644 0 645 1 646 2 647 1 648 2 649 15 650 1 651 3 652 15 653 13 654 7 655 2 656 0 657 16 658 0 659 1 660 13 661 10 662 5 663 17 664 0 665 0 666 13 667 3 668 1 669 3 670 2 671 5 672 6 673 15 674 3 675 12 676 9 677 12 678 2 679 2 680 1 681 13 682 12 683 1 684 5 685 12 686 0 687 0 688 5 689 2 690 0 691 13 692 0 693 3 694 3 695 9 696 9 697 0 698 13 699 3 700 12 701 2 702 0 703 3 704 0 705 12 706 4 707 15 708 18 709 13 710 3 711 12 712 7 713 7 714 2 715 0 716 7 717 2 718 13 719 17 720 5 721 12 722 7 723 4 724 0 725 20 726 2 727 5 728 3 729 6 730 0 731 23 732 0 733 10 734 0 735 2 736 0 737 2 738 2 739 0 740 0 741 13 742 15 743 2 744 15 745 22 746 2 747 13 748 2 749 2 750 2 751 7 752 5 753 1 754 2 755 2 756 0 757 9 758 11 759 5 760 12 761 1 762 20 763 4 764 0 765 3 766 11 767 3 768 13 769 5 770 5 771 5 772 3 773 12 774 7 775 7 776 14 777 0 778 6 779 0 780 10 781 0 782 12 783 2 784 5 785 9 786 5 787 1 788 2 789 3 790 1 791 0 792 0 793 9 794 13 795 5 796 11 797 0 798 10 799 0 800 18 801 2 802 8 803 2 804 9 805 15 806 1 807 13 808 5 809 2 810 18 811 12 812 13 813 3 814 7 815 6 816 2 817 13 818 15 819 8 820 2 821 13 822 13 823 10 824 4 825 7 826 0 827 13 828 6 829 6 830 15 831 0 832 12 833 3 834 12 835 17 836 0 837 0 838 0 839 0 840 18 841 5 842 2 843 2 844 7 845 14 846 4 847 0 848 0 849 0 850 0 851 0 852 9 853 1 854 1 855 2 856 13 857 10 858 0 859 6 860 20 861 2 862 5 863 10 864 1 865 4 866 12 867 5 868 13 869 0 870 20 871 10 872 0 873 13 874 1 875 0 876 2 877 0 878 2 879 13 880 9 881 4 882 13 883 12 884 9 885 2 886 2 887 0 888 15 889 2 890 2 891 1 892 10 893 5 894 9 895 5 896 6 897 2 898 10 899 8 900 12 901 1 902 6 903 0 904 5 905 12 906 13 907 9 908 12 909 10 910 12 911 21 912 13 913 11 914 6 915 17 916 10 917 6 918 1 919 10 920 15 921 0 922 0 923 13 924 6 925 7 926 6 927 3 928 0 929 13 930 3 931 13 932 8 933 1 934 15 935 3 936 2 937 3 938 13 939 21 940 2 941 0 942 0 943 15 944 5 945 1 946 3 947 1 948 0 949 6 950 12 951 2 952 1 953 0 954 10 955 11 956 19 957 12 958 7 959 12 960 3 961 12 962 5 963 0 964 6 965 2 966 5 967 22 968 5 969 1 970 4 971 18 972 13 973 12 974 3 975 7 976 13 977 0 978 2 979 2 980 2 981 2 982 1 983 1 984 13 985 5 986 9 987 13 988 17 989 0 990 2 991 0 992 13 993 3 994 0 995 2 996 11 997 13 998 7 999 4 1000 0 1001 9 1002 2 1003 2 1004 18 1005 0 1006 2 1007 0 1008 6 1009 12 1010 9 1011 13 1012 10 1013 3 1014 2 1015 12 1016 9 1017 9 1018 5 1019 20 1020 1 1021 0 1022 5 1023 7 1024 5 1025 6 1026 23 1027 0 1028 0 1029 0 1030 9 1031 12 1032 8 1033 13 1034 0 1035 23 1036 1 1037 7 1038 20 1039 4 1040 3 1041 1 1042 12 1043 12 1044 5 1045 12 1046 12 1047 6 1048 0 1049 0 1050 0 1051 4 1052 9 1053 6 1054 15 1055 7 1056 1 1057 6 1058 7 1059 2 1060 19 1061 6 1062 2 1063 12 1064 10 1065 2 1066 3 1067 15 1068 0 1069 6 1070 1 1071 0 1072 0 1073 13 1074 1 1075 4 1076 6 1077 2 1078 2 1079 9 1080 0 1081 13 1082 6 1083 0 1084 4 1085 15 1086 12 1087 0 1088 2 1089 12 1090 6 1091 0 1092 12 1093 13 1094 9 1095 20 1096 9 1097 12 1098 6 1099 0 1100 9 1101 3 1102 5 1103 2 1104 2 1105 3 1106 18 1107 11 1108 5 1109 0 1110 7 1111 3 1112 0 1113 13 1114 5 1115 5 1116 13 1117 12 1118 12 1119 12 1120 0 1121 7 1122 1 1123 13 1124 1 1125 0 1126 7 1127 2 1128 10 1129 13 1130 11 1131 0 1132 1 1133 0 1134 17 1135 4 1136 13 1137 0 1138 0 1139 2 1140 0 1141 0 1142 15 1143 2 1144 0 1145 7 1146 3 1147 4 1148 1 1149 2 1150 12 1151 20 1152 13 1153 2 1154 2 1155 12 1156 2 1157 13 1158 13 1159 12 1160 4 1161 0 1162 19 1163 15 1164 2 1165 2 1166 0 1167 0 1168 17 1169 6 1170 12 1171 12 1172 12 1173 2 1174 5 1175 10 1176 5 1177 7 1178 0 1179 13 1180 2 1181 7 1182 3 1183 2 1184 18 1185 6 1186 5 1187 0 1188 0 1189 5 1190 3 1191 20 1192 2 1193 5 1194 1 1195 6 1196 0 1197 4 1198 7 1199 12 1200 1 1201 1 1202 2 1203 6 1204 0 1205 1 1206 12 1207 12 1208 5 1209 5 1210 3 1211 0 1212 9 1213 0 1214 21 1215 4 1216 2 1217 3 1218 5 1219 2 1220 12 1221 1 1222 15 1223 5 1224 5 1225 8 1226 13 1227 5 1228 0 1229 7 1230 5 1231 2 1232 0 1233 1 1234 4 1235 16 1236 6 1237 20 1238 0 1239 3 1240 2 1241 0 1242 1 1243 7 1244 2 1245 11 1246 14 1247 3 1248 0 1249 6 1250 7 1251 0 1252 4 1253 2 1254 17 1255 6 1256 5 1257 13 1258 4 1259 2 1260 12 1261 0 1262 0 1263 1 1264 7 1265 2 1266 2 1267 0 1268 2 1269 10 1270 6 1271 10 1272 13 1273 12 1274 0 1275 0 1276 0 1277 6 1278 4 1279 6 1280 6 1281 15 1282 1 1283 5 1284 5 1285 13 1286 1 1287 12 1288 0 1289 2 1290 0 1291 0 1292 10 1293 0 1294 13 1295 5 1296 7 1297 1 1298 2 1299 0 1300 9 1301 1 1302 9 1303 2 1304 2 1305 6 1306 6 1307 1 1308 15 1309 6 1310 1 1311 9 1312 0 1313 2 1314 15 1315 3 1316 13 1317 1 1318 15 1319 5 1320 10 1321 1 1322 5 1323 12 1324 1 1325 13 1326 2 1327 2 1328 5 1329 15 1330 3 1331 0 1332 5 1333 2 1334 0 1335 3 1336 1 1337 3 1338 20 1339 0 1340 13 1341 4 1342 2 1343 0 1344 12 1345 2 1346 2 1347 5 1348 12 1349 12 1350 13 1351 3 1352 2 1353 20 1354 4 1355 5 1356 0 1357 6 1358 7 1359 10 1360 2 1361 1 1362 13 1363 18 1364 0 1365 5 1366 8 1367 3 1368 18 1369 9 1370 0 1371 0 1372 1 1373 3 1374 0 1375 2 1376 17 1377 4 1378 0 1379 12 1380 1 1381 5 1382 0 1383 12 1384 15 1385 2 1386 2 1387 0 1388 1 1389 10 1390 12 1391 1 1392 0 1393 2 1394 0 1395 13 1396 0 1397 0 1398 18 1399 3 1400 0 1401 0 1402 17 1403 6 1404 0 1405 14 1406 1 1407 12 1408 5 1409 11 1410 0 1411 2 1412 0 1413 1 1414 13 1415 1 1416 7 1417 3 1418 9 1419 0 1420 7 1421 2 1422 2 1423 9 1424 13 1425 12 1426 18 1427 0 1428 13 1429 1 1430 10 1431 2 1432 0 1433 0 1434 6 1435 0 1436 2 1437 1 1438 5 1439 15 1440 0 1441 16 1442 2 1443 2 1444 4 1445 2 1446 12 1447 0 1448 1 1449 6 1450 6 1451 13 1452 9 1453 6 1454 15 1455 1 1456 15 1457 4 1458 2 1459 22 1460 5 1461 1 1462 0 1463 17 1464 12 1465 2 1466 1 1467 0 1468 1 1469 2 1470 0 1471 5 1472 5 1473 5 1474 3 1475 0 1476 3 1477 3 1478 9 1479 15 1480 13 1481 9 1482 12 1483 11 1484 0 1485 0 1486 1 1487 2 1488 2 1489 15 1490 1 1491 13 1492 0 1493 1 1494 17 1495 3 1496 0 1497 13 1498 15 1499 0 1500 13 1501 11 1502 1 1503 0 1504 12 1505 0 1506 0 1507 2 1508 13 1509 5 1510 7 1511 1 1512 12 1513 0 1514 5 1515 0 1516 9 1517 5 1518 1 1519 1 1520 1 1521 1 1522 0 1523 22 1524 24 1525 9 1526 2 1527 2 1528 23 1529 1 1530 0 1531 7 1532 11 1533 2 1534 10 1535 6 1536 13 1537 1 1538 2 1539 5 1540 1 1541 5 1542 4 1543 0 1544 7 1545 13 1546 7 1547 13 1548 13 1549 2 1550 16 1551 2 1552 14 1553 0 1554 1 1555 2 1556 13 1557 5 1558 9 1559 12 1560 12 1561 6 1562 2 1563 2 1564 0 1565 12 1566 0 1567 13 1568 4 1569 5 1570 0 1571 0 1572 20 1573 11 1574 2 1575 1 1576 5 1577 0 1578 0 1579 18 1580 0 1581 12 1582 0 1583 1 1584 10 1585 0 1586 6 1587 1 1588 0 1589 5 1590 1 1591 4 1592 15 1593 12 1594 20 1595 12 1596 5 1597 5 1598 2 1599 7 1600 0 1601 20 1602 13 1603 4 1604 20 1605 0 1606 7 1607 1 1608 19 1609 3 1610 3 1611 1 1612 0 1613 24 1614 1 1615 0 1616 12 1617 7 1618 0 1619 0 1620 9 1621 20 1622 1 1623 3 1624 1 1625 17 1626 3 1627 9 1628 0 1629 12 1630 24 1631 0 1632 18 1633 11 1634 1 1635 1 1636 6 1637 12 1638 0 1639 4 1640 18 1641 5 1642 9 1643 11 1644 24 1645 4 1646 2 1647 24 1648 24 1649 5 1650 12 1651 1 1652 5 1653 0 1654 2 1655 0 1656 12 1657 12 1658 12 1659 5 1660 0 1661 1 1662 3 1663 7 1664 0 1665 0 1666 2 1667 6 1668 0 1669 5 1670 6 1671 13 1672 12 1673 17 1674 2 1675 2 1676 0 1677 5 1678 18 1679 0 1680 0 1681 0 1682 2 1683 0 1684 20 1685 9 1686 7 1687 0 1688 1 1689 0 1690 0 1691 0 1692 16 1693 5 1694 0 1695 0 1696 2 1697 20 1698 0 1699 1 1700 13 1701 5 1702 12 1703 0 1704 2 1705 13 1706 0 1707 15 1708 0 1709 1 1710 2 1711 3 1712 9 1713 0 1714 18 1715 11 1716 4 1717 0 1718 0 1719 0 1720 5 1721 0 1722 2 1723 4 1724 2 1725 5 1726 20 1727 12 1728 5 1729 5 1730 0 1731 2 1732 18 1733 7 1734 2 1735 0 1736 0 1737 0 1738 2 1739 15 1740 11 1741 5 1742 1 1743 0 1744 2 1745 0 1746 13 1747 24 1748 5 1749 0 1750 4 1751 12 1752 1 1753 15 1754 0 1755 24 1756 2 1757 2 1758 0 1759 12 1760 0 1761 1 1762 0 1763 1 1764 3 1765 5 1766 0 1767 4 1768 15 1769 5 1770 0 1771 2 1772 13 1773 12 1774 12 1775 0 1776 8 1777 12 1778 12 1779 13 1780 4 1781 3 1782 5 1783 3 1784 18 1785 0 1786 13 1787 24 1788 3 1789 0 1790 6 1791 2 1792 15 1793 5 1794 15 1795 9 1796 0 1797 13 1798 1 1799 4 1800 9 1801 2 1802 0 1803 24 1804 6 1805 17 1806 1 1807 1 1808 17 1809 2 1810 0 1811 9 1812 10 1813 0 1814 5 1815 1 1816 0 1817 24 1818 18 1819 0 1820 12 1821 13 1822 13 1823 15 1824 12 1825 2 1826 2 1827 10 1828 1 1829 9 1830 3 1831 0 1832 3 1833 1 1834 10 1835 24 1836 6 1837 0 1838 13 1839 4 1840 1 1841 0 1842 9 1843 5 1844 2 1845 4 1846 2 1847 0 1848 7 1849 16 1850 9 1851 5 1852 0 1853 5 1854 24 1855 12 1856 2 1857 0 1858 2 1859 0 1860 0 1861 0 1862 12 1863 13 1864 12 1865 0 1866 2 1867 1 1868 20 1869 13 1870 5 1871 4 1872 12 1873 3 1874 15 1875 1 1876 0 1877 1 1878 0 1879 6 1880 2 1881 2 1882 11 1883 0 1884 2 1885 7 1886 10 1887 1 1888 2 1889 15 1890 0 1891 0 1892 14 1893 0 1894 18 1895 2 1896 1 1897 1 1898 13 1899 22 1900 2 1901 1 1902 0 1903 1 1904 13 1905 4 1906 12 1907 1 1908 0 1909 2 1910 5 1911 0 1912 2 1913 13 1914 6 1915 0 1916 1 1917 6 1918 7 1919 0 1920 13 1921 3 1922 12 1923 0 1924 22 1925 7 1926 12 1927 12 1928 0 1929 0 1930 2 1931 1 1932 1 1933 6 1934 2 1935 0 1936 2 1937 3 1938 0 1939 12 1940 0 1941 10 1942 2 1943 5 1944 17 1945 0 1946 2 1947 4 1948 4 1949 4 1950 2 1951 1 1952 1 1953 24 1954 12 1955 22 1956 1 1957 1 1958 20 1959 12 1960 1 1961 0 1962 12 1963 0 1964 2 1965 1 1966 2 1967 0 1968 0 1969 7 1970 15 1971 2 1972 9 1973 3 1974 1 1975 12 1976 0 1977 2 1978 13 1979 1 1980 5 1981 12 1982 2 1983 24 1984 1 1985 0 1986 2 1987 2 1988 0 1989 12 1990 0 1991 0 1992 1 1993 4 1994 24 1995 15 1996 12 1997 1 1998 12 1999 5 2000 5 2001 22 2002 0 2003 3 2004 0 2005 0 2006 24 2007 13 2008 24 2009 12 2010 12 2011 2 2012 13 2013 2 2014 9 2015 16 2016 1 2017 13 2018 4 2019 19 2020 1 2021 1 2022 7 2023 4 2024 4 2025 13 2026 0 2027 5 2028 0 2029 4 2030 2 2031 13 2032 5 2033 12 2034 12 2035 13 2036 12 2037 13 2038 23 2039 5 2040 5 2041 0 2042 5 2043 0 2044 10 2045 10 2046 5 2047 2 2048 5 2049 11 2050 24 2051 2 2052 1 2053 2 2054 0 2055 21 2056 11 2057 5 2058 6 2059 4 2060 0 2061 1 2062 7 2063 0 2064 13 2065 2 2066 5 2067 15 2068 13 2069 2 2070 1 2071 2 2072 5 2073 2 2074 0 2075 18 2076 2 2077 7 2078 1 2079 13 2080 5 2081 0 2082 2 2083 4 2084 0 2085 24 2086 0 2087 4 2088 0 2089 2 2090 4 2091 13 2092 13 2093 13 2094 4 2095 23 2096 22 2097 4 2098 2 2099 6 2100 1 2101 0 2102 15 2103 0 2104 1 2105 0 2106 0 2107 18 2108 10 2109 1 2110 2 2111 5 2112 2 2113 0 2114 20 2115 13 2116 2 2117 0 2118 7 2119 2 2120 7 2121 0 2122 2 2123 0 2124 13 2125 5 2126 0 2127 0 2128 20 2129 13 2130 1 2131 5 2132 20 2133 3 2134 9 2135 6 2136 13 2137 0 2138 13 2139 0 2140 0 2141 13 2142 12 2143 0 2144 0 2145 2 2146 2 2147 13 2148 24 2149 2 2150 3 2151 17 2152 20 2153 13 2154 4 2155 18 2156 3 2157 4 2158 2 2159 12 2160 1 2161 6 2162 13 2163 12 2164 13 2165 12 2166 12 2167 9 2168 9 2169 2 2170 0 2171 22 2172 2 2173 0 2174 17 2175 13 2176 2 2177 13 2178 5 2179 0 2180 12 2181 0 2182 12 2183 0 2184 10 2185 2 2186 9 2187 5 2188 9 2189 13 2190 15 2191 8 2192 0 2193 12 2194 13 2195 0 2196 13 2197 5 2198 2 2199 17 2200 1 2201 22 2202 11 2203 1 2204 0 2205 12 2206 13 2207 24 2208 1 2209 13 2210 1 2211 4 2212 0 2213 5 2214 1 2215 1 2216 1 2217 12 2218 13 2219 0 2220 15 2221 13 2222 4 2223 1 2224 22 2225 1 2226 17 2227 2 2228 10 2229 12 2230 5 2231 2 2232 1 2233 2 2234 5 2235 10 2236 7 2237 0 2238 0 2239 0 2240 0 2241 2 2242 15 2243 13 2244 3 2245 18 2246 7 2247 1 2248 12 2249 0 2250 12 2251 0 2252 2 2253 2 2254 0 2255 2 2256 13 2257 9 2258 15 2259 8 2260 0 2261 1 2262 12 2263 12 2264 12 2265 24 2266 5 2267 24 2268 1 2269 2 2270 16 2271 4 2272 1 2273 6 2274 12 2275 3 2276 0 2277 0 2278 0 2279 19 2280 2 2281 15 2282 6 2283 4 2284 7 2285 12 2286 1 2287 3 2288 1 2289 0 2290 24 2291 2 2292 1 2293 0 2294 17 2295 12 2296 12 2297 20 2298 3 2299 2 2300 6 2301 0 2302 24 2303 1 2304 1 2305 1 2306 1 2307 0 2308 1 2309 19 2310 1 2311 7 2312 2 2313 5 2314 1 2315 9 2316 2 2317 1 2318 12 2319 9 2320 0 2321 0 2322 3 2323 17 2324 2 2325 1 2326 11 2327 0 2328 12 2329 13 2330 2 2331 7 2332 2 2333 13 2334 0 2335 1 2336 0 2337 11 2338 1 2339 4 2340 0 2341 2 2342 2 2343 7 2344 12 2345 1 2346 0 2347 13 2348 12 2349 6 2350 2 2351 10 2352 0 2353 0 2354 0 2355 17 2356 10 2357 2 2358 3 2359 13 2360 3 2361 12 2362 6 2363 5 2364 2 2365 1 2366 1 2367 2 2368 2 2369 0 2370 1 2371 5 2372 19 2373 7 2374 7 2375 2 2376 24 2377 1 2378 13 2379 0 2380 2 2381 5 2382 2 2383 11 2384 2 2385 15 2386 5 2387 11 2388 4 2389 10 2390 23 2391 17 2392 9 2393 2 2394 0 2395 5 2396 4 2397 5 2398 11 2399 2 2400 0 2401 24 2402 12 2403 2 2404 1 2405 5 2406 5 2407 0 2408 2 2409 1 2410 4 2411 2 2412 2 2413 13 2414 13 2415 6 2416 12 2417 12 2418 5 2419 15 2420 1 2421 16 2422 24 2423 0 2424 9 2425 13 2426 0 2427 1 2428 13 2429 2 2430 2 2431 0 2432 1 2433 0 2434 0 2435 24 2436 18 2437 15 2438 4 2439 1 2440 0 2441 2 2442 0 2443 0 2444 5 2445 13 2446 1 2447 2 2448 9 2449 5 2450 0 2451 12 2452 2 2453 24 2454 13 2455 1 2456 7 2457 7 2458 2 2459 15 2460 0 2461 2 2462 24 2463 20 2464 22 2465 12 2466 0 2467 2 2468 0 2469 10 2470 0 2471 0 2472 13 2473 13 2474 0 2475 15 2476 0 2477 2 2478 10 2479 4 2480 15 2481 13 2482 15 2483 5 2484 12 2485 3 2486 2 2487 12 2488 10 2489 2 2490 15 2491 0 2492 4 2493 2 2494 1 2495 0 2496 5 2497 0 2498 12 2499 6 2500 5 2501 15 2502 0 2503 0 2504 5 2505 3 2506 0 2507 6 2508 8 2509 1 2510 13 2511 1 2512 1 2513 0 2514 24 2515 0 2516 15 2517 1 2518 0 2519 0 2520 2 2521 7 2522 1 2523 0 2524 0 2525 9 2526 0 2527 0 2528 18 2529 2 2530 1 2531 3 2532 2 2533 12 2534 7 2535 0 2536 2 2537 0 2538 20 2539 0 2540 2 2541 0 2542 20 2543 1 2544 0 2545 0 2546 0 2547 2 2548 15 2549 0 2550 0 2551 2 2552 13 2553 4 2554 0 2555 2 2556 6 2557 0 2558 1 2559 7 2560 0 2561 18 2562 5 2563 5 2564 5 2565 3 2566 2 2567 0 2568 2 2569 0 2570 20 2571 9 2572 2 2573 4 2574 3 2575 5 2576 2 2577 5 2578 13 2579 5 2580 12 2581 12 2582 3 2583 17 2584 2 2585 1 2586 0 2587 17 2588 5 2589 12 2590 0 2591 5 2592 12 2593 5 2594 0 2595 2 2596 12 2597 12 2598 4 2599 13 2600 22 2601 7 2602 0 2603 1 2604 13 2605 3 2606 12 2607 5 2608 6 2609 13 2610 20 2611 3 2612 0 2613 24 2614 5 2615 12 2616 1 2617 1 2618 5 2619 1 2620 11 2621 0 2622 0 2623 13 2624 2 2625 2 2626 4 2627 22 2628 24 2629 12 2630 2 2631 4 2632 13 2633 0 2634 12 2635 13 2636 0 2637 5 2638 12 2639 1 2640 7 2641 20 2642 2 2643 7 2644 0 2645 0 2646 2 2647 0 2648 7 2649 20 2650 15 2651 24 2652 1 2653 6 2654 1 2655 2 2656 2 2657 0 2658 0 2659 0 2660 1 2661 2 2662 24 2663 10 2664 0 2665 12 2666 15 2667 0 2668 0 2669 13 2670 6 2671 12 2672 12 2673 4 2674 0 2675 9 2676 5 2677 0 2678 1 2679 2 2680 6 2681 5 2682 24 2683 1 2684 2 2685 18 2686 1 2687 10 2688 0 2689 4 2690 1 2691 0 2692 4 2693 9 2694 5 2695 24 2696 24 2697 4 2698 0 2699 5 2700 0 2701 1 2702 6 2703 15 2704 0 2705 13 2706 5 2707 15 2708 0 2709 12 2710 2 2711 13 2712 0 2713 1 2714 0 2715 0 2716 22 2717 1 2718 12 2719 13 2720 13 2721 15 2722 23 2723 22 2724 13 2725 18 2726 0 2727 12 2728 5 2729 2 2730 0 2731 19 2732 3 2733 0 2734 5 2735 4 2736 1 2737 13 2738 2 2739 0 2740 2 2741 7 2742 9 2743 15 2744 2 2745 13 2746 2 2747 3 2748 9 2749 5 2750 12 2751 12 2752 0 2753 1 2754 12 2755 12 2756 7 2757 5 2758 4 2759 2 2760 1 2761 18 2762 2 2763 1 2764 2 2765 13 2766 2 2767 2 2768 15 2769 13 2770 12 2771 2 2772 0 2773 12 2774 0 2775 5 2776 7 2777 5 2778 1 2779 0 2780 4 2781 9 2782 2 2783 4 2784 7 2785 9 2786 12 2787 1 2788 1 2789 0 2790 17 2791 12 2792 0 2793 24 2794 2 2795 15 2796 0 2797 24 2798 0 2799 0 2800 1 2801 0 2802 12 2803 20 2804 4 2805 13 2806 0 2807 1 2808 12 2809 2 2810 0 2811 24 2812 0 2813 2 2814 9 2815 12 2816 2 2817 0 2818 12 2819 13 2820 0 2821 7 2822 3 2823 8 2824 1 2825 5 2826 1 2827 2 2828 0 2829 18 2830 12 2831 5 2832 2 2833 0 2834 24 2835 2 2836 6 2837 24 2838 13 2839 3 2840 9 2841 0 2842 9 2843 15 2844 24 2845 11 2846 24 2847 9 2848 2 2849 0 2850 17 2851 24 2852 5 2853 13 2854 0 2855 2 2856 2 2857 1 2858 5 2859 4 2860 12 2861 9 2862 1 2863 0 2864 2 2865 1 2866 3 2867 7 2868 1 2869 0 2870 9 2871 0 2872 1 2873 9 2874 13 2875 1 2876 7 2877 0 2878 2 2879 0 2880 0 2881 15 2882 18 2883 12 2884 0 2885 1 2886 13 2887 9 2888 4 2889 0 2890 2 2891 0 2892 1 2893 0 2894 0 2895 13 2896 0 2897 0 2898 12 2899 1 2900 1 2901 0 2902 2 2903 15 2904 13 2905 0 2906 12 2907 3 2908 12 2909 0 2910 12 2911 2 2912 0 2913 15 2914 9 2915 2 2916 0 2917 1 2918 7 2919 0 2920 13 2921 3 2922 2 2923 4 2924 4 2925 1 2926 13 2927 15 2928 5 2929 8 2930 0 2931 18 2932 1 2933 2 2934 0 2935 6 2936 0 2937 23 2938 7 2939 18 2940 2 2941 5 2942 7 2943 5 2944 24 2945 2 2946 0 2947 0 2948 11 2949 1 2950 6 2951 2 2952 2 2953 2 2954 13 2955 5 2956 15 2957 22 2958 12 2959 15 2960 0 2961 6 2962 2 2963 9 2964 5 2965 2 2966 6 2967 2 2968 24 2969 4 2970 0 2971 2 2972 4 2973 3 2974 2 2975 1 2976 2 2977 4 2978 0 2979 9 2980 2 2981 2 2982 5 2983 9 2984 2 2985 24 2986 15 2987 2 2988 0 2989 1 2990 0 2991 1 2992 5 2993 0 2994 12 2995 5 2996 7 2997 24 2998 0 2999 0 3000 12 3001 0 3002 10 3003 5 3004 1 3005 24 3006 1 3007 4 3008 15 3009 15 3010 9 3011 24 3012 20 3013 3 3014 2 3015 13 3016 0 3017 4 3018 4 3019 8 3020 5 3021 5 3022 4 3023 0 3024 2 3025 2 3026 0 3027 5 3028 0 3029 22 3030 2 3031 2 3032 0 3033 0 3034 10 3035 24 3036 2 3037 5 3038 15 3039 20 3040 1 3041 20 3042 20 3043 12 3044 5 3045 12 3046 4 3047 0 3048 0 3049 5 3050 2 3051 2 3052 4 3053 0 3054 5 3055 4 3056 0 3057 24 3058 1 3059 0 3060 4 3061 7 3062 12 3063 20 3064 0 3065 6 3066 24 3067 0 3068 2 3069 0 3070 15 3071 15 3072 12 3073 13 3074 1 3075 0 3076 5 3077 0 3078 0 3079 2 3080 0 3081 0 3082 0 3083 1 3084 6 3085 24 3086 1 3087 5 3088 15 3089 8 3090 2 3091 1 3092 22 3093 0 3094 13 3095 12 3096 24 3097 12 3098 4 3099 0 3100 5 3101 10 3102 2 3103 5 3104 4 3105 7 3106 10 3107 16 3108 1 3109 2 3110 7 3111 0 3112 2 3113 5 3114 11 3115 0 3116 24 3117 6 3118 2 3119 10 3120 0 3121 4 3122 1 3123 7 3124 24 3125 2 3126 5 3127 1 3128 2 3129 10 3130 12 3131 2 3132 1 3133 12 3134 1 3135 0 3136 1 3137 2 3138 15 3139 5 3140 3 3141 0 3142 15 3143 9 3144 22 3145 6 3146 0 3147 0 3148 2 3149 2 3150 5 3151 2 3152 2 3153 2 3154 9 3155 1 3156 15 3157 4 3158 1 3159 0 3160 0 3161 2 3162 13 3163 1 3164 5 3165 20 3166 2 3167 0 3168 0 3169 2 3170 10 3171 12 3172 2 3173 4 3174 3 3175 0 3176 7 3177 7 3178 15 3179 18 3180 2 3181 24 3182 15 3183 24 3184 12 3185 13 3186 8 3187 1 3188 11 3189 21 3190 1 3191 2 3192 2 3193 5 3194 1 3195 12 3196 12 3197 1 3198 8 3199 2 3200 18 3201 2 3202 20 3203 0 3204 5 3205 5 3206 12 3207 5 3208 24 3209 15 3210 24 3211 5 3212 3 3213 24 3214 2 3215 2 3216 24 3217 12 3218 0 3219 1 3220 24 3221 24 3222 2 3223 4 3224 6 3225 0 3226 2 3227 0 3228 5 3229 1 3230 0 3231 0 3232 2 3233 12 3234 2 3235 0 3236 15 3237 1 3238 0 3239 12 3240 9 3241 2 3242 24 3243 3 3244 12 3245 1 3246 12 3247 2 3248 2 3249 7 3250 5 3251 5 3252 0 3253 0 3254 5 3255 0 3256 15 3257 20 3258 12 3259 0 3260 5 3261 2 3262 24 3263 0 3264 3 3265 4 3266 18 3267 1 3268 4 3269 0 3270 9 3271 10 3272 24 3273 4 3274 24 3275 24 3276 2 3277 0 3278 13 3279 1 3280 7 3281 7 3282 2 3283 0 3284 0 3285 12 3286 3 3287 5 3288 5 3289 2 3290 0 3291 3 3292 6 3293 2 3294 5 3295 0 3296 12 3297 7 3298 12 3299 24 3300 15 3301 15 3302 0 3303 0 3304 24 3305 3 3306 2 3307 12 3308 0 3309 1 3310 7 3311 1 3312 9 3313 5 3314 12 3315 2 3316 1 3317 5 3318 6 3319 0 3320 1 3321 0 3322 12 3323 2 3324 1 3325 0 3326 0 3327 13 3328 12 3329 13 3330 24 3331 13 3332 4 3333 18 3334 1 3335 1 3336 24 3337 24 3338 1 3339 1 3340 2 3341 2 3342 22 3343 2 3344 2 3345 3 3346 0 3347 2 3348 24 3349 2 3350 1 3351 12 3352 7 3353 1 3354 2 3355 0 3356 5 3357 2 3358 18 3359 15 3360 0 3361 8 3362 24 3363 12 3364 5 3365 0 3366 0 3367 2 3368 3 3369 2 3370 3 3371 1 3372 1 3373 1 3374 1 3375 17 3376 12 3377 12 3378 0 3379 5 3380 2 3381 2 3382 5 3383 4 3384 5 3385 5 3386 15 3387 24 3388 0 3389 2 3390 17 3391 2 3392 0 3393 5 3394 15 3395 2 3396 24 3397 0 3398 9 3399 1 3400 15 3401 2 3402 5 3403 4 3404 5 3405 12 3406 0 3407 0 3408 12 3409 1 3410 6 3411 2 3412 2 3413 7 3414 1 3415 0 3416 1 3417 2 3418 0 3419 1 3420 14 3421 24 3422 2 3423 0 3424 12 3425 6 3426 0 3427 0 3428 9 3429 2 3430 0 3431 1 3432 24 3433 16 3434 0 3435 2 3436 2 3437 2 3438 24 3439 1 3440 0 3441 8 3442 14 3443 7 3444 4 3445 2 3446 7 3447 0 3448 0 3449 2 3450 0 3451 0 3452 2 3453 2 3454 2 3455 13 3456 0 3457 15 3458 24 3459 5 3460 0 3461 20 3462 0 3463 20 3464 1 3465 2 3466 24 3467 5 3468 0 3469 24 3470 0 3471 3 3472 13 3473 2 3474 0 3475 1 3476 1 3477 5 3478 20 3479 0 3480 2 3481 24 3482 12 3483 2 3484 2 3485 17 3486 1 3487 5 3488 1 3489 2 3490 17 3491 13 3492 22 3493 6 3494 2 3495 20 3496 1 3497 10 3498 19 3499 4 3500 0 3501 20 3502 15 3503 19 3504 1 3505 9 3506 0 3507 2 3508 22 3509 0 3510 0 3511 7 3512 0 3513 0 3514 0 3515 13 3516 1 3517 0 3518 5 3519 0 3520 9 3521 13 3522 2 3523 2 3524 0 3525 2 3526 0 3527 0 3528 0 3529 1 3530 13 3531 13 3532 12 3533 3 3534 1 3535 9 3536 0 3537 12 3538 0 3539 5 3540 18 3541 1 3542 1 3543 2 3544 13 3545 2 3546 0 3547 1 3548 1 3549 1 3550 6 3551 15 3552 3 3553 3 3554 0 3555 0 3556 2 3557 24 3558 0 3559 15 3560 1 3561 2 3562 2 3563 17 3564 1 3565 0 3566 24 3567 12 3568 2 3569 24 3570 4 3571 5 3572 12 3573 1 3574 5 3575 2 3576 4 3577 12 3578 1 3579 2 3580 10 3581 6 3582 3 3583 13 3584 10 3585 7 3586 0 3587 1 3588 2 3589 1 3590 6 3591 0 3592 1 3593 15 3594 24 3595 2 3596 5 3597 0 3598 5 3599 22 3600 22 3601 1 3602 6 3603 13 3604 3 3605 2 3606 24 3607 5 3608 5 3609 13 3610 5 3611 1 3612 0 3613 5 3614 1 3615 20 3616 2 3617 4 3618 0 3619 12 3620 0 3621 18 3622 2 3623 12 3624 0 3625 5 3626 13 3627 0 3628 23 3629 5 3630 1 3631 4 3632 12 3633 13 3634 24 3635 12 3636 1 3637 2 3638 4 3639 13 3640 24 3641 5 3642 6 3643 24 3644 3 3645 24 3646 1 3647 24 3648 12 3649 0 3650 11 3651 13 3652 3 3653 20 3654 12 3655 10 3656 2 3657 1 3658 4 3659 9 3660 4 3661 2 3662 2 3663 2 3664 12 3665 1 3666 0 3667 24 3668 2 3669 4 3670 2 3671 13 3672 1 3673 13 3674 13 3675 0 3676 0 3677 9 3678 15 3679 3 3680 17 3681 1 3682 9 3683 2 3684 1 3685 13 3686 6 3687 17 3688 12 3689 2 3690 12 3691 9 3692 22 3693 6 3694 16 3695 12 3696 13 3697 1 3698 13 3699 4 3700 24 3701 9 3702 24 3703 12 3704 1 3705 16 3706 24 3707 14 3708 0 3709 1 3710 4 3711 10 3712 18 3713 1 3714 7 3715 12 3716 1 3717 0 3718 12 3719 2 3720 12 3721 0 3722 5 3723 0 3724 0 3725 5 3726 0 3727 12 3728 5 3729 24 3730 6 3731 0 3732 8 3733 2 3734 2 3735 5 3736 15 3737 1 3738 3 3739 2 3740 2 3741 2 3742 1 3743 5 3744 2 3745 4 3746 2 3747 0 3748 1 3749 12 3750 24 3751 15 3752 13 3753 12 3754 2 3755 1 3756 15 3757 3 3758 1 3759 0 3760 1 3761 10 3762 1 3763 4 3764 0 3765 13 3766 13 3767 5 3768 0 3769 17 3770 10 3771 1 3772 24 3773 0 3774 0 3775 1 3776 1 3777 13 3778 1 3779 14 3780 13 3781 9 3782 9 3783 9 3784 0 3785 4 3786 0 3787 12 3788 7 3789 15 3790 24 3791 0 3792 2 3793 1 3794 9 3795 2 3796 12 3797 7 3798 4 3799 5 3800 12 3801 0 3802 5 3803 4 3804 2 3805 2 3806 12 3807 0 3808 19 3809 17 3810 2 3811 1 3812 2 3813 2 3814 0 3815 0 3816 2 3817 9 3818 24 3819 2 3820 13 3821 2 3822 7 3823 0 3824 1 3825 5 3826 1 3827 5 3828 1 3829 5 3830 3 3831 9 3832 24 3833 3 3834 1 3835 5 3836 1 3837 3 3838 15 3839 1 3840 0 3841 0 3842 6 3843 0 3844 0 3845 10 3846 0 3847 2 3848 1 3849 0 3850 18 3851 12 3852 4 3853 2 3854 12 3855 0 3856 13 3857 5 3858 1 3859 0 3860 5 3861 10 3862 2 3863 0 3864 6 3865 0 3866 12 3867 8 3868 24 3869 5 3870 3 3871 0 3872 0 3873 1 3874 0 3875 24 3876 5 3877 2 3878 10 3879 19 3880 15 3881 1 3882 24 3883 0 3884 0 3885 10 3886 1 3887 13 3888 6 3889 7 3890 1 3891 5 3892 0 3893 24 3894 0 3895 3 3896 1 3897 0 3898 15 3899 0 3900 12 3901 0 3902 0 3903 12 3904 24 3905 2 3906 0 3907 0 3908 13 3909 1 3910 5 3911 13 3912 3 3913 0 3914 18 3915 1 3916 1 3917 0 3918 12 3919 7 3920 24 3921 24 3922 2 3923 4 3924 12 3925 17 3926 12 3927 0 3928 2 3929 13 3930 0 3931 7 3932 0 3933 2 3934 15 3935 6 3936 24 3937 13 3938 7 3939 1 3940 17 3941 1 3942 7 3943 12 3944 9 3945 1 3946 8 3947 24 3948 6 3949 0 3950 12 3951 7 3952 0 3953 2 3954 2 3955 1 3956 5 3957 1 3958 12 3959 0 3960 24 3961 0 3962 12 3963 12 3964 0 3965 0 3966 0 3967 17 3968 13 3969 0 3970 1 3971 19 3972 2 3973 10 3974 12 3975 22 3976 0 3977 0 3978 12 3979 1 3980 12 3981 24 3982 12 3983 24 3984 12 3985 0 3986 4 3987 3 3988 2 3989 1 3990 1 3991 1 3992 0 3993 20 3994 2 3995 1 3996 2 3997 1 3998 2 3999 0 4000 1 4001 20 4002 2 4003 3 4004 0 4005 1 4006 24 4007 0 4008 1 4009 0 4010 0 4011 18 4012 6 4013 13 4014 1 4015 1 4016 24 4017 2 4018 2 4019 0 4020 1 4021 5 4022 5 4023 0 4024 9 4025 24 4026 1 4027 24 4028 4 4029 2 4030 2 4031 12 4032 0 4033 13 4034 1 4035 13 4036 1 4037 5 4038 9 4039 6 4040 20 4041 1 4042 24 4043 0 4044 24 4045 13 4046 0 4047 0 4048 1 4049 0 4050 13 4051 23 4052 2 4053 5 4054 24 4055 13 4056 12 4057 0 4058 0 4059 8 4060 1 4061 0 4062 0 4063 0 4064 0 4065 5 4066 15 4067 1 4068 0 4069 12 4070 0 4071 1 4072 24 4073 24 4074 5 4075 4 4076 0 4077 1 4078 2 4079 1 4080 6 4081 24 4082 2 4083 0 4084 1 4085 0 4086 15 4087 9 4088 0 4089 2 4090 0 4091 0 4092 1 4093 6 4094 2 4095 4 4096 12 4097 3 4098 1 4099 1 4100 5 4101 2 4102 0 4103 2 4104 2 4105 3 4106 2 4107 0 4108 4 4109 2 4110 2 4111 3 4112 24 4113 3 4114 24 4115 1 4116 2 4117 2 4118 9 4119 0 4120 0 4121 3 4122 24 4123 10 4124 6 4125 0 4126 0 4127 15 4128 12 4129 2 4130 24 4131 2 4132 0 4133 1 4134 0 4135 1 4136 2 4137 13 4138 5 4139 13 4140 0 4141 0 4142 2 4143 5 4144 0 4145 2 4146 2 4147 0 4148 17 4149 24 4150 1 4151 9 4152 2 4153 5 4154 0 4155 0 4156 0 4157 0 4158 1 4159 2 4160 24 4161 15 4162 0 4163 0 4164 0 4165 2 4166 11 4167 2 4168 7 4169 2 4170 1 4171 9 4172 0 4173 0 4174 24 4175 2 4176 2 4177 24 4178 24 4179 2 4180 22 4181 5 4182 0 4183 9 4184 0 4185 6 4186 1 4187 0 4188 2 4189 1 4190 12 4191 24 4192 7 4193 0 4194 0 4195 12 4196 6 4197 24 4198 24 4199 1 4200 10 4201 12 4202 2 4203 2 4204 1 4205 17 4206 7 4207 0 4208 2 4209 24 4210 1 4211 2 4212 13 4213 6 4214 2 4215 1 4216 1 4217 2 4218 0 4219 15 4220 2 4221 2 4222 6 4223 3 4224 2 4225 4 4226 2 4227 4 4228 2 4229 8 4230 13 4231 24 4232 22 4233 0 4234 2 4235 3 4236 5 4237 9 4238 13 4239 12 4240 0 4241 16 4242 1 4243 5 4244 24 4245 2 4246 6 4247 23 4248 24 4249 1 4250 0 4251 1 4252 4 4253 4 4254 0 4255 0 4256 10 4257 2 4258 0 4259 0 4260 1 4261 12 4262 0 4263 2 4264 2 4265 24 4266 5 4267 0 4268 24 4269 24 4270 2 4271 6 4272 0 4273 2 4274 17 4275 0 4276 15 4277 5 4278 24 4279 2 4280 2 4281 11 4282 6 4283 12 4284 2 4285 2 4286 2 4287 7 4288 2 4289 0 4290 12 4291 3 4292 0 4293 13 4294 0 4295 5 4296 1 4297 24 4298 0 4299 9 4300 2 4301 18 4302 0 4303 12 4304 0 4305 11 4306 2 4307 13 4308 1 4309 13 4310 24 4311 24 4312 24 4313 4 4314 0 4315 2 4316 5 4317 23 4318 2 4319 5 4320 18 4321 3 4322 6 4323 7 4324 20 4325 13 4326 2 4327 15 4328 17 4329 17 4330 2 4331 2 4332 24 4333 24 4334 11 4335 2 4336 17 4337 4 4338 20 4339 13 4340 2 4341 13 4342 1 4343 12 4344 2 4345 2 4346 12 4347 2 4348 1 4349 7 4350 0 4351 13 4352 1 4353 0 4354 24 4355 3 4356 2 4357 24 4358 2 4359 24 4360 0 4361 4 4362 5 4363 2 4364 1 4365 24 4366 5 4367 10 4368 22 4369 12 4370 17 4371 2 4372 0 4373 0 4374 24 4375 0 4376 2 4377 2 4378 0 4379 0 4380 6 4381 24 4382 2 4383 7 4384 1 4385 0 4386 1 4387 5 4388 12 4389 4 4390 2 4391 2 4392 24 4393 1 4394 0 4395 0 4396 2 4397 13 4398 1 4399 1 4400 13 4401 2 4402 1 4403 1 4404 6 4405 24 4406 13 4407 12 4408 16 4409 0 4410 24 4411 0 4412 17 4413 0 4414 5 4415 5 4416 0 4417 0 4418 20 4419 2 4420 2 4421 2 4422 2 4423 0 4424 12 4425 24 4426 2 4427 2 4428 3 4429 2 4430 1 4431 12 4432 24 4433 0 4434 0 4435 12 4436 1 4437 0 4438 7 4439 24 4440 8 4441 10 4442 1 4443 2 4444 0 4445 24 4446 3 4447 24 4448 15 4449 0 4450 1 4451 2 4452 22 4453 1 4454 3 4455 1 4456 13 4457 5 4458 4 4459 1 4460 2 4461 12 4462 5 4463 2 4464 0 4465 24 4466 3 4467 0 4468 0 4469 24 4470 5 4471 7 4472 2 4473 2 4474 6 4475 24 4476 1 4477 24 4478 7 4479 2 4480 1 4481 1 4482 0 4483 10 4484 12 4485 24 4486 0 4487 0 4488 2 4489 2 4490 0 4491 10 4492 24 4493 1 4494 4 4495 1 4496 1 4497 4 4498 1 4499 12 4500 0 4501 2 4502 22 4503 24 4504 0 4505 22 4506 24 4507 0 4508 1 4509 19 4510 0 4511 2 4512 0 4513 0 4514 0 4515 12 4516 5 4517 1 4518 2 4519 24 4520 24 4521 0 4522 3 4523 0 4524 2 4525 0 4526 0 4527 24 4528 2 4529 2 4530 2 4531 0 4532 24 4533 0 4534 3 4535 0 4536 1 4537 2 4538 0 4539 0 4540 0 4541 0 4542 0 4543 1 4544 2 4545 2 4546 24 4547 0 4548 0 4549 0 4550 0 4551 2 4552 0 4553 2 4554 0 4555 0 4556 0 4557 4 4558 24 4559 11 4560 4 4561 24 4562 24 4563 13 4564 4 4565 0 4566 13 4567 0 4568 1 4569 0 4570 9 4571 4 4572 6 4573 22 4574 0 4575 24 4576 9 4577 0 4578 11 4579 0 4580 2 4581 5 4582 0 4583 2 4584 0 4585 0 4586 15 4587 2 4588 3 4589 0 4590 2 4591 3 4592 24 4593 24 4594 5 4595 24 4596 3 4597 7 4598 2 4599 0 4600 3 4601 0 4602 4 4603 22 4604 6 4605 24 4606 24 4607 3 4608 2 4609 11 4610 0 4611 8 4612 3 4613 0 4614 2 4615 24 4616 2 4617 9 4618 2 4619 1 4620 5 4621 8 4622 2 4623 4 4624 24 4625 5 4626 2 4627 22 4628 0 4629 24 4630 24 4631 1 4632 14 4633 0 4634 2 4635 8 4636 13 4637 7 4638 1 4639 4 4640 6 4641 3 4642 2 4643 4 4644 7 4645 2 4646 2 4647 1 4648 2 4649 3 4650 5 4651 24 4652 2 4653 1 4654 0 4655 0 4656 0 4657 13 4658 22 4659 2 4660 0 4661 0 4662 14 4663 6 4664 21 4665 10 4666 0 4667 10 4668 4 4669 4 4670 3 4671 2 4672 1 4673 2 4674 0 4675 0 4676 24 4677 3 4678 17 4679 11 4680 2 4681 24 4682 0 4683 0 4684 0 4685 3 4686 24 4687 2 4688 0 4689 17 4690 0 4691 0 4692 9 4693 2 4694 4 4695 3 4696 20 4697 2 4698 22 4699 2 4700 5 4701 2 4702 24 4703 0 4704 2 4705 24 4706 2 4707 9 4708 0 4709 3 4710 7 4711 1 4712 0 4713 2 4714 2 4715 2 ================================================ FILE: youtube-8m-ensemble/.vimrc ================================================ set tabstop=2 set shiftwidth=2 set expandtab set autoindent ================================================ FILE: youtube-8m-ensemble/CONTRIBUTING.md ================================================ # How to contribute We are accepting patches and contributions to this project. To set expectations, this project is primarily intended to be a flexible starting point for researchers working with the YouTube-8M dataset. As such, we would like to keep it simple. We are more likely to accept small bug fixes and optimizations, and less likely to accept patches which add significant complexity. For the latter type of contribution, we recommend creating a Github fork of the project instead. If you would like to contribute, there are a few small guidelines you need to follow. ## Contributor License Agreement Contributions to any Google project must be accompanied by a Contributor License Agreement. This is necessary because you own the copyright to your changes, even after your contribution becomes part of this project. So this agreement simply gives us permission to use and redistribute your contributions as part of the project. Head over to to see your current agreements on file or to sign a new one. You generally only need to submit a CLA once, so if you've already submitted one (even if it was for a different project), you probably don't need to do it again. ## Code reviews All submissions, including submissions by project members, require review. We use GitHub pull requests for this purpose. Consult [GitHub Help] for more information on using pull requests. [GitHub Help]: https://help.github.com/articles/about-pull-requests/ ================================================ FILE: youtube-8m-ensemble/LICENSE ================================================ Apache License Version 2.0, January 2004 http://www.apache.org/licenses/ TERMS AND CONDITIONS FOR USE, REPRODUCTION, AND DISTRIBUTION 1. Definitions. "License" shall mean the terms and conditions for use, reproduction, and distribution as defined by Sections 1 through 9 of this document. "Licensor" shall mean the copyright owner or entity authorized by the copyright owner that is granting the License. "Legal Entity" shall mean the union of the acting entity and all other entities that control, are controlled by, or are under common control with that entity. For the purposes of this definition, "control" means (i) the power, direct or indirect, to cause the direction or management of such entity, whether by contract or otherwise, or (ii) ownership of fifty percent (50%) or more of the outstanding shares, or (iii) beneficial ownership of such entity. "You" (or "Your") shall mean an individual or Legal Entity exercising permissions granted by this License. "Source" form shall mean the preferred form for making modifications, including but not limited to software source code, documentation source, and configuration files. "Object" form shall mean any form resulting from mechanical transformation or translation of a Source form, including but not limited to compiled object code, generated documentation, and conversions to other media types. "Work" shall mean the work of authorship, whether in Source or Object form, made available under the License, as indicated by a copyright notice that is included in or attached to the work (an example is provided in the Appendix below). "Derivative Works" shall mean any work, whether in Source or Object form, that is based on (or derived from) the Work and for which the editorial revisions, annotations, elaborations, or other modifications represent, as a whole, an original work of authorship. For the purposes of this License, Derivative Works shall not include works that remain separable from, or merely link (or bind by name) to the interfaces of, the Work and Derivative Works thereof. "Contribution" shall mean any work of authorship, including the original version of the Work and any modifications or additions to that Work or Derivative Works thereof, that is intentionally submitted to Licensor for inclusion in the Work by the copyright owner or by an individual or Legal Entity authorized to submit on behalf of the copyright owner. For the purposes of this definition, "submitted" means any form of electronic, verbal, or written communication sent to the Licensor or its representatives, including but not limited to communication on electronic mailing lists, source code control systems, and issue tracking systems that are managed by, or on behalf of, the Licensor for the purpose of discussing and improving the Work, but excluding communication that is conspicuously marked or otherwise designated in writing by the copyright owner as "Not a Contribution." "Contributor" shall mean Licensor and any individual or Legal Entity on behalf of whom a Contribution has been received by Licensor and subsequently incorporated within the Work. 2. Grant of Copyright License. Subject to the terms and conditions of this License, each Contributor hereby grants to You a perpetual, worldwide, non-exclusive, no-charge, royalty-free, irrevocable copyright license to reproduce, prepare Derivative Works of, publicly display, publicly perform, sublicense, and distribute the Work and such Derivative Works in Source or Object form. 3. Grant of Patent License. Subject to the terms and conditions of this License, each Contributor hereby grants to You a perpetual, worldwide, non-exclusive, no-charge, royalty-free, irrevocable (except as stated in this section) patent license to make, have made, use, offer to sell, sell, import, and otherwise transfer the Work, where such license applies only to those patent claims licensable by such Contributor that are necessarily infringed by their Contribution(s) alone or by combination of their Contribution(s) with the Work to which such Contribution(s) was submitted. If You institute patent litigation against any entity (including a cross-claim or counterclaim in a lawsuit) alleging that the Work or a Contribution incorporated within the Work constitutes direct or contributory patent infringement, then any patent licenses granted to You under this License for that Work shall terminate as of the date such litigation is filed. 4. Redistribution. You may reproduce and distribute copies of the Work or Derivative Works thereof in any medium, with or without modifications, and in Source or Object form, provided that You meet the following conditions: (a) You must give any other recipients of the Work or Derivative Works a copy of this License; and (b) You must cause any modified files to carry prominent notices stating that You changed the files; and (c) You must retain, in the Source form of any Derivative Works that You distribute, all copyright, patent, trademark, and attribution notices from the Source form of the Work, excluding those notices that do not pertain to any part of the Derivative Works; and (d) If the Work includes a "NOTICE" text file as part of its distribution, then any Derivative Works that You distribute must include a readable copy of the attribution notices contained within such NOTICE file, excluding those notices that do not pertain to any part of the Derivative Works, in at least one of the following places: within a NOTICE text file distributed as part of the Derivative Works; within the Source form or documentation, if provided along with the Derivative Works; or, within a display generated by the Derivative Works, if and wherever such third-party notices normally appear. The contents of the NOTICE file are for informational purposes only and do not modify the License. You may add Your own attribution notices within Derivative Works that You distribute, alongside or as an addendum to the NOTICE text from the Work, provided that such additional attribution notices cannot be construed as modifying the License. You may add Your own copyright statement to Your modifications and may provide additional or different license terms and conditions for use, reproduction, or distribution of Your modifications, or for any such Derivative Works as a whole, provided Your use, reproduction, and distribution of the Work otherwise complies with the conditions stated in this License. 5. Submission of Contributions. Unless You explicitly state otherwise, any Contribution intentionally submitted for inclusion in the Work by You to the Licensor shall be under the terms and conditions of this License, without any additional terms or conditions. Notwithstanding the above, nothing herein shall supersede or modify the terms of any separate license agreement you may have executed with Licensor regarding such Contributions. 6. Trademarks. This License does not grant permission to use the trade names, trademarks, service marks, or product names of the Licensor, except as required for reasonable and customary use in describing the origin of the Work and reproducing the content of the NOTICE file. 7. Disclaimer of Warranty. Unless required by applicable law or agreed to in writing, Licensor provides the Work (and each Contributor provides its Contributions) on an "AS IS" BASIS, WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied, including, without limitation, any warranties or conditions of TITLE, NON-INFRINGEMENT, MERCHANTABILITY, or FITNESS FOR A PARTICULAR PURPOSE. You are solely responsible for determining the appropriateness of using or redistributing the Work and assume any risks associated with Your exercise of permissions under this License. 8. Limitation of Liability. In no event and under no legal theory, whether in tort (including negligence), contract, or otherwise, unless required by applicable law (such as deliberate and grossly negligent acts) or agreed to in writing, shall any Contributor be liable to You for damages, including any direct, indirect, special, incidental, or consequential damages of any character arising as a result of this License or out of the use or inability to use the Work (including but not limited to damages for loss of goodwill, work stoppage, computer failure or malfunction, or any and all other commercial damages or losses), even if such Contributor has been advised of the possibility of such damages. 9. Accepting Warranty or Additional Liability. While redistributing the Work or Derivative Works thereof, You may choose to offer, and charge a fee for, acceptance of support, warranty, indemnity, or other liability obligations and/or rights consistent with this License. However, in accepting such obligations, You may act only on Your own behalf and on Your sole responsibility, not on behalf of any other Contributor, and only if You agree to indemnify, defend, and hold each Contributor harmless for any liability incurred by, or claims asserted against, such Contributor by reason of your accepting any such warranty or additional liability. END OF TERMS AND CONDITIONS APPENDIX: How to apply the Apache License to your work. To apply the Apache License to your work, attach the following boilerplate notice, with the fields enclosed by brackets "[]" replaced with your own identifying information. (Don't include the brackets!) The text should be enclosed in the appropriate comment syntax for the file format. We also recommend that a file or class name and description of purpose be included on the same "printed page" as the copyright notice for easier identification within third-party archives. Copyright [yyyy] [name of copyright owner] Licensed under the Apache License, Version 2.0 (the "License"); you may not use this file except in compliance with the License. You may obtain a copy of the License at http://www.apache.org/licenses/LICENSE-2.0 Unless required by applicable law or agreed to in writing, software distributed under the License is distributed on an "AS IS" BASIS, WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied. See the License for the specific language governing permissions and limitations under the License. ================================================ FILE: youtube-8m-ensemble/README.md ================================================ # YouTube-8M Tensorflow Starter Code This repo contains starter code for training and evaluating machine learning models over the [YouTube-8M](https://research.google.com/youtube8m/) dataset. The code gives an end-to-end working example for reading the dataset, training a TensorFlow model, and evaluating the performance of the model. Out of the box, you can train several [model architectures](#overview-of-models) over either frame-level or video-level features. The code can easily be extended to train your own custom-defined models. It is possible to train and evaluate on YouTube-8M in two ways: on Google Cloud or on your own machine. This README provides instructions for both. ## Table of Contents * [Running on Google's Cloud Machine Learning Platform](#running-on-googles-cloud-machine-learning-platform) * [Requirements](#requirements) * [Testing Locally](#testing-locally) * [Training on the Cloud over Video-Level Features](#training-on-video-level-features) * [Evaluation and Inference](#evaluation-and-inference) * [Accessing Files on Google Cloud](#accessing-files-on-google-cloud) * [Using Frame-Level Features](#using-frame-level-features) * [Using Audio Features](#using-audio-features) * [Running on Your Own Machine](#running-on-your-own-machine) * [Requirements](#requirements-1) * [Training on Video-Level Features](#training-on-video-level-features-1) * [Evaluation and Inference](#evaluation-and-inference-1) * [Using Frame-Level Features](#using-frame-level-features-1) * [Using Audio Features](#using-audio-features-1) * [Ground-Truth Label Files](#ground-truth-label-files) * [Overview of Models](#overview-of-models) * [Video-Level Models](#video-level-models) * [Frame-Level Models](#frame-level-models) * [Overview of Files](#overview-of-files) * [Training](#training) * [Evaluation](#evaluation) * [Inference](#inference) * [Misc](#misc) * [About This Project](#about-this-project) ## Running on Google's Cloud Machine Learning Platform ### Requirements This option requires you to have an appropriately configured Google Cloud Platform account. To create and configure your account, please make sure you follow the instructions [here](https://cloud.google.com/ml/docs/how-tos/getting-set-up). If you are participating in the Google Cloud & YouTube-8M Video Understanding Challenge hosted on [kaggle](https://www.kaggle.com/c/youtube8m), see [these instructions](https://www.kaggle.com/c/youtube8m#getting-started-with-google-cloud) instead. Please also verify that you have Python 2.7+ and Tensorflow 1.0.0 or higher installed by running the following commands: ```sh python --version python -c 'import tensorflow as tf; print(tf.__version__)' ``` ### Testing Locally All gcloud commands should be done from the directory *immediately above* the source code. You should be able to see the source code directory if you run 'ls'. As you are developing your own models, you will want to test them quickly to flush out simple problems without having to submit them to the cloud. You can use the `gcloud beta ml local` set of commands for that. Here is an example command line for video-level training: ```sh gcloud beta ml local train \ --package-path=youtube-8m --module-name=youtube-8m.train -- \ --train_data_pattern='gs://youtube8m-ml/1/video_level/train/train*.tfrecord' \ --train_dir=/tmp/yt8m_train --start_new_model ``` You might want to download some training shards locally to speed things up and allow you to work offline. The command below will copy 10 out of the 4096 training data files to the current directory. ```sh # Downloads 55MB of data. gsutil cp gs://us.data.yt8m.org/1/video_level/train/traina[0-9].tfrecord . ``` Once you download the files, you can point the job to them using the 'train_data_pattern' argument (i.e. instead of pointing to the "gs://..." files, you point to the local files). Once your model is working locally, you can scale up on the Cloud which is described below. ### Training on the Cloud over Video-Level Features The following commands will train a model on Google Cloud over video-level features. ```sh BUCKET_NAME=gs://${USER}_yt8m_train_bucket # (One Time) Create a storage bucket to store training logs and checkpoints. gsutil mb -l us-east1 $BUCKET_NAME # Submit the training job. JOB_NAME=yt8m_train_$(date +%Y%m%d_%H%M%S); gcloud --verbosity=debug beta ml jobs \ submit training $JOB_NAME \ --package-path=youtube-8m --module-name=youtube-8m.train \ --staging-bucket=$BUCKET_NAME --region=us-east1 \ --config=youtube-8m/cloudml-gpu.yaml \ -- --train_data_pattern='gs://youtube8m-ml-us-east1/1/video_level/train/train*.tfrecord' \ --train_dir=$BUCKET_NAME/yt8m_train_video_level_logistic_model ``` In the 'gsutil' command above, the 'package-path' flag refers to the directory containing the 'train.py' script and more generally the python package which should be deployed to the cloud worker. The module-name refers to the specific python script which should be executed (in this case the train module). It may take several minutes before the job starts running on Google Cloud. When it starts you will see outputs like the following: ``` training step 270| Hit@1: 0.68 PERR: 0.52 Loss: 638.453 training step 271| Hit@1: 0.66 PERR: 0.49 Loss: 635.537 training step 272| Hit@1: 0.70 PERR: 0.52 Loss: 637.564 ``` At this point you can disconnect your console by pressing "ctrl-c". The model will continue to train indefinitely in the Cloud. Later, you can check on its progress or halt the job by visiting the [Google Cloud ML Jobs console](https://console.cloud.google.com/ml/jobs). You can train many jobs at once and use tensorboard to compare their performance visually. ```sh tensorboard --logdir=$BUCKET_NAME --port=8080 ``` Once tensorboard is running, you can access it at the following url: [http://localhost:8080](http://localhost:8080). If you are using Google Cloud Shell, you can instead click the Web Preview button on the upper left corner of the Cloud Shell window and select "Preview on port 8080". This will bring up a new browser tab with the Tensorboard view. ### Evaluation and Inference Here's how to evaluate a model on the validation dataset: ```sh JOB_TO_EVAL=yt8m_train_video_level_logistic_model JOB_NAME=yt8m_eval_$(date +%Y%m%d_%H%M%S); gcloud --verbosity=debug beta ml jobs \ submit training $JOB_NAME \ --package-path=youtube-8m --module-name=youtube-8m.eval \ --staging-bucket=$BUCKET_NAME --region=us-east1 \ --config=youtube-8m/cloudml-gpu.yaml \ -- --eval_data_pattern='gs://youtube8m-ml-us-east1/1/video_level/validate/validate*.tfrecord' \ --train_dir=$BUCKET_NAME/${JOB_TO_EVAL} --run_once=True ``` And here's how to perform inference with a model on the test set: ```sh JOB_TO_EVAL=yt8m_train_video_level_logistic_model JOB_NAME=yt8m_inference_$(date +%Y%m%d_%H%M%S); gcloud --verbosity=debug beta ml jobs \ submit training $JOB_NAME \ --package-path=youtube-8m --module-name=youtube-8m.inference \ --staging-bucket=$BUCKET_NAME --region=us-east1 \ --config=youtube-8m/cloudml-gpu.yaml \ -- --input_data_pattern='gs://youtube8m-ml/1/video_level/test/test*.tfrecord' \ --train_dir=$BUCKET_NAME/${JOB_TO_EVAL} \ --output_file=$BUCKET_NAME/${JOB_TO_EVAL}/predictions.csv ``` Note the confusing use of 'training' in the above gcloud commands. Despite the name, the 'training' argument really just offers a cloud hosted python/tensorflow service. From the point of view of the Cloud Platform, there is no distinction between our training and inference jobs. The Cloud ML platform also offers specialized functionality for prediction with Tensorflow models, but discussing that is beyond the scope of this readme. Once these job starts executing you will see outputs similar to the following for the evaluation code: ``` examples_processed: 1024 | global_step 447044 | Batch Hit@1: 0.782 | Batch PERR: 0.637 | Batch Loss: 7.821 | Examples_per_sec: 834.658 ``` and the following for the inference code: ``` num examples processed: 8192 elapsed seconds: 14.85 ``` ### Accessing Files on Google Cloud You can browse the storage buckets you created on Google Cloud, for example, to access the trained models, prediction CSV files, etc. by visiting the [Google Cloud storage browser](https://console.cloud.google.com/storage/browser). Alternatively, you can use the 'gsutil' command to download the files directly. For example, to download the output of the inference code from the previous section to your local machine, run: ``` gsutil cp $BUCKET_NAME/${JOB_TO_EVAL}/predictions.csv . ``` ### Using Frame-Level Features Append ```sh --frame_features=True --model=FrameLevelLogisticModel --feature_names="rgb" \ --feature_sizes="1024" --batch_size=128 \ --train_dir=$BUCKET_NAME/yt8m_train_frame_level_logistic_model ``` to the 'gcloud' commands given above, and change 'video_level' in paths to 'frame_level'. Here is a sample command to kick-off a frame-level job: ```sh JOB_NAME=yt8m_train_$(date +%Y%m%d_%H%M%S); gcloud --verbosity=debug beta ml jobs \ submit training $JOB_NAME \ --package-path=youtube-8m --module-name=youtube-8m.train \ --staging-bucket=$BUCKET_NAME --region=us-east1 \ --config=youtube-8m/cloudml-gpu.yaml \ -- --train_data_pattern='gs://youtube8m-ml-us-east1/1/frame_level/train/train*.tfrecord' \ --frame_features=True --model=FrameLevelLogisticModel --feature_names="rgb" \ --feature_sizes="1024" --batch_size=128 \ --train_dir=$BUCKET_NAME/yt8m_train_frame_level_logistic_model ``` The 'FrameLevelLogisticModel' is designed to provide equivalent results to a logistic model trained over the video-level features. Please look at the 'video_level_models.py' or 'frame_level_models.py' files to see how to implement your own models. ### Using Audio Features The feature files (both Frame-Level and Video-Level) contain two sets of features: 1) visual and 2) audio. The code defaults to using the visual features only, but it is possible to use audio features instead of (or besides) visual features. To specify the (combination of) features to use you must set `--feature_names` and `--feature_sizes` flags. The visual and audio features are called 'rgb' and 'audio' and have 1024 and 128 dimensions, respectively. The two flags take a comma-separated list of values in string. For example, to use audio-visual Video-Level features the flags must be set as follows: ``` --feature_names="mean_rgb, mean_audio" --feature_sizes="1024, 128" ``` Similarly, to use audio-visual Frame-Level features use: ``` --feature_names="rgb, audio" --feature_sizes="1024, 128" ``` **NOTE:** Make sure the set of features and the order in which the appear in the lists provided to the two flags above match. Also, the order must match when running training, evaluation, or inference. ## Running on Your Own Machine ### Requirements The starter code requires Tensorflow. If you haven't installed it yet, follow the instructions on [tensorflow.org](https://www.tensorflow.org/install/). This code has been tested with Tensorflow 1.0.0. Going forward, we will continue to target the latest released version of Tensorflow. Please verify that you have Python 2.7+ and Tensorflow 1.0.0 or higher installed by running the following commands: ```sh python --version python -c 'import tensorflow as tf; print(tf.__version__)' ``` You can find complete instructions for downloading the dataset on the [YouTube-8M website](https://research.google.com/youtube8m/download.html). We recommend downloading the smaller video-level features dataset first when getting started. To do that, run: ``` mkdir -p features; cd features curl data.yt8m.org/download.py | partition=1/video_level/train mirror=us python ``` This will download the full set of video level features, which takes up 31GB of space. If you are located outside of North America, you should change the flag 'mirror' to 'eu' for Europe or 'asia' for Asia to speed up the transfer of the files. Change 'train' to 'validate'/'test' and re-run the command to download the other splits of the dataset. Change 'video_level' to 'frame_level' to download the frame-level features. The complete frame-level features take about 1.71TB of space. You can set the environment variable 'shard' to 'm,n' to download only m/n-th of the data. For example, to download 1/100-th of the frame-level features from the training set, run: ``` curl data.yt8m.org/download.py | shard=1,100 partition=1/frame_level/train mirror=us python ``` ### Training on Video-Level Features To start training a logistic model on the video-level features, run ```sh MODEL_DIR=/tmp/yt8m python train.py --train_data_pattern='/path/to/features/train*.tfrecord' --train_dir=$MODEL_DIR/video_level_logistic_model ``` Since the dataset is sharded into 4096 individual files, we use a wildcard (\*) to represent all of those files. By default, the training code will frequently write _checkpoint_ files (i.e. values of all trainable parameters, at the current training iteration). These will be written to the `--train_dir`. If you re-use a `--train_dir`, the trainer will first restore the latest checkpoint written in that directory. This only works if the architecture of the checkpoint matches the graph created by the training code. If you are in active development/debugging phase, consider adding `--start_new_model` flag to your run configuration. ### Evaluation and Inference To evaluate the model, run ```sh python eval.py --eval_data_pattern='/path/to/features/validate*.tfrecord' --train_dir=$MODEL_DIR/video_level_logistic_model --run_once=True ``` As the model is training or evaluating, you can view the results on tensorboard by running ```sh tensorboard --logdir=$MODEL_DIR ``` and navigating to http://localhost:6006 in your web browser. When you are happy with your model, you can generate a csv file of predictions from it by running ```sh python inference.py --output_file=$MODEL_DIR/video_level_logistic_model/predictions.csv --input_data_pattern='/path/to/features/test*.tfrecord' --train_dir=$MODEL_DIR/video_level_logistic_model ``` This will output the top 20 predicted labels from the model for every example to 'predictions.csv'. ### Using Frame-Level Features Follow the same instructions as above, appending `--frame_features=True --model=FrameLevelLogisticModel --feature_names="rgb" --feature_sizes="1024" --train_dir=$MODEL_DIR/frame_level_logistic_model` for the 'train.py', 'eval.py', and 'inference.py' scripts. The 'FrameLevelLogisticModel' is designed to provide equivalent results to a logistic model trained over the video-level features. Please look at the 'models.py' file to see how to implement your own models. ### Using Audio Features See [Using Audio Features](#using-audio-features) section above. ### Ground-Truth Label Files We also provide CSV files containing the ground-truth label information of the 'train' and 'validation' partitions of the dataset. These files can be downloaded using 'gsutil' command: ``` gsutil cp gs://us.data.yt8m.org/1/ground_truth_labels/train_labels.csv /destination/folder/ gsutil cp gs://us.data.yt8m.org/1/ground_truth_labels/validate_labels.csv /destination/folder/ ``` or directly using the following links: * [http://us.data.yt8m.org/1/ground_truth_labels/train_labels.csv](http://us.data.yt8m.org/1/ground_truth_labels/train_labels.csv) * [http://us.data.yt8m.org/1/ground_truth_labels/validate_labels.csv](http://us.data.yt8m.org/1/ground_truth_labels/validate_labels.csv) Each line in the files starts with the video id and is followed by the list of ground-truth labels corresponding to that video. For example, for a video with id 'VIDEO_ID' and two lables 'LABLE1' and 'LABEL2' we store the following line: ``` VIDEO_ID,LABEL1 LABEL2 ``` ## Overview of Models This sample code contains implementations of the models given in the [YouTube-8M technical report](https://arxiv.org/abs/1609.08675). ### Video-Level Models * `LogisticModel`: Linear projection of the output features into the label space, followed by a sigmoid function to convert logit values to probabilities. * `MoeModel`: A per-class softmax distribution over a configurable number of logistic classifiers. One of the classifiers in the mixture is not trained, and always predicts 0. ### Frame-Level Models * `LstmModel`: Processes the features for each frame using a multi-layered LSTM neural net. The final internal state of the LSTM is input to a video-level model for classification. Note that you will need to change the learning rate to 0.001 when using this model. * `DbofModel`: Projects the features for each frame into a higher dimensional 'clustering' space, pools across frames in that space, and then uses a video-level model to classify the now aggregated features. * `FrameLevelLogisticModel`: Equivalent to 'LogisticModel', but performs average-pooling on the fly over frame-level features rather than using pre-aggregated features. ## Overview of Files ### Training * `train.py`: The primary script for training models. * `losses.py`: Contains definitions for loss functions. * `models.py`: Contains the base class for defining a model. * `video_level_models.py`: Contains definitions for models that take aggregated features as input. * `frame_level_models.py`: Contains definitions for models that take frame- level features as input. * `model_util.py`: Contains functions that are of general utility for implementing models. * `readers.py`: Contains definitions for the Video dataset and Frame dataset readers. ### Evaluation * `eval.py`: The primary script for evaluating models. * `eval_util.py`: Provides a class that calculates all evaluation metrics. * `average_precision_calculator.py`: Functions for calculating average precision. * `mean_average_precision_calculator.py`: Functions for calculating mean average precision. ### Inference * `inference.py`: Generates an output file containing predictions of the model over a set of videos. ### Misc * `README.md`: This documentation. * `utils.py`: Common functions. ## About This Project This project is meant help people quickly get started working with the [YouTube-8M](https://research.google.com/youtube8m/) dataset. This is not an official Google product. ================================================ FILE: youtube-8m-ensemble/__init__.py ================================================ # Copyright 2016 Google Inc. All Rights Reserved. # # Licensed under the Apache License, Version 2.0 (the "License"); # you may not use this file except in compliance with the License. # You may obtain a copy of the License at # # http://www.apache.org/licenses/LICENSE-2.0 # # Unless required by applicable law or agreed to in writing, software # distributed under the License is distributed on an "AS-IS" BASIS, # WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied. # See the License for the specific language governing permissions and # limitations under the License. ================================================ FILE: youtube-8m-ensemble/all_ensemble_models/.vimrc ================================================ set tabstop=2 set shiftwidth=2 set expandtab set autoindent ================================================ FILE: youtube-8m-ensemble/all_ensemble_models/__init__.py ================================================ from logistic_model import * from moe_model import * from attention_moe_model import * from attention_moe_matrix_model import * from input_moe_model import * from linear_regression_model import * from matrix_regression_model import * from attention_matrix_model import * from attention_linear_model import * from attention_linmatrix_model import * from attention_rectified_linear_model import * from nonunit_matrix_regression_model import * from mean_model import * from deep_combine_chain_model import * ================================================ FILE: youtube-8m-ensemble/all_ensemble_models/attention_linear_model.py ================================================ import math import models import tensorflow as tf import utils from tensorflow import flags import tensorflow.contrib.slim as slim FLAGS = flags.FLAGS class AttentionLinearModel(models.BaseModel): def create_model(self, model_input, vocab_size, num_mixtures=None, l2_penalty=1e-8, sub_scope="", original_input=None, **unused_params): num_methods = model_input.get_shape().as_list()[-1] num_features = model_input.get_shape().as_list()[-2] num_mixtures = FLAGS.moe_num_mixtures # gating coefficients original_input = tf.nn.l2_normalize(original_input, dim=1) mean_output = tf.reduce_mean(model_input, axis=2) ## batch_size x moe_num_mixtures gate_activations = slim.fully_connected( tf.concat([original_input, mean_output], axis=1), num_mixtures, activation_fn=tf.nn.softmax, weights_regularizer=slim.l2_regularizer(l2_penalty), scope="gates"+sub_scope) # matrix weight_var = tf.get_variable("ensemble_weight", shape=[num_mixtures, num_methods], regularizer=slim.l2_regularizer(l2_penalty)) # weight gated_weight = tf.einsum("ij,jk->ik", gate_activations, weight_var) weight = tf.nn.softmax(gated_weight) # weighted output output = tf.einsum("ik,ijk->ij", weight, model_input) return {"predictions": output} ================================================ FILE: youtube-8m-ensemble/all_ensemble_models/attention_linmatrix_model.py ================================================ import math import models import tensorflow as tf import utils from tensorflow import flags import tensorflow.contrib.slim as slim FLAGS = flags.FLAGS class AttentionLinmatrixModel(models.BaseModel): def create_model(self, model_input, vocab_size, num_mixtures=None, l2_penalty=1e-8, sub_scope="", original_input=None, **unused_params): num_methods = model_input.get_shape().as_list()[-1] num_features = model_input.get_shape().as_list()[-2] num_mixtures = FLAGS.moe_num_mixtures attention_matrix_rank = FLAGS.attention_matrix_rank # gating coefficients original_input = tf.nn.l2_normalize(original_input, dim=1) mean_output = tf.reduce_mean(model_input, axis=2) ## batch_size x moe_num_mixtures gate_activations = slim.fully_connected( tf.concat([original_input, mean_output], axis=1), num_mixtures, activation_fn=tf.nn.softmax, weights_regularizer=slim.l2_regularizer(l2_penalty), scope="gates"+sub_scope) # matrix b = tf.get_variable("ensemble_bias", shape=[num_mixtures,1,1]) weight_var = tf.get_variable("ensemble_weight", shape=[num_mixtures, 1, num_methods], regularizer=slim.l2_regularizer(l2_penalty)) weight_x = tf.get_variable("ensemble_weightx", shape=[num_mixtures, num_features, attention_matrix_rank], regularizer=slim.l2_regularizer(l2_penalty)) weight_y = tf.get_variable("ensemble_weighty", shape=[num_mixtures, attention_matrix_rank, num_methods], regularizer=slim.l2_regularizer(l2_penalty)) ## moe_num_mixtures x num_features x num_methods weight_xy = tf.einsum("ijl,ilk->ijk", weight_x, weight_y) + weight_var + b # weight gated_weight_xy = tf.einsum("ij,jkl->ikl", gate_activations, weight_xy) weight = tf.nn.softmax(gated_weight_xy) # weighted output output = tf.reduce_sum(weight * model_input, axis=2) return {"predictions": output} ================================================ FILE: youtube-8m-ensemble/all_ensemble_models/attention_matrix_model.py ================================================ import math import models import tensorflow as tf import utils from tensorflow import flags import tensorflow.contrib.slim as slim FLAGS = flags.FLAGS class AttentionMatrixModel(models.BaseModel): def create_model(self, model_input, vocab_size, num_mixtures=None, l2_penalty=1e-8, sub_scope="", original_input=None, **unused_params): num_methods = model_input.get_shape().as_list()[-1] num_features = model_input.get_shape().as_list()[-2] num_mixtures = FLAGS.moe_num_mixtures attention_matrix_rank = FLAGS.attention_matrix_rank # gating coefficients original_input = tf.nn.l2_normalize(original_input, dim=1) mean_output = tf.reduce_mean(model_input, axis=2) ## batch_size x moe_num_mixtures gate_activations = slim.fully_connected( tf.concat([original_input, mean_output], axis=1), num_mixtures, activation_fn=tf.nn.softmax, weights_regularizer=slim.l2_regularizer(l2_penalty), scope="gates"+sub_scope) # matrix weight_x = tf.get_variable("ensemble_weightx", shape=[num_mixtures, num_features, attention_matrix_rank], regularizer=slim.l2_regularizer(l2_penalty)) weight_y = tf.get_variable("ensemble_weighty", shape=[num_mixtures, attention_matrix_rank, num_methods], regularizer=slim.l2_regularizer(l2_penalty)) ## moe_num_mixtures x num_features x num_methods weight_xy = tf.einsum("ijl,ilk->ijk", weight_x, weight_y) # weight gated_weight_xy = tf.einsum("ij,jkl->ikl", gate_activations, weight_xy) weight = tf.nn.softmax(gated_weight_xy) # weighted output output = tf.reduce_sum(weight * model_input, axis=2) return {"predictions": output} ================================================ FILE: youtube-8m-ensemble/all_ensemble_models/attention_moe_matrix_model.py ================================================ import math import models import tensorflow as tf import utils from tensorflow import flags import tensorflow.contrib.slim as slim FLAGS = flags.FLAGS class AttentionMoeMatrixModel(models.BaseModel): def create_model(self, model_input, vocab_size, num_mixtures=None, l2_penalty=1e-8, sub_scope="", original_input=None, **unused_params): num_relu = FLAGS.attention_relu_cells num_methods = model_input.get_shape().as_list()[-1] num_features = model_input.get_shape().as_list()[-2] num_mixtures = FLAGS.moe_num_mixtures attention_matrix_rank = FLAGS.attention_matrix_rank # gating coefficients mean_input = tf.reduce_mean(model_input, axis=2) std_input = tf.reduce_sum( tf.square(model_input - tf.expand_dims(mean_input, dim=2)), axis=2) / (num_methods - 1) # relu original_relu = self.relu(original_input, num_relu, sub_scope="origin") mean_relu = self.relu(mean_input, num_relu, sub_scope="mean") std_relu = self.relu(std_input, num_relu, sub_scope="std") # normalize original_relu = tf.nn.l2_normalize(original_relu) mean_relu = tf.nn.l2_normalize(mean_relu) std_relu = tf.nn.l2_normalize(std_relu) ## batch_size x moe_num_mixtures gate_activations = slim.fully_connected( tf.concat([original_relu, mean_relu, std_relu], axis=1), num_mixtures, activation_fn=tf.nn.softmax, weights_regularizer=slim.l2_regularizer(l2_penalty), scope="gates"+sub_scope) # matrix weight_x = tf.get_variable("ensemble_weightx", shape=[num_mixtures, num_features, attention_matrix_rank], regularizer=slim.l2_regularizer(l2_penalty)) weight_y = tf.get_variable("ensemble_weighty", shape=[num_mixtures, attention_matrix_rank, num_methods], regularizer=slim.l2_regularizer(l2_penalty)) ## moe_num_mixtures x num_features x num_methods weight_xy = tf.einsum("ijl,ilk->ijk", weight_x, weight_y) # weight gated_weight_xy = tf.einsum("ij,jkl->ikl", gate_activations, weight_xy) weight = tf.nn.softmax(gated_weight_xy) # weighted output output = tf.reduce_sum(weight * model_input, axis=2) return {"predictions": output} def relu(self, model_input, relu_cells, l2_penalty=1e-8, sub_scope=""): sub_activation = slim.fully_connected( model_input, relu_cells, activation_fn=tf.nn.relu, weights_regularizer=slim.l2_regularizer(l2_penalty), scope="relu-"+sub_scope) return sub_activation ================================================ FILE: youtube-8m-ensemble/all_ensemble_models/attention_moe_model.py ================================================ import math import models import tensorflow as tf import utils from tensorflow import flags import tensorflow.contrib.slim as slim FLAGS = flags.FLAGS class AttentionMoeModel(models.BaseModel): def create_model(self, model_input, vocab_size, num_mixtures=None, l2_penalty=1e-8, sub_scope="", original_input=None, **unused_params): num_relu = FLAGS.attention_relu_cells num_methods = model_input.get_shape().as_list()[-1] num_features = model_input.get_shape().as_list()[-2] original_input = tf.nn.l2_normalize(original_input, dim=1) model_input_list = tf.unstack(model_input, axis=2) relu_units = [self.relu(original_input, num_relu, sub_scope="input")] i = 0 for mi in model_input_list: relu_units.append(self.relu(mi, num_relu, sub_scope="sub"+str(i))) i += 1 gate_activations = slim.fully_connected( tf.concat(relu_units, axis=1), num_methods, activation_fn=None, biases_initializer=None, weights_regularizer=slim.l2_regularizer(l2_penalty), scope="gate") gate = tf.nn.softmax(gate_activations) output = tf.einsum("ijk,ik->ij", model_input, gate) return {"predictions": output} def relu(self, model_input, relu_cells, l2_penalty=1e-8, sub_scope=""): sub_activation = slim.fully_connected( model_input, relu_cells, activation_fn=tf.nn.relu, weights_regularizer=slim.l2_regularizer(l2_penalty), scope="relu-"+sub_scope) return sub_activation ================================================ FILE: youtube-8m-ensemble/all_ensemble_models/attention_rectified_linear_model.py ================================================ import math import models import tensorflow as tf import utils from tensorflow import flags import tensorflow.contrib.slim as slim FLAGS = flags.FLAGS class AttentionRectifiedLinearModel(models.BaseModel): def create_model(self, model_input, vocab_size, num_mixtures=None, l2_penalty=1e-8, sub_scope="", original_input=None, **unused_params): num_methods = model_input.get_shape().as_list()[-1] num_features = model_input.get_shape().as_list()[-2] num_mixtures = FLAGS.moe_num_mixtures # gating coefficients original_input = tf.nn.l2_normalize(original_input, dim=1) mean_output = tf.reduce_mean(model_input, axis=2) ## batch_size x moe_num_mixtures gate_activations = slim.fully_connected( tf.concat([original_input, mean_output], axis=1), num_mixtures, activation_fn=tf.nn.softmax, weights_regularizer=slim.l2_regularizer(l2_penalty), scope="gates"+sub_scope) # matrix weight_var = tf.get_variable("ensemble_weight", shape=[num_mixtures, num_methods], regularizer=slim.l2_regularizer(l2_penalty)) # weight gated_weight = tf.einsum("ij,jk->ik", gate_activations, weight_var) rl_gated_weight = tf.nn.relu(gated_weight) + 1e-9 sum_gated_weight = tf.reduce_sum(rl_gated_weight, axis=1, keep_dims=True) weight = rel_gated_weight / sum_gated_weight # weighted output output = tf.einsum("ik,ijk->ij", weight, model_input) return {"predictions": output} ================================================ FILE: youtube-8m-ensemble/all_ensemble_models/deep_combine_chain_model.py ================================================ import math import models import tensorflow as tf import utils from tensorflow import flags import tensorflow.contrib.slim as slim FLAGS = flags.FLAGS class DeepCombineChainModel(models.BaseModel): def create_model(self, model_input, vocab_size, num_mixtures=None, l2_penalty=1e-8, sub_scope="", original_input=None, **unused_params): num_methods = model_input.get_shape().as_list()[-1] num_features = model_input.get_shape().as_list()[-2] num_mixtures = FLAGS.moe_num_mixtures attention_matrix_rank = FLAGS.attention_matrix_rank relu_cells = FLAGS.deep_chain_relu_cells # mean_output mean_output = tf.reduce_mean(model_input, axis=2) mean_relu = slim.fully_connected( mean_output, relu_cells, activation_fn=tf.nn.relu, weights_regularizer=slim.l2_regularizer(l2_penalty), reuse=True, scope=sub_scope+"transform") permuted_input = tf.transpose(model_input, perm=[0,2,1]) input_relu = slim.fully_connected( permuted_input, relu_cells, activation_fn=tf.nn.relu, weights_regularizer=slim.l2_regularizer(l2_penalty), reuse=True, scope=sub_scope+"transform") input_relu_list = tf.unstack(input_relu, axis=1) predictions = [] discounted_prediction = mean_output prev_relu = mean_relu for i in xrange(num_methods): cur_relu = input_relu_list[i] cur_prediction = self.sub_moe( model_input=tf.concat([prev_relu, cur_relu], axis=1), vocab_size=vocab_size, sub_scope="predict-%d"%i) discounted_prediction = (discounted_prediction + cur_prediction) / 2 # if not the last one if i + 1 < num_methods: predictions.append(cur_prediction) new_relu = slim.fully_connected( cur_prediction, relu_cells, activation_fn=tf.nn.relu, weights_regularizer=slim.l2_regularizer(l2_penalty), reuse=True, scope=sub_scope+"compress") final_prediction = discounted_prediction support_prediction = tf.reduce_mean(tf.stack(predictions, axis=0), axis=0) return {"predictions": final_prediction, "support_predictions": support_prediction} def sub_moe(self, model_input, vocab_size, num_mixtures=None, l2_penalty=1e-8, sub_scope="", **unused_params): num_mixtures = num_mixtures or FLAGS.moe_num_mixtures gate_activations = slim.fully_connected( model_input, vocab_size * (num_mixtures + 1), activation_fn=None, biases_initializer=None, weights_regularizer=slim.l2_regularizer(l2_penalty), scope="gates-"+sub_scope) expert_activations = slim.fully_connected( model_input, vocab_size * num_mixtures, activation_fn=None, weights_regularizer=slim.l2_regularizer(l2_penalty), scope="experts-"+sub_scope) gating_distribution = tf.nn.softmax(tf.reshape( gate_activations, [-1, num_mixtures + 1])) # (Batch * #Labels) x (num_mixtures + 1) expert_distribution = tf.nn.sigmoid(tf.reshape( expert_activations, [-1, num_mixtures])) # (Batch * #Labels) x num_mixtures final_probabilities_by_class_and_batch = tf.reduce_sum( gating_distribution[:, :num_mixtures] * expert_distribution, 1) final_probabilities = tf.reshape(final_probabilities_by_class_and_batch, [-1, vocab_size]) return final_probabilities ================================================ FILE: youtube-8m-ensemble/all_ensemble_models/input_moe_model.py ================================================ import math import models import tensorflow as tf import utils from tensorflow import flags import tensorflow.contrib.slim as slim FLAGS = flags.FLAGS class InputMoeModel(models.BaseModel): """A softmax over a mixture of logistic models (with L2 regularization).""" def create_model(self, model_input, vocab_size, num_mixtures=None, l2_penalty=1e-8, sub_scope="", original_input=None, **unused_params): num_methods = model_input.get_shape().as_list()[-1] num_features = model_input.get_shape().as_list()[-2] original_input = tf.nn.l2_normalize(original_input, dim=1) gate_activations = slim.fully_connected( original_input, num_methods, activation_fn=tf.nn.softmax, weights_regularizer=slim.l2_regularizer(l2_penalty), scope="gates"+sub_scope) output = tf.einsum("ijk,ik->ij", model_input, gate_activations) return {"predictions": output} ================================================ FILE: youtube-8m-ensemble/all_ensemble_models/linear_regression_model.py ================================================ import math import models import tensorflow as tf import utils from tensorflow import flags import tensorflow.contrib.slim as slim FLAGS = flags.FLAGS class LinearRegressionModel(models.BaseModel): """Logistic model with L2 regularization.""" def create_model(self, model_input, vocab_size, l2_penalty=1e-8, original_input=None, **unused_params): """Creates a linear regression model. Args: model_input: 'batch' x 'num_features' x 'num_methods' matrix of input features. vocab_size: The number of classes in the dataset. Returns: A dictionary with a tensor containing the probability predictions of the model in the 'predictions' key. The dimensions of the tensor are batch_size x num_classes.""" num_methods = model_input.get_shape().as_list()[-1] weight = tf.get_variable("ensemble_weight", shape=[num_methods], regularizer=slim.l2_regularizer(l2_penalty)) weight = tf.nn.softmax(weight) output = tf.einsum("ijk,k->ij", model_input, weight) return {"predictions": output} ================================================ FILE: youtube-8m-ensemble/all_ensemble_models/logistic_model.py ================================================ import math import models import tensorflow as tf import utils from tensorflow import flags import tensorflow.contrib.slim as slim FLAGS = flags.FLAGS class LogisticModel(models.BaseModel): """Logistic model with L2 regularization.""" def create_model(self, model_input, vocab_size, l2_penalty=1e-8, original_input=None, **unused_params): """Creates a logistic model. Args: model_input: 'batch' x 'num_features' matrix of input features. vocab_size: The number of classes in the dataset. Returns: A dictionary with a tensor containing the probability predictions of the model in the 'predictions' key. The dimensions of the tensor are batch_size x num_classes.""" output = slim.fully_connected( model_input, vocab_size, activation_fn=tf.nn.sigmoid, weights_regularizer=slim.l2_regularizer(l2_penalty)) return {"predictions": output} ================================================ FILE: youtube-8m-ensemble/all_ensemble_models/matrix_regression_model.py ================================================ import math import models import tensorflow as tf import utils from tensorflow import flags import tensorflow.contrib.slim as slim FLAGS = flags.FLAGS class MatrixRegressionModel(models.BaseModel): """Logistic model with L2 regularization.""" def create_model(self, model_input, vocab_size, l2_penalty=1e-8, original_input=None, **unused_params): """Creates a matrix regression model. Args: model_input: 'batch' x 'num_features' x 'num_methods' matrix of input features. vocab_size: The number of classes in the dataset. Returns: A dictionary with a tensor containing the probability predictions of the model in the 'predictions' key. The dimensions of the tensor are batch_size x num_classes.""" num_features = model_input.get_shape().as_list()[-2] num_methods = model_input.get_shape().as_list()[-1] weight1d = tf.get_variable("ensemble_weight1d", shape=[num_methods], regularizer=slim.l2_regularizer(l2_penalty)) weight2d = tf.get_variable("ensemble_weight2d", shape=[num_features, num_methods], regularizer=slim.l2_regularizer(10 * l2_penalty)) weight = tf.nn.softmax(tf.einsum("ij,j->ij", weight2d, weight1d), dim=-1) output = tf.einsum("ijk,jk->ij", model_input, weight) return {"predictions": output} ================================================ FILE: youtube-8m-ensemble/all_ensemble_models/mean_model.py ================================================ import math import models import tensorflow as tf import utils from tensorflow import flags import tensorflow.contrib.slim as slim FLAGS = flags.FLAGS class MeanModel(models.BaseModel): """Mean model.""" def create_model(self, model_input, **unused_params): """Creates a logistic model. model_input: 'batch' x 'num_features' x 'num_methods' matrix of input features. vocab_size: The number of classes in the dataset. Returns: A dictionary with a tensor containing the probability predictions of the model in the 'predictions' key. The dimensions of the tensor are batch_size x num_classes.""" output = tf.reduce_mean(model_input, axis=2) return {"predictions": output} ================================================ FILE: youtube-8m-ensemble/all_ensemble_models/moe_model.py ================================================ import math import models import tensorflow as tf import utils from tensorflow import flags import tensorflow.contrib.slim as slim FLAGS = flags.FLAGS class MoeModel(models.BaseModel): """A softmax over a mixture of logistic models (with L2 regularization).""" def create_model(self, model_input, vocab_size, num_mixtures=None, l2_penalty=1e-8, sub_scope="", original_input=None, **unused_params): """Creates a Mixture of (Logistic) Experts model. The model consists of a per-class softmax distribution over a configurable number of logistic classifiers. One of the classifiers in the mixture is not trained, and always predicts 0. Args: model_input: 'batch_size' x 'num_features' matrix of input features. vocab_size: The number of classes in the dataset. num_mixtures: The number of mixtures (excluding a dummy 'expert' that always predicts the non-existence of an entity). l2_penalty: How much to penalize the squared magnitudes of parameter values. Returns: A dictionary with a tensor containing the probability predictions of the model in the 'predictions' key. The dimensions of the tensor are batch_size x num_classes. """ num_methods = model_input.get_shape().as_list()[-1] num_features = model_input.get_shape().as_list()[-2] flat_input = tf.reshape(model_input, shape=[-1,num_features * num_methods]) tensor_weight = tf.get_variable("tensor_weight", shape=[num_features, num_methods, num_methods], regularizer=slim.l2_regularizer(l2_penalty)) tensor_bias = tf.get_variable("tensor_bias", shape=[num_features, num_methods], initializer=tf.zeros_initializer(), regularizer=slim.l2_regularizer(l2_penalty)) gate_activations = tf.einsum("ijk,jkl->ijl", model_input, tensor_weight) \ + tf.expand_dims(tensor_bias, dim=0) output = tf.reduce_sum(model_input * tf.nn.softmax(gate_activations), axis=2) return {"predictions": output} ================================================ FILE: youtube-8m-ensemble/all_ensemble_models/nonunit_matrix_regression_model.py ================================================ import math import models import tensorflow as tf import utils from tensorflow import flags import tensorflow.contrib.slim as slim FLAGS = flags.FLAGS class NonunitMatrixRegressionModel(models.BaseModel): """Logistic model with L2 regularization.""" def create_model(self, model_input, vocab_size, l2_penalty=1e-8, original_input=None, epsilon=1e-5, **unused_params): """Creates a non-unified matrix regression model. Args: model_input: 'batch' x 'num_features' x 'num_methods' matrix of input features. vocab_size: The number of classes in the dataset. Returns: A dictionary with a tensor containing the probability predictions of the model in the 'predictions' key. The dimensions of the tensor are batch_size x num_classes.""" num_features = model_input.get_shape().as_list()[-2] num_methods = model_input.get_shape().as_list()[-1] log_model_input = tf.stop_gradient(tf.log((epsilon + model_input) / (1.0 + epsilon - model_input))) weight = tf.get_variable("ensemble_weight", shape=[num_features, num_methods], regularizer=slim.l2_regularizer(l2_penalty)) weight = tf.nn.softmax(weight) output = tf.nn.sigmoid(tf.einsum("ijk,jk->ij", log_model_input, weight)) return {"predictions": output} ================================================ FILE: youtube-8m-ensemble/average_precision_calculator.py ================================================ # Copyright 2016 Google Inc. All Rights Reserved. # # Licensed under the Apache License, Version 2.0 (the "License"); # you may not use this file except in compliance with the License. # You may obtain a copy of the License at # # http://www.apache.org/licenses/LICENSE-2.0 # # Unless required by applicable law or agreed to in writing, software # distributed under the License is distributed on an "AS-IS" BASIS, # WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied. # See the License for the specific language governing permissions and # limitations under the License. """Calculate or keep track of the interpolated average precision. It provides an interface for calculating interpolated average precision for an entire list or the top-n ranked items. For the definition of the (non-)interpolated average precision: http://trec.nist.gov/pubs/trec15/appendices/CE.MEASURES06.pdf Example usages: 1) Use it as a static function call to directly calculate average precision for a short ranked list in the memory. ``` import random p = np.array([random.random() for _ in xrange(10)]) a = np.array([random.choice([0, 1]) for _ in xrange(10)]) ap = average_precision_calculator.AveragePrecisionCalculator.ap(p, a) ``` 2) Use it as an object for long ranked list that cannot be stored in memory or the case where partial predictions can be observed at a time (Tensorflow predictions). In this case, we first call the function accumulate many times to process parts of the ranked list. After processing all the parts, we call peek_interpolated_ap_at_n. ``` p1 = np.array([random.random() for _ in xrange(5)]) a1 = np.array([random.choice([0, 1]) for _ in xrange(5)]) p2 = np.array([random.random() for _ in xrange(5)]) a2 = np.array([random.choice([0, 1]) for _ in xrange(5)]) # interpolated average precision at 10 using 1000 break points calculator = average_precision_calculator.AveragePrecisionCalculator(10) calculator.accumulate(p1, a1) calculator.accumulate(p2, a2) ap3 = calculator.peek_ap_at_n() ``` """ import heapq import random import numbers import numpy class AveragePrecisionCalculator(object): """Calculate the average precision and average precision at n.""" def __init__(self, top_n=None): """Construct an AveragePrecisionCalculator to calculate average precision. This class is used to calculate the average precision for a single label. Args: top_n: A positive Integer specifying the average precision at n, or None to use all provided data points. Raises: ValueError: An error occurred when the top_n is not a positive integer. """ if not ((isinstance(top_n, int) and top_n >= 0) or top_n is None): raise ValueError("top_n must be a positive integer or None.") self._top_n = top_n # average precision at n self._total_positives = 0 # total number of positives have seen self._heap = [] # max heap of (prediction, actual) @property def heap_size(self): """Gets the heap size maintained in the class.""" return len(self._heap) @property def num_accumulated_positives(self): """Gets the number of positive samples that have been accumulated.""" return self._total_positives def accumulate(self, predictions, actuals, num_positives=None): """Accumulate the predictions and their ground truth labels. After the function call, we may call peek_ap_at_n to actually calculate the average precision. Note predictions and actuals must have the same shape. Args: predictions: a list storing the prediction scores. actuals: a list storing the ground truth labels. Any value larger than 0 will be treated as positives, otherwise as negatives. num_positives = If the 'predictions' and 'actuals' inputs aren't complete, then it's possible some true positives were missed in them. In that case, you can provide 'num_positives' in order to accurately track recall. Raises: ValueError: An error occurred when the format of the input is not the numpy 1-D array or the shape of predictions and actuals does not match. """ if len(predictions) != len(actuals): raise ValueError("the shape of predictions and actuals does not match.") if not num_positives is None: if not isinstance(num_positives, numbers.Number) or num_positives < 0: raise ValueError("'num_positives' was provided but it wan't a nonzero number.") if not num_positives is None: self._total_positives += num_positives else: self._total_positives += numpy.size(numpy.where(actuals > 0)) topk = self._top_n heap = self._heap for i in range(numpy.size(predictions)): if topk is None or len(heap) < topk: heapq.heappush(heap, (predictions[i], actuals[i])) else: if predictions[i] > heap[0][0]: # heap[0] is the smallest heapq.heappop(heap) heapq.heappush(heap, (predictions[i], actuals[i])) def clear(self): """Clear the accumulated predictions.""" self._heap = [] self._total_positives = 0 def peek_ap_at_n(self): """Peek the non-interpolated average precision at n. Returns: The non-interpolated average precision at n (default 0). If n is larger than the length of the ranked list, the average precision will be returned. """ if self.heap_size <= 0: return 0 predlists = numpy.array(list(zip(*self._heap))) ap = self.ap_at_n(predlists[0], predlists[1], n=self._top_n, total_num_positives=self._total_positives) return ap @staticmethod def ap(predictions, actuals): """Calculate the non-interpolated average precision. Args: predictions: a numpy 1-D array storing the sparse prediction scores. actuals: a numpy 1-D array storing the ground truth labels. Any value larger than 0 will be treated as positives, otherwise as negatives. Returns: The non-interpolated average precision at n. If n is larger than the length of the ranked list, the average precision will be returned. Raises: ValueError: An error occurred when the format of the input is not the numpy 1-D array or the shape of predictions and actuals does not match. """ return AveragePrecisionCalculator.ap_at_n(predictions, actuals, n=None) @staticmethod def ap_at_n(predictions, actuals, n=20, total_num_positives=None): """Calculate the non-interpolated average precision. Args: predictions: a numpy 1-D array storing the sparse prediction scores. actuals: a numpy 1-D array storing the ground truth labels. Any value larger than 0 will be treated as positives, otherwise as negatives. n: the top n items to be considered in ap@n. total_num_positives : (optionally) you can specify the number of total positive in the list. If specified, it will be used in calculation. Returns: The non-interpolated average precision at n. If n is larger than the length of the ranked list, the average precision will be returned. Raises: ValueError: An error occurred when 1) the format of the input is not the numpy 1-D array; 2) the shape of predictions and actuals does not match; 3) the input n is not a positive integer. """ if len(predictions) != len(actuals): raise ValueError("the shape of predictions and actuals does not match.") if n is not None: if not isinstance(n, int) or n <= 0: raise ValueError("n must be 'None' or a positive integer." " It was '%s'." % n) ap = 0.0 predictions = numpy.array(predictions) actuals = numpy.array(actuals) # add a shuffler to avoid overestimating the ap predictions, actuals = AveragePrecisionCalculator._shuffle(predictions, actuals) sortidx = sorted( range(len(predictions)), key=lambda k: predictions[k], reverse=True) if total_num_positives is None: numpos = numpy.size(numpy.where(actuals > 0)) else: numpos = total_num_positives if numpos == 0: return 0 if n is not None: numpos = min(numpos, n) delta_recall = 1.0 / numpos poscount = 0.0 # calculate the ap r = len(sortidx) if n is not None: r = min(r, n) for i in range(r): if actuals[sortidx[i]] > 0: poscount += 1 ap += poscount / (i + 1) * delta_recall return ap @staticmethod def _shuffle(predictions, actuals): random.seed(0) suffidx = random.sample(range(len(predictions)), len(predictions)) predictions = predictions[suffidx] actuals = actuals[suffidx] return predictions, actuals @staticmethod def _zero_one_normalize(predictions, epsilon=1e-7): """Normalize the predictions to the range between 0.0 and 1.0. For some predictions like SVM predictions, we need to normalize them before calculate the interpolated average precision. The normalization will not change the rank in the original list and thus won't change the average precision. Args: predictions: a numpy 1-D array storing the sparse prediction scores. epsilon: a small constant to avoid denominator being zero. Returns: The normalized prediction. """ denominator = numpy.max(predictions) - numpy.min(predictions) ret = (predictions - numpy.min(predictions)) / numpy.max(denominator, epsilon) return ret ================================================ FILE: youtube-8m-ensemble/check_distillation.py ================================================ # Copyright 2016 Google Inc. All Rights Reserved. # # Licensed under the Apache License, Version 2.0 (the "License"); # you may not use this file except in compliance with the License. # You may obtain a copy of the License at # # http://www.apache.org/licenses/LICENSE-2.0 # # Unless required by applicable law or agreed to in writing, software # distributed under the License is distributed on an "AS-IS" BASIS, # WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied. # See the License for the specific language governing permissions and # limitations under the License. """Binary for evaluating Tensorflow models on the YouTube-8M dataset.""" import time import numpy as np import eval_util import losses import ensemble_level_models import readers import tensorflow as tf from tensorflow import app from tensorflow import flags from tensorflow import gfile from tensorflow import logging import utils FLAGS = flags.FLAGS if __name__ == "__main__": # Dataset flags. flags.DEFINE_string( "eval_data_patterns", "", "File globs defining the evaluation dataset in tensorflow.SequenceExample format.") flags.DEFINE_string( "input_data_pattern", None, "File globs for original model input.") flags.DEFINE_string("feature_names", "predictions", "Name of the feature " "to use for training.") flags.DEFINE_string("feature_sizes", "4716", "Length of the feature vectors.") # Model flags. flags.DEFINE_integer("batch_size", 1024, "How many examples to process per batch.") def find_class_by_name(name, modules): """Searches the provided modules for the named class and returns it.""" modules = [getattr(module, name, None) for module in modules] return next(a for a in modules if a) def get_input_evaluation_tensors(reader, data_pattern, batch_size=256): logging.info("Using batch size of " + str(batch_size) + " for evaluation.") with tf.name_scope("eval_input"): files = gfile.Glob(data_pattern) if not files: print data_pattern, files raise IOError("Unable to find the evaluation files.") logging.info("number of evaluation files: " + str(len(files))) files.sort() filename_queue = tf.train.string_input_producer( files, shuffle=False, num_epochs=1) eval_data = reader.prepare_reader(filename_queue) return tf.train.batch( eval_data, batch_size=batch_size, capacity=3 * batch_size, allow_smaller_final_batch=True, enqueue_many=True) def build_graph(all_readers, input_reader, input_data_pattern, all_eval_data_patterns, batch_size=256): original_video_id, original_input, unused_labels_batch, unused_num_frames = ( get_input_evaluation_tensors( input_reader, input_data_pattern, batch_size=batch_size)) video_id_equal_tensors = [] model_input_tensor = None input_distance_tensors = [] model_label_tensor = tf.cast(unused_labels_batch, dtype=tf.float32) label_distance_tensors = [] for reader, data_pattern in zip(all_readers, all_eval_data_patterns): video_id, model_input_raw, labels_batch, unused_num_frames = ( get_input_evaluation_tensors( reader, data_pattern, batch_size=batch_size)) video_id_equal_tensors.append(tf.reduce_sum(tf.cast(tf.not_equal(original_video_id, video_id), dtype=tf.float32))) input_distance_tensors.append(tf.reduce_mean(tf.reduce_sum(tf.square(original_input - model_input_raw), axis=1))) labels_batch = tf.cast(labels_batch, dtype=tf.float32) x = model_input_raw y = labels_batch ce = - y * tf.log(x + 1e-7) - (1.0 - y) * tf.log(1.0 + 1e-7 - x) label_distance_tensors.append(tf.reduce_mean(tf.reduce_sum(ce, axis=1))) video_id_equal_tensor = tf.stack(video_id_equal_tensors) input_distance_tensor = tf.stack(input_distance_tensors) label_distance_tensor = tf.stack(label_distance_tensors) tf.add_to_collection("model_input", model_input_tensor) tf.add_to_collection("video_id_equal", video_id_equal_tensor) tf.add_to_collection("input_distance", input_distance_tensor) tf.add_to_collection("label_distance", label_distance_tensor) def check_loop(model_input, video_id_equal, input_distance, label_distance, all_patterns): with tf.Session() as sess: sess.run([tf.local_variables_initializer()]) # Start the queue runners. coord = tf.train.Coordinator() try: threads = [] for qr in tf.get_collection(tf.GraphKeys.QUEUE_RUNNERS): threads.extend(qr.create_threads( sess, coord=coord, daemon=True, start=True)) examples_processed = 0 while not coord.should_stop(): batch_start_time = time.time() model_input_val, video_id_equal_val, input_distance_val, label_distance_val = sess.run([model_input, video_id_equal, input_distance, label_distance]) print "model_input.max", np.max(model_input_val) print "model_input.min", np.min(model_input_val) print "input_distance_val", input_distance_val print "label_distance_val", label_distance_val for i in xrange(video_id_equal_val.shape[0]): if video_id_equal_val[i] > 0: print "%d discrepancies in %s" % (int(video_id_equal_val[i]), all_patterns[i]) seconds_per_batch = time.time() - batch_start_time example_per_second = video_id_equal_val.shape[0] / seconds_per_batch examples_processed += video_id_equal_val.shape[0] logging.info("examples_processed: %d", examples_processed) except tf.errors.OutOfRangeError as e: logging.info( "Done with batched inference. Now calculating global performance " "metrics.") except Exception as e: # pylint: disable=broad-except logging.info("Unexpected exception: " + str(e)) coord.request_stop(e) coord.request_stop() coord.join(threads, stop_grace_period_secs=10) return global_step_val def check_video_id(): tf.set_random_seed(0) # for reproducibility with tf.Graph().as_default(): # convert feature_names and feature_sizes to lists of values feature_names, feature_sizes = utils.GetListOfFeatureNamesAndSizes( FLAGS.feature_names, FLAGS.feature_sizes) # prepare a reader for each single model prediction result all_readers = [] all_patterns = FLAGS.eval_data_patterns all_patterns = map(lambda x: x.strip(), all_patterns.strip().strip(",").split(",")) for i in xrange(len(all_patterns)): reader = readers.EnsembleReader( feature_names=feature_names, feature_sizes=feature_sizes) all_readers.append(reader) input_reader = None input_data_pattern = None if FLAGS.input_data_pattern is not None: input_reader = readers.EnsembleReader( feature_names=["mean_rgb","mean_audio"], feature_sizes=[1024,128]) input_data_pattern = FLAGS.input_data_pattern if FLAGS.eval_data_patterns is "": raise IOError("'eval_data_patterns' was not specified. " + "Nothing to evaluate.") build_graph( all_readers=all_readers, input_reader=input_reader, input_data_pattern=input_data_pattern, all_eval_data_patterns=all_patterns, batch_size=FLAGS.batch_size) logging.info("built evaluation graph") video_id_equal = tf.get_collection("video_id_equal")[0] model_input = tf.get_collection("model_input")[0] input_distance = tf.get_collection("input_distance")[0] label_distance = tf.get_collection("label_distance")[0] check_loop(model_input, video_id_equal, input_distance, label_distance, all_patterns) def main(unused_argv): logging.set_verbosity(tf.logging.INFO) print("tensorflow version: %s" % tf.__version__) check_video_id() if __name__ == "__main__": app.run() ================================================ FILE: youtube-8m-ensemble/check_video_id.py ================================================ # Copyright 2016 Google Inc. All Rights Reserved. # # Licensed under the Apache License, Version 2.0 (the "License"); # you may not use this file except in compliance with the License. # You may obtain a copy of the License at # # http://www.apache.org/licenses/LICENSE-2.0 # # Unless required by applicable law or agreed to in writing, software # distributed under the License is distributed on an "AS-IS" BASIS, # WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied. # See the License for the specific language governing permissions and # limitations under the License. """Binary for evaluating Tensorflow models on the YouTube-8M dataset.""" import time import eval_util import losses import ensemble_level_models import readers import tensorflow as tf from tensorflow import app from tensorflow import flags from tensorflow import gfile from tensorflow import logging import utils FLAGS = flags.FLAGS if __name__ == "__main__": # Dataset flags. flags.DEFINE_string("model_checkpoint_path", "", "The file to load the model files from. ") flags.DEFINE_string("train_dir", "/tmp/yt8m/", "The directory to write the result in. ") flags.DEFINE_string( "eval_data_patterns", "", "File globs defining the evaluation dataset in tensorflow.SequenceExample format.") flags.DEFINE_string( "input_data_pattern", None, "File globs for original model input.") flags.DEFINE_string("feature_names", "predictions", "Name of the feature " "to use for training.") flags.DEFINE_string("feature_sizes", "4716", "Length of the feature vectors.") # Model flags. flags.DEFINE_string( "model", "LinearRegressionModel", "Which architecture to use for the model.") flags.DEFINE_integer("batch_size", 1024, "How many examples to process per batch.") flags.DEFINE_string("label_loss", "CrossEntropyLoss", "Loss computed on validation data") # Other flags. flags.DEFINE_boolean("run_once", True, "Whether to run eval only once.") flags.DEFINE_boolean("echo_gap", False, "Whether to echo GAP at the end.") flags.DEFINE_integer("top_k", 20, "How many predictions to output per video.") def find_class_by_name(name, modules): """Searches the provided modules for the named class and returns it.""" modules = [getattr(module, name, None) for module in modules] return next(a for a in modules if a) def get_input_evaluation_tensors(reader, data_pattern, batch_size=256): logging.info("Using batch size of " + str(batch_size) + " for evaluation.") with tf.name_scope("eval_input"): files = gfile.Glob(data_pattern) if not files: print data_pattern, files raise IOError("Unable to find the evaluation files.") logging.info("number of evaluation files: " + str(len(files))) files.sort() filename_queue = tf.train.string_input_producer( files, shuffle=False, num_epochs=1) eval_data = reader.prepare_reader(filename_queue) return tf.train.batch( eval_data, batch_size=batch_size, capacity=3 * batch_size, allow_smaller_final_batch=True, enqueue_many=True) def build_graph(all_readers, input_reader, input_data_pattern, all_eval_data_patterns, batch_size=256): original_video_id, original_input, unused_labels_batch, unused_num_frames = ( get_input_evaluation_tensors( input_reader, input_data_pattern, batch_size=batch_size)) video_id_equal_tensors = [] model_input_tensor = None input_distance_tensors = [] for reader, data_pattern in zip(all_readers, all_eval_data_patterns): video_id, model_input_raw, labels_batch, unused_num_frames = ( get_input_evaluation_tensors( reader, data_pattern, batch_size=batch_size)) video_id_equal_tensors.append(tf.reduce_sum(tf.cast(tf.not_equal(original_video_id, video_id), dtype=tf.float32))) if model_input_tensor is None: model_input_tensor = model_input_raw input_distance_tensors.append(tf.reduce_mean(tf.reduce_sum(tf.square(model_input_tensor - model_input_raw), axis=1))) video_id_equal_tensor = tf.stack(video_id_equal_tensors) input_distance_tensor = tf.stack(input_distance_tensors) tf.add_to_collection("video_id_equal", video_id_equal_tensor) tf.add_to_collection("input_distance", input_distance_tensor) def check_loop(video_id_equal, input_distance, all_patterns): with tf.Session() as sess: sess.run([tf.local_variables_initializer()]) # Start the queue runners. coord = tf.train.Coordinator() try: threads = [] for qr in tf.get_collection(tf.GraphKeys.QUEUE_RUNNERS): threads.extend(qr.create_threads( sess, coord=coord, daemon=True, start=True)) examples_processed = 0 while not coord.should_stop(): batch_start_time = time.time() video_id_equal_val, input_distance_val = sess.run([video_id_equal, input_distance]) print input_distance_val for i in xrange(video_id_equal_val.shape[0]): if video_id_equal_val[i] > 0: print "%d discrepancies in %s" % (int(video_id_equal_val[i]), all_patterns[i]) seconds_per_batch = time.time() - batch_start_time example_per_second = video_id_equal_val.shape[0] / seconds_per_batch examples_processed += video_id_equal_val.shape[0] logging.info("examples_processed: %d", examples_processed) except tf.errors.OutOfRangeError as e: logging.info( "Done with batched inference. Now calculating global performance " "metrics.") except Exception as e: # pylint: disable=broad-except logging.info("Unexpected exception: " + str(e)) coord.request_stop(e) coord.request_stop() coord.join(threads, stop_grace_period_secs=10) return global_step_val def check_video_id(): tf.set_random_seed(0) # for reproducibility with tf.Graph().as_default(): # convert feature_names and feature_sizes to lists of values feature_names, feature_sizes = utils.GetListOfFeatureNamesAndSizes( FLAGS.feature_names, FLAGS.feature_sizes) # prepare a reader for each single model prediction result all_readers = [] all_patterns = FLAGS.eval_data_patterns all_patterns = map(lambda x: x.strip(), all_patterns.strip().strip(",").split(",")) for i in xrange(len(all_patterns)): reader = readers.EnsembleReader( feature_names=feature_names, feature_sizes=feature_sizes) all_readers.append(reader) input_reader = None input_data_pattern = None if FLAGS.input_data_pattern is not None: input_reader = readers.EnsembleReader( feature_names=["mean_rgb","mean_audio"], feature_sizes=[1024,128]) input_data_pattern = FLAGS.input_data_pattern if FLAGS.eval_data_patterns is "": raise IOError("'eval_data_patterns' was not specified. " + "Nothing to evaluate.") build_graph( all_readers=all_readers, input_reader=input_reader, input_data_pattern=input_data_pattern, all_eval_data_patterns=all_patterns, batch_size=FLAGS.batch_size) logging.info("built evaluation graph") video_id_equal = tf.get_collection("video_id_equal")[0] input_distance = tf.get_collection("input_distance")[0] check_loop(video_id_equal, input_distance, all_patterns) def main(unused_argv): logging.set_verbosity(tf.logging.INFO) print("tensorflow version: %s" % tf.__version__) check_video_id() if __name__ == "__main__": app.run() ================================================ FILE: youtube-8m-ensemble/check_video_id_match.py ================================================ # Copyright 2016 Google Inc. All Rights Reserved. # # Licensed under the Apache License, Version 2.0 (the "License"); # you may not use this file except in compliance with the License. # You may obtain a copy of the License at # # http://www.apache.org/licenses/LICENSE-2.0 # # Unless required by applicable law or agreed to in writing, software # distributed under the License is distributed on an "AS-IS" BASIS, # WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied. # See the License for the specific language governing permissions and # limitations under the License. """Binary for evaluating Tensorflow models on the YouTube-8M dataset.""" import time import eval_util import losses import ensemble_level_models import readers import tensorflow as tf from tensorflow import app from tensorflow import flags from tensorflow import gfile from tensorflow import logging import utils FLAGS = flags.FLAGS if __name__ == "__main__": # Dataset flags. flags.DEFINE_string("model_checkpoint_path", "", "The file to load the model files from. ") flags.DEFINE_string("train_dir", "/tmp/yt8m/", "The directory to write the result in. ") flags.DEFINE_string( "eval_data_patterns", "", "File globs defining the evaluation dataset in tensorflow.SequenceExample format.") flags.DEFINE_string( "input_data_pattern", None, "File globs for original model input.") flags.DEFINE_string("feature_names", "predictions", "Name of the feature " "to use for training.") flags.DEFINE_string("feature_sizes", "4716", "Length of the feature vectors.") # Model flags. flags.DEFINE_string( "model", "LinearRegressionModel", "Which architecture to use for the model.") flags.DEFINE_integer("batch_size", 1024, "How many examples to process per batch.") flags.DEFINE_string("label_loss", "CrossEntropyLoss", "Loss computed on validation data") # Other flags. flags.DEFINE_boolean("run_once", True, "Whether to run eval only once.") flags.DEFINE_boolean("echo_gap", False, "Whether to echo GAP at the end.") flags.DEFINE_integer("top_k", 20, "How many predictions to output per video.") def find_class_by_name(name, modules): """Searches the provided modules for the named class and returns it.""" modules = [getattr(module, name, None) for module in modules] return next(a for a in modules if a) def get_input_evaluation_tensors(reader, data_pattern, batch_size=256): logging.info("Using batch size of " + str(batch_size) + " for evaluation.") with tf.name_scope("eval_input"): files = gfile.Glob(data_pattern) if not files: print data_pattern, files raise IOError("Unable to find the evaluation files.") logging.info("number of evaluation files: " + str(len(files))) files.sort() filename_queue = tf.train.string_input_producer( files, shuffle=False, num_epochs=1) eval_data = reader.prepare_reader(filename_queue) return tf.train.batch( eval_data, batch_size=batch_size, capacity=3 * FLAGS.batch_size, allow_smaller_final_batch=True, enqueue_many=True) def build_graph(all_readers, input_reader, input_data_pattern, all_eval_data_patterns, batch_size=256): original_video_id, original_input, unused_labels_batch, unused_num_frames = ( get_input_evaluation_tensors( input_reader, input_data_pattern, batch_size=batch_size)) video_id_notequal_tensors = [] model_input_tensor = None input_distance_tensors = [] for reader, data_pattern in zip(all_readers, all_eval_data_patterns): video_id, model_input_raw, labels_batch, unused_num_frames = ( get_input_evaluation_tensors( reader, data_pattern, batch_size=batch_size)) video_id_notequal_tensors.append(tf.reduce_sum(tf.cast(tf.not_equal(original_video_id, video_id), dtype=tf.float32))) if model_input_tensor is None: model_input_tensor = model_input_raw input_distance_tensors.append(tf.reduce_mean(tf.reduce_sum(tf.square(model_input_tensor - model_input_raw), axis=1))) video_id_mismatch_tensor = tf.stack(video_id_notequal_tensors) input_distance_tensor = tf.stack(input_distance_tensors) actual_batch_size = tf.shape(original_video_id)[0] tf.add_to_collection("video_id_mismatch", video_id_mismatch_tensor) tf.add_to_collection("input_distance", input_distance_tensor) tf.add_to_collection("actual_batch_size", actual_batch_size) def check_loop(video_id_mismatch, input_distance, actual_batch_size, all_patterns): count_mismatch = [0] * len(all_patterns) sum_distance = [0.0] * len(all_patterns) with tf.Session() as sess: sess.run([tf.local_variables_initializer()]) # Start the queue runners. start_time = time.time() coord = tf.train.Coordinator() try: threads = [] for qr in tf.get_collection(tf.GraphKeys.QUEUE_RUNNERS): threads.extend(qr.create_threads( sess, coord=coord, daemon=True, start=True)) examples_processed = 0 j = 0 while not coord.should_stop(): batch_start_time = time.time() video_id_mismatch_val, input_distance_val, batch_size_val = sess.run([video_id_mismatch, input_distance, actual_batch_size]) for i in xrange(len(all_patterns)): count_mismatch[i] += int(video_id_mismatch_val[i]) sum_distance[i] += float(input_distance_val[i]) print "mismatch", count_mismatch seconds_per_batch = time.time() - batch_start_time example_per_second = batch_size_val / seconds_per_batch examples_processed += batch_size_val logging.info("examples_processed: %d, time elapsed %f s", examples_processed, time.time()-start_time) j += 1 if j % 100 == 0: print "---------------- SUMMARY -------------------" print "COUNT \t DIST \t PATTERNS" for i in xrange(len(all_patterns)): print count_mismatch[i], "\t", sum_distance[i], "\t", all_patterns[i] except tf.errors.OutOfRangeError as e: logging.info( "Done with batched inference. Now calculating global performance " "metrics.") except Exception as e: # pylint: disable=broad-except logging.info("Unexpected exception: " + str(e)) coord.request_stop(e) coord.request_stop() coord.join(threads, stop_grace_period_secs=10) print "---------------- FINAL SUMMARY -------------------" print "COUNT \t DIST \t PATTERNS" for i in xrange(len(all_patterns)): print count_mismatch[i], "\t", sum_distance[i], "\t", all_patterns[i] def check_video_id(): tf.set_random_seed(0) # for reproducibility with tf.Graph().as_default(): # convert feature_names and feature_sizes to lists of values feature_names, feature_sizes = utils.GetListOfFeatureNamesAndSizes( FLAGS.feature_names, FLAGS.feature_sizes) # prepare a reader for each single model prediction result all_readers = [] all_patterns = FLAGS.eval_data_patterns all_patterns = map(lambda x: x.strip(), all_patterns.strip().strip(",").split(",")) for i in xrange(len(all_patterns)): reader = readers.EnsembleReader( feature_names=feature_names, feature_sizes=feature_sizes) all_readers.append(reader) input_reader = None input_data_pattern = None if FLAGS.input_data_pattern is not None: input_reader = readers.EnsembleReader( feature_names=["mean_rgb","mean_audio"], feature_sizes=[1024,128]) # input_reader = readers.EnsembleReader( # feature_names=["input"], feature_sizes=[1024+128]) input_data_pattern = FLAGS.input_data_pattern if FLAGS.eval_data_patterns is "": raise IOError("'eval_data_patterns' was not specified. " + "Nothing to evaluate.") build_graph( all_readers=all_readers, input_reader=input_reader, input_data_pattern=input_data_pattern, all_eval_data_patterns=all_patterns, batch_size=FLAGS.batch_size) logging.info("built evaluation graph") video_id_mismatch = tf.get_collection("video_id_mismatch")[0] input_distance = tf.get_collection("input_distance")[0] actual_batch_size = tf.get_collection("actual_batch_size")[0] check_loop(video_id_mismatch, input_distance, actual_batch_size, all_patterns) def main(unused_argv): logging.set_verbosity(tf.logging.INFO) print("tensorflow version: %s" % tf.__version__) check_video_id() if __name__ == "__main__": app.run() ================================================ FILE: youtube-8m-ensemble/cloudml-gpu-distributed.yaml ================================================ trainingInput: runtimeVersion: "1.0" scaleTier: CUSTOM masterType: standard_gpu workerCount: 2 workerType: standard_gpu parameterServerCount: 2 parameterServerType: standard ================================================ FILE: youtube-8m-ensemble/cloudml-gpu.yaml ================================================ trainingInput: scaleTier: CUSTOM # standard_gpu provides 1 GPU. Change to complex_model_m_gpu for 4 GPUs masterType: standard_gpu runtimeVersion: "1.0" ================================================ FILE: youtube-8m-ensemble/data_augmentation.py ================================================ from tensorflow import flags flags.DEFINE_string("data_augmenter", "DefaultAugmenter", "how to preprocess feature, defaults to identical, which means no transform") flags.DEFINE_float("input_noise_level", 0.2, "the standard deviation of normal noise added to input") from all_data_augmentation import * ================================================ FILE: youtube-8m-ensemble/ensemble_command.example ================================================ # bash [script] [model_name] [config] # use a different model name each time, make it understandable # create a new conf file, add it to git repository bash ensemble_scripts/train-mean_model.sh ensemble_mean_model_10 ensemble_scripts/ensemble_no3.conf bash ensemble_scripts/eval-mean_model.sh ensemble_mean_model_10 ensemble_scripts/ensemble_no3.conf bash ensemble_scripts/infer-mean_model.sh ensemble_mean_model_10 ensemble_scripts/ensemble_no3.conf ================================================ FILE: youtube-8m-ensemble/ensemble_level_models.py ================================================ # Copyright 2016 Google Inc. All Rights Reserved. # # Licensed under the Apache License, Version 2.0 (the "License"); # you may not use this file except in compliance with the License. # You may obtain a copy of the License at # # http://www.apache.org/licenses/LICENSE-2.0 # # Unless required by applicable law or agreed to in writing, software # distributed under the License is distributed on an "AS-IS" BASIS, # WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied. # See the License for the specific language governing permissions and # limitations under the License. """Contains model definitions.""" from tensorflow import flags FLAGS = flags.FLAGS flags.DEFINE_integer( "moe_num_mixtures", 2, "The number of mixtures (excluding the dummy 'expert') used for MoeModel.") flags.DEFINE_integer( "deep_chain_layers", 3, "The number of layers used for DeepChainModel") flags.DEFINE_integer( "deep_chain_relu_cells", 200, "The number of relu cells used for DeepChainModel") flags.DEFINE_string( "deep_chain_relu_type", "relu", "The type of relu cells used for DeepChainModel (options are elu and relu)") flags.DEFINE_bool( "deep_chain_use_length", False, "The number of relu cells used for DeepChainModel") flags.DEFINE_integer( "hidden_chain_layers", 4, "The number of layers used for HiddenChainModel") flags.DEFINE_integer( "hidden_chain_relu_cells", 256, "The number of relu cells used for HiddenChainModel") flags.DEFINE_integer( "attention_matrix_rank", 3, "The rank of weight matrix used for AttentionMatrixModel.") flags.DEFINE_integer( "attention_relu_cells", 128, "The rank of weight matrix used for AttentionMatrixModel.") import sys from os.path import dirname if dirname(__file__) not in sys.path: sys.path.append(dirname(__file__)) from all_ensemble_models import * ================================================ FILE: youtube-8m-ensemble/ensemble_scripts/.vimrc ================================================ set tabstop=2 set shiftwidth=2 set expandtab set autoindent ================================================ FILE: youtube-8m-ensemble/ensemble_scripts/after_submission_no1.conf ================================================ video_relabel_combine_chain video_very_deep_combine_chain lstmmemory_cell1024_layer2_moe8 lstmmemory_cell2048_layer2_moe4 lstmparallelfinaloutput1024_moe8 lstmparallelmemory1024_moe8 lstmgate_cell1024_layer1_moe8 lstmglu2_cell1024_layer1_moe8 lstmbiglu_cell1024_layer1_moe8 biunilstm1024_moe4 lstm_random_mean_moe8 lstm_shortlayers_moe8 framehop_lstm cnnlstmmemory1024_moe8 attentionlstm_moe4 lstmattlstm1024_moe8 lstm_attention8_max lstm2_attention8_max lstm_positional_attention8max cnn_deep_combine_chain deep_cnn_deep_combine_chain lstm_cnn_deep_combine_chain multilstmmemory1024_moe4_deep_chain lstmmem1024_layer2_moe4_deep_combine_chain_add_length multires_lstm_deep_combine_chain lstm_gate_multiscale4_moe4 lstm_multiscale4_moe4 ================================================ FILE: youtube-8m-ensemble/ensemble_scripts/after_submission_no2.conf ================================================ video_relabel_combine_chain video_very_deep_combine_chain lstmmemory_cell1024_layer2_moe8 lstmmemory_cell2048_layer2_moe4 lstmparallelfinaloutput1024_moe8 lstmparallelmemory1024_moe8 lstmgate_cell1024_layer1_moe8 lstmglu2_cell1024_layer1_moe8 lstmbiglu_cell1024_layer1_moe8 biunilstm1024_moe4 lstm_random_mean_moe8 lstm_shortlayers_moe8 framehop_lstm cnnlstmmemory1024_moe8 attentionlstm_moe4 lstmattlstm1024_moe8 lstm_attention8_max lstm2_attention8_max lstm_positional_attention8max cnn_deep_combine_chain deep_cnn_deep_combine_chain lstm_cnn_deep_combine_chain multilstmmemory1024_moe4_deep_chain lstmmem1024_layer2_moe4_deep_combine_chain_add_length multires_lstm_deep_combine_chain lstm_gate_multiscale4_moe4 lstm_multiscale4_moe4 video_dcc_bagging video_cc_structure_bagging/ensemble_matrix_model video_dcc_boosting_discardhopeless/ensemble_matrix_model video_dcc_boosting/ensemble_matrix_model video_dcc_boosting_weightclip/ensemble_matrix_model cnn_deep_combine_chain_bagging/ensemble_matrix_model cnn_deep_combine_chain_boosting/ensemble_attention_matrix_model lstmattention8max_bagging/ensemble_matrix_model lstmattention8max_boosting/ensemble_attention_matrix_model lstmparalleloutput_bagging/ensemble_mean_model lstmparalleloutput_boosting_weightclip/ensemble_matrix_model ================================================ FILE: youtube-8m-ensemble/ensemble_scripts/after_submission_no3.conf ================================================ video_relabel_combine_chain video_very_deep_combine_chain lstmmemory_cell1024_layer2_moe8 lstmmemory_cell2048_layer2_moe4 lstmparallelfinaloutput1024_moe8 lstmparallelmemory1024_moe8 lstmgate_cell1024_layer1_moe8 lstmglu2_cell1024_layer1_moe8 lstmbiglu_cell1024_layer1_moe8 biunilstm1024_moe4 lstm_random_mean_moe8 lstm_shortlayers_moe8 framehop_lstm cnnlstmmemory1024_moe8 attentionlstm_moe4 lstmattlstm1024_moe8 lstm_attention8_max lstm2_attention8_max lstm_positional_attention8max cnn_deep_combine_chain deep_cnn_deep_combine_chain lstm_cnn_deep_combine_chain multilstmmemory1024_moe4_deep_chain lstmmem1024_layer2_moe4_deep_combine_chain_add_length multires_lstm_deep_combine_chain lstm_gate_multiscale4_moe4 lstm_multiscale4_moe4 video_dcc_bagging video_cc_structure_bagging/ensemble_matrix_model video_dcc_boosting_discardhopeless/ensemble_matrix_model video_dcc_boosting/ensemble_matrix_model video_dcc_boosting_weightclip/ensemble_matrix_model cnn_deep_combine_chain_bagging/ensemble_matrix_model cnn_deep_combine_chain_boosting/ensemble_attention_matrix_model lstmattention8max_bagging/ensemble_matrix_model lstmattention8max_boosting/ensemble_attention_matrix_model lstmparalleloutput_bagging/ensemble_mean_model lstmparalleloutput_boosting_weightclip/ensemble_matrix_model distillation_video_dcc_noise distillation_cnn_dcc_boosting/sub_model_1 distillation_lstmcnn_dcc_boosting/sub_model_1 distillation_lstmparalleloutput_boosting/sub_model_1 distillation_multilstm_dcc_boosting/sub_model_1 distillation_multiscale_cnnlstm_boosting/sub_model_1 distillation_video_dcc_bagging/ensemble_matrix_model distillation_video_dcc_boosting/ensemble_matrix_model ================================================ FILE: youtube-8m-ensemble/ensemble_scripts/after_submission_no4.conf ================================================ video_relabel_combine_chain video_very_deep_combine_chain lstmmemory_cell1024_layer2_moe8 lstmmemory_cell2048_layer2_moe4 lstmparallelfinaloutput1024_moe8 lstmparallelmemory1024_moe8 lstmgate_cell1024_layer1_moe8 lstmglu2_cell1024_layer1_moe8 lstmbiglu_cell1024_layer1_moe8 biunilstm1024_moe4 lstm_random_mean_moe8 lstm_shortlayers_moe8 framehop_lstm cnnlstmmemory1024_moe8 attentionlstm_moe4 lstmattlstm1024_moe8 lstm_attention8_max lstm2_attention8_max lstm_positional_attention8max cnn_deep_combine_chain deep_cnn_deep_combine_chain lstm_cnn_deep_combine_chain multilstmmemory1024_moe4_deep_chain lstmmem1024_layer2_moe4_deep_combine_chain_add_length multires_lstm_deep_combine_chain lstm_gate_multiscale4_moe4 lstm_multiscale4_moe4 video_dcc_bagging video_cc_structure_bagging/ensemble_matrix_model video_dcc_boosting_discardhopeless/ensemble_matrix_model video_dcc_boosting/ensemble_matrix_model video_dcc_boosting_weightclip/ensemble_matrix_model cnn_deep_combine_chain_bagging/ensemble_matrix_model cnn_deep_combine_chain_boosting/ensemble_attention_matrix_model lstmattention8max_bagging/ensemble_matrix_model lstmattention8max_boosting/ensemble_attention_matrix_model lstmparalleloutput_bagging/ensemble_mean_model lstmparalleloutput_boosting_weightclip/ensemble_matrix_model distillation_video_dcc_noise distillation_cnn_dcc_boosting/sub_model_1 distillation_lstmcnn_dcc_boosting/sub_model_1 distillation_lstmparalleloutput_boosting/sub_model_1 distillation_multilstm_dcc_boosting/sub_model_1 distillation_multiscale_cnnlstm_boosting/sub_model_1 distillation_video_dcc_bagging/ensemble_matrix_model distillation_video_dcc_boosting/ensemble_matrix_model distillchain_video_dcc distillchain_video_norm_moe8 distillchain_lstm_gate_moe8 distillchain_cnn_dcc distillchain_cnndcc_layer2moe4_ensemble distillchain_lstm_attention8max distillchain_lstm_moe8 distillchain_lstm_multiscale_layer2_moe8 distillchain_lstm_multiscale_layer4_moe8 distillchain_lstmcnn_dcc distillsplit_lstm_gate_moe8 distillchain_v2_video_dcc distillchain_v2_lstmattention8max distillchain_v2_lstmcnn_dcc distillchain_v2_lstm_gate_moe8 distillchain_v2_lstm_glu2_moe8 distillchain_v2_multilstm_dcc distillchain_v2_multiscale_cnnlstm distillchain_v2_lstmparalleloutput distillchain_v2_boost_lstmparalleloutput videochain/distillchain_v2_video_dcc_1 videochain/distillchain_v2_video_dcc_2 videochain/distillchain_v2_video_dcc_3 videochain/distillchain_v2_video_dcc_4 videochain/distillchain_v2_video_dcc_5 hybridchain/distillchain_v2_video_dcc hybridchain/distillchain_v2_cnn_dcc hybridchain/distillchain_v2_lstmparalleloutput ================================================ FILE: youtube-8m-ensemble/ensemble_scripts/auto-preensemble-deep_combine_chain_model.sh ================================================ model=$1 conf=$2 part=$3 DEFAULT_GPU_ID=0 if [ -z ${CUDA_VISIBLE_DEVICES+x} ]; then GPU_ID=$DEFAULT_GPU_ID echo "set CUDA_VISIBLE_DEVICES to default('$GPU_ID')" else GPU_ID=$CUDA_VISIBLE_DEVICES echo "set CUDA_VISIBLE_DEVICES to external('$GPU_ID')" fi test_path=/Youtube-8M/model_predictions/${part} test_data_patterns="" for d in $(cat $conf); do test_data_patterns="${test_path}/${d}/*.tfrecord${test_data_patterns:+,$test_data_patterns}" done echo "$test_data_patterns" CUDA_VISIBLE_DEVICES="$GPU_ID" python inference-pre-ensemble.py \ --output_dir="/Youtube-8M/model_predictions/${part}/${model}" \ --train_dir="../model/${model}" \ --input_data_patterns="$test_data_patterns" \ --model="DeepCombineChainModel" \ --deep_chain_relu_cells=128 \ --moe_num_mixtures=2 \ --batch_size=512 \ --file_size=4096 ================================================ FILE: youtube-8m-ensemble/ensemble_scripts/auto-preensemble-matrix_model.sh ================================================ model=$1 conf=$2 part=$3 postfix=$4 DEFAULT_GPU_ID=0 if [ -z ${CUDA_VISIBLE_DEVICES+x} ]; then GPU_ID=$DEFAULT_GPU_ID echo "set CUDA_VISIBLE_DEVICES to default('$GPU_ID')" else GPU_ID=$CUDA_VISIBLE_DEVICES echo "set CUDA_VISIBLE_DEVICES to external('$GPU_ID')" fi test_path=/Youtube-8M/model_predictions${postfix}/${part} test_data_patterns="" for d in $(cat $conf); do test_data_patterns="${test_path}/${d}/*.tfrecord${test_data_patterns:+,$test_data_patterns}" done echo "$test_data_patterns" CUDA_VISIBLE_DEVICES="$GPU_ID" python inference-pre-ensemble.py \ --output_dir="/Youtube-8M/model_predictions/${part}/${model}" \ --train_dir="../model/${model}" \ --input_data_patterns="$test_data_patterns" \ --model="MatrixRegressionModel" \ --batch_size=1024 \ --file_size=4096 ================================================ FILE: youtube-8m-ensemble/ensemble_scripts/check-video_id.sh ================================================ conf=$1 part=$2 validate_path=/Youtube-8M/model_predictions/${part} validate_data_patterns="" for d in $(cat $conf); do validate_data_patterns="${validate_path}/${d}/*.tfrecord${validate_data_patterns:+,$validate_data_patterns}" done echo "$validate_data_patterns" input_data_pattern="/Youtube-8M/data/video/${part}/*.tfrecord" CUDA_VISIBLE_DEVICES="" python check_video_id.py \ --input_data_pattern=$input_data_pattern \ --eval_data_patterns="$validate_data_patterns" ================================================ FILE: youtube-8m-ensemble/ensemble_scripts/check-video_id_match.sh ================================================ conf=$1 part=$2 validate_path=/Youtube-8M/model_predictions_local/${part} validate_data_patterns="" for d in $(cat $conf); do validate_data_patterns="${validate_path}/${d}/*.tfrecord${validate_data_patterns:+,$validate_data_patterns}" done echo "$validate_data_patterns" input_data_pattern="/Youtube-8M/data/video/${part}/*.tfrecord" #input_data_pattern="/Youtube-8M/model_predictions/${part}/model_input/*.tfrecord" CUDA_VISIBLE_DEVICES="" python check_video_id_match.py \ --input_data_pattern=$input_data_pattern \ --eval_data_patterns="$validate_data_patterns" ================================================ FILE: youtube-8m-ensemble/ensemble_scripts/combine-tfrecords-frame-v2.sh ================================================ #!/bin/bash file_num_mod=$1 data_path=/Youtube-8M/model_predictions_x32/train input_data_pattern="/Youtube-8M/data/frame/train/*.tfrecord" prediction_data_pattern="${data_path}/distillation/ensemble_v2_matrix_model/prediction*.tfrecord" CUDA_VISIBLE_DEVICES="" python inference-combine-tfrecords-frame.py \ --output_dir="/Youtube-8M/distillation_v2/frame/train" \ --input_data_pattern="$input_data_pattern" \ --prediction_data_pattern="$prediction_data_pattern" \ --file_num_mod=$file_num_mod \ --batch_size=128 \ --file_size=1024 ================================================ FILE: youtube-8m-ensemble/ensemble_scripts/combine-tfrecords-frame.sh ================================================ #!/bin/bash data_path=/Youtube-8M/model_predictions/train input_data_pattern="/Youtube-8M/data/frame/train/train*.tfrecord" prediction_data_pattern="${data_path}/distillation/ensemble_mean_model/prediction*.tfrecord" CUDA_VISIBLE_DEVICES="" python inference-combine-tfrecords-frame.py \ --output_dir="/Youtube-8M/distillation/frame/train" \ --input_data_pattern="$input_data_pattern" \ --prediction_data_pattern="$prediction_data_pattern" \ --batch_size=128 \ --file_size=1024 ================================================ FILE: youtube-8m-ensemble/ensemble_scripts/combine-tfrecords-video-v2.sh ================================================ #!/bin/bash data_path=/Youtube-8M/model_predictions_local/train input_data_pattern="/Youtube-8M/data/video/train/*.tfrecord" prediction_data_pattern="${data_path}/distillation/ensemble_v2_matrix_model/prediction*.tfrecord" CUDA_VISIBLE_DEVICES="" python inference-combine-tfrecords-video.py \ --output_dir="/Youtube-8M/distillation_v2/video/train" \ --input_data_pattern="$input_data_pattern" \ --prediction_data_pattern="$prediction_data_pattern" \ --batch_size=1024 \ --file_size=4096 ================================================ FILE: youtube-8m-ensemble/ensemble_scripts/combine-tfrecords-video.sh ================================================ #!/bin/bash data_path=/Youtube-8M/model_predictions/train input_data_pattern="/Youtube-8M/data/video/train/train*.tfrecord" prediction_data_pattern="${data_path}/distillation/ensemble_mean_model/prediction*.tfrecord" CUDA_VISIBLE_DEVICES="" python inference-combine-tfrecords-video.py \ --output_dir="/Youtube-8M/distillation/video/train" \ --input_data_pattern="$input_data_pattern" \ --prediction_data_pattern="$prediction_data_pattern" \ --batch_size=1024 \ --file_size=4096 ================================================ FILE: youtube-8m-ensemble/ensemble_scripts/ensemble_no1.conf ================================================ lstmmem1024_layer2_moe4_deep_combine_chain_add_length lstmmemory_cell2048_layer2_moe4 video_very_deep_combine_chain ================================================ FILE: youtube-8m-ensemble/ensemble_scripts/ensemble_no10.conf ================================================ attentionlstm_moe4 biunilstm1024_moe4 cnn_deep_combine_chain cnnlstmmemory1024_moe8 deep_cnn_deep_combine_chain framehop_lstm lstm2_attention8_max lstm_attention8_max lstmattlstm1024_moe8 lstm_cnn_deep_combine_chain lstmmem1024_layer2_moe4_deep_combine_chain_add_length lstmmemory_cell1024_layer2_moe8 lstmmemory_cell2048_layer2_moe4 lstmparallelfinaloutput1024_moe8 lstmparallelmemory1024_moe8 lstm_positional_attention8max lstm_random_mean_moe8 lstm_shortlayers_moe8 multilstmmemory1024_moe4_deep_chain multires_lstm_deep_combine_chain video_relabel_combine_chain video_very_deep_combine_chain distillation_video_dcc_noise video_cc_structure_bagging/ensemble_matrix_model video_dcc_bagging video_dcc_boosting/ensemble_matrix_model video_dcc_boosting_discardhopeless/ensemble_matrix_model video_dcc_boosting_weightclip/ensemble_matrix_model cnn_deep_combine_chain_bagging/ensemble_matrix_model cnn_deep_combine_chain_boosting/ensemble_matrix_model lstmattention8max_bagging/ensemble_matrix_model lstmattention8max_boosting/ensemble4_matrix_model lstmparalleloutput_bagging/ensemble_mean_model lstmparalleloutput_boosting_weightclip/ensemble_matrix_model ================================================ FILE: youtube-8m-ensemble/ensemble_scripts/ensemble_no11.conf ================================================ attentionlstm_moe4 biunilstm1024_moe4 cnn_deep_combine_chain cnn_deep_combine_chain_bagging/ensemble_matrix_model cnn_deep_combine_chain_boosting/ensemble_matrix_model cnnlstmmemory1024_moe8 deep_cnn_deep_combine_chain distillation_video_dcc_bagging/ensemble_matrix_model distillation_video_dcc_boosting/ensemble_matrix_model distillation_video_dcc_noise framehop_lstm lstm2_attention8_max lstm_attention8_max lstmattention8max_bagging/ensemble_matrix_model lstmattention8max_boosting/ensemble4_matrix_model lstmattlstm1024_moe8 lstm_cnn_deep_combine_chain lstmmem1024_layer2_moe4_deep_combine_chain_add_length lstmmemory_cell1024_layer2_moe8 lstmmemory_cell2048_layer2_moe4 lstmparallelfinaloutput1024_moe8 lstmparallelmemory1024_moe8 lstmparalleloutput_bagging/ensemble_mean_model lstmparalleloutput_boosting_weightclip/ensemble_matrix_model lstm_positional_attention8max lstm_random_mean_moe8 lstm_shortlayers_moe8 multilstmmemory1024_moe4_deep_chain multires_lstm_deep_combine_chain video_cc_structure_bagging/ensemble_matrix_model video_dcc_bagging video_dcc_boosting_discardhopeless/ensemble_matrix_model video_dcc_boosting/ensemble_matrix_model video_dcc_boosting_weightclip/ensemble_matrix_model video_relabel_combine_chain video_very_deep_combine_chain ================================================ FILE: youtube-8m-ensemble/ensemble_scripts/ensemble_no12.conf ================================================ attentionlstm_moe4 biunilstm1024_moe4 cnn_deep_combine_chain cnn_deep_combine_chain_bagging/ensemble_matrix_model cnn_deep_combine_chain_boosting/ensemble_matrix_model cnnlstmmemory1024_moe8 deep_cnn_deep_combine_chain distillation_video_dcc_bagging/ensemble_matrix_model distillation_video_dcc_boosting/ensemble_matrix_model distillation_video_dcc_noise framehop_lstm lstm2_attention8_max lstm_attention8_max lstmattention8max_bagging/ensemble_matrix_model lstmattention8max_boosting/ensemble4_matrix_model lstmattlstm1024_moe8 lstm_cnn_deep_combine_chain lstmgate_cell1024_layer1_moe8 lstmglu2_cell1024_layer1_moe8 lstmmem1024_layer2_moe4_deep_combine_chain_add_length lstmmemory_cell1024_layer2_moe8 lstmmemory_cell2048_layer2_moe4 lstm_multiscale4_moe4 lstmparallelfinaloutput1024_moe8 lstmparallelmemory1024_moe8 lstmparalleloutput_bagging/ensemble_mean_model lstmparalleloutput_boosting_weightclip/ensemble_matrix_model lstm_positional_attention8max lstm_random_mean_moe8 lstm_shortlayers_moe8 multilstmmemory1024_moe4_deep_chain multires_lstm_deep_combine_chain video_cc_structure_bagging/ensemble_matrix_model video_dcc_bagging video_dcc_boosting_discardhopeless/ensemble_matrix_model video_dcc_boosting/ensemble_matrix_model video_dcc_boosting_weightclip/ensemble_matrix_model video_relabel_combine_chain video_very_deep_combine_chain ================================================ FILE: youtube-8m-ensemble/ensemble_scripts/ensemble_no13.conf ================================================ attentionlstm_moe4 biunilstm1024_moe4 cnn_deep_combine_chain cnn_deep_combine_chain_bagging/ensemble_matrix_model cnn_deep_combine_chain_boosting/ensemble_matrix_model cnn_deep_combine_chain_boosting/ensemble_attention_matrix_model cnnlstmmemory1024_moe8 deep_cnn_deep_combine_chain distillation_video_dcc_bagging/ensemble_matrix_model distillation_video_dcc_boosting/ensemble_matrix_model distillation_video_dcc_noise framehop_lstm lstm2_attention8_max lstm_attention8_max lstmattention8max_bagging/ensemble_matrix_model lstmattention8max_boosting/ensemble4_matrix_model lstmattention8max_boosting/sub_model_1 lstmattention8max_boosting/sub_model_2 lstmattention8max_boosting/sub_model_3 lstmattention8max_boosting/sub_model_4 lstmattention8max_boosting/sub_model_5 lstmattention8max_boosting/sub_model_6 lstmattention8max_boosting/sub_model_7 lstmattlstm1024_moe8 lstm_cnn_deep_combine_chain lstmgate_cell1024_layer1_moe8 lstmglu2_cell1024_layer1_moe8 lstmmem1024_layer2_moe4_deep_combine_chain_add_length lstmmemory_cell1024_layer2_moe8 lstmmemory_cell2048_layer2_moe4 lstm_multiscale4_moe4 lstmparallelfinaloutput1024_moe8 lstmparallelmemory1024_moe8 lstmparalleloutput_bagging/ensemble_mean_model lstmparalleloutput_boosting_weightclip/ensemble_matrix_model lstm_positional_attention8max lstm_random_mean_moe8 lstm_shortlayers_moe8 multilstmmemory1024_moe4_deep_chain multires_lstm_deep_combine_chain video_cc_structure_bagging/ensemble_matrix_model video_cc_structure_bagging/video_knowledge_combine_chain video_cc_structure_bagging/video_notzero_combine_chain video_cc_structure_bagging/video_softmax_combine_chain video_cc_structure_bagging/video_weight_combine_chain video_dcc_bagging video_dcc_boosting_discardhopeless/ensemble_matrix_model video_dcc_boosting_discardhopeless/sub_model_1 video_dcc_boosting_discardhopeless/sub_model_2 video_dcc_boosting_discardhopeless/sub_model_3 video_dcc_boosting_discardhopeless/sub_model_4 video_dcc_boosting_discardhopeless/sub_model_5 video_dcc_boosting_discardhopeless/sub_model_6 video_dcc_boosting_discardhopeless/sub_model_7 video_dcc_boosting_discardhopeless/sub_model_8 video_dcc_boosting/ensemble_matrix_model video_dcc_boosting_weightclip/ensemble_matrix_model video_relabel_combine_chain video_very_deep_combine_chain ================================================ FILE: youtube-8m-ensemble/ensemble_scripts/ensemble_no14.conf ================================================ attentionlstm_moe4 biunilstm1024_moe4 cnn_deep_combine_chain cnn_deep_combine_chain_bagging/ensemble_matrix_model cnn_deep_combine_chain_boosting/ensemble_attention_matrix_model cnnlstmmemory1024_moe8 deep_cnn_deep_combine_chain distillation_video_dcc_bagging/ensemble_matrix_model distillation_video_dcc_boosting/ensemble_matrix_model distillation_video_dcc_noise framehop_lstm lstm2_attention8_max lstm_attention8_max lstmattention8max_bagging/ensemble_matrix_model lstmattention8max_boosting/ensemble_attention_matrix_model lstmattlstm1024_moe8 lstm_cnn_deep_combine_chain lstmgate_cell1024_layer1_moe8 lstmglu2_cell1024_layer1_moe8 lstmmem1024_layer2_moe4_deep_combine_chain_add_length lstmmemory_cell1024_layer2_moe8 lstmmemory_cell2048_layer2_moe4 lstm_multiscale4_moe4 lstmparallelfinaloutput1024_moe8 lstmparallelmemory1024_moe8 lstmparalleloutput_bagging/ensemble_mean_model lstmparalleloutput_boosting_weightclip/ensemble_matrix_model lstm_positional_attention8max lstm_random_mean_moe8 lstm_shortlayers_moe8 multilstmmemory1024_moe4_deep_chain multires_lstm_deep_combine_chain video_cc_structure_bagging/ensemble_matrix_model video_dcc_bagging video_dcc_boosting_discardhopeless/ensemble_matrix_model video_dcc_boosting/ensemble_matrix_model video_dcc_boosting_weightclip/ensemble_matrix_model video_relabel_combine_chain video_very_deep_combine_chain distillation_cnn_dcc_boosting/sub_model_1 distillation_lstmcnn_dcc_boosting/sub_model_1 distillation_lstmparalleloutput_boosting/sub_model_1 distillation_multilstm_dcc_boosting/sub_model_1 distillation_multiscale_cnnlstm_boosting/sub_model_1 ================================================ FILE: youtube-8m-ensemble/ensemble_scripts/ensemble_no15.conf ================================================ attentionlstm_moe4 biunilstm1024_moe4 cnn_deep_combine_chain cnn_deep_combine_chain_bagging/ensemble_matrix_model cnn_deep_combine_chain_boosting/ensemble_attention_matrix_model cnnlstmmemory1024_moe8 deep_cnn_deep_combine_chain distillation_video_dcc_bagging/ensemble_matrix_model distillation_video_dcc_boosting/ensemble_matrix_model distillation_video_dcc_noise framehop_lstm lstm2_attention8_max lstm_attention8_max lstmattention8max_bagging/ensemble_matrix_model lstmattention8max_boosting/ensemble_attention_matrix_model lstmattlstm1024_moe8 lstm_cnn_deep_combine_chain lstmgate_cell1024_layer1_moe8 lstmglu2_cell1024_layer1_moe8 lstmmem1024_layer2_moe4_deep_combine_chain_add_length lstmmemory_cell1024_layer2_moe8 lstmmemory_cell2048_layer2_moe4 lstm_multiscale4_moe4 lstmparallelfinaloutput1024_moe8 lstmparallelmemory1024_moe8 lstmparalleloutput_bagging/ensemble_mean_model lstmparalleloutput_boosting_weightclip/ensemble_matrix_model lstm_positional_attention8max lstm_random_mean_moe8 lstm_shortlayers_moe8 multilstmmemory1024_moe4_deep_chain multires_lstm_deep_combine_chain video_cc_structure_bagging/ensemble_matrix_model video_dcc_bagging video_dcc_boosting_discardhopeless/ensemble_matrix_model video_dcc_boosting/ensemble_matrix_model video_dcc_boosting_weightclip/ensemble_matrix_model video_relabel_combine_chain video_very_deep_combine_chain distillation_cnn_dcc_boosting/sub_model_1 distillation_lstmcnn_dcc_boosting/sub_model_1 distillation_lstmparalleloutput_boosting/sub_model_1 distillation_multilstm_dcc_boosting/sub_model_1 distillation_multiscale_cnnlstm_boosting/sub_model_1 video_logistic_model video_moe16_model dbof_model frame_seg_model ================================================ FILE: youtube-8m-ensemble/ensemble_scripts/ensemble_no16.conf ================================================ attentionlstm_moe4 biunilstm1024_moe4 cnn_deep_combine_chain cnn_deep_combine_chain_bagging/ensemble_matrix_model cnn_deep_combine_chain_boosting/ensemble_attention_matrix_model cnnlstmmemory1024_moe8 deep_cnn_deep_combine_chain distillation_video_dcc_bagging/ensemble_matrix_model distillation_video_dcc_boosting/ensemble_matrix_model distillation_video_dcc_noise framehop_lstm lstm2_attention8_max lstm_attention8_max lstmattention8max_bagging/ensemble_matrix_model lstmattention8max_boosting/ensemble_attention_matrix_model lstmattlstm1024_moe8 lstm_cnn_deep_combine_chain lstmgate_cell1024_layer1_moe8 lstmglu2_cell1024_layer1_moe8 lstmmem1024_layer2_moe4_deep_combine_chain_add_length lstmmemory_cell1024_layer2_moe8 lstmmemory_cell2048_layer2_moe4 lstm_multiscale4_moe4 lstmparallelfinaloutput1024_moe8 lstmparallelmemory1024_moe8 lstmparalleloutput_bagging/ensemble_mean_model lstmparalleloutput_boosting_weightclip/ensemble_matrix_model lstm_positional_attention8max lstm_random_mean_moe8 lstm_shortlayers_moe8 multilstmmemory1024_moe4_deep_chain multires_lstm_deep_combine_chain video_cc_structure_bagging/ensemble_matrix_model video_dcc_bagging video_dcc_boosting_discardhopeless/ensemble_matrix_model video_dcc_boosting/ensemble_matrix_model video_dcc_boosting_weightclip/ensemble_matrix_model video_relabel_combine_chain video_very_deep_combine_chain distillation_cnn_dcc_boosting/sub_model_1 distillation_lstmcnn_dcc_boosting/sub_model_1 distillation_lstmparalleloutput_boosting/sub_model_1 distillation_multilstm_dcc_boosting/sub_model_1 distillation_multiscale_cnnlstm_boosting/sub_model_1 distillchain_lstm_gate_moe8 distillchain_video_dcc ================================================ FILE: youtube-8m-ensemble/ensemble_scripts/ensemble_no17.conf ================================================ cnn_deep_combine_chain cnn_deep_combine_chain_bagging/ensemble_matrix_model distillation_cnn_dcc_boosting/sub_model_1 distillation_lstmcnn_dcc_boosting/sub_model_1 distillation_multilstm_dcc_boosting/sub_model_1 framehop_lstm lstm_attention8_max lstm_cnn_deep_combine_chain lstm_positional_attention8max lstmattention8max_bagging/ensemble_matrix_model lstmattention8max_boosting/ensemble_attention_matrix_model lstmbiglu_cell1024_layer1_moe8 lstmglu2_cell1024_layer1_moe8 lstmmem1024_layer2_moe4_deep_combine_chain_add_length lstmparallelfinaloutput1024_moe8 lstmparallelmemory1024_moe8 multilstmmemory1024_moe4_deep_chain video_dcc_boosting/ensemble_matrix_model video_dcc_boosting_discardhopeless/ensemble_matrix_model video_dcc_boosting_weightclip/ensemble_matrix_model distillchain_cnn_dcc distillchain_cnndcc_layer2moe4 distillchain_lstm_attention8max distillchain_lstm_gate_moe8 distillchain_lstm_moe8 distillchain_lstm_multiscale_layer2_moe8 distillchain_lstm_multiscale_layer4_moe8 distillchain_video_dcc ================================================ FILE: youtube-8m-ensemble/ensemble_scripts/ensemble_no18.conf ================================================ attentionlstm_moe4 biunilstm1024_moe4 cnn_deep_combine_chain cnn_deep_combine_chain_bagging/ensemble_matrix_model cnn_deep_combine_chain_boosting/ensemble_attention_matrix_model cnnlstmmemory1024_moe8 deep_cnn_deep_combine_chain distillation_video_dcc_bagging/ensemble_matrix_model distillation_video_dcc_boosting/ensemble_matrix_model distillation_video_dcc_noise framehop_lstm lstm2_attention8_max lstm_attention8_max lstmattention8max_bagging/ensemble_matrix_model lstmattention8max_boosting/ensemble_attention_matrix_model lstmattlstm1024_moe8 lstm_cnn_deep_combine_chain lstmgate_cell1024_layer1_moe8 lstmglu2_cell1024_layer1_moe8 lstmmem1024_layer2_moe4_deep_combine_chain_add_length lstmmemory_cell1024_layer2_moe8 lstmmemory_cell2048_layer2_moe4 lstm_multiscale4_moe4 lstmparallelfinaloutput1024_moe8 lstmparallelmemory1024_moe8 lstmparalleloutput_bagging/ensemble_mean_model lstmparalleloutput_boosting_weightclip/ensemble_matrix_model lstm_positional_attention8max lstm_random_mean_moe8 lstm_shortlayers_moe8 multilstmmemory1024_moe4_deep_chain multires_lstm_deep_combine_chain video_cc_structure_bagging/ensemble_matrix_model video_dcc_bagging video_dcc_boosting_discardhopeless/ensemble_matrix_model video_dcc_boosting/ensemble_matrix_model video_dcc_boosting_weightclip/ensemble_matrix_model video_relabel_combine_chain video_very_deep_combine_chain distillation_cnn_dcc_boosting/sub_model_1 distillation_lstmcnn_dcc_boosting/sub_model_1 distillation_lstmparalleloutput_boosting/sub_model_1 distillation_multilstm_dcc_boosting/sub_model_1 distillation_multiscale_cnnlstm_boosting/sub_model_1 distillchain_lstm_gate_moe8 distillchain_video_dcc distillchain_cnn_dcc distillchain_cnndcc_layer2moe4_ensemble distillchain_lstm_attention8max distillchain_lstm_moe8 distillchain_lstm_multiscale_layer2_moe8 distillchain_lstm_multiscale_layer4_moe8 ================================================ FILE: youtube-8m-ensemble/ensemble_scripts/ensemble_no19.conf ================================================ attentionlstm_moe4 biunilstm1024_moe4 cnn_deep_combine_chain cnn_deep_combine_chain_bagging/ensemble_matrix_model cnn_deep_combine_chain_boosting/ensemble_attention_matrix_model cnnlstmmemory1024_moe8 deep_cnn_deep_combine_chain distillation_video_dcc_bagging/ensemble_matrix_model distillation_video_dcc_boosting/ensemble_matrix_model distillation_video_dcc_noise framehop_lstm lstm2_attention8_max lstm_attention8_max lstmattention8max_bagging/ensemble_matrix_model lstmattention8max_boosting/ensemble_attention_matrix_model lstmattlstm1024_moe8 lstm_cnn_deep_combine_chain lstmgate_cell1024_layer1_moe8 lstmglu2_cell1024_layer1_moe8 lstmmem1024_layer2_moe4_deep_combine_chain_add_length lstmmemory_cell1024_layer2_moe8 lstmmemory_cell2048_layer2_moe4 lstm_multiscale4_moe4 lstmparallelfinaloutput1024_moe8 lstmparallelmemory1024_moe8 lstmparalleloutput_bagging/ensemble_mean_model lstmparalleloutput_boosting_weightclip/ensemble_matrix_model lstm_positional_attention8max lstm_random_mean_moe8 lstm_shortlayers_moe8 multilstmmemory1024_moe4_deep_chain multires_lstm_deep_combine_chain video_cc_structure_bagging/ensemble_matrix_model video_dcc_bagging video_dcc_boosting_discardhopeless/ensemble_matrix_model video_dcc_boosting/ensemble_matrix_model video_dcc_boosting_weightclip/ensemble_matrix_model video_relabel_combine_chain video_very_deep_combine_chain distillation_cnn_dcc_boosting/sub_model_1 distillation_lstmcnn_dcc_boosting/sub_model_1 distillation_lstmparalleloutput_boosting/sub_model_1 distillation_multilstm_dcc_boosting/sub_model_1 distillation_multiscale_cnnlstm_boosting/sub_model_1 distillchain_lstm_gate_moe8 distillchain_video_dcc distillchain_cnn_dcc distillchain_cnndcc_layer2moe4_ensemble distillchain_lstm_attention8max distillchain_lstm_moe8 distillchain_lstm_multiscale_layer2_moe8 distillchain_lstm_multiscale_layer4_moe8 lstm_gate_multiscale4_moe4 lstmbiglu_cell1024_layer1_moe8 distillchain_lstmcnn_dcc distillchain_v2_video_dcc distillsplit_lstm_gate_moe8 ================================================ FILE: youtube-8m-ensemble/ensemble_scripts/ensemble_no2.conf ================================================ cnn_deep_combine_chain deep_cnn_deep_combine_chain lstm_attention8_max lstmmem1024_layer2_moe4_deep_combine_chain_add_length lstmmemory_cell1024_layer2_moe8 lstmmemory_cell2048_layer2_moe4 lstmparallelmemory1024_moe8 multilstmmemory1024_moe4_deep_chain video_very_deep_combine_chain ================================================ FILE: youtube-8m-ensemble/ensemble_scripts/ensemble_no20.conf ================================================ cnn_deep_combine_chain cnn_deep_combine_chain_bagging/ensemble_matrix_model distillation_cnn_dcc_boosting/sub_model_1 distillation_lstmcnn_dcc_boosting/sub_model_1 distillation_multilstm_dcc_boosting/sub_model_1 framehop_lstm lstm_attention8_max lstm_cnn_deep_combine_chain lstm_positional_attention8max lstmattention8max_bagging/ensemble_matrix_model lstmattention8max_boosting/ensemble_attention_matrix_model lstmbiglu_cell1024_layer1_moe8 lstmglu2_cell1024_layer1_moe8 lstmmem1024_layer2_moe4_deep_combine_chain_add_length lstmparallelfinaloutput1024_moe8 lstmparallelmemory1024_moe8 multilstmmemory1024_moe4_deep_chain video_dcc_boosting/ensemble_matrix_model video_dcc_boosting_discardhopeless/ensemble_matrix_model video_dcc_boosting_weightclip/ensemble_matrix_model distillchain_cnn_dcc distillchain_cnndcc_layer2moe4_ensemble distillchain_lstm_attention8max distillchain_lstm_gate_moe8 distillchain_lstm_moe8 distillchain_lstm_multiscale_layer2_moe8 distillchain_lstm_multiscale_layer4_moe8 distillchain_video_dcc lstm_gate_multiscale4_moe4 distillchain_lstmcnn_dcc distillchain_v2_video_dcc distillsplit_lstm_gate_moe8 distillchain_v2_lstmattention8max distillchain_v2_lstmcnn_dcc distillchain_v2_lstm_gate_moe8 distillchain_v2_lstm_glu2_moe8 distillchain_v2_multilstm_dcc distillchain_v2_multiscale_cnnlstm distillchain_v2_lstmparalleloutput distillchain_v2_boost_lstmparalleloutput videochain/distillchain_v2_video_dcc_1 videochain/distillchain_v2_video_dcc_2 videochain/distillchain_v2_video_dcc_3 videochain/distillchain_v2_video_dcc_4 videochain/distillchain_v2_video_dcc_5 hybridchain/distillchain_v2_cnn_dcc hybridchain/distillchain_v2_lstmparalleloutput hybridchain/distillchain_v2_video_dcc ================================================ FILE: youtube-8m-ensemble/ensemble_scripts/ensemble_no21.conf ================================================ attentionlstm_moe4 biunilstm1024_moe4 cnn_deep_combine_chain cnn_deep_combine_chain_bagging/ensemble_matrix_model cnn_deep_combine_chain_boosting/ensemble_attention_matrix_model cnnlstmmemory1024_moe8 deep_cnn_deep_combine_chain distillation_video_dcc_bagging/ensemble_matrix_model distillation_video_dcc_boosting/ensemble_matrix_model distillation_video_dcc_noise framehop_lstm lstm2_attention8_max lstm_attention8_max lstmattention8max_bagging/ensemble_matrix_model lstmattention8max_boosting/ensemble_attention_matrix_model lstmattlstm1024_moe8 lstm_cnn_deep_combine_chain lstmgate_cell1024_layer1_moe8 lstmglu2_cell1024_layer1_moe8 lstmmem1024_layer2_moe4_deep_combine_chain_add_length lstmmemory_cell1024_layer2_moe8 lstmmemory_cell2048_layer2_moe4 lstm_multiscale4_moe4 lstmparallelfinaloutput1024_moe8 lstmparallelmemory1024_moe8 lstmparalleloutput_bagging/ensemble_mean_model lstmparalleloutput_boosting_weightclip/ensemble_matrix_model lstm_positional_attention8max lstm_random_mean_moe8 lstm_shortlayers_moe8 multilstmmemory1024_moe4_deep_chain multires_lstm_deep_combine_chain video_cc_structure_bagging/ensemble_matrix_model video_dcc_bagging video_dcc_boosting_discardhopeless/ensemble_matrix_model video_dcc_boosting/ensemble_matrix_model video_dcc_boosting_weightclip/ensemble_matrix_model video_relabel_combine_chain video_very_deep_combine_chain distillation_cnn_dcc_boosting/sub_model_1 distillation_lstmcnn_dcc_boosting/sub_model_1 distillation_lstmparalleloutput_boosting/sub_model_1 distillation_multilstm_dcc_boosting/sub_model_1 distillation_multiscale_cnnlstm_boosting/sub_model_1 distillchain_lstm_gate_moe8 distillchain_video_dcc distillchain_cnn_dcc distillchain_cnndcc_layer2moe4_ensemble distillchain_lstm_attention8max distillchain_lstm_moe8 distillchain_lstm_multiscale_layer2_moe8 distillchain_lstm_multiscale_layer4_moe8 lstm_gate_multiscale4_moe4 lstmbiglu_cell1024_layer1_moe8 distillchain_lstmcnn_dcc distillchain_v2_video_dcc distillsplit_lstm_gate_moe8 distillchain_v2_lstmattention8max distillchain_v2_lstmcnn_dcc distillchain_v2_lstm_gate_moe8 distillchain_v2_lstm_glu2_moe8 distillchain_v2_multilstm_dcc distillchain_v2_multiscale_cnnlstm videochain/distillchain_v2_video_dcc_1 videochain/distillchain_v2_video_dcc_2 videochain/distillchain_v2_video_dcc_3 videochain/distillchain_v2_video_dcc_4 videochain/distillchain_v2_video_dcc_5 hybridchain/distillchain_v2_cnn_dcc hybridchain/distillchain_v2_lstmparalleloutput hybridchain/distillchain_v2_video_dcc distillchain_v2_lstmparalleloutput distillchain_v2_boost_lstmparalleloutput distillchain_video_norm_moe8 ================================================ FILE: youtube-8m-ensemble/ensemble_scripts/ensemble_no3.conf ================================================ attentionlstm_moe4 biunilstm1024_moe4 cnn_deep_combine_chain cnnlstmmemory1024_moe8 deep_cnn_deep_combine_chain framehop_lstm lstm_attention8_max lstmattlstm1024_moe8 lstmmem1024_layer2_moe4_deep_combine_chain_add_length lstmmemory_cell1024_layer2_moe8 lstmmemory_cell2048_layer2_moe4 lstmparallelfinaloutput1024_moe8 lstmparallelmemory1024_moe8 multilstmmemory1024_moe4_deep_chain video_very_deep_combine_chain ================================================ FILE: youtube-8m-ensemble/ensemble_scripts/ensemble_no4.conf ================================================ attentionlstm_moe4 biunilstm1024_moe4 cnn_deep_combine_chain cnnlstmmemory1024_moe8 deep_cnn_deep_combine_chain framehop_lstm lstm_attention8_max lstmattlstm1024_moe8 lstmmem1024_layer2_moe4_deep_combine_chain_add_length lstmmemory_cell1024_layer2_moe8 lstmmemory_cell2048_layer2_moe4 lstmparallelfinaloutput1024_moe8 lstmparallelmemory1024_moe8 multilstmmemory1024_moe4_deep_chain video_very_deep_combine_chain lstm_shortlayers_moe8 video_dcc_bagging ================================================ FILE: youtube-8m-ensemble/ensemble_scripts/ensemble_no5.conf ================================================ cnn_deep_combine_chain cnnlstmmemory1024_moe8 deep_cnn_deep_combine_chain framehop_lstm lstm_attention8_max lstm_cnn_deep_combine_chain lstmmem1024_layer2_moe4_deep_combine_chain_add_length lstmparallelfinaloutput1024_moe8 lstmparallelmemory1024_moe8 multilstmmemory1024_moe4_deep_chain multires_lstm_deep_combine_chain video_dcc_bagging video_very_deep_combine_chain ================================================ FILE: youtube-8m-ensemble/ensemble_scripts/ensemble_no6.conf ================================================ attentionlstm_moe4 biunilstm1024_moe4 cnn_deep_combine_chain cnnlstmmemory1024_moe8 deep_cnn_deep_combine_chain framehop_lstm lstm_attention8_max lstmattlstm1024_moe8 lstm_cnn_deep_combine_chain lstmmem1024_layer2_moe4_deep_combine_chain_add_length lstmmemory_cell1024_layer2_moe8 lstmmemory_cell2048_layer2_moe4 lstmparallelfinaloutput1024_moe8 lstmparallelmemory1024_moe8 lstm_shortlayers_moe8 multilstmmemory1024_moe4_deep_chain multires_lstm_deep_combine_chain video_dcc_bagging video_very_deep_combine_chain ================================================ FILE: youtube-8m-ensemble/ensemble_scripts/ensemble_no7.conf ================================================ attentionlstm_moe4 biunilstm1024_moe4 cnn_deep_combine_chain cnnlstmmemory1024_moe8 deep_cnn_deep_combine_chain framehop_lstm lstm2_attention8_max lstm_attention8_max lstm_random_mean_moe8 lstmattlstm1024_moe8 lstm_cnn_deep_combine_chain lstmmem1024_layer2_moe4_deep_combine_chain_add_length lstmmemory_cell1024_layer2_moe8 lstmmemory_cell2048_layer2_moe4 lstmparallelfinaloutput1024_moe8 lstmparallelmemory1024_moe8 lstm_shortlayers_moe8 multilstmmemory1024_moe4_deep_chain multires_lstm_deep_combine_chain video_very_deep_combine_chain video_dcc_bagging video_cc_structure_bagging/ensemble_matrix_model video_dcc_boosting/ensemble_matrix_model video_dcc_boosting_discardhopeless/ensemble_matrix_model video_dcc_boosting_weightclip/ensemble_matrix_model cnn_deep_combine_chain_bagging/ensemble_matrix_model lstmparalleloutput_bagging/ensemble_mean_model ================================================ FILE: youtube-8m-ensemble/ensemble_scripts/ensemble_no8.conf ================================================ cnn_deep_combine_chain cnn_deep_combine_chain_bagging/ensemble_matrix_model cnnlstmmemory1024_moe8 framehop_lstm lstm_attention8_max lstm_cnn_deep_combine_chain lstmmem1024_layer2_moe4_deep_combine_chain_add_length lstmparallelfinaloutput1024_moe8 lstmparallelmemory1024_moe8 multilstmmemory1024_moe4_deep_chain multires_lstm_deep_combine_chain video_cc_structure_bagging/ensemble_matrix_model video_dcc_boosting/ensemble_matrix_model video_dcc_boosting_discardhopeless/ensemble_matrix_model video_dcc_boosting_weightclip/ensemble_matrix_model lstmparalleloutput_bagging/ensemble_mean_model ================================================ FILE: youtube-8m-ensemble/ensemble_scripts/ensemble_no9.conf ================================================ attentionlstm_moe4 biunilstm1024_moe4 cnn_deep_combine_chain cnnlstmmemory1024_moe8 deep_cnn_deep_combine_chain framehop_lstm lstm2_attention8_max lstm_attention8_max lstm_random_mean_moe8 lstmattlstm1024_moe8 lstm_cnn_deep_combine_chain lstmmem1024_layer2_moe4_deep_combine_chain_add_length lstmmemory_cell1024_layer2_moe8 lstmmemory_cell2048_layer2_moe4 lstmparallelfinaloutput1024_moe8 lstmparallelmemory1024_moe8 lstm_shortlayers_moe8 multilstmmemory1024_moe4_deep_chain multires_lstm_deep_combine_chain video_very_deep_combine_chain video_relabel_combine_chain video_dcc_bagging video_cc_structure_bagging/ensemble_matrix_model video_dcc_boosting/ensemble_matrix_model video_dcc_boosting_discardhopeless/ensemble_matrix_model video_dcc_boosting_weightclip/ensemble_matrix_model cnn_deep_combine_chain_bagging/ensemble_matrix_model lstmparalleloutput_bagging/ensemble_mean_model lstmattention8max_bagging/ensemble_matrix_model ================================================ FILE: youtube-8m-ensemble/ensemble_scripts/eval-attention_linear_model.sh ================================================ model=$1 conf=$2 start=$3 DEFAULT_GPU_ID=0 if [ -z ${CUDA_VISIBLE_DEVICES+x} ]; then GPU_ID=$DEFAULT_GPU_ID echo "set CUDA_VISIBLE_DEVICES to default('$GPU_ID')" else GPU_ID=$CUDA_VISIBLE_DEVICES echo "set CUDA_VISIBLE_DEVICES to external('$GPU_ID')" fi validate_path=/Youtube-8M/model_predictions/ensemble_validate validate_data_patterns="" for d in $(cat $conf); do validate_data_patterns="${validate_path}/${d}/*.tfrecord${validate_data_patterns:+,$validate_data_patterns}" done echo "$validate_data_patterns" input_data_pattern="${validate_path}/model_input/*.tfrecord" EVERY=300 DIR="$(pwd)" MODEL_DIR="${DIR}/../model/${model}" \ for checkpoint in $(cd $MODEL_DIR && python ${DIR}/training_utils/select.py $EVERY); do echo $checkpoint; if [ $checkpoint -gt $start ]; then echo $checkpoint; CUDA_VISIBLE_DEVICES="$GPU_ID" python eval.py \ --model_checkpoint_path="../model/${model}/model.ckpt-${checkpoint}" \ --train_dir="../model/${model}" \ --model="AttentionLinearModel" \ --moe_num_mixtures=16 \ --eval_data_patterns="$validate_data_patterns" \ --input_data_pattern="$input_data_pattern" fi done ================================================ FILE: youtube-8m-ensemble/ensemble_scripts/eval-attention_linmatrix_model.sh ================================================ model=$1 conf=$2 start=$3 DEFAULT_GPU_ID=0 if [ -z ${CUDA_VISIBLE_DEVICES+x} ]; then GPU_ID=$DEFAULT_GPU_ID echo "set CUDA_VISIBLE_DEVICES to default('$GPU_ID')" else GPU_ID=$CUDA_VISIBLE_DEVICES echo "set CUDA_VISIBLE_DEVICES to external('$GPU_ID')" fi validate_path=/Youtube-8M/model_predictions/ensemble_validate validate_data_patterns="" for d in $(cat $conf); do validate_data_patterns="${validate_path}/${d}/*.tfrecord${validate_data_patterns:+,$validate_data_patterns}" done echo "$validate_data_patterns" input_data_pattern="${validate_path}/model_input/*.tfrecord" EVERY=100 DIR="$(pwd)" MODEL_DIR="${DIR}/../model/${model}" \ for checkpoint in $(cd $MODEL_DIR && python ${DIR}/training_utils/select.py $EVERY); do echo $checkpoint; if [ $checkpoint -gt $start ]; then echo $checkpoint; CUDA_VISIBLE_DEVICES="$GPU_ID" python eval.py \ --model_checkpoint_path="../model/${model}/model.ckpt-${checkpoint}" \ --train_dir="../model/${model}" \ --model="AttentionLinmatrixModel" \ --moe_num_mixtures=16 \ --attention_matrix_rank=4 \ --batch_size=1024 \ --eval_data_patterns="$validate_data_patterns" \ --input_data_pattern="$input_data_pattern" fi done ================================================ FILE: youtube-8m-ensemble/ensemble_scripts/eval-attention_matrix_model.sh ================================================ model=$1 conf=$2 start=$3 DEFAULT_GPU_ID=0 if [ -z ${CUDA_VISIBLE_DEVICES+x} ]; then GPU_ID=$DEFAULT_GPU_ID echo "set CUDA_VISIBLE_DEVICES to default('$GPU_ID')" else GPU_ID=$CUDA_VISIBLE_DEVICES echo "set CUDA_VISIBLE_DEVICES to external('$GPU_ID')" fi validate_path=/Youtube-8M/model_predictions/ensemble_validate validate_data_patterns="" for d in $(cat $conf); do validate_data_patterns="${validate_path}/${d}/*.tfrecord${validate_data_patterns:+,$validate_data_patterns}" done echo "$validate_data_patterns" input_data_pattern="${validate_path}/model_input/*.tfrecord" EVERY=200 DIR="$(pwd)" MODEL_DIR="${DIR}/../model/${model}" \ for checkpoint in $(cd $MODEL_DIR && python ${DIR}/training_utils/select.py $EVERY); do echo $checkpoint; if [ $checkpoint -gt $start ]; then echo $checkpoint; CUDA_VISIBLE_DEVICES="$GPU_ID" python eval.py \ --model_checkpoint_path="../model/${model}/model.ckpt-${checkpoint}" \ --train_dir="../model/${model}" \ --model="AttentionMatrixModel" \ --moe_num_mixtures=4 \ --attention_matrix_rank=8 \ --eval_data_patterns="$validate_data_patterns" \ --input_data_pattern="$input_data_pattern" fi done ================================================ FILE: youtube-8m-ensemble/ensemble_scripts/eval-attention_moe_matrix_model.sh ================================================ model=$1 conf=$2 DEFAULT_GPU_ID=0 if [ -z ${CUDA_VISIBLE_DEVICES+x} ]; then GPU_ID=$DEFAULT_GPU_ID echo "set CUDA_VISIBLE_DEVICES to default('$GPU_ID')" else GPU_ID=$CUDA_VISIBLE_DEVICES echo "set CUDA_VISIBLE_DEVICES to external('$GPU_ID')" fi validate_path=/Youtube-8M/model_predictions/ensemble_validate validate_data_patterns="" for d in $(cat $conf); do validate_data_patterns="${validate_path}/${d}/*.tfrecord${validate_data_patterns:+,$validate_data_patterns}" done echo "$validate_data_patterns" input_data_pattern="${validate_path}/model_input/*.tfrecord" EVERY=200 start=0 DIR="$(pwd)" MODEL_DIR="${DIR}/../model/${model}" \ for checkpoint in $(cd $MODEL_DIR && python ${DIR}/training_utils/select.py $EVERY); do echo $checkpoint; if [ $checkpoint -gt $start ]; then echo $checkpoint; CUDA_VISIBLE_DEVICES="$GPU_ID" python eval.py \ --model_checkpoint_path="../model/${model}/model.ckpt-${checkpoint}" \ --train_dir="../model/${model}" \ --model="AttentionMatrixModel" \ --moe_num_mixtures=4 \ --attention_relu_cells=256 \ --attention_matrix_rank=4 \ --eval_data_patterns="$validate_data_patterns" \ --input_data_pattern="$input_data_pattern" fi done ================================================ FILE: youtube-8m-ensemble/ensemble_scripts/eval-attention_moe_model.sh ================================================ model=$1 conf=$2 DEFAULT_GPU_ID=0 if [ -z ${CUDA_VISIBLE_DEVICES+x} ]; then GPU_ID=$DEFAULT_GPU_ID echo "set CUDA_VISIBLE_DEVICES to default('$GPU_ID')" else GPU_ID=$CUDA_VISIBLE_DEVICES echo "set CUDA_VISIBLE_DEVICES to external('$GPU_ID')" fi validate_path=/Youtube-8M/model_predictions/ensemble_validate validate_data_patterns="" for d in $(cat $conf); do validate_data_patterns="${validate_path}/${d}/*.tfrecord${validate_data_patterns:+,$validate_data_patterns}" done echo "$validate_data_patterns" input_data_pattern="${validate_path}/model_input/*.tfrecord" EVERY=200 start=0 DIR="$(pwd)" MODEL_DIR="${DIR}/../model/${model}" \ for checkpoint in $(cd $MODEL_DIR && python ${DIR}/training_utils/select.py $EVERY); do echo $checkpoint; if [ $checkpoint -gt $start ]; then echo $checkpoint; CUDA_VISIBLE_DEVICES="$GPU_ID" python eval.py \ --model_checkpoint_path="../model/${model}/model.ckpt-${checkpoint}" \ --train_dir="../model/${model}" \ --model=AttentionMoeModel \ --moe_num_mixtures=8 \ --attention_relu_cells=128 \ --eval_data_patterns="$validate_data_patterns" \ --input_data_pattern="$input_data_pattern" fi done ================================================ FILE: youtube-8m-ensemble/ensemble_scripts/eval-attention_rectified_linear_model.sh ================================================ model=$1 conf=$2 start=$3 DEFAULT_GPU_ID=0 if [ -z ${CUDA_VISIBLE_DEVICES+x} ]; then GPU_ID=$DEFAULT_GPU_ID echo "set CUDA_VISIBLE_DEVICES to default('$GPU_ID')" else GPU_ID=$CUDA_VISIBLE_DEVICES echo "set CUDA_VISIBLE_DEVICES to external('$GPU_ID')" fi validate_path=/Youtube-8M/model_predictions/ensemble_validate validate_data_patterns="" for d in $(cat $conf); do validate_data_patterns="${validate_path}/${d}/*.tfrecord${validate_data_patterns:+,$validate_data_patterns}" done echo "$validate_data_patterns" input_data_pattern="${validate_path}/model_input/*.tfrecord" EVERY=300 DIR="$(pwd)" MODEL_DIR="${DIR}/../model/${model}" \ for checkpoint in $(cd $MODEL_DIR && python ${DIR}/training_utils/select.py $EVERY); do echo $checkpoint; if [ $checkpoint -gt $start ]; then echo $checkpoint; CUDA_VISIBLE_DEVICES="$GPU_ID" python eval.py \ --model_checkpoint_path="../model/${model}/model.ckpt-${checkpoint}" \ --train_dir="../model/${model}" \ --model="AttentionRectifiedLinearModel" \ --moe_num_mixtures=16 \ --eval_data_patterns="$validate_data_patterns" \ --input_data_pattern="$input_data_pattern" fi done ================================================ FILE: youtube-8m-ensemble/ensemble_scripts/eval-deep_combine_chain_model.sh ================================================ model=$1 conf=$2 DEFAULT_GPU_ID=0 if [ -z ${CUDA_VISIBLE_DEVICES+x} ]; then GPU_ID=$DEFAULT_GPU_ID echo "set CUDA_VISIBLE_DEVICES to default('$GPU_ID')" else GPU_ID=$CUDA_VISIBLE_DEVICES echo "set CUDA_VISIBLE_DEVICES to external('$GPU_ID')" fi validate_path=/Youtube-8M/model_predictions/ensemble_validate validate_data_patterns="" labels="" for d in $(cat $conf); do validate_data_patterns="${validate_path}/${d}/*.tfrecord${validate_data_patterns:+,$validate_data_patterns}" labels="label,${labels:+,$labels}" done echo "$validate_data_patterns" input_data_pattern="${validate_path}/model_input/*.tfrecord" EVERY=100 start=-1 DIR="$(pwd)" MODEL_DIR="${DIR}/../model/${model}" \ for checkpoint in $(cd $MODEL_DIR && python ${DIR}/training_utils/select.py $EVERY); do echo $checkpoint; if [ $checkpoint -gt $start ]; then echo $checkpoint; CUDA_VISIBLE_DEVICES="$GPU_ID" python eval.py \ --model_checkpoint_path="../model/${model}/model.ckpt-${checkpoint}" \ --train_dir="../model/${model}" \ --model="DeepCombineChainModel" \ --deep_chain_relu_cells=128 \ --moe_num_mixtures=2 \ --eval_data_patterns="$validate_data_patterns" \ --batch_size=1024 \ --input_data_pattern="$input_data_pattern" fi done ================================================ FILE: youtube-8m-ensemble/ensemble_scripts/eval-input_moe_model.sh ================================================ model=$1 conf=$2 DEFAULT_GPU_ID=0 if [ -z ${CUDA_VISIBLE_DEVICES+x} ]; then GPU_ID=$DEFAULT_GPU_ID echo "set CUDA_VISIBLE_DEVICES to default('$GPU_ID')" else GPU_ID=$CUDA_VISIBLE_DEVICES echo "set CUDA_VISIBLE_DEVICES to external('$GPU_ID')" fi validate_path=/Youtube-8M/model_predictions/ensemble_validate validate_data_patterns="" for d in $(cat $conf); do validate_data_patterns="${validate_path}/${d}/*.tfrecord${validate_data_patterns:+,$validate_data_patterns}" done echo "$validate_data_patterns" input_data_pattern="${validate_path}/model_input/*.tfrecord" EVERY=100 start=-1 DIR="$(pwd)" MODEL_DIR="${DIR}/../model/${model}" \ for checkpoint in $(cd $MODEL_DIR && python ${DIR}/training_utils/select.py $EVERY); do echo $checkpoint; if [ $checkpoint -gt $start ]; then echo $checkpoint; CUDA_VISIBLE_DEVICES="$GPU_ID" python eval.py \ --model_checkpoint_path="../model/${model}/model.ckpt-${checkpoint}" \ --train_dir="../model/${model}" \ --input_data_pattern=$input_data_pattern \ --model="InputMoeModel" \ --eval_data_patterns="$validate_data_patterns" fi done ================================================ FILE: youtube-8m-ensemble/ensemble_scripts/eval-linear_model.sh ================================================ model=$1 conf=$2 DEFAULT_GPU_ID=0 if [ -z ${CUDA_VISIBLE_DEVICES+x} ]; then GPU_ID=$DEFAULT_GPU_ID echo "set CUDA_VISIBLE_DEVICES to default('$GPU_ID')" else GPU_ID=$CUDA_VISIBLE_DEVICES echo "set CUDA_VISIBLE_DEVICES to external('$GPU_ID')" fi validate_path=/Youtube-8M/model_predictions/ensemble_validate validate_data_patterns="" for d in $(cat $conf); do validate_data_patterns="${validate_path}/${d}/*.tfrecord${validate_data_patterns:+,$validate_data_patterns}" done echo "$validate_data_patterns" EVERY=100 start=-1 DIR="$(pwd)" MODEL_DIR="${DIR}/../model/${model}" \ for checkpoint in $(cd $MODEL_DIR && python ${DIR}/training_utils/select.py $EVERY); do echo $checkpoint; if [ $checkpoint -gt $start ]; then echo $checkpoint; CUDA_VISIBLE_DEVICES="$GPU_ID" python eval.py \ --model_checkpoint_path="../model/${model}/model.ckpt-${checkpoint}" \ --train_dir="../model/${model}" \ --model="LinearRegressionModel" \ --eval_data_patterns="$validate_data_patterns" fi done ================================================ FILE: youtube-8m-ensemble/ensemble_scripts/eval-matrix_model.sh ================================================ model=$1 conf=$2 postfix=$3 DEFAULT_GPU_ID=0 if [ -z ${CUDA_VISIBLE_DEVICES+x} ]; then GPU_ID=$DEFAULT_GPU_ID echo "set CUDA_VISIBLE_DEVICES to default('$GPU_ID')" else GPU_ID=$CUDA_VISIBLE_DEVICES echo "set CUDA_VISIBLE_DEVICES to external('$GPU_ID')" fi validate_path=/Youtube-8M/model_predictions${postfix}/ensemble_validate validate_data_patterns="" for d in $(cat $conf); do validate_data_patterns="${validate_path}/${d}/*.tfrecord${validate_data_patterns:+,$validate_data_patterns}" done echo "$validate_data_patterns" EVERY=200 start=-1 DIR="$(pwd)" MODEL_DIR="${DIR}/../model/${model}" \ for checkpoint in $(cd $MODEL_DIR && python ${DIR}/training_utils/select.py $EVERY); do echo $checkpoint; if [ $checkpoint -gt $start ]; then echo $checkpoint; CUDA_VISIBLE_DEVICES="$GPU_ID" python eval.py \ --model_checkpoint_path="../model/${model}/model.ckpt-${checkpoint}" \ --train_dir="../model/${model}" \ --model="MatrixRegressionModel" \ --eval_data_patterns="$validate_data_patterns" fi done ================================================ FILE: youtube-8m-ensemble/ensemble_scripts/eval-mean_model.sh ================================================ model=$1 conf=$2 DEFAULT_GPU_ID=0 if [ -z ${CUDA_VISIBLE_DEVICES+x} ]; then GPU_ID=$DEFAULT_GPU_ID echo "set CUDA_VISIBLE_DEVICES to default('$GPU_ID')" else GPU_ID=$CUDA_VISIBLE_DEVICES echo "set CUDA_VISIBLE_DEVICES to external('$GPU_ID')" fi validate_path=/Youtube-8M/model_predictions/ensemble_validate validate_data_patterns="" for d in $(cat $conf); do validate_data_patterns="${validate_path}/${d}/*.tfrecord${validate_data_patterns:+,$validate_data_patterns}" done echo "$validate_data_patterns" CUDA_VISIBLE_DEVICES="$GPU_ID" python eval.py \ --model_checkpoint_path="../model/${model}/model.ckpt-0" \ --train_dir="../model/${model}" \ --model="MeanModel" \ --eval_data_patterns="$validate_data_patterns" ================================================ FILE: youtube-8m-ensemble/ensemble_scripts/eval-moe_model.sh ================================================ model=$1 conf=$2 DEFAULT_GPU_ID=0 if [ -z ${CUDA_VISIBLE_DEVICES+x} ]; then GPU_ID=$DEFAULT_GPU_ID echo "set CUDA_VISIBLE_DEVICES to default('$GPU_ID')" else GPU_ID=$CUDA_VISIBLE_DEVICES echo "set CUDA_VISIBLE_DEVICES to external('$GPU_ID')" fi validate_path=/Youtube-8M/model_predictions/ensemble_validate validate_data_patterns="" for d in $(cat $conf); do validate_data_patterns="${validate_path}/${d}/*.tfrecord${validate_data_patterns:+,$validate_data_patterns}" done echo "$validate_data_patterns" EVERY=100 start=-1 DIR="$(pwd)" MODEL_DIR="${DIR}/../model/${model}" \ for checkpoint in $(cd $MODEL_DIR && python ${DIR}/training_utils/select.py $EVERY); do echo $checkpoint; if [ $checkpoint -gt $start ]; then echo $checkpoint; CUDA_VISIBLE_DEVICES="$GPU_ID" python eval.py \ --model_checkpoint_path="../model/${model}/model.ckpt-${checkpoint}" \ --train_dir="../model/${model}" \ --model="MoeModel" \ --eval_data_patterns="$validate_data_patterns" fi done ================================================ FILE: youtube-8m-ensemble/ensemble_scripts/eval-nonunit_matrix_model.sh ================================================ model=$1 conf=$2 DEFAULT_GPU_ID=0 if [ -z ${CUDA_VISIBLE_DEVICES+x} ]; then GPU_ID=$DEFAULT_GPU_ID echo "set CUDA_VISIBLE_DEVICES to default('$GPU_ID')" else GPU_ID=$CUDA_VISIBLE_DEVICES echo "set CUDA_VISIBLE_DEVICES to external('$GPU_ID')" fi validate_path=/Youtube-8M/model_predictions/ensemble_validate validate_data_patterns="" for d in $(cat $conf); do validate_data_patterns="${validate_path}/${d}/*.tfrecord${validate_data_patterns:+,$validate_data_patterns}" done echo "$validate_data_patterns" EVERY=100 start=-1 DIR="$(pwd)" MODEL_DIR="${DIR}/../model/${model}" \ for checkpoint in $(cd $MODEL_DIR && python ${DIR}/training_utils/select.py $EVERY); do echo $checkpoint; if [ $checkpoint -gt $start ]; then echo $checkpoint; CUDA_VISIBLE_DEVICES="$GPU_ID" python eval.py \ --model_checkpoint_path="../model/${model}/model.ckpt-${checkpoint}" \ --train_dir="../model/${model}" \ --model="NonunitMatrixRegressionModel" \ --eval_data_patterns="$validate_data_patterns" fi done ================================================ FILE: youtube-8m-ensemble/ensemble_scripts/explore-mean_model.log ================================================ single model ------------------------------------------------ cnn_deep_combine_chain GAP = 0.816665221681 ------------------------------------------------ lstm_attention8_max GAP = 0.815108872388 ------------------------------------------------ lstmparallelmemory1024_moe8 GAP = 0.814833331691 ------------------------------------------------ deep_cnn_deep_combine_chain GAP = 0.814806325097 ------------------------------------------------ lstmmem1024_layer2_moe4_deep_combine_chain_add_length GAP = 0.814129286481 ------------------------------------------------ lstmmemory_cell2048_layer2_moe4 GAP = 0.813540507351 ------------------------------------------------ lstmmemory_cell1024_layer2_moe8 GAP = 0.812080163135 ------------------------------------------------ video_very_deep_combine_chain GAP = 0.810174420122 ------------------------------------------------ 2-model ensemble ------------------------------------------------ cnn_deep_combine_chain,deep_cnn_deep_combine_chain GAP = 0.82557031245 ------------------------------------------------ cnn_deep_combine_chain,lstm_attention8_max GAP = 0.829659120727 ------------------------------------------------ cnn_deep_combine_chain,lstmmem1024_layer2_moe4_deep_combine_chain_add_length GAP = 0.828572770795 ------------------------------------------------ cnn_deep_combine_chain,lstmmemory_cell1024_layer2_moe8 GAP = 0.827447547713 ------------------------------------------------ cnn_deep_combine_chain,lstmmemory_cell2048_layer2_moe4 GAP = 0.828015668957 ------------------------------------------------ cnn_deep_combine_chain,lstmparallelmemory1024_moe8 GAP = 0.829346968501 ------------------------------------------------ cnn_deep_combine_chain,video_very_deep_combine_chain GAP = 0.825294576305 ------------------------------------------------ deep_cnn_deep_combine_chain,lstm_attention8_max GAP = 0.827991520743 ------------------------------------------------ deep_cnn_deep_combine_chain,lstmmem1024_layer2_moe4_deep_combine_chain_add_length GAP = 0.827050978978 ------------------------------------------------ deep_cnn_deep_combine_chain,lstmmemory_cell1024_layer2_moe8 GAP = 0.825639821236 ------------------------------------------------ deep_cnn_deep_combine_chain,lstmmemory_cell2048_layer2_moe4 GAP = 0.826148618017 ------------------------------------------------ deep_cnn_deep_combine_chain,lstmparallelmemory1024_moe8 GAP = 0.827756382817 ------------------------------------------------ deep_cnn_deep_combine_chain,video_very_deep_combine_chain GAP = 0.823699502378 ------------------------------------------------ lstm_attention8_max,lstmmem1024_layer2_moe4_deep_combine_chain_add_length GAP = 0.827966037686 ------------------------------------------------ lstm_attention8_max,lstmmemory_cell1024_layer2_moe8 GAP = 0.82594497815 ------------------------------------------------ lstm_attention8_max,lstmmemory_cell2048_layer2_moe4 GAP = 0.826350996945 ------------------------------------------------ lstm_attention8_max,lstmparallelmemory1024_moe8 GAP = 0.828111399229 ------------------------------------------------ lstm_attention8_max,video_very_deep_combine_chain GAP = 0.827831471962 ------------------------------------------------ lstmmem1024_layer2_moe4_deep_combine_chain_add_length,lstmmemory_cell1024_layer2_moe8 GAP = 0.824264654816 ------------------------------------------------ lstmmem1024_layer2_moe4_deep_combine_chain_add_length,lstmmemory_cell2048_layer2_moe4 GAP = 0.824553122942 ------------------------------------------------ lstmmem1024_layer2_moe4_deep_combine_chain_add_length,lstmparallelmemory1024_moe8 GAP = 0.826524738997 ------------------------------------------------ lstmmem1024_layer2_moe4_deep_combine_chain_add_length,video_very_deep_combine_chain GAP = 0.826541745861 ------------------------------------------------ lstmmemory_cell1024_layer2_moe8,lstmmemory_cell2048_layer2_moe4 GAP = 0.822700768554 ------------------------------------------------ lstmmemory_cell1024_layer2_moe8,lstmparallelmemory1024_moe8 GAP = 0.824428614492 ------------------------------------------------ lstmmemory_cell1024_layer2_moe8,video_very_deep_combine_chain GAP = 0.825377747168 ------------------------------------------------ lstmmemory_cell2048_layer2_moe4,lstmparallelmemory1024_moe8 GAP = 0.825320116688 ------------------------------------------------ lstmmemory_cell2048_layer2_moe4,video_very_deep_combine_chain GAP = 0.825870841013 ------------------------------------------------ lstmparallelmemory1024_moe8,video_very_deep_combine_chain GAP = 0.827311197228 ------------------------------------------------ 3-model ensemble ------------------------------------------------ cnn_deep_combine_chain,deep_cnn_deep_combine_chain,lstm_attention8_max GAP = 0.832191332449 ------------------------------------------------ cnn_deep_combine_chain,deep_cnn_deep_combine_chain,lstmmem1024_layer2_moe4_deep_combine_chain_add_length GAP = 0.831426452795 ------------------------------------------------ cnn_deep_combine_chain,deep_cnn_deep_combine_chain,lstmmemory_cell1024_layer2_moe8 GAP = 0.830459884161 ------------------------------------------------ cnn_deep_combine_chain,deep_cnn_deep_combine_chain,lstmmemory_cell2048_layer2_moe4 GAP = 0.830804693389 ------------------------------------------------ cnn_deep_combine_chain,deep_cnn_deep_combine_chain,lstmparallelmemory1024_moe8 GAP = 0.832026266634 ------------------------------------------------ cnn_deep_combine_chain,deep_cnn_deep_combine_chain,video_very_deep_combine_chain GAP = 0.828845214499 ------------------------------------------------ cnn_deep_combine_chain,lstm_attention8_max,lstmmem1024_layer2_moe4_deep_combine_chain_add_length GAP = 0.833621054469 ------------------------------------------------ cnn_deep_combine_chain,lstm_attention8_max,lstmmemory_cell1024_layer2_moe8 GAP = 0.832483129822 ------------------------------------------------ cnn_deep_combine_chain,lstm_attention8_max,lstmmemory_cell2048_layer2_moe4 GAP = 0.832721553048 ------------------------------------------------ cnn_deep_combine_chain,lstm_attention8_max,lstmparallelmemory1024_moe8 GAP = 0.834000141252 ------------------------------------------------ cnn_deep_combine_chain,lstm_attention8_max,video_very_deep_combine_chain GAP = 0.832594974463 ------------------------------------------------ cnn_deep_combine_chain,lstmmem1024_layer2_moe4_deep_combine_chain_add_length,lstmmemory_cell1024_layer2_moe8 GAP = 0.83140605405 ------------------------------------------------ cnn_deep_combine_chain,lstmmem1024_layer2_moe4_deep_combine_chain_add_length,lstmmemory_cell2048_layer2_moe4 GAP = 0.831612457968 ------------------------------------------------ cnn_deep_combine_chain,lstmmem1024_layer2_moe4_deep_combine_chain_add_length,lstmparallelmemory1024_moe8 GAP = 0.832955557705 ------------------------------------------------ cnn_deep_combine_chain,lstmmem1024_layer2_moe4_deep_combine_chain_add_length,video_very_deep_combine_chain GAP = 0.831666234693 ------------------------------------------------ cnn_deep_combine_chain,lstmmemory_cell1024_layer2_moe8,lstmmemory_cell2048_layer2_moe4 GAP = 0.830441764733 ------------------------------------------------ cnn_deep_combine_chain,lstmmemory_cell1024_layer2_moe8,lstmparallelmemory1024_moe8 GAP = 0.831766254279 ------------------------------------------------ cnn_deep_combine_chain,lstmmemory_cell1024_layer2_moe8,video_very_deep_combine_chain GAP = 0.830901832056 ------------------------------------------------ cnn_deep_combine_chain,lstmmemory_cell2048_layer2_moe4,lstmparallelmemory1024_moe8 GAP = 0.832169199739 ------------------------------------------------ cnn_deep_combine_chain,lstmmemory_cell2048_layer2_moe4,video_very_deep_combine_chain GAP = 0.831109764339 ------------------------------------------------ cnn_deep_combine_chain,lstmparallelmemory1024_moe8,video_very_deep_combine_chain GAP = 0.83223587122 ------------------------------------------------ deep_cnn_deep_combine_chain,lstm_attention8_max,lstmmem1024_layer2_moe4_deep_combine_chain_add_length GAP = 0.832349778772 ------------------------------------------------ deep_cnn_deep_combine_chain,lstm_attention8_max,lstmmemory_cell1024_layer2_moe8 GAP = 0.831052849226 ------------------------------------------------ deep_cnn_deep_combine_chain,lstm_attention8_max,lstmmemory_cell2048_layer2_moe4 GAP = 0.831305454699 ------------------------------------------------ deep_cnn_deep_combine_chain,lstm_attention8_max,lstmparallelmemory1024_moe8 GAP = 0.832756750266 ------------------------------------------------ deep_cnn_deep_combine_chain,lstm_attention8_max,video_very_deep_combine_chain GAP = 0.831369689002 ------------------------------------------------ deep_cnn_deep_combine_chain,lstmmem1024_layer2_moe4_deep_combine_chain_add_length,lstmmemory_cell1024_layer2_moe8 GAP = 0.829913304808 ------------------------------------------------ deep_cnn_deep_combine_chain,lstmmem1024_layer2_moe4_deep_combine_chain_add_length,lstmmemory_cell2048_layer2_moe4 GAP = 0.830247896269 ------------------------------------------------ deep_cnn_deep_combine_chain,lstmmem1024_layer2_moe4_deep_combine_chain_add_length,lstmparallelmemory1024_moe8 GAP = 0.831736842166 ------------------------------------------------ deep_cnn_deep_combine_chain,lstmmem1024_layer2_moe4_deep_combine_chain_add_length,video_very_deep_combine_chain GAP = 0.830453715972 ------------------------------------------------ deep_cnn_deep_combine_chain,lstmmemory_cell1024_layer2_moe8,lstmmemory_cell2048_layer2_moe4 GAP = 0.828878078355 ------------------------------------------------ deep_cnn_deep_combine_chain,lstmmemory_cell1024_layer2_moe8,lstmparallelmemory1024_moe8 GAP = 0.830298620071 ------------------------------------------------ deep_cnn_deep_combine_chain,lstmmemory_cell1024_layer2_moe8,video_very_deep_combine_chain GAP = 0.829553175272 ------------------------------------------------ deep_cnn_deep_combine_chain,lstmmemory_cell2048_layer2_moe4,lstmparallelmemory1024_moe8 GAP = 0.830743752891 ------------------------------------------------ deep_cnn_deep_combine_chain,lstmmemory_cell2048_layer2_moe4,video_very_deep_combine_chain GAP = 0.829768283623 ------------------------------------------------ deep_cnn_deep_combine_chain,lstmparallelmemory1024_moe8,video_very_deep_combine_chain GAP = 0.831023711068 ------------------------------------------------ lstm_attention8_max,lstmmem1024_layer2_moe4_deep_combine_chain_add_length,lstmmemory_cell1024_layer2_moe8 GAP = 0.830592277718 ------------------------------------------------ lstm_attention8_max,lstmmem1024_layer2_moe4_deep_combine_chain_add_length,lstmmemory_cell2048_layer2_moe4 GAP = 0.830746937407 ------------------------------------------------ lstm_attention8_max,lstmmem1024_layer2_moe4_deep_combine_chain_add_length,lstmparallelmemory1024_moe8 GAP = 0.832236858925 ------------------------------------------------ lstm_attention8_max,lstmmem1024_layer2_moe4_deep_combine_chain_add_length,video_very_deep_combine_chain GAP = 0.832650348923 ------------------------------------------------ lstm_attention8_max,lstmmemory_cell1024_layer2_moe8,lstmmemory_cell2048_layer2_moe4 GAP = 0.829139497202 ------------------------------------------------ lstm_attention8_max,lstmmemory_cell1024_layer2_moe8,lstmparallelmemory1024_moe8 GAP = 0.830648119982 ------------------------------------------------ lstm_attention8_max,lstmmemory_cell1024_layer2_moe8,video_very_deep_combine_chain GAP = 0.831435923278 ------------------------------------------------ lstm_attention8_max,lstmmemory_cell2048_layer2_moe4,lstmparallelmemory1024_moe8 GAP = 0.831018565226 ------------------------------------------------ lstm_attention8_max,lstmmemory_cell2048_layer2_moe4,video_very_deep_combine_chain GAP = 0.831602818518 ------------------------------------------------ lstm_attention8_max,lstmparallelmemory1024_moe8,video_very_deep_combine_chain GAP = 0.833058032219 ------------------------------------------------ lstmmem1024_layer2_moe4_deep_combine_chain_add_length,lstmmemory_cell1024_layer2_moe8,lstmmemory_cell2048_layer2_moe4 GAP = 0.827740923928 ------------------------------------------------ lstmmem1024_layer2_moe4_deep_combine_chain_add_length,lstmmemory_cell1024_layer2_moe8,lstmparallelmemory1024_moe8 GAP = 0.829327455006 ------------------------------------------------ lstmmem1024_layer2_moe4_deep_combine_chain_add_length,lstmmemory_cell1024_layer2_moe8,video_very_deep_combine_chain GAP = 0.830225595389 ------------------------------------------------ lstmmem1024_layer2_moe4_deep_combine_chain_add_length,lstmmemory_cell2048_layer2_moe4,lstmparallelmemory1024_moe8 GAP = 0.829626608004 ------------------------------------------------ lstmmem1024_layer2_moe4_deep_combine_chain_add_length,lstmmemory_cell2048_layer2_moe4,video_very_deep_combine_chain GAP = 0.830460642345 ------------------------------------------------ lstmmem1024_layer2_moe4_deep_combine_chain_add_length,lstmparallelmemory1024_moe8,video_very_deep_combine_chain GAP = 0.831778438641 ------------------------------------------------ lstmmemory_cell1024_layer2_moe8,lstmmemory_cell2048_layer2_moe4,lstmparallelmemory1024_moe8 GAP = 0.828104386062 ------------------------------------------------ lstmmemory_cell1024_layer2_moe8,lstmmemory_cell2048_layer2_moe4,video_very_deep_combine_chain GAP = 0.829167967486 ------------------------------------------------ lstmmemory_cell1024_layer2_moe8,lstmparallelmemory1024_moe8,video_very_deep_combine_chain GAP = 0.830592592778 ------------------------------------------------ lstmmemory_cell2048_layer2_moe4,lstmparallelmemory1024_moe8,video_very_deep_combine_chain GAP = 0.830995957639 ------------------------------------------------ 4-model ensemble ------------------------------------------------ cnn_deep_combine_chain,deep_cnn_deep_combine_chain,lstm_attention8_max,lstmmem1024_layer2_moe4_deep_combine_chain_add_length GAP = 0.834859637855 ------------------------------------------------ cnn_deep_combine_chain,deep_cnn_deep_combine_chain,lstm_attention8_max,lstmmemory_cell1024_layer2_moe8 GAP = 0.833895260202 ------------------------------------------------ cnn_deep_combine_chain,deep_cnn_deep_combine_chain,lstm_attention8_max,lstmmemory_cell2048_layer2_moe4 GAP = 0.834045908801 ------------------------------------------------ cnn_deep_combine_chain,deep_cnn_deep_combine_chain,lstm_attention8_max,lstmparallelmemory1024_moe8 GAP = 0.835235503218 ------------------------------------------------ cnn_deep_combine_chain,deep_cnn_deep_combine_chain,lstm_attention8_max,video_very_deep_combine_chain GAP = 0.833723731659 ------------------------------------------------ cnn_deep_combine_chain,deep_cnn_deep_combine_chain,lstmmem1024_layer2_moe4_deep_combine_chain_add_length,lstmmemory_cell1024_layer2_moe8 GAP = 0.833104488978 ------------------------------------------------ cnn_deep_combine_chain,deep_cnn_deep_combine_chain,lstmmem1024_layer2_moe4_deep_combine_chain_add_length,lstmmemory_cell2048_layer2_moe4 GAP = 0.833324163897 ------------------------------------------------ cnn_deep_combine_chain,deep_cnn_deep_combine_chain,lstmmem1024_layer2_moe4_deep_combine_chain_add_length,lstmparallelmemory1024_moe8 GAP = 0.834484898844 ------------------------------------------------ cnn_deep_combine_chain,deep_cnn_deep_combine_chain,lstmmem1024_layer2_moe4_deep_combine_chain_add_length,video_very_deep_combine_chain GAP = 0.832930582133 ------------------------------------------------ cnn_deep_combine_chain,deep_cnn_deep_combine_chain,lstmmemory_cell1024_layer2_moe8,lstmmemory_cell2048_layer2_moe4 GAP = 0.832377724469 ------------------------------------------------ cnn_deep_combine_chain,deep_cnn_deep_combine_chain,lstmmemory_cell1024_layer2_moe8,lstmparallelmemory1024_moe8 GAP = 0.833414957585 ------------------------------------------------ cnn_deep_combine_chain,deep_cnn_deep_combine_chain,lstmmemory_cell1024_layer2_moe8,video_very_deep_combine_chain GAP = 0.832301392579 ------------------------------------------------ cnn_deep_combine_chain,deep_cnn_deep_combine_chain,lstmmemory_cell2048_layer2_moe4,lstmparallelmemory1024_moe8 GAP = 0.833769221192 ------------------------------------------------ cnn_deep_combine_chain,deep_cnn_deep_combine_chain,lstmmemory_cell2048_layer2_moe4,video_very_deep_combine_chain GAP = 0.832408546437 ------------------------------------------------ cnn_deep_combine_chain,deep_cnn_deep_combine_chain,lstmparallelmemory1024_moe8,video_very_deep_combine_chain GAP = 0.833437630091 ------------------------------------------------ cnn_deep_combine_chain,lstm_attention8_max,lstmmem1024_layer2_moe4_deep_combine_chain_add_length,lstmmemory_cell1024_layer2_moe8 GAP = 0.834464961259 ------------------------------------------------ cnn_deep_combine_chain,lstm_attention8_max,lstmmem1024_layer2_moe4_deep_combine_chain_add_length,lstmmemory_cell2048_layer2_moe4 GAP = 0.834659214213 ------------------------------------------------ cnn_deep_combine_chain,lstm_attention8_max,lstmmem1024_layer2_moe4_deep_combine_chain_add_length,lstmparallelmemory1024_moe8 GAP = 0.835789192715 ------------------------------------------------ cnn_deep_combine_chain,lstm_attention8_max,lstmmem1024_layer2_moe4_deep_combine_chain_add_length,video_very_deep_combine_chain GAP = 0.835273660124 ------------------------------------------------ cnn_deep_combine_chain,lstm_attention8_max,lstmmemory_cell1024_layer2_moe8,lstmmemory_cell2048_layer2_moe4 GAP = 0.833592380025 ------------------------------------------------ cnn_deep_combine_chain,lstm_attention8_max,lstmmemory_cell1024_layer2_moe8,lstmparallelmemory1024_moe8 GAP = 0.834699040434 ------------------------------------------------ cnn_deep_combine_chain,lstm_attention8_max,lstmmemory_cell1024_layer2_moe8,video_very_deep_combine_chain GAP = 0.834399178828 ------------------------------------------------ cnn_deep_combine_chain,lstm_attention8_max,lstmmemory_cell2048_layer2_moe4,lstmparallelmemory1024_moe8 GAP = 0.834979500288 ------------------------------------------------ cnn_deep_combine_chain,lstm_attention8_max,lstmmemory_cell2048_layer2_moe4,video_very_deep_combine_chain GAP = 0.834599103937 ------------------------------------------------ cnn_deep_combine_chain,lstm_attention8_max,lstmparallelmemory1024_moe8,video_very_deep_combine_chain GAP = 0.835657070379 ------------------------------------------------ cnn_deep_combine_chain,lstmmem1024_layer2_moe4_deep_combine_chain_add_length,lstmmemory_cell1024_layer2_moe8,lstmmemory_cell2048_layer2_moe4 GAP = 0.832603654742 ------------------------------------------------ cnn_deep_combine_chain,lstmmem1024_layer2_moe4_deep_combine_chain_add_length,lstmmemory_cell1024_layer2_moe8,lstmparallelmemory1024_moe8 GAP = 0.8337961468 ------------------------------------------------ cnn_deep_combine_chain,lstmmem1024_layer2_moe4_deep_combine_chain_add_length,lstmmemory_cell1024_layer2_moe8,video_very_deep_combine_chain GAP = 0.833522575937 ------------------------------------------------ cnn_deep_combine_chain,lstmmem1024_layer2_moe4_deep_combine_chain_add_length,lstmmemory_cell2048_layer2_moe4,lstmparallelmemory1024_moe8 GAP = 0.833995288321 ------------------------------------------------ cnn_deep_combine_chain,lstmmem1024_layer2_moe4_deep_combine_chain_add_length,lstmmemory_cell2048_layer2_moe4,video_very_deep_combine_chain GAP = 0.833691549239 ------------------------------------------------ cnn_deep_combine_chain,lstmmem1024_layer2_moe4_deep_combine_chain_add_length,lstmparallelmemory1024_moe8,video_very_deep_combine_chain GAP = 0.834677051398 ------------------------------------------------ cnn_deep_combine_chain,lstmmemory_cell1024_layer2_moe8,lstmmemory_cell2048_layer2_moe4,lstmparallelmemory1024_moe8 GAP = 0.832988258644 ------------------------------------------------ cnn_deep_combine_chain,lstmmemory_cell1024_layer2_moe8,lstmmemory_cell2048_layer2_moe4,video_very_deep_combine_chain GAP = 0.83279736014 ------------------------------------------------ cnn_deep_combine_chain,lstmmemory_cell1024_layer2_moe8,lstmparallelmemory1024_moe8,video_very_deep_combine_chain GAP = 0.833836412621 ------------------------------------------------ cnn_deep_combine_chain,lstmmemory_cell2048_layer2_moe4,lstmparallelmemory1024_moe8,video_very_deep_combine_chain GAP = 0.834129839666 ------------------------------------------------ deep_cnn_deep_combine_chain,lstm_attention8_max,lstmmem1024_layer2_moe4_deep_combine_chain_add_length,lstmmemory_cell1024_layer2_moe8 GAP = 0.833393442149 ------------------------------------------------ deep_cnn_deep_combine_chain,lstm_attention8_max,lstmmem1024_layer2_moe4_deep_combine_chain_add_length,lstmmemory_cell2048_layer2_moe4 GAP = 0.833546049331 ------------------------------------------------ deep_cnn_deep_combine_chain,lstm_attention8_max,lstmmem1024_layer2_moe4_deep_combine_chain_add_length,lstmparallelmemory1024_moe8 GAP = 0.834839753828 ------------------------------------------------ deep_cnn_deep_combine_chain,lstm_attention8_max,lstmmem1024_layer2_moe4_deep_combine_chain_add_length,video_very_deep_combine_chain GAP = 0.83420746933 ------------------------------------------------ deep_cnn_deep_combine_chain,lstm_attention8_max,lstmmemory_cell1024_layer2_moe8,lstmmemory_cell2048_layer2_moe4 GAP = 0.83240292504 ------------------------------------------------ deep_cnn_deep_combine_chain,lstm_attention8_max,lstmmemory_cell1024_layer2_moe8,lstmparallelmemory1024_moe8 GAP = 0.833633371649 ------------------------------------------------ deep_cnn_deep_combine_chain,lstm_attention8_max,lstmmemory_cell1024_layer2_moe8,video_very_deep_combine_chain GAP = 0.833366664258 ------------------------------------------------ deep_cnn_deep_combine_chain,lstm_attention8_max,lstmmemory_cell2048_layer2_moe4,lstmparallelmemory1024_moe8 GAP = 0.833897307363 ------------------------------------------------ deep_cnn_deep_combine_chain,lstm_attention8_max,lstmmemory_cell2048_layer2_moe4,video_very_deep_combine_chain GAP = 0.833519613265 ------------------------------------------------ deep_cnn_deep_combine_chain,lstm_attention8_max,lstmparallelmemory1024_moe8,video_very_deep_combine_chain GAP = 0.834663565256 ------------------------------------------------ deep_cnn_deep_combine_chain,lstmmem1024_layer2_moe4_deep_combine_chain_add_length,lstmmemory_cell1024_layer2_moe8,lstmmemory_cell2048_layer2_moe4 GAP = 0.831446232504 ------------------------------------------------ deep_cnn_deep_combine_chain,lstmmem1024_layer2_moe4_deep_combine_chain_add_length,lstmmemory_cell1024_layer2_moe8,lstmparallelmemory1024_moe8 GAP = 0.832594788307 ------------------------------------------------ deep_cnn_deep_combine_chain,lstmmem1024_layer2_moe4_deep_combine_chain_add_length,lstmmemory_cell1024_layer2_moe8,video_very_deep_combine_chain GAP = 0.832457962265 ------------------------------------------------ deep_cnn_deep_combine_chain,lstmmem1024_layer2_moe4_deep_combine_chain_add_length,lstmmemory_cell2048_layer2_moe4,lstmparallelmemory1024_moe8 GAP = 0.832881953446 ------------------------------------------------ deep_cnn_deep_combine_chain,lstmmem1024_layer2_moe4_deep_combine_chain_add_length,lstmmemory_cell2048_layer2_moe4,video_very_deep_combine_chain GAP = 0.832662090456 ------------------------------------------------ deep_cnn_deep_combine_chain,lstmmem1024_layer2_moe4_deep_combine_chain_add_length,lstmparallelmemory1024_moe8,video_very_deep_combine_chain GAP = 0.833702886165 ------------------------------------------------ deep_cnn_deep_combine_chain,lstmmemory_cell1024_layer2_moe8,lstmmemory_cell2048_layer2_moe4,lstmparallelmemory1024_moe8 GAP = 0.831723349274 ------------------------------------------------ deep_cnn_deep_combine_chain,lstmmemory_cell1024_layer2_moe8,lstmmemory_cell2048_layer2_moe4,video_very_deep_combine_chain GAP = 0.831664565904 ------------------------------------------------ deep_cnn_deep_combine_chain,lstmmemory_cell1024_layer2_moe8,lstmparallelmemory1024_moe8,video_very_deep_combine_chain GAP = 0.832780128708 ------------------------------------------------ deep_cnn_deep_combine_chain,lstmmemory_cell2048_layer2_moe4,lstmparallelmemory1024_moe8,video_very_deep_combine_chain GAP = 0.83309748339 ------------------------------------------------ lstm_attention8_max,lstmmem1024_layer2_moe4_deep_combine_chain_add_length,lstmmemory_cell1024_layer2_moe8,lstmmemory_cell2048_layer2_moe4 GAP = 0.831784972382 ------------------------------------------------ lstm_attention8_max,lstmmem1024_layer2_moe4_deep_combine_chain_add_length,lstmmemory_cell1024_layer2_moe8,lstmparallelmemory1024_moe8 GAP = 0.83302089599 ------------------------------------------------ lstm_attention8_max,lstmmem1024_layer2_moe4_deep_combine_chain_add_length,lstmmemory_cell1024_layer2_moe8,video_very_deep_combine_chain GAP = 0.833818492121 ------------------------------------------------ lstm_attention8_max,lstmmem1024_layer2_moe4_deep_combine_chain_add_length,lstmmemory_cell2048_layer2_moe4,lstmparallelmemory1024_moe8 GAP = 0.833296927263 ------------------------------------------------ lstm_attention8_max,lstmmem1024_layer2_moe4_deep_combine_chain_add_length,lstmmemory_cell2048_layer2_moe4,video_very_deep_combine_chain GAP = 0.833948837411 ------------------------------------------------ lstm_attention8_max,lstmmem1024_layer2_moe4_deep_combine_chain_add_length,lstmparallelmemory1024_moe8,video_very_deep_combine_chain GAP = 0.835128538546 ------------------------------------------------ lstm_attention8_max,lstmmemory_cell1024_layer2_moe8,lstmmemory_cell2048_layer2_moe4,lstmparallelmemory1024_moe8 GAP = 0.831896833011 ------------------------------------------------ lstm_attention8_max,lstmmemory_cell1024_layer2_moe8,lstmmemory_cell2048_layer2_moe4,video_very_deep_combine_chain GAP = 0.832843246048 ------------------------------------------------ lstm_attention8_max,lstmmemory_cell1024_layer2_moe8,lstmparallelmemory1024_moe8,video_very_deep_combine_chain GAP = 0.833988784393 ------------------------------------------------ lstm_attention8_max,lstmmemory_cell2048_layer2_moe4,lstmparallelmemory1024_moe8,video_very_deep_combine_chain GAP = 0.834228084116 ------------------------------------------------ lstmmem1024_layer2_moe4_deep_combine_chain_add_length,lstmmemory_cell1024_layer2_moe8,lstmmemory_cell2048_layer2_moe4,lstmparallelmemory1024_moe8 GAP = 0.830787155159 ------------------------------------------------ lstmmem1024_layer2_moe4_deep_combine_chain_add_length,lstmmemory_cell1024_layer2_moe8,lstmmemory_cell2048_layer2_moe4,video_very_deep_combine_chain GAP = 0.831791382924 ------------------------------------------------ lstmmem1024_layer2_moe4_deep_combine_chain_add_length,lstmmemory_cell1024_layer2_moe8,lstmparallelmemory1024_moe8,video_very_deep_combine_chain GAP = 0.832990647204 ------------------------------------------------ lstmmem1024_layer2_moe4_deep_combine_chain_add_length,lstmmemory_cell2048_layer2_moe4,lstmparallelmemory1024_moe8,video_very_deep_combine_chain GAP = 0.833219017873 ------------------------------------------------ lstmmemory_cell1024_layer2_moe8,lstmmemory_cell2048_layer2_moe4,lstmparallelmemory1024_moe8,video_very_deep_combine_chain GAP = 0.832099590832 ------------------------------------------------ 5-model ensemble ------------------------------------------------ cnn_deep_combine_chain,deep_cnn_deep_combine_chain,lstm_attention8_max,lstmmem1024_layer2_moe4_deep_combine_chain_add_length,lstmmemory_cell1024_layer2_moe8 GAP = 0.835438033957 ------------------------------------------------ cnn_deep_combine_chain,deep_cnn_deep_combine_chain,lstm_attention8_max,lstmmem1024_layer2_moe4_deep_combine_chain_add_length,lstmmemory_cell2048_layer2_moe4 GAP = 0.835605498079 ------------------------------------------------ cnn_deep_combine_chain,deep_cnn_deep_combine_chain,lstm_attention8_max,lstmmem1024_layer2_moe4_deep_combine_chain_add_length,lstmparallelmemory1024_moe8 GAP = 0.83653394275 ------------------------------------------------ cnn_deep_combine_chain,deep_cnn_deep_combine_chain,lstm_attention8_max,lstmmem1024_layer2_moe4_deep_combine_chain_add_length,video_very_deep_combine_chain GAP = 0.835745783944 ------------------------------------------------ cnn_deep_combine_chain,deep_cnn_deep_combine_chain,lstm_attention8_max,lstmmemory_cell1024_layer2_moe8,lstmmemory_cell2048_layer2_moe4 GAP = 0.834686137206 ------------------------------------------------ cnn_deep_combine_chain,deep_cnn_deep_combine_chain,lstm_attention8_max,lstmmemory_cell1024_layer2_moe8,lstmparallelmemory1024_moe8 GAP = 0.835670112826 ------------------------------------------------ cnn_deep_combine_chain,deep_cnn_deep_combine_chain,lstm_attention8_max,lstmmemory_cell1024_layer2_moe8,video_very_deep_combine_chain GAP = 0.835047895241 ------------------------------------------------ cnn_deep_combine_chain,deep_cnn_deep_combine_chain,lstm_attention8_max,lstmmemory_cell2048_layer2_moe4,lstmparallelmemory1024_moe8 GAP = 0.835889967275 ------------------------------------------------ cnn_deep_combine_chain,deep_cnn_deep_combine_chain,lstm_attention8_max,lstmmemory_cell2048_layer2_moe4,video_very_deep_combine_chain GAP = 0.835210150381 ------------------------------------------------ cnn_deep_combine_chain,deep_cnn_deep_combine_chain,lstm_attention8_max,lstmparallelmemory1024_moe8,video_very_deep_combine_chain GAP = 0.836093418248 ------------------------------------------------ cnn_deep_combine_chain,deep_cnn_deep_combine_chain,lstmmem1024_layer2_moe4_deep_combine_chain_add_length,lstmmemory_cell1024_layer2_moe8,lstmmemory_cell2048_layer2_moe4 GAP = 0.83401135364 ------------------------------------------------ cnn_deep_combine_chain,deep_cnn_deep_combine_chain,lstmmem1024_layer2_moe4_deep_combine_chain_add_length,lstmmemory_cell1024_layer2_moe8,lstmparallelmemory1024_moe8 GAP = 0.834938997204 ------------------------------------------------ cnn_deep_combine_chain,deep_cnn_deep_combine_chain,lstmmem1024_layer2_moe4_deep_combine_chain_add_length,lstmmemory_cell1024_layer2_moe8,video_very_deep_combine_chain GAP = 0.834311398082 ------------------------------------------------ cnn_deep_combine_chain,deep_cnn_deep_combine_chain,lstmmem1024_layer2_moe4_deep_combine_chain_add_length,lstmmemory_cell2048_layer2_moe4,lstmparallelmemory1024_moe8 GAP = 0.835178958848 ------------------------------------------------ cnn_deep_combine_chain,deep_cnn_deep_combine_chain,lstmmem1024_layer2_moe4_deep_combine_chain_add_length,lstmmemory_cell2048_layer2_moe4,video_very_deep_combine_chain GAP = 0.834508517888 ------------------------------------------------ cnn_deep_combine_chain,deep_cnn_deep_combine_chain,lstmmem1024_layer2_moe4_deep_combine_chain_add_length,lstmparallelmemory1024_moe8,video_very_deep_combine_chain GAP = 0.835322108623 ------------------------------------------------ cnn_deep_combine_chain,deep_cnn_deep_combine_chain,lstmmemory_cell1024_layer2_moe8,lstmmemory_cell2048_layer2_moe4,lstmparallelmemory1024_moe8 GAP = 0.834318844694 ------------------------------------------------ cnn_deep_combine_chain,deep_cnn_deep_combine_chain,lstmmemory_cell1024_layer2_moe8,lstmmemory_cell2048_layer2_moe4,video_very_deep_combine_chain GAP = 0.833795003243 ------------------------------------------------ cnn_deep_combine_chain,deep_cnn_deep_combine_chain,lstmmemory_cell1024_layer2_moe8,lstmparallelmemory1024_moe8,video_very_deep_combine_chain GAP = 0.834649121681 ------------------------------------------------ cnn_deep_combine_chain,deep_cnn_deep_combine_chain,lstmmemory_cell2048_layer2_moe4,lstmparallelmemory1024_moe8,video_very_deep_combine_chain GAP = 0.834845499789 ------------------------------------------------ cnn_deep_combine_chain,lstm_attention8_max,lstmmem1024_layer2_moe4_deep_combine_chain_add_length,lstmmemory_cell1024_layer2_moe8,lstmmemory_cell2048_layer2_moe4 GAP = 0.83493975204 ------------------------------------------------ cnn_deep_combine_chain,lstm_attention8_max,lstmmem1024_layer2_moe4_deep_combine_chain_add_length,lstmmemory_cell1024_layer2_moe8,lstmparallelmemory1024_moe8 GAP = 0.835876294898 ------------------------------------------------ cnn_deep_combine_chain,lstm_attention8_max,lstmmem1024_layer2_moe4_deep_combine_chain_add_length,lstmmemory_cell1024_layer2_moe8,video_very_deep_combine_chain GAP = 0.835856824821 ------------------------------------------------ cnn_deep_combine_chain,lstm_attention8_max,lstmmem1024_layer2_moe4_deep_combine_chain_add_length,lstmmemory_cell2048_layer2_moe4,lstmparallelmemory1024_moe8 GAP = 0.836099640517 ------------------------------------------------ cnn_deep_combine_chain,lstm_attention8_max,lstmmem1024_layer2_moe4_deep_combine_chain_add_length,lstmmemory_cell2048_layer2_moe4,video_very_deep_combine_chain GAP = 0.836003869651 ------------------------------------------------ cnn_deep_combine_chain,lstm_attention8_max,lstmmem1024_layer2_moe4_deep_combine_chain_add_length,lstmparallelmemory1024_moe8,video_very_deep_combine_chain GAP = 0.836959694784 ------------------------------------------------ cnn_deep_combine_chain,lstm_attention8_max,lstmmemory_cell1024_layer2_moe8,lstmmemory_cell2048_layer2_moe4,lstmparallelmemory1024_moe8 GAP = 0.835127020411 ------------------------------------------------ cnn_deep_combine_chain,lstm_attention8_max,lstmmemory_cell1024_layer2_moe8,lstmmemory_cell2048_layer2_moe4,video_very_deep_combine_chain GAP = 0.835186988207 ------------------------------------------------ cnn_deep_combine_chain,lstm_attention8_max,lstmmemory_cell1024_layer2_moe8,lstmparallelmemory1024_moe8,video_very_deep_combine_chain GAP = 0.836084071621 ------------------------------------------------ cnn_deep_combine_chain,lstm_attention8_max,lstmmemory_cell2048_layer2_moe4,lstmparallelmemory1024_moe8,video_very_deep_combine_chain GAP = 0.836365387451 ------------------------------------------------ cnn_deep_combine_chain,lstmmem1024_layer2_moe4_deep_combine_chain_add_length,lstmmemory_cell1024_layer2_moe8,lstmmemory_cell2048_layer2_moe4,lstmparallelmemory1024_moe8 GAP = 0.834314811777 ------------------------------------------------ cnn_deep_combine_chain,lstmmem1024_layer2_moe4_deep_combine_chain_add_length,lstmmemory_cell1024_layer2_moe8,lstmmemory_cell2048_layer2_moe4,video_very_deep_combine_chain GAP = 0.834380440941 ------------------------------------------------ cnn_deep_combine_chain,lstmmem1024_layer2_moe4_deep_combine_chain_add_length,lstmmemory_cell1024_layer2_moe8,lstmparallelmemory1024_moe8,video_very_deep_combine_chain GAP = 0.835262275125 ------------------------------------------------ cnn_deep_combine_chain,lstmmem1024_layer2_moe4_deep_combine_chain_add_length,lstmmemory_cell2048_layer2_moe4,lstmparallelmemory1024_moe8,video_very_deep_combine_chain GAP = 0.835462501109 ------------------------------------------------ cnn_deep_combine_chain,lstmmemory_cell1024_layer2_moe8,lstmmemory_cell2048_layer2_moe4,lstmparallelmemory1024_moe8,video_very_deep_combine_chain GAP = 0.834682844519 ------------------------------------------------ deep_cnn_deep_combine_chain,lstm_attention8_max,lstmmem1024_layer2_moe4_deep_combine_chain_add_length,lstmmemory_cell1024_layer2_moe8,lstmmemory_cell2048_layer2_moe4 GAP = 0.834000181534 ------------------------------------------------ deep_cnn_deep_combine_chain,lstm_attention8_max,lstmmem1024_layer2_moe4_deep_combine_chain_add_length,lstmmemory_cell1024_layer2_moe8,lstmparallelmemory1024_moe8 GAP = 0.834996923723 ------------------------------------------------ deep_cnn_deep_combine_chain,lstm_attention8_max,lstmmem1024_layer2_moe4_deep_combine_chain_add_length,lstmmemory_cell1024_layer2_moe8,video_very_deep_combine_chain GAP = 0.835026558331 ------------------------------------------------ deep_cnn_deep_combine_chain,lstm_attention8_max,lstmmem1024_layer2_moe4_deep_combine_chain_add_length,lstmmemory_cell2048_layer2_moe4,lstmparallelmemory1024_moe8 GAP = 0.835168142367 ------------------------------------------------ deep_cnn_deep_combine_chain,lstm_attention8_max,lstmmem1024_layer2_moe4_deep_combine_chain_add_length,lstmmemory_cell2048_layer2_moe4,video_very_deep_combine_chain GAP = 0.835154802734 ------------------------------------------------ deep_cnn_deep_combine_chain,lstm_attention8_max,lstmmem1024_layer2_moe4_deep_combine_chain_add_length,lstmparallelmemory1024_moe8,video_very_deep_combine_chain GAP = 0.836113458024 ------------------------------------------------ deep_cnn_deep_combine_chain,lstm_attention8_max,lstmmemory_cell1024_layer2_moe8,lstmmemory_cell2048_layer2_moe4,lstmparallelmemory1024_moe8 GAP = 0.834178068594 ------------------------------------------------ deep_cnn_deep_combine_chain,lstm_attention8_max,lstmmemory_cell1024_layer2_moe8,lstmmemory_cell2048_layer2_moe4,video_very_deep_combine_chain GAP = 0.834322025109 ------------------------------------------------ deep_cnn_deep_combine_chain,lstm_attention8_max,lstmmemory_cell1024_layer2_moe8,lstmparallelmemory1024_moe8,video_very_deep_combine_chain GAP = 0.835242509217 ------------------------------------------------ deep_cnn_deep_combine_chain,lstm_attention8_max,lstmmemory_cell2048_layer2_moe4,lstmparallelmemory1024_moe8,video_very_deep_combine_chain GAP = 0.835421384341 ------------------------------------------------ deep_cnn_deep_combine_chain,lstmmem1024_layer2_moe4_deep_combine_chain_add_length,lstmmemory_cell1024_layer2_moe8,lstmmemory_cell2048_layer2_moe4,lstmparallelmemory1024_moe8 GAP = 0.833302862578 ------------------------------------------------ deep_cnn_deep_combine_chain,lstmmem1024_layer2_moe4_deep_combine_chain_add_length,lstmmemory_cell1024_layer2_moe8,lstmmemory_cell2048_layer2_moe4,video_very_deep_combine_chain GAP = 0.833484620513 ------------------------------------------------ deep_cnn_deep_combine_chain,lstmmem1024_layer2_moe4_deep_combine_chain_add_length,lstmmemory_cell1024_layer2_moe8,lstmparallelmemory1024_moe8,video_very_deep_combine_chain GAP = 0.834398472225 ------------------------------------------------ deep_cnn_deep_combine_chain,lstmmem1024_layer2_moe4_deep_combine_chain_add_length,lstmmemory_cell2048_layer2_moe4,lstmparallelmemory1024_moe8,video_very_deep_combine_chain GAP = 0.834618836206 ------------------------------------------------ deep_cnn_deep_combine_chain,lstmmemory_cell1024_layer2_moe8,lstmmemory_cell2048_layer2_moe4,lstmparallelmemory1024_moe8,video_very_deep_combine_chain GAP = 0.833736799505 ------------------------------------------------ lstm_attention8_max,lstmmem1024_layer2_moe4_deep_combine_chain_add_length,lstmmemory_cell1024_layer2_moe8,lstmmemory_cell2048_layer2_moe4,lstmparallelmemory1024_moe8 GAP = 0.833542783923 ------------------------------------------------ lstm_attention8_max,lstmmem1024_layer2_moe4_deep_combine_chain_add_length,lstmmemory_cell1024_layer2_moe8,lstmmemory_cell2048_layer2_moe4,video_very_deep_combine_chain GAP = 0.834372993093 ------------------------------------------------ lstm_attention8_max,lstmmem1024_layer2_moe4_deep_combine_chain_add_length,lstmmemory_cell1024_layer2_moe8,lstmparallelmemory1024_moe8,video_very_deep_combine_chain GAP = 0.835390081231 ------------------------------------------------ lstm_attention8_max,lstmmem1024_layer2_moe4_deep_combine_chain_add_length,lstmmemory_cell2048_layer2_moe4,lstmparallelmemory1024_moe8,video_very_deep_combine_chain GAP = 0.835597107314 ------------------------------------------------ lstm_attention8_max,lstmmemory_cell1024_layer2_moe8,lstmmemory_cell2048_layer2_moe4,lstmparallelmemory1024_moe8,video_very_deep_combine_chain GAP = 0.83455284016 ------------------------------------------------ lstmmem1024_layer2_moe4_deep_combine_chain_add_length,lstmmemory_cell1024_layer2_moe8,lstmmemory_cell2048_layer2_moe4,lstmparallelmemory1024_moe8,video_very_deep_combine_chain GAP = 0.833675261416 ------------------------------------------------ 6-model ensemble ------------------------------------------------ cnn_deep_combine_chain,deep_cnn_deep_combine_chain,lstm_attention8_max,lstmmem1024_layer2_moe4_deep_combine_chain_add_length,lstmmemory_cell1024_layer2_moe8,lstmmemory_cell2048_layer2_moe4 GAP = 0.835764820291 ------------------------------------------------ cnn_deep_combine_chain,deep_cnn_deep_combine_chain,lstm_attention8_max,lstmmem1024_layer2_moe4_deep_combine_chain_add_length,lstmmemory_cell1024_layer2_moe8,lstmparallelmemory1024_moe8 GAP = 0.836573100548 ------------------------------------------------ cnn_deep_combine_chain,deep_cnn_deep_combine_chain,lstm_attention8_max,lstmmem1024_layer2_moe4_deep_combine_chain_add_length,lstmmemory_cell1024_layer2_moe8,video_very_deep_combine_chain GAP = 0.836286868542 ------------------------------------------------ cnn_deep_combine_chain,deep_cnn_deep_combine_chain,lstm_attention8_max,lstmmem1024_layer2_moe4_deep_combine_chain_add_length,lstmmemory_cell2048_layer2_moe4,lstmparallelmemory1024_moe8 GAP = 0.836789577149 ------------------------------------------------ cnn_deep_combine_chain,deep_cnn_deep_combine_chain,lstm_attention8_max,lstmmem1024_layer2_moe4_deep_combine_chain_add_length,lstmmemory_cell2048_layer2_moe4,video_very_deep_combine_chain GAP = 0.836410907788 ------------------------------------------------ cnn_deep_combine_chain,deep_cnn_deep_combine_chain,lstm_attention8_max,lstmmem1024_layer2_moe4_deep_combine_chain_add_length,lstmparallelmemory1024_moe8,video_very_deep_combine_chain GAP = 0.837230111667 ------------------------------------------------ cnn_deep_combine_chain,deep_cnn_deep_combine_chain,lstm_attention8_max,lstmmemory_cell1024_layer2_moe8,lstmmemory_cell2048_layer2_moe4,lstmparallelmemory1024_moe8 GAP = 0.835967761151 ------------------------------------------------ cnn_deep_combine_chain,deep_cnn_deep_combine_chain,lstm_attention8_max,lstmmemory_cell1024_layer2_moe8,lstmmemory_cell2048_layer2_moe4,video_very_deep_combine_chain GAP = 0.835742351815 ------------------------------------------------ cnn_deep_combine_chain,deep_cnn_deep_combine_chain,lstm_attention8_max,lstmmemory_cell1024_layer2_moe8,lstmparallelmemory1024_moe8,video_very_deep_combine_chain GAP = 0.836527025619 ------------------------------------------------ cnn_deep_combine_chain,deep_cnn_deep_combine_chain,lstm_attention8_max,lstmmemory_cell2048_layer2_moe4,lstmparallelmemory1024_moe8,video_very_deep_combine_chain GAP = 0.836708452803 ------------------------------------------------ cnn_deep_combine_chain,deep_cnn_deep_combine_chain,lstmmem1024_layer2_moe4_deep_combine_chain_add_length,lstmmemory_cell1024_layer2_moe8,lstmmemory_cell2048_layer2_moe4,lstmparallelmemory1024_moe8 GAP = 0.835354805374 ------------------------------------------------ cnn_deep_combine_chain,deep_cnn_deep_combine_chain,lstmmem1024_layer2_moe4_deep_combine_chain_add_length,lstmmemory_cell1024_layer2_moe8,lstmmemory_cell2048_layer2_moe4,video_very_deep_combine_chain GAP = 0.835042792794 ------------------------------------------------ cnn_deep_combine_chain,deep_cnn_deep_combine_chain,lstmmem1024_layer2_moe4_deep_combine_chain_add_length,lstmmemory_cell1024_layer2_moe8,lstmparallelmemory1024_moe8,video_very_deep_combine_chain GAP = 0.835826695583 ------------------------------------------------ cnn_deep_combine_chain,deep_cnn_deep_combine_chain,lstmmem1024_layer2_moe4_deep_combine_chain_add_length,lstmmemory_cell2048_layer2_moe4,lstmparallelmemory1024_moe8,video_very_deep_combine_chain GAP = 0.836076030996 ------------------------------------------------ cnn_deep_combine_chain,deep_cnn_deep_combine_chain,lstmmemory_cell1024_layer2_moe8,lstmmemory_cell2048_layer2_moe4,lstmparallelmemory1024_moe8,video_very_deep_combine_chain GAP = 0.835345677749 ------------------------------------------------ cnn_deep_combine_chain,lstm_attention8_max,lstmmem1024_layer2_moe4_deep_combine_chain_add_length,lstmmemory_cell1024_layer2_moe8,lstmmemory_cell2048_layer2_moe4,lstmparallelmemory1024_moe8 GAP = 0.835999247145 ------------------------------------------------ cnn_deep_combine_chain,lstm_attention8_max,lstmmem1024_layer2_moe4_deep_combine_chain_add_length,lstmmemory_cell1024_layer2_moe8,lstmmemory_cell2048_layer2_moe4,video_very_deep_combine_chain GAP = 0.836183473417 ------------------------------------------------ cnn_deep_combine_chain,lstm_attention8_max,lstmmem1024_layer2_moe4_deep_combine_chain_add_length,lstmmemory_cell1024_layer2_moe8,lstmparallelmemory1024_moe8,video_very_deep_combine_chain GAP = 0.837008192923 ------------------------------------------------ cnn_deep_combine_chain,lstm_attention8_max,lstmmem1024_layer2_moe4_deep_combine_chain_add_length,lstmmemory_cell2048_layer2_moe4,lstmparallelmemory1024_moe8,video_very_deep_combine_chain GAP = 0.837117755989 ------------------------------------------------ cnn_deep_combine_chain,lstm_attention8_max,lstmmemory_cell1024_layer2_moe8,lstmmemory_cell2048_layer2_moe4,lstmparallelmemory1024_moe8,video_very_deep_combine_chain GAP = 0.836389579777 ------------------------------------------------ cnn_deep_combine_chain,lstmmem1024_layer2_moe4_deep_combine_chain_add_length,lstmmemory_cell1024_layer2_moe8,lstmmemory_cell2048_layer2_moe4,lstmparallelmemory1024_moe8,video_very_deep_combine_chain GAP = 0.835662725883 ------------------------------------------------ deep_cnn_deep_combine_chain,lstm_attention8_max,lstmmem1024_layer2_moe4_deep_combine_chain_add_length,lstmmemory_cell1024_layer2_moe8,lstmmemory_cell2048_layer2_moe4,lstmparallelmemory1024_moe8 GAP = 0.835175538634 ------------------------------------------------ deep_cnn_deep_combine_chain,lstm_attention8_max,lstmmem1024_layer2_moe4_deep_combine_chain_add_length,lstmmemory_cell1024_layer2_moe8,lstmmemory_cell2048_layer2_moe4,video_very_deep_combine_chain GAP = 0.835437516935 ------------------------------------------------ deep_cnn_deep_combine_chain,lstm_attention8_max,lstmmem1024_layer2_moe4_deep_combine_chain_add_length,lstmmemory_cell1024_layer2_moe8,lstmparallelmemory1024_moe8,video_very_deep_combine_chain GAP = 0.836246224116 ------------------------------------------------ deep_cnn_deep_combine_chain,lstm_attention8_max,lstmmem1024_layer2_moe4_deep_combine_chain_add_length,lstmmemory_cell2048_layer2_moe4,lstmparallelmemory1024_moe8,video_very_deep_combine_chain GAP = 0.83637430754 ------------------------------------------------ deep_cnn_deep_combine_chain,lstm_attention8_max,lstmmemory_cell1024_layer2_moe8,lstmmemory_cell2048_layer2_moe4,lstmparallelmemory1024_moe8,video_very_deep_combine_chain GAP = 0.835617024019 ------------------------------------------------ deep_cnn_deep_combine_chain,lstmmem1024_layer2_moe4_deep_combine_chain_add_length,lstmmemory_cell1024_layer2_moe8,lstmmemory_cell2048_layer2_moe4,lstmparallelmemory1024_moe8,video_very_deep_combine_chain GAP = 0.834914528567 ------------------------------------------------ lstm_attention8_max,lstmmem1024_layer2_moe4_deep_combine_chain_add_length,lstmmemory_cell1024_layer2_moe8,lstmmemory_cell2048_layer2_moe4,lstmparallelmemory1024_moe8,video_very_deep_combine_chain GAP = 0.835543683299 ------------------------------------------------ 7-model ensemble ------------------------------------------------ cnn_deep_combine_chain,deep_cnn_deep_combine_chain,lstm_attention8_max,lstmmem1024_layer2_moe4_deep_combine_chain_add_length,lstmmemory_cell1024_layer2_moe8,lstmmemory_cell2048_layer2_moe4,lstmparallelmemory1024_moe8 GAP = 0.836664222772 ------------------------------------------------ cnn_deep_combine_chain,deep_cnn_deep_combine_chain,lstm_attention8_max,lstmmem1024_layer2_moe4_deep_combine_chain_add_length,lstmmemory_cell1024_layer2_moe8,lstmmemory_cell2048_layer2_moe4,video_very_deep_combine_chain GAP = 0.83658908908 ------------------------------------------------ cnn_deep_combine_chain,deep_cnn_deep_combine_chain,lstm_attention8_max,lstmmem1024_layer2_moe4_deep_combine_chain_add_length,lstmmemory_cell1024_layer2_moe8,lstmparallelmemory1024_moe8,video_very_deep_combine_chain GAP = 0.837284897399 ------------------------------------------------ cnn_deep_combine_chain,deep_cnn_deep_combine_chain,lstm_attention8_max,lstmmem1024_layer2_moe4_deep_combine_chain_add_length,lstmmemory_cell2048_layer2_moe4,lstmparallelmemory1024_moe8,video_very_deep_combine_chain GAP = 0.837397359344 ------------------------------------------------ cnn_deep_combine_chain,deep_cnn_deep_combine_chain,lstm_attention8_max,lstmmemory_cell1024_layer2_moe8,lstmmemory_cell2048_layer2_moe4,lstmparallelmemory1024_moe8,video_very_deep_combine_chain GAP = 0.836838205684 ------------------------------------------------ cnn_deep_combine_chain,deep_cnn_deep_combine_chain,lstmmem1024_layer2_moe4_deep_combine_chain_add_length,lstmmemory_cell1024_layer2_moe8,lstmmemory_cell2048_layer2_moe4,lstmparallelmemory1024_moe8,video_very_deep_combine_chain GAP = 0.836228065763 ------------------------------------------------ cnn_deep_combine_chain,lstm_attention8_max,lstmmem1024_layer2_moe4_deep_combine_chain_add_length,lstmmemory_cell1024_layer2_moe8,lstmmemory_cell2048_layer2_moe4,lstmparallelmemory1024_moe8,video_very_deep_combine_chain GAP = 0.837045884874 ------------------------------------------------ deep_cnn_deep_combine_chain,lstm_attention8_max,lstmmem1024_layer2_moe4_deep_combine_chain_add_length,lstmmemory_cell1024_layer2_moe8,lstmmemory_cell2048_layer2_moe4,lstmparallelmemory1024_moe8,video_very_deep_combine_chain GAP = 0.836382105823 ------------------------------------------------ 8-model ensemble ------------------------------------------------ cnn_deep_combine_chain,deep_cnn_deep_combine_chain,lstm_attention8_max,lstmmem1024_layer2_moe4_deep_combine_chain_add_length,lstmmemory_cell1024_layer2_moe8,lstmmemory_cell2048_layer2_moe4,lstmparallelmemory1024_moe8,video_very_deep_combine_chain GAP = 0.837375440061 ------------------------------------------------ ================================================ FILE: youtube-8m-ensemble/ensemble_scripts/explore-mean_model.sh ================================================ train_path=/Youtube-8M/model_predictions/ensemble_train for d in $(ls $train_path | sort); do train_data_patterns="${train_path}/${d}/*.tfrecord" echo "$d" CUDA_VISIBLE_DEVICES=0 python eval.py \ --model_checkpoint_path="../model/mean_model/model.ckpt-0" \ --train_dir="../model/mean_model" \ --model="MeanModel" \ --echo_gap=True \ --eval_data_patterns=$train_data_patterns | tail -n 1 echo "------------------------------------------------" done for d1 in $(ls $train_path | sort); do for d2 in $(ls $train_path | sort); do if [[ "$d1" < "$d2" ]]; then train_data_patterns="${train_path}/${d1}/*.tfrecord,${train_path}/${d2}/*.tfrecord" echo "$d1,$d2" CUDA_VISIBLE_DEVICES=0 python eval.py \ --model_checkpoint_path="../model/mean_model/model.ckpt-0" \ --train_dir="../model/mean_model" \ --model="MeanModel" \ --echo_gap=True \ --eval_data_patterns=$train_data_patterns | tail -n 1 echo "------------------------------------------------" fi done done for d1 in $(ls $train_path | sort); do for d2 in $(ls $train_path | sort); do if [[ "$d1" < "$d2" ]]; then for d3 in $(ls $train_path | sort); do if [[ "$d2" < "$d3" ]]; then train_data_patterns="${train_path}/${d1}/*.tfrecord,${train_path}/${d2}/*.tfrecord,${train_path}/${d3}/*.tfrecord" echo "$d1,$d2,$d3" CUDA_VISIBLE_DEVICES=0 python eval.py \ --model_checkpoint_path="../model/mean_model/model.ckpt-0" \ --train_dir="../model/mean_model" \ --model="MeanModel" \ --echo_gap=True \ --eval_data_patterns=$train_data_patterns | tail -n 1 echo "------------------------------------------------" fi done fi done done for d1 in $(ls $train_path | sort); do for d2 in $(ls $train_path | sort); do if [[ "$d1" < "$d2" ]]; then for d3 in $(ls $train_path | sort); do if [[ "$d2" < "$d3" ]]; then for d4 in $(ls $train_path | sort); do if [[ "$d3" < "$d4" ]]; then train_data_patterns="${train_path}/${d1}/*.tfrecord,${train_path}/${d2}/*.tfrecord,${train_path}/${d3}/*.tfrecord,${train_path}/${d4}/*.tfrecord" echo "$d1,$d2,$d3,$d4" CUDA_VISIBLE_DEVICES=0 python eval.py \ --model_checkpoint_path="../model/mean_model/model.ckpt-0" \ --train_dir="../model/mean_model" \ --model="MeanModel" \ --echo_gap=True \ --eval_data_patterns=$train_data_patterns | tail -n 1 echo "------------------------------------------------" fi done fi done fi done done for d1 in $(ls $train_path | sort); do for d2 in $(ls $train_path | sort); do if [[ "$d1" < "$d2" ]]; then for d3 in $(ls $train_path | sort); do if [[ "$d2" < "$d3" ]]; then for d4 in $(ls $train_path | sort); do if [[ "$d3" < "$d4" ]]; then for d5 in $(ls $train_path | sort); do if [[ "$d4" < "$d5" ]]; then train_data_patterns="${train_path}/${d1}/*.tfrecord,${train_path}/${d2}/*.tfrecord,${train_path}/${d3}/*.tfrecord,${train_path}/${d4}/*.tfrecord,${train_path}/${d5}/*.tfrecord" echo "$d1,$d2,$d3,$d4,$d5" CUDA_VISIBLE_DEVICES=0 python eval.py \ --model_checkpoint_path="../model/mean_model/model.ckpt-0" \ --train_dir="../model/mean_model" \ --model="MeanModel" \ --echo_gap=True \ --eval_data_patterns=$train_data_patterns | tail -n 1 echo "------------------------------------------------" fi done fi done fi done fi done done for d1 in $(ls $train_path | sort); do for d2 in $(ls $train_path | sort); do if [[ "$d1" < "$d2" ]]; then for d3 in $(ls $train_path | sort); do if [[ "$d2" < "$d3" ]]; then for d4 in $(ls $train_path | sort); do if [[ "$d3" < "$d4" ]]; then for d5 in $(ls $train_path | sort); do if [[ "$d4" < "$d5" ]]; then for d6 in $(ls $train_path | sort); do if [[ "$d5" < "$d6" ]]; then train_data_patterns="${train_path}/${d1}/*.tfrecord,${train_path}/${d2}/*.tfrecord,${train_path}/${d3}/*.tfrecord,${train_path}/${d4}/*.tfrecord,${train_path}/${d5}/*.tfrecord,${train_path}/${d6}/*.tfrecord" echo "$d1,$d2,$d3,$d4,$d5,$d6" CUDA_VISIBLE_DEVICES=0 python eval.py \ --model_checkpoint_path="../model/mean_model/model.ckpt-0" \ --train_dir="../model/mean_model" \ --model="MeanModel" \ --echo_gap=True \ --eval_data_patterns=$train_data_patterns | tail -n 1 echo "------------------------------------------------" fi done fi done fi done fi done fi done done ================================================ FILE: youtube-8m-ensemble/ensemble_scripts/final_submission.conf ================================================ # This is not a real configuration file that you could use with ensemble_scripts # This is merely for the sake of understanding # The final configuration is ensemble_no21.conf, use it instead # video baseline models 1. video_relabel_combine_chain video level combine chain model, add the most confident tag predicted by the current model to the labels when loss<10.0 youtube-8m-zhangteng.video_level_models.MoeMix4Model #moe_mixtures is 4, #combination_layers is 3, #chain_relu_cells is 100 GAP in validate1 is 0.8088 at step 14863 2. video_very_deep_combine_chain # lstm baseline models 3. lstmmemory_cell1024_layer2_moe8 4. lstmmemory_cell2048_layer2_moe4 5. lstmparallelfinaloutput1024_moe8 6. lstmparallelmemory1024_moe8 7. lstmgate_cell1024_layer1_moe8 lstm and moe model, using scalar input and forget gate in lstm cell youtube-8m-zhangteng.frame_level_models.LstmGateModel #moe_mixtures is 8, #lstm_cells is 1024, #lstm_layers is 1 GAP in validate1 is 0.8148 at step 151107 8. lstmglu2_cell1024_layer1_moe8 lstm and moe model, using scalar input and forget gate in lstm cell, add forward immediate input memory (with its own input and forget gate) to lstm cell youtube-8m-zhangteng.frame_level_models.LstmGlu2Model #moe_mixtures is 8, #lstm_cells is 1024, #lstm_layers is 1 GAP in validate1 is 0.8152 at step 132334 9. lstmbiglu_cell1024_layer1_moe8 lstm and moe model, using scalar input and forget gate in lstm cell, add bi-directional immediate input memory (with its own input and forget gate) to lstm cell youtube-8m-zhangteng.frame_level_models.LstmBigluModel #moe_mixtures is 8, #lstm_cells is 1024, #lstm_layers is 1 GAP in validate1 is 0.8131 at step 130983 10. biunilstm1024_moe4 11. lstm_random_mean_moe8 lstm and moe model, randomly select half of all frames to create new samples to train youtube-8m-zhangteng.frame_level_models.LstmRandomModel + youtube-8m-zhangteng.video_level_models.MoeModel #moe_mixtures is 8, #lstm_cells is 1024, #lstm_layers is 2, augmentation times for each sample is 16 when eval and inference GAP in validate1 is 0.8145 at step 145850 12. lstm_shortlayers_moe8 lstm and moe model, divide sample into several clips with fixed frames, use one lstm to model intra-clip structure, use another lstm to model inter-clip structure youtube-8m-zhangteng.frame_level_models.LstmLayerModel + youtube-8m-zhangteng.video_level_models.MoeModel #moe_mixtures is 8, #lstm_cells is 1024, #lstm_layers is 2, #clip_length_frames is 10 GAP in validate1 is 0.8145 at step 107188 13. framehop_lstm 14. cnnlstmmemory1024_moe8 # attention models 15. attentionlstm_moe4 16. lstmattlstm1024_moe8 17. lstm_attention8_max lstm and moe model, use frame inputs and lstm outputs to calculate weight of each frame, then get the attention vector using a weighted sum of lstm outputs, then make predictions using moe model, repeat several times, take the maximum prediction of these times as the final prediction youtube-8m-zhangteng.frame_level_models.LstmExtendModel + youtube-8m-zhangteng.video_level_models.MoeExtendModel #moe_mixtures is 8, #lstm_cells is 1024, #lstm_layers is 2, #attention_repeat_times is 8 GAP in validate1 is 0.8177 at step 181785 18. lstm2_attention8_max lstm and moe model, use frame inputs and one lstm outputs to calculate weight of each frame, then get the attention vector using a weighted sum of another lstm outputs, then make predictions using moe model, repeat several times, take the maximum prediction of these times as the final prediction youtube-8m-zhangteng.frame_level_models.InputExtendModel + youtube-8m-zhangteng.video_level_models.MoeExtendModel #moe_mixtures is 8, #lstm_cells is 1024, #lstm_layers is 2, #attention_repeat_times is 4 GAP in validate1 is 0.8106 at step 197005 19. lstm_positional_attention8max # multiple sub model joined by deep combine chain 20. cnn_deep_combine_chain 21. deep_cnn_deep_combine_chain 22. lstm_cnn_deep_combine_chain 23. multilstmmemory1024_moe4_deep_chain 24. lstmmem1024_layer2_moe4_deep_combine_chain_add_length 25. multires_lstm_deep_combine_chain 26. lstm_gate_multiscale4_moe4 cnn-lstm and moe model, use a deep cnn to get multiscale representions at different layers, then use a lstm described in model #7 and moe model at each layer to make predictions, take the mean prediction of these layers as the final prediction youtube-8m-zhangteng.frame_level_models.LstmMultiscale2Model #moe_mixtures is 4, #cnn_layers is 4, #cnn_cells is [256,256,512], #cnn_fields is [1,2,3] , #lstm_cells is 1024, #lstm_layers is 1 GAP in validate1 is 0.8166 at step 193803 27. lstm_multiscale4_moe4 cnn-lstm and moe model, use a deep cnn to get multiscale representions at different layers, then use a basic lstm and moe model at each layer to make predictions, take the mean prediction of these layers as the final prediction youtube-8m-zhangteng.frame_level_models.LstmMultiscaleModel #moe_mixtures is 4, #cnn_layers is 4, #cnn_cells is [256,256,512], #cnn_fields is [1,2,3] , #lstm_cells is 1024, #lstm_layers is 1 GAP in validate1 is 0.8223 at step 231916 # bagging and boosting models 28. video_dcc_bagging 29. video_cc_structure_bagging/ensemble_matrix_model the video_cc_structure_bagging model is composed of four sub models: 29.1 video_notzero_combine_chain video level combine chain model, use both support_predictions and (1-support_predictions) as label dependence at each combination layer youtube-8m-zhangteng.video_level_models.MoeMix4Model #moe_mixtures is 4, #combination_layers is 3, #chain_relu_cells is 100 GAP in validate1 is 0.8099 at step 14668 29.2 video_weight_combine_chain video level combine chain model, multiply loss of positive labels by 10, while multiply loss of negative labels by 1 youtube-8m-zhangteng.video_level_models.MoeMix4Model #moe_mixtures is 4, #combination_layers is 3, #chain_relu_cells is 100 GAP in validate1 is 0.8079 at step 17415 29.3 video_knowledge_combine_chain video level combine chain model, use both support_predictions and top-level verticals of support_predictions as label dependence at each combination layer youtube-8m-zhangteng.video_level_models.MoeKnowledgeModel #moe_mixtures is 4, #combination_layers is 3, #chain_relu_cells is 100 GAP in validate1 is 0.8108 at step 9606 29.4 video_softmax_combine_chain video level combine chain model, use sigmiod model to predict the first 3000 labels, and use softmax model to predict the rest 1716 labels youtube-8m-zhangteng.video_level_models.MoeSoftmaxModel #moe_mixtures is 4, #combination_layers is 3, #chain_relu_cells is 100 GAP in validate1 is 0.8102 at step 9501 GAP of ensemble_matrix_model in validate2 is 30. video_dcc_boosting_discardhopeless/ensemble_matrix_model 31. video_dcc_boosting/ensemble_matrix_model 32. video_dcc_boosting_weightclip/ensemble_matrix_model 33. cnn_deep_combine_chain_bagging/ensemble_matrix_model 34. cnn_deep_combine_chain_boosting/ensemble_attention_matrix_model 35. lstmattention8max_bagging/ensemble_matrix_model 36. lstmattention8max_boosting/ensemble_attention_matrix_model 37. lstmparalleloutput_bagging/ensemble_mean_model 38. lstmparalleloutput_boosting_weightclip/ensemble_matrix_model # distillation models 39. distillation_video_dcc_noise 40. distillation_cnn_dcc_boosting/sub_model_1 41. distillation_lstmcnn_dcc_boosting/sub_model_1 42. distillation_lstmparalleloutput_boosting/sub_model_1 43. distillation_multilstm_dcc_boosting/sub_model_1 44. distillation_multiscale_cnnlstm_boosting/sub_model_1 45. distillation_video_dcc_bagging/ensemble_matrix_model 46. distillation_video_dcc_boosting/ensemble_matrix_model # cascade models 47. distillchain_video_dcc 48. distillchain_video_norm_moe8 video level moe model, use ditill labels as label dependence, normalize the label dependence with the summation of ditill labels youtube-8m-zhangteng.video_level_models.MoeDistillChainNorm2Model #moe_mixtures is 8, #chain_relu_cells is 256 GAP in validate1 is 0.8260 at step 13412 49. distillchain_lstm_gate_moe8 lstm and moe model, use lstm cell described in model #7 in frame level, use ditill labels as label dependence in video level youtube-8m-zhangteng.frame_level_models.LstmGateModel + youtube-8m-zhangteng.video_level_models.MoeDistillChainModel #moe_mixtures is 8, #chain_relu_cells is 256, #lstm_cells is 1024, #lstm_layers is 1 GAP in validate1 is 0.8293 at step 63344 50. distillchain_cnn_dcc 51. distillchain_cnndcc_layer2moe4 cnn and deep combine chain model, use ditill labels as label dependence only in the first combination layer youtube-8m-zhangteng.frame_level_models.CnnDCCDistillChainModel #moe_mixtures is 4, #chain_relu_cells is 256, #combination_layers=2 GAP in validate1 is 0.8241 at step 188273 52. distillchain_lstm_attention8max lstm and moe model as described in model #17, use ditill labels as label dependence on each attention output vector youtube-8m-zhangteng.frame_level_models.LstmExtendModel + youtube-8m-zhangteng.video_level_models.MoeExtendDistillChainModel #moe_mixtures is 8, #chain_relu_cells is 256, #attention_repeat_times is 8, #lstm_cells is 1024, #lstm_layers is 2 GAP in validate1 is 0.8298 at step 74796 53. distillchain_lstm_moe8 lstm and moe model as described in model #3, use ditill labels as label dependence in video level youtube-8m-zhangteng.frame_level_models.LstmModel + youtube-8m-zhangteng.video_level_models.MoeDistillChainModel #moe_mixtures is 8, #chain_relu_cells is 256, #lstm_cells is 1024, #lstm_layers is 2 GAP in validate1 is 0.8283 at step 49873 54. distillchain_lstm_multiscale_layer2_moe8 cnn-lstm and moe model as described in model #27, use ditill labels as label dependence in each cnn layer youtube-8m-zhangteng.frame_level_models.LstmMultiscaleDitillChainModel #moe_mixtures is 4, #chain_relu_cells is 256, #cnn_layers is 2, #cnn_cells is [256,256,512], #cnn_fields is [1,2,3] , #lstm_cells is 1024, #lstm_layers is 1 GAP in validate1 is 0.8313 at step 103725 55. distillchain_lstm_multiscale_layer4_moe8 cnn-lstm and moe model as described in model #27, use ditill labels as label dependence in each cnn layer youtube-8m-zhangteng.frame_level_models.LstmMultiscaleDitillChainModel #moe_mixtures is 4, #chain_relu_cells is 256, #cnn_layers is 4, #cnn_cells is [196,196,392], #cnn_fields is [1,2,3] , #lstm_cells is 1024, #lstm_layers is 1 GAP in validate1 is 0.8320 at step 102004 56. distillchain_lstmcnn_dcc 57. distillsplit_lstm_gate_moe8 lstm and moe model, use lstm cell described in model #7 in frame level, use the last 3716 ditill labels as label dependence in video level to redict the rest 1000 labels, concat the 3716 ditill labels and 1000 predicted labels as the final predictions youtube-8m-zhangteng.frame_level_models.LstmGateModel + youtube-8m-zhangteng.video_level_models.MoeDistillSplitModel #moe_mixtures is 8, #chain_relu_cells is 256, #lstm_cells is 1024, #lstm_layers is 1 GAP in validate1 is 0.8257 at step 130570 58. distillchain_v2_video_dcc 59. distillchain_v2_lstmattention8max 60. distillchain_v2_lstmcnn_dcc 61. distillchain_v2_lstm_gate_moe8 lstm and moe model as described in model #49, but on a new distillation training dataset youtube-8m-zhangteng.frame_level_models.LstmGateModel + youtube-8m-zhangteng.video_level_models.MoeDistillChainModel #moe_mixtures is 8, #chain_relu_cells is 256, #lstm_cells is 1024, #lstm_layers is 1 GAP in validate1 is 0.8273 at step 73400 62. distillchain_v2_lstm_glu2_moe8 lstm and moe model as described in model #49, but use lstm cell as described in model #8, and on a new distillation training dataset youtube-8m-zhangteng.frame_level_models.LstmGlu2Model + youtube-8m-zhangteng.video_level_models.MoeDistillChainModel #moe_mixtures is 8, #chain_relu_cells is 256, #lstm_cells is 1024, #lstm_layers is 1 GAP in validate1 is 0.8280 at step 72525 63. distillchain_v2_multilstm_dcc 64. distillchain_v2_multiscale_cnnlstm 65. distillchain_v2_lstmparalleloutput 66. distillchain_v2_boost_lstmparalleloutput 67. videochain/distillchain_v2_video_dcc_1 68. videochain/distillchain_v2_video_dcc_2 69. videochain/distillchain_v2_video_dcc_3 70. videochain/distillchain_v2_video_dcc_4 71. videochain/distillchain_v2_video_dcc_5 72. hybridchain/distillchain_v2_video_dcc 73. hybridchain/distillchain_v2_cnn_dcc 74. hybridchain/distillchain_v2_lstmparalleloutput ================================================ FILE: youtube-8m-ensemble/ensemble_scripts/infer-attention_linear_model.sh ================================================ model=$1 conf=$2 checkpoint=$3 test_path=/Youtube-8M/model_predictions/test test_data_patterns="" for d in $(cat $conf); do test_data_patterns="${test_path}/${d}/*.tfrecord${test_data_patterns:+,$test_data_patterns}" done echo "$test_data_patterns" input_data_pattern="${test_path}/model_input/*.tfrecord" #CUDA_VISIBLE_DEVICES=0 python inference.py \ python inference.py \ --model_checkpoint_path="../model/${model}/model.ckpt-${checkpoint}" \ --output_file="../model/${model}/predictions.${model}.csv" \ --model="AttentionLinearModel" \ --moe_num_mixtures=16 \ --batch_size=1024 \ --input_data_pattern="$input_data_pattern" \ --input_data_patterns="$test_data_patterns" ================================================ FILE: youtube-8m-ensemble/ensemble_scripts/infer-attention_linmatrix_model.sh ================================================ model=$1 conf=$2 checkpoint=$3 test_path=/Youtube-8M/model_predictions/test test_data_patterns="" for d in $(cat $conf); do test_data_patterns="${test_path}/${d}/*.tfrecord${test_data_patterns:+,$test_data_patterns}" done echo "$test_data_patterns" input_data_pattern="${test_path}/model_input/*.tfrecord" #CUDA_VISIBLE_DEVICES=0 python inference.py \ python inference.py \ --model_checkpoint_path="../model/${model}/model.ckpt-${checkpoint}" \ --output_file="../model/${model}/predictions.${model}.csv" \ --model="AttentionLinmatrixModel" \ --moe_num_mixtures=16 \ --attention_matrix_rank=4 \ --batch_size=1024 \ --input_data_pattern="$input_data_pattern" \ --input_data_patterns="$test_data_patterns" ================================================ FILE: youtube-8m-ensemble/ensemble_scripts/infer-attention_matrix_model.sh ================================================ model=$1 conf=$2 checkpoint=$3 test_path=/Youtube-8M/model_predictions/test test_data_patterns="" for d in $(cat $conf); do test_data_patterns="${test_path}/${d}/*.tfrecord${test_data_patterns:+,$test_data_patterns}" done echo "$test_data_patterns" input_data_pattern="${test_path}/model_input/*.tfrecord" #CUDA_VISIBLE_DEVICES=0 python inference.py \ python inference.py \ --model_checkpoint_path="../model/${model}/model.ckpt-${checkpoint}" \ --output_file="../model/${model}/predictions.${model}.csv" \ --model="AttentionMatrixModel" \ --moe_num_mixtures=4 \ --attention_matrix_rank=8 \ --batch_size=1024 \ --input_data_pattern="$input_data_pattern" \ --input_data_patterns="$test_data_patterns" ================================================ FILE: youtube-8m-ensemble/ensemble_scripts/infer-attention_moe_matrix_model.sh ================================================ model=$1 conf=$2 checkpoint=$3 test_path=/Youtube-8M/model_predictions/test test_data_patterns="" for d in $(cat $conf); do test_data_patterns="${test_path}/${d}/*.tfrecord${test_data_patterns:+,$test_data_patterns}" done echo "$test_data_patterns" input_data_pattern="${test_path}/model_input/*.tfrecord" #CUDA_VISIBLE_DEVICES=0 python inference.py \ python inference.py \ --model_checkpoint_path="../model/${model}/model.ckpt-${checkpoint}" \ --output_file="../model/${model}/predictions.${model}.csv" \ --model="AttentionMatrixModel" \ --moe_num_mixtures=4 \ --attention_relu_cells=256 \ --attention_matrix_rank=4 \ --batch_size=1024 \ --input_data_pattern="$input_data_pattern" \ --input_data_patterns="$test_data_patterns" ================================================ FILE: youtube-8m-ensemble/ensemble_scripts/infer-attention_moe_model.sh ================================================ model=$1 conf=$2 checkpoint=$3 test_path=/Youtube-8M/model_predictions/test test_data_patterns="" for d in $(cat $conf); do test_data_patterns="${test_path}/${d}/*.tfrecord${test_data_patterns:+,$test_data_patterns}" done echo "$test_data_patterns" input_data_pattern="${test_path}/model_input/*.tfrecord" #CUDA_VISIBLE_DEVICES=0 python inference.py \ python inference.py \ --model_checkpoint_path="../model/${model}/model.ckpt-${checkpoint}" \ --output_file="../model/${model}/predictions.${model}.csv" \ --model=AttentionMoeModel \ --moe_num_mixtures=8 \ --attention_relu_cells=128 \ --batch_size=1024 \ --input_data_pattern="$input_data_pattern" \ --input_data_patterns="$test_data_patterns" ================================================ FILE: youtube-8m-ensemble/ensemble_scripts/infer-attention_rectified_linear_model.sh ================================================ model=$1 conf=$2 checkpoint=$3 test_path=/Youtube-8M/model_predictions/test test_data_patterns="" for d in $(cat $conf); do test_data_patterns="${test_path}/${d}/*.tfrecord${test_data_patterns:+,$test_data_patterns}" done echo "$test_data_patterns" input_data_pattern="${test_path}/model_input/*.tfrecord" #CUDA_VISIBLE_DEVICES=0 python inference.py \ python inference.py \ --model_checkpoint_path="../model/${model}/model.ckpt-${checkpoint}" \ --output_file="../model/${model}/predictions.${model}.csv" \ --model="AttentionRectifiedLinearModel" \ --moe_num_mixtures=16 \ --batch_size=1024 \ --input_data_pattern="$input_data_pattern" \ --input_data_patterns="$test_data_patterns" ================================================ FILE: youtube-8m-ensemble/ensemble_scripts/infer-linear_model.sh ================================================ model=$1 conf=$2 checkpoint=$3 DEFAULT_GPU_ID=0 if [ -z ${CUDA_VISIBLE_DEVICES+x} ]; then GPU_ID=$DEFAULT_GPU_ID echo "set CUDA_VISIBLE_DEVICES to default('$GPU_ID')" else GPU_ID=$CUDA_VISIBLE_DEVICES echo "set CUDA_VISIBLE_DEVICES to external('$GPU_ID')" fi test_path=/Youtube-8M/model_predictions/test test_data_patterns="" for d in $(cat $conf); do test_data_patterns="${test_path}/${d}/*.tfrecord${test_data_patterns:+,$test_data_patterns}" done echo "$test_data_patterns" CUDA_VISIBLE_DEVICES=${GPU_ID} python inference.py \ --model_checkpoint_path="../model/${model}/model.ckpt-$checkpoint" \ --output_file="../model/${model}/predictions.csv" \ --model="LinearRegressionModel" \ --input_data_patterns="$test_data_patterns" ================================================ FILE: youtube-8m-ensemble/ensemble_scripts/infer-matrix_model.sh ================================================ model=$1 conf=$2 checkpoint=$3 DEFAULT_GPU_ID=0 if [ -z ${CUDA_VISIBLE_DEVICES+x} ]; then GPU_ID=$DEFAULT_GPU_ID echo "set CUDA_VISIBLE_DEVICES to default('$GPU_ID')" else GPU_ID=$CUDA_VISIBLE_DEVICES echo "set CUDA_VISIBLE_DEVICES to external('$GPU_ID')" fi test_path=/Youtube-8M/model_predictions/test test_data_patterns="" for d in $(cat $conf); do test_data_patterns="${test_path}/${d}/*.tfrecord${test_data_patterns:+,$test_data_patterns}" done echo "$test_data_patterns" CUDA_VISIBLE_DEVICES="$GPU_ID" python inference.py \ --model_checkpoint_path="../model/${model}/model.ckpt-${checkpoint}" \ --output_file="../model/${model}/predictions.csv" \ --model="MatrixRegressionModel" \ --batch_size=1024 \ --input_data_patterns="$test_data_patterns" ================================================ FILE: youtube-8m-ensemble/ensemble_scripts/infer-mean_model.sh ================================================ model=$1 conf=$2 DEFAULT_GPU_ID=0 if [ -z ${CUDA_VISIBLE_DEVICES+x} ]; then GPU_ID=$DEFAULT_GPU_ID echo "set CUDA_VISIBLE_DEVICES to default('$GPU_ID')" else GPU_ID=$CUDA_VISIBLE_DEVICES echo "set CUDA_VISIBLE_DEVICES to external('$GPU_ID')" fi test_path=/Youtube-8M/model_predictions/test test_data_patterns="" for d in $(cat $conf); do test_data_patterns="${test_path}/${d}/*.tfrecord${test_data_patterns:+,$test_data_patterns}" done echo "$test_data_patterns" CUDA_VISIBLE_DEVICES="$GPU_ID" python inference.py \ --model_checkpoint_path="../model/${model}/model.ckpt-0" \ --output_file="../model/${model}/predictions.csv" \ --model="MeanModel" \ --input_data_patterns="$test_data_patterns" ================================================ FILE: youtube-8m-ensemble/ensemble_scripts/infer-moe_model.sh ================================================ model=$1 conf=$2 DEFAULT_GPU_ID=0 if [ -z ${CUDA_VISIBLE_DEVICES+x} ]; then GPU_ID=$DEFAULT_GPU_ID echo "set CUDA_VISIBLE_DEVICES to default('$GPU_ID')" else GPU_ID=$CUDA_VISIBLE_DEVICES echo "set CUDA_VISIBLE_DEVICES to external('$GPU_ID')" fi test_path=/Youtube-8M/model_predictions/test test_data_patterns="" for d in $(cat $conf); do test_data_patterns="${test_path}/${d}/*.tfrecord${test_data_patterns:+,$test_data_patterns}" done echo "$test_data_patterns" CUDA_VISIBLE_DEVICES="$GPU_ID" python inference.py \ --model_checkpoint_path="../model/${model}/model.ckpt-2203" \ --output_file="../model/${model}/predictions.${model}.csv" \ --model="MoeModel" \ --input_data_patterns="$test_data_patterns" ================================================ FILE: youtube-8m-ensemble/ensemble_scripts/make-bagging-of-ensembles.sh ================================================ #!/bin/bash DEFAULT_GPU_ID=0 if [ -z ${CUDA_VISIBLE_DEVICES+x} ]; then GPU_ID=$DEFAULT_GPU_ID echo "set CUDA_VISIBLE_DEVICES to default('$GPU_ID')" else GPU_ID=$CUDA_VISIBLE_DEVICES echo "set CUDA_VISIBLE_DEVICES to external('$GPU_ID')" fi # base_model or sub_model_1 or sub_model_2 or so on model_type="$1" model_name="boe_ensemble_matrix_model_no10" MODEL_DIR="../model/${model_name}" main_conf="ensemble_scripts/ensemble_no10.conf" # generate vocab file vocab_file="resources/validate.video_id.vocab" default_freq_file="resources/validate.video_id.freq" if [ ! -f $vocab_file ]; then cd resources wget http://us.data.yt8m.org/1/ground_truth_labels/validate_labels.csv echo "OOV" > validate.video_id.vocab cat validate_labels.csv | cut -d ',' -f 1 >> validate.video_id.vocab cd .. fi vocab_checksum=$(md5sum $vocab_file | cut -d ' ' -f 1) if [ "$vocab_checksum" == "4a0c959434bb540f926d17a492161ca8" ]; then echo $vocab_file is valid else echo $vocab_file is corrupted exit 1 fi if [ ! -f $default_freq_file ]; then cat $vocab_file | awk '{print 1}' > $default_freq_file fi rm ${MODEL_DIR}/ensemble.conf mkdir $MODEL_DIR for i in {1..8}; do # sub model model_type="sub_model_${i}" sub_model_dir="${MODEL_DIR}/${model_type}" sub_conf="${sub_model_dir}/ensemble.conf" sub_freq="${sub_model_dir}/validate.video_id.freq" if [ ! -d $sub_model_dir ]; then mkdir -p $sub_model_dir # generate freq file python training_utils/sample_freq.py \ --video_id_file="$vocab_file" \ --output_freq_file="$sub_freq" # generate conf file python training_utils/sample_conf.py \ --main_conf_file="$main_conf" \ --sub_conf_file="$sub_conf" # train data patterns for ensemble training train_path=/Youtube-8M/model_predictions/ensemble_train sub_train_data_patterns="" for d in $(cat $sub_conf); do sub_train_data_patterns="${train_path}/${d}/*.tfrecord${sub_train_data_patterns:+,$sub_train_data_patterns}" done echo "$sub_train_data_patterns" # train N models with re-weighted samples CUDA_VISIBLE_DEVICES="$GPU_ID" python train.py \ --train_dir="$sub_model_dir" \ --train_data_patterns="$sub_train_data_patterns" \ --reweight=True \ --sample_vocab_file="$vocab_file" \ --sample_freq_file="$sub_freq" \ --model=MatrixRegressionModel \ --keep_checkpoint_every_n_hours=0.1 \ --batch_size=1024 \ --num_epochs=2 # inference-pre-ensemble for part in test ensemble_validate ensemble_train; do output_dir="/Youtube-8M/model_predictions/${part}/${model_name}/${model_type}" if [ ! -d $output_dir ]; then # test data patterns test_path=/Youtube-8M/model_predictions/${part} sub_test_data_patterns="" for d in $(cat $sub_conf); do sub_test_data_patterns="${test_path}/${d}/*.tfrecord${sub_test_data_patterns:+,$sub_test_data_patterns}" done echo "$sub_test_data_patterns" CUDA_VISIBLE_DEVICES="$GPU_ID" python inference-pre-ensemble.py \ --output_dir="$output_dir" \ --train_dir="$sub_model_dir" \ --input_data_patterns="$sub_test_data_patterns" \ --model="MatrixRegressionModel" \ --batch_size=1024 \ --file_size=4096 fi done fi echo "${model_name}/${model_type}" >> ${MODEL_DIR}/ensemble.conf done # on ensemble server bash ensemble_scripts/train-attention_matrix_model.sh ${model_name}/ensemble_attention_matrix_model ${MODEL_DIR}/ensemble.conf bash ensemble_scripts/eval-attention_matrix_model.sh ${model_name}/ensemble_attention_matrix_model ${MODEL_DIR}/ensemble.conf # bash ensemble_scripts/infer-mean_model.sh ${model_name}/ensemble_mean_model ${MODEL_DIR}/ensemble.conf ================================================ FILE: youtube-8m-ensemble/ensemble_scripts/make-virtual-groups.sh ================================================ #!/bin/bash group_dir="../model/virtual_grouping" for file in $(ls $group_dir); do if [[ $file =~ ^virtual_group ]] && [[ $file =~ conf$ ]]; then group_name="${file:0:-5}" model="virtual_grouping/${group_name}" model_dir="${group_dir}/${group_name}" if [ ! -d $model_dir ]; then conf="${group_dir}/${group_name}.conf" echo "training $model ..." bash ensemble_scripts/train-matrix_model.sh $model $conf bash ensemble_scripts/eval-matrix_model.sh $model $conf fi for part in ensemble_train ensemble_validate test; do output_dir="/Youtube-8M/model_predictions/${part}/${model}" if [ ! -d $output_dir ]; then bash ensemble_scripts/auto-preensemble-matrix_model.sh $model $conf $part fi done fi done model_dir="${group_dir}/ensemble_matrix_model" if [ ! -d $model_dir ]; then model="virtual_grouping/ensemble_matrix_model" conf="${group_dir}/ensemble.conf" echo "training $model ..." bash ensemble_scripts/train-matrix_model.sh $model $conf bash ensemble_scripts/eval-matrix_model.sh $model $conf fi ================================================ FILE: youtube-8m-ensemble/ensemble_scripts/preensemble-attention_matrix_model.sh ================================================ model=$1 conf=$2 part=$3 checkpoint=$4 DEFAULT_GPU_ID=0 if [ -z ${CUDA_VISIBLE_DEVICES+x} ]; then GPU_ID=$DEFAULT_GPU_ID echo "set CUDA_VISIBLE_DEVICES to default('$GPU_ID')" else GPU_ID=$CUDA_VISIBLE_DEVICES echo "set CUDA_VISIBLE_DEVICES to external('$GPU_ID')" fi test_path=/Youtube-8M/model_predictions/${part} test_data_patterns="" for d in $(cat $conf); do test_data_patterns="${test_path}/${d}/*.tfrecord${test_data_patterns:+,$test_data_patterns}" done echo "$test_data_patterns" input_data_pattern="${test_path}/model_input/*.tfrecord" CUDA_VISIBLE_DEVICES="$GPU_ID" python inference-pre-ensemble.py \ --output_dir="/Youtube-8M/model_predictions/${part}/${model}" \ --model_checkpoint_path="../model/${model}/model.ckpt-${checkpoint}" \ --input_data_patterns="$test_data_patterns" \ --input_data_pattern="$input_data_pattern" \ --model="AttentionMatrixModel" \ --moe_num_mixtures=4 \ --attention_matrix_rank=8 \ --batch_size=1024 \ --file_size=4096 ================================================ FILE: youtube-8m-ensemble/ensemble_scripts/preensemble-matrix_model.sh ================================================ model=$1 conf=$2 part=$3 checkpoint=$4 DEFAULT_GPU_ID=0 if [ -z ${CUDA_VISIBLE_DEVICES+x} ]; then GPU_ID=$DEFAULT_GPU_ID echo "set CUDA_VISIBLE_DEVICES to default('$GPU_ID')" else GPU_ID=$CUDA_VISIBLE_DEVICES echo "set CUDA_VISIBLE_DEVICES to external('$GPU_ID')" fi test_path=/Youtube-8M/model_predictions/${part} test_data_patterns="" for d in $(cat $conf); do test_data_patterns="${test_path}/${d}/*.tfrecord${test_data_patterns:+,$test_data_patterns}" done echo "$test_data_patterns" CUDA_VISIBLE_DEVICES="$GPU_ID" python inference-pre-ensemble.py \ --output_dir="/Youtube-8M/model_predictions/${part}/${model}" \ --model_checkpoint_path="../model/${model}/model.ckpt-${checkpoint}" \ --input_data_patterns="$test_data_patterns" \ --model="MatrixRegressionModel" \ --batch_size=1024 \ --file_size=4096 ================================================ FILE: youtube-8m-ensemble/ensemble_scripts/preensemble-mean_model.sh ================================================ model=$1 conf=$2 part=$3 DEFAULT_GPU_ID=0 if [ -z ${CUDA_VISIBLE_DEVICES+x} ]; then GPU_ID=$DEFAULT_GPU_ID echo "set CUDA_VISIBLE_DEVICES to default('$GPU_ID')" else GPU_ID=$CUDA_VISIBLE_DEVICES echo "set CUDA_VISIBLE_DEVICES to external('$GPU_ID')" fi test_path=/Youtube-8M/model_predictions/${part} test_data_patterns="" for d in $(cat $conf); do test_data_patterns="${test_path}/${d}/*.tfrecord${test_data_patterns:+,$test_data_patterns}" done echo "$test_data_patterns" CUDA_VISIBLE_DEVICES="$GPU_ID" python inference-pre-ensemble.py \ --output_dir="/Youtube-8M/model_predictions/${part}/${model}" \ --model_checkpoint_path="../model/${model}/model.ckpt-0" \ --input_data_patterns="$test_data_patterns" \ --model="MeanModel" \ --batch_size=1024 \ --file_size=4096 ================================================ FILE: youtube-8m-ensemble/ensemble_scripts/train-attention_linear_model.sh ================================================ model=$1 conf=$2 DEFAULT_GPU_ID=0 if [ -z ${CUDA_VISIBLE_DEVICES+x} ]; then GPU_ID=$DEFAULT_GPU_ID echo "set CUDA_VISIBLE_DEVICES to default('$GPU_ID')" else GPU_ID=$CUDA_VISIBLE_DEVICES echo "set CUDA_VISIBLE_DEVICES to external('$GPU_ID')" fi train_path=/Youtube-8M/model_predictions/ensemble_train train_data_patterns="" for d in $(cat $conf); do train_data_patterns="${train_path}/${d}/*.tfrecord${train_data_patterns:+,$train_data_patterns}" done echo $train_data_patterns input_data_pattern="${train_path}/model_input/*.tfrecord" CUDA_VISIBLE_DEVICES="$GPU_ID" python train.py \ --train_dir="../model/${model}" \ --train_data_patterns="$train_data_patterns" \ --input_data_pattern="$input_data_pattern" \ --model=AttentionLinearModel \ --base_learning_rate=0.01 \ --moe_num_mixtures=16 \ --keep_checkpoint_every_n_hours=0.1 \ --batch_size=1024 \ --num_epochs=4 ================================================ FILE: youtube-8m-ensemble/ensemble_scripts/train-attention_linmatrix_model.sh ================================================ model=$1 conf=$2 DEFAULT_GPU_ID=0 if [ -z ${CUDA_VISIBLE_DEVICES+x} ]; then GPU_ID=$DEFAULT_GPU_ID echo "set CUDA_VISIBLE_DEVICES to default('$GPU_ID')" else GPU_ID=$CUDA_VISIBLE_DEVICES echo "set CUDA_VISIBLE_DEVICES to external('$GPU_ID')" fi train_path=/Youtube-8M/model_predictions/ensemble_train train_data_patterns="" for d in $(cat $conf); do train_data_patterns="${train_path}/${d}/*.tfrecord${train_data_patterns:+,$train_data_patterns}" done echo $train_data_patterns input_data_pattern="${train_path}/model_input/*.tfrecord" CUDA_VISIBLE_DEVICES="$GPU_ID" python train.py \ --train_dir="../model/${model}" \ --train_data_patterns="$train_data_patterns" \ --input_data_pattern="$input_data_pattern" \ --model=AttentionLinmatrixModel \ --base_learning_rate=0.004 \ --moe_num_mixtures=16 \ --attention_matrix_rank=4 \ --keep_checkpoint_every_n_hours=0.1 \ --batch_size=1024 \ --num_epochs=4 ================================================ FILE: youtube-8m-ensemble/ensemble_scripts/train-attention_matrix_model.sh ================================================ model=$1 conf=$2 DEFAULT_GPU_ID=0 if [ -z ${CUDA_VISIBLE_DEVICES+x} ]; then GPU_ID=$DEFAULT_GPU_ID echo "set CUDA_VISIBLE_DEVICES to default('$GPU_ID')" else GPU_ID=$CUDA_VISIBLE_DEVICES echo "set CUDA_VISIBLE_DEVICES to external('$GPU_ID')" fi train_path=/Youtube-8M/model_predictions/ensemble_train train_data_patterns="" for d in $(cat $conf); do train_data_patterns="${train_path}/${d}/*.tfrecord${train_data_patterns:+,$train_data_patterns}" done echo $train_data_patterns input_data_pattern="${train_path}/model_input/*.tfrecord" CUDA_VISIBLE_DEVICES="$GPU_ID" python train.py \ --train_dir="../model/${model}" \ --train_data_patterns="$train_data_patterns" \ --input_data_pattern="$input_data_pattern" \ --model=AttentionMatrixModel \ --base_learning_rate=0.004 \ --moe_num_mixtures=4 \ --attention_matrix_rank=8 \ --keep_checkpoint_every_n_hours=0.1 \ --batch_size=1024 \ --num_epochs=4 ================================================ FILE: youtube-8m-ensemble/ensemble_scripts/train-attention_moe_matrix_model.sh ================================================ model=$1 conf=$2 DEFAULT_GPU_ID=0 if [ -z ${CUDA_VISIBLE_DEVICES+x} ]; then GPU_ID=$DEFAULT_GPU_ID echo "set CUDA_VISIBLE_DEVICES to default('$GPU_ID')" else GPU_ID=$CUDA_VISIBLE_DEVICES echo "set CUDA_VISIBLE_DEVICES to external('$GPU_ID')" fi train_path=/Youtube-8M/model_predictions/ensemble_train train_data_patterns="" for d in $(cat $conf); do train_data_patterns="${train_path}/${d}/*.tfrecord${train_data_patterns:+,$train_data_patterns}" done echo $train_data_patterns input_data_pattern="${train_path}/model_input/*.tfrecord" CUDA_VISIBLE_DEVICES="$GPU_ID" python train.py \ --train_dir="../model/${model}" \ --train_data_patterns="$train_data_patterns" \ --input_data_pattern="$input_data_pattern" \ --model=AttentionMatrixModel \ --base_learning_rate=0.01 \ --moe_num_mixtures=4 \ --attention_relu_cells=256 \ --attention_matrix_rank=4 \ --keep_checkpoint_every_n_hours=0.1 \ --batch_size=1024 \ --num_epochs=4 ================================================ FILE: youtube-8m-ensemble/ensemble_scripts/train-attention_moe_model.sh ================================================ model=$1 conf=$2 DEFAULT_GPU_ID=0 if [ -z ${CUDA_VISIBLE_DEVICES+x} ]; then GPU_ID=$DEFAULT_GPU_ID echo "set CUDA_VISIBLE_DEVICES to default('$GPU_ID')" else GPU_ID=$CUDA_VISIBLE_DEVICES echo "set CUDA_VISIBLE_DEVICES to external('$GPU_ID')" fi train_path=/Youtube-8M/model_predictions/ensemble_train train_data_patterns="" for d in $(cat $conf); do train_data_patterns="${train_path}/${d}/*.tfrecord${train_data_patterns:+,$train_data_patterns}" done echo $train_data_patterns input_data_pattern="${train_path}/model_input/*.tfrecord" CUDA_VISIBLE_DEVICES="$GPU_ID" python train.py \ --train_dir="../model/${model}" \ --train_data_patterns="$train_data_patterns" \ --input_data_pattern="$input_data_pattern" \ --model=AttentionMoeModel \ --moe_num_mixtures=8 \ --attention_relu_cells=128 \ --base_learning_rate=0.01 \ --keep_checkpoint_every_n_hours=0.1 \ --batch_size=1024 \ --num_epochs=4 ================================================ FILE: youtube-8m-ensemble/ensemble_scripts/train-attention_rectified_linear_model.sh ================================================ model=$1 conf=$2 DEFAULT_GPU_ID=0 if [ -z ${CUDA_VISIBLE_DEVICES+x} ]; then GPU_ID=$DEFAULT_GPU_ID echo "set CUDA_VISIBLE_DEVICES to default('$GPU_ID')" else GPU_ID=$CUDA_VISIBLE_DEVICES echo "set CUDA_VISIBLE_DEVICES to external('$GPU_ID')" fi train_path=/Youtube-8M/model_predictions/ensemble_train train_data_patterns="" for d in $(cat $conf); do train_data_patterns="${train_path}/${d}/*.tfrecord${train_data_patterns:+,$train_data_patterns}" done echo $train_data_patterns input_data_pattern="${train_path}/model_input/*.tfrecord" CUDA_VISIBLE_DEVICES="$GPU_ID" python train.py \ --train_dir="../model/${model}" \ --train_data_patterns="$train_data_patterns" \ --input_data_pattern="$input_data_pattern" \ --model=AttentionRectifiedLinearModel \ --base_learning_rate=0.01 \ --moe_num_mixtures=16 \ --keep_checkpoint_every_n_hours=0.1 \ --batch_size=1024 \ --num_epochs=4 ================================================ FILE: youtube-8m-ensemble/ensemble_scripts/train-deep_combine_chain_model.sh ================================================ model=$1 conf=$2 DEFAULT_GPU_ID=0 if [ -z ${CUDA_VISIBLE_DEVICES+x} ]; then GPU_ID=$DEFAULT_GPU_ID echo "set CUDA_VISIBLE_DEVICES to default('$GPU_ID')" else GPU_ID=$CUDA_VISIBLE_DEVICES echo "set CUDA_VISIBLE_DEVICES to external('$GPU_ID')" fi train_path=/Youtube-8M/model_predictions/ensemble_train train_data_patterns="" labels="" for d in $(cat $conf); do train_data_patterns="${train_path}/${d}/*.tfrecord${train_data_patterns:+,$train_data_patterns}" labels="label,${labels:+,$labels}" done echo $train_data_patterns input_data_pattern="${train_path}/model_input/*.tfrecord" CUDA_VISIBLE_DEVICES="$GPU_ID" python train.py \ --train_dir="../model/${model}" \ --train_data_patterns="$train_data_patterns" \ --input_data_pattern="$input_data_pattern" \ --model="DeepCombineChainModel" \ --deep_chain_relu_cells=128 \ --moe_num_mixtures=2 \ --multitask=True \ --label_loss=MultiTaskCrossEntropyLoss \ --support_loss_percent=0.2 \ --support_type="label" \ --base_learning_rate=0.001 \ --keep_checkpoint_every_n_hours=0.1 \ --batch_size=128 \ --num_epochs=4 ================================================ FILE: youtube-8m-ensemble/ensemble_scripts/train-input_moe_model.sh ================================================ model=$1 conf=$2 DEFAULT_GPU_ID=0 if [ -z ${CUDA_VISIBLE_DEVICES+x} ]; then GPU_ID=$DEFAULT_GPU_ID echo "set CUDA_VISIBLE_DEVICES to default('$GPU_ID')" else GPU_ID=$CUDA_VISIBLE_DEVICES echo "set CUDA_VISIBLE_DEVICES to external('$GPU_ID')" fi train_path=/Youtube-8M/model_predictions/ensemble_train train_data_patterns="" for d in $(cat $conf); do train_data_patterns="${train_path}/${d}/*.tfrecord${train_data_patterns:+,$train_data_patterns}" done echo $train_data_patterns input_data_pattern="${train_path}/model_input/*.tfrecord" CUDA_VISIBLE_DEVICES="$GPU_ID" python train.py \ --train_dir="../model/${model}" \ --train_data_patterns="$train_data_patterns" \ --input_data_pattern="$input_data_pattern" \ --model=InputMoeModel \ --keep_checkpoint_every_n_hours=0.1 \ --batch_size=1024 \ --num_epochs=5 ================================================ FILE: youtube-8m-ensemble/ensemble_scripts/train-linear_model.sh ================================================ model=$1 conf=$2 DEFAULT_GPU_ID=0 if [ -z ${CUDA_VISIBLE_DEVICES+x} ]; then GPU_ID=$DEFAULT_GPU_ID echo "set CUDA_VISIBLE_DEVICES to default('$GPU_ID')" else GPU_ID=$CUDA_VISIBLE_DEVICES echo "set CUDA_VISIBLE_DEVICES to external('$GPU_ID')" fi train_path=/Youtube-8M/model_predictions/ensemble_train train_data_patterns="" for d in $(cat $conf); do train_data_patterns="${train_path}/${d}/*.tfrecord${train_data_patterns:+,$train_data_patterns}" done echo $train_data_patterns CUDA_VISIBLE_DEVICES="$GPU_ID" python train.py \ --train_dir="../model/${model}" \ --train_data_patterns="$train_data_patterns" \ --model=LinearRegressionModel \ --keep_checkpoint_every_n_hours=0.1 \ --batch_size=1024 \ --num_epochs=1 ================================================ FILE: youtube-8m-ensemble/ensemble_scripts/train-matrix_model.sh ================================================ model=$1 conf=$2 postfix=$3 DEFAULT_GPU_ID=0 if [ -z ${CUDA_VISIBLE_DEVICES+x} ]; then GPU_ID=$DEFAULT_GPU_ID echo "set CUDA_VISIBLE_DEVICES to default('$GPU_ID')" else GPU_ID=$CUDA_VISIBLE_DEVICES echo "set CUDA_VISIBLE_DEVICES to external('$GPU_ID')" fi train_path=/Youtube-8M/model_predictions${postfix}/ensemble_train train_data_patterns="" for d in $(cat $conf); do train_data_patterns="${train_path}/${d}/*.tfrecord${train_data_patterns:+,$train_data_patterns}" done echo "$train_data_patterns" CUDA_VISIBLE_DEVICES="$GPU_ID" python train.py \ --train_dir="../model/${model}" \ --train_data_patterns="$train_data_patterns" \ --model=MatrixRegressionModel \ --keep_checkpoint_every_n_hours=0.1 \ --batch_size=1024 \ --num_epochs=2 ================================================ FILE: youtube-8m-ensemble/ensemble_scripts/train-matrix_model_lr.sh ================================================ model=$1 conf=$2 learn_rate=$3 DEFAULT_GPU_ID=0 if [ -z ${CUDA_VISIBLE_DEVICES+x} ]; then GPU_ID=$DEFAULT_GPU_ID echo "set CUDA_VISIBLE_DEVICES to default('$GPU_ID')" else GPU_ID=$CUDA_VISIBLE_DEVICES echo "set CUDA_VISIBLE_DEVICES to external('$GPU_ID')" fi train_path=/Youtube-8M/model_predictions/ensemble_train train_data_patterns="" for d in $(cat $conf); do train_data_patterns="${train_path}/${d}/*.tfrecord${train_data_patterns:+,$train_data_patterns}" done echo "$train_data_patterns" CUDA_VISIBLE_DEVICES="$GPU_ID" python train.py \ --train_dir="../model/${model}" \ --train_data_patterns="$train_data_patterns" \ --model=MatrixRegressionModel \ --base_learning_rate=$learn_rate \ --keep_checkpoint_every_n_hours=0.1 \ --batch_size=1024 \ --num_epochs=2 ================================================ FILE: youtube-8m-ensemble/ensemble_scripts/train-mean_model.sh ================================================ model=$1 conf=$2 DEFAULT_GPU_ID=0 if [ -z ${CUDA_VISIBLE_DEVICES+x} ]; then GPU_ID=$DEFAULT_GPU_ID echo "set CUDA_VISIBLE_DEVICES to default('$GPU_ID')" else GPU_ID=$CUDA_VISIBLE_DEVICES echo "set CUDA_VISIBLE_DEVICES to external('$GPU_ID')" fi train_path=/Youtube-8M/model_predictions/ensemble_train train_data_patterns="" for d in $(cat $conf); do train_data_patterns="${train_path}/${d}/*.tfrecord${train_data_patterns:+,$train_data_patterns}" done echo $train_data_patterns CUDA_VISIBLE_DEVICES="$GPU_ID" python train.py \ --train_dir="../model/${model}" \ --train_data_patterns="$train_data_patterns" \ --model=MeanModel \ --training=False \ --num_epochs=1 ================================================ FILE: youtube-8m-ensemble/ensemble_scripts/train-moe_model.sh ================================================ model=$1 conf=$2 DEFAULT_GPU_ID=0 if [ -z ${CUDA_VISIBLE_DEVICES+x} ]; then GPU_ID=$DEFAULT_GPU_ID echo "set CUDA_VISIBLE_DEVICES to default('$GPU_ID')" else GPU_ID=$CUDA_VISIBLE_DEVICES echo "set CUDA_VISIBLE_DEVICES to external('$GPU_ID')" fi train_path=/Youtube-8M/model_predictions/ensemble_train train_data_patterns="" for d in $(cat $conf); do train_data_patterns="${train_path}/${d}/*.tfrecord${train_data_patterns:+,$train_data_patterns}" done echo $train_data_patterns CUDA_VISIBLE_DEVICES="$GPU_ID" python train.py \ --train_dir="../model/${model}" \ --train_data_patterns="$train_data_patterns" \ --model=MoeModel \ --keep_checkpoint_every_n_hours=0.1 \ --batch_size=1024 \ --num_epochs=5 ================================================ FILE: youtube-8m-ensemble/ensemble_scripts/train-nonunit_matrix_model.sh ================================================ model=$1 conf=$2 DEFAULT_GPU_ID=0 if [ -z ${CUDA_VISIBLE_DEVICES+x} ]; then GPU_ID=$DEFAULT_GPU_ID echo "set CUDA_VISIBLE_DEVICES to default('$GPU_ID')" else GPU_ID=$CUDA_VISIBLE_DEVICES echo "set CUDA_VISIBLE_DEVICES to external('$GPU_ID')" fi train_path=/Youtube-8M/model_predictions/ensemble_train train_data_patterns="" for d in $(cat $conf); do train_data_patterns="${train_path}/${d}/*.tfrecord${train_data_patterns:+,$train_data_patterns}" done echo $train_data_patterns CUDA_VISIBLE_DEVICES="$GPU_ID" python train.py \ --train_dir="../model/${model}" \ --train_data_patterns="$train_data_patterns" \ --model=NonunitMatrixRegressionModel \ --keep_checkpoint_every_n_hours=0.1 \ --batch_size=1024 \ --num_epochs=5 ================================================ FILE: youtube-8m-ensemble/eval.py ================================================ # Copyright 2016 Google Inc. All Rights Reserved. # # Licensed under the Apache License, Version 2.0 (the "License"); # you may not use this file except in compliance with the License. # You may obtain a copy of the License at # # http://www.apache.org/licenses/LICENSE-2.0 # # Unless required by applicable law or agreed to in writing, software # distributed under the License is distributed on an "AS-IS" BASIS, # WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied. # See the License for the specific language governing permissions and # limitations under the License. """Binary for evaluating Tensorflow models on the YouTube-8M dataset.""" import time import eval_util import losses import ensemble_level_models import readers import tensorflow as tf from tensorflow import app from tensorflow import flags from tensorflow import gfile from tensorflow import logging import utils FLAGS = flags.FLAGS if __name__ == "__main__": # Dataset flags. flags.DEFINE_string("model_checkpoint_path", "", "The file to load the model files from. ") flags.DEFINE_string("train_dir", "/tmp/yt8m/", "The directory to write the result in. ") flags.DEFINE_string( "eval_data_patterns", "", "File globs defining the evaluation dataset in tensorflow.SequenceExample format.") flags.DEFINE_string( "input_data_pattern", None, "File globs for original model input.") flags.DEFINE_string("feature_names", "predictions", "Name of the feature " "to use for training.") flags.DEFINE_string("feature_sizes", "4716", "Length of the feature vectors.") # Model flags. flags.DEFINE_string( "model", "LinearRegressionModel", "Which architecture to use for the model.") flags.DEFINE_integer("batch_size", 256, "How many examples to process per batch.") flags.DEFINE_string("label_loss", "CrossEntropyLoss", "Loss computed on validation data") # Other flags. flags.DEFINE_boolean("run_once", True, "Whether to run eval only once.") flags.DEFINE_boolean("echo_gap", False, "Whether to echo GAP at the end.") flags.DEFINE_integer("top_k", 20, "How many predictions to output per video.") def find_class_by_name(name, modules): """Searches the provided modules for the named class and returns it.""" modules = [getattr(module, name, None) for module in modules] return next(a for a in modules if a) def get_input_evaluation_tensors(reader, data_pattern, batch_size=256): logging.info("Using batch size of " + str(batch_size) + " for evaluation.") with tf.name_scope("eval_input"): files = gfile.Glob(data_pattern) if not files: print data_pattern, files raise IOError("Unable to find the evaluation files.") logging.info("number of evaluation files: " + str(len(files))) files.sort() filename_queue = tf.train.string_input_producer( files, shuffle=False, num_epochs=1) eval_data = reader.prepare_reader(filename_queue) return tf.train.batch( eval_data, batch_size=batch_size, capacity=3 * batch_size, allow_smaller_final_batch=True, enqueue_many=True) def build_graph(all_readers, input_reader, all_eval_data_patterns, input_data_pattern, model, label_loss_fn, batch_size=256): """Creates the Tensorflow graph for evaluation. Args: all_readers: The data file reader. It should inherit from BaseReader. model: The core model (e.g. logistic or neural net). It should inherit from BaseModel. all_eval_data_patterns: glob path to the evaluation data files. label_loss_fn: What kind of loss to apply to the model. It should inherit from BaseLoss. batch_size: How many examples to process at a time. """ global_step = tf.Variable(0, trainable=False, name="global_step") model_input_raw_tensors = [] labels_batch_tensor = None video_id_batch = None for reader, data_pattern in zip(all_readers, all_eval_data_patterns): unused_video_id, model_input_raw, labels_batch, unused_num_frames = ( get_input_evaluation_tensors( reader, data_pattern, batch_size=batch_size)) if labels_batch_tensor is None: labels_batch_tensor = labels_batch if video_id_batch is None: video_id_batch = unused_video_id model_input_raw_tensors.append(tf.expand_dims(model_input_raw, axis=2)) original_input = None if input_data_pattern is not None: unused_video_id, original_input, unused_labels_batch, unused_num_frames = ( get_input_evaluation_tensors( input_reader, input_data_pattern, batch_size=batch_size)) model_input = tf.concat(model_input_raw_tensors, axis=2) labels_batch = labels_batch_tensor with tf.name_scope("model"): result = model.create_model(model_input, labels=labels_batch, vocab_size=reader.num_classes, original_input=original_input, is_training=False) predictions = result["predictions"] tf.summary.histogram("model_activations", predictions) if "loss" in result.keys(): label_loss = result["loss"] else: label_loss = label_loss_fn.calculate_loss(predictions, labels_batch) tf.add_to_collection("global_step", global_step) tf.add_to_collection("loss", label_loss) tf.add_to_collection("predictions", predictions) tf.add_to_collection("input_batch", model_input) tf.add_to_collection("video_id_batch", video_id_batch) tf.add_to_collection("labels", tf.cast(labels_batch, tf.float32)) tf.add_to_collection("summary_op", tf.summary.merge_all()) def evaluation_loop(video_id_batch, prediction_batch, label_batch, loss, summary_op, saver, summary_writer, evl_metrics, last_global_step_val): """Run the evaluation loop once. Args: video_id_batch: a tensor of video ids mini-batch. prediction_batch: a tensor of predictions mini-batch. label_batch: a tensor of label_batch mini-batch. loss: a tensor of loss for the examples in the mini-batch. summary_op: a tensor which runs the tensorboard summary operations. saver: a tensorflow saver to restore the model. summary_writer: a tensorflow summary_writer evl_metrics: an EvaluationMetrics object. last_global_step_val: the global step used in the previous evaluation. Returns: The global_step used in the latest model. """ global_step_val = -1 with tf.Session() as sess: checkpoint = FLAGS.model_checkpoint_path if checkpoint: logging.info("Loading checkpoint for eval: " + checkpoint) # Restores from checkpoint saver.restore(sess, checkpoint) # Assuming model_checkpoint_path looks something like: # /my-favorite-path/yt8m_train/model.ckpt-0, extract global_step from it. global_step_val = checkpoint.split("/")[-1].split("-")[-1] else: logging.info("No checkpoint file found.") return global_step_val sess.run([tf.local_variables_initializer()]) # Start the queue runners. fetches = [video_id_batch, prediction_batch, label_batch, loss, summary_op] coord = tf.train.Coordinator() try: threads = [] for qr in tf.get_collection(tf.GraphKeys.QUEUE_RUNNERS): threads.extend(qr.create_threads( sess, coord=coord, daemon=True, start=True)) logging.info("enter eval_once loop global_step_val = %s. ", global_step_val) evl_metrics.clear() examples_processed = 0 while not coord.should_stop(): batch_start_time = time.time() _, predictions_val, labels_val, loss_val, summary_val = sess.run(fetches) seconds_per_batch = time.time() - batch_start_time example_per_second = labels_val.shape[0] / seconds_per_batch examples_processed += labels_val.shape[0] iteration_info_dict = evl_metrics.accumulate(predictions_val, labels_val, loss_val) iteration_info_dict["examples_per_second"] = example_per_second iterinfo = utils.AddGlobalStepSummary( summary_writer, global_step_val, iteration_info_dict, summary_scope="Eval") logging.info("examples_processed: %d | %s", examples_processed, iterinfo) except tf.errors.OutOfRangeError as e: logging.info( "Done with batched inference. Now calculating global performance " "metrics.") # calculate the metrics for the entire epoch epoch_info_dict = evl_metrics.get() epoch_info_dict["epoch_id"] = global_step_val summary_writer.add_summary(summary_val, global_step_val) epochinfo = utils.AddEpochSummary( summary_writer, global_step_val, epoch_info_dict, summary_scope="Eval") if FLAGS.echo_gap: print "GAP =", epoch_info_dict["gap"] logging.info(epochinfo) evl_metrics.clear() except Exception as e: # pylint: disable=broad-except logging.info("Unexpected exception: " + str(e)) coord.request_stop(e) coord.request_stop() coord.join(threads, stop_grace_period_secs=10) return global_step_val def evaluate(): tf.set_random_seed(0) # for reproducibility with tf.Graph().as_default(): # convert feature_names and feature_sizes to lists of values feature_names, feature_sizes = utils.GetListOfFeatureNamesAndSizes( FLAGS.feature_names, FLAGS.feature_sizes) # prepare a reader for each single model prediction result all_readers = [] all_patterns = FLAGS.eval_data_patterns all_patterns = map(lambda x: x.strip(), all_patterns.strip().strip(",").split(",")) for i in xrange(len(all_patterns)): reader = readers.EnsembleReader( feature_names=feature_names, feature_sizes=feature_sizes) all_readers.append(reader) input_reader = None input_data_pattern = None if FLAGS.input_data_pattern is not None: input_reader = readers.EnsembleReader( feature_names=["input"], feature_sizes=[1024+128]) input_data_pattern = FLAGS.input_data_pattern # find the model model = find_class_by_name(FLAGS.model, [ensemble_level_models])() label_loss_fn = find_class_by_name(FLAGS.label_loss, [losses])() if FLAGS.eval_data_patterns is "": raise IOError("'eval_data_patterns' was not specified. " + "Nothing to evaluate.") build_graph( all_readers=all_readers, input_reader=input_reader, all_eval_data_patterns=all_patterns, input_data_pattern=input_data_pattern, model=model, label_loss_fn=label_loss_fn, batch_size=FLAGS.batch_size) logging.info("built evaluation graph") video_id_batch = tf.get_collection("video_id_batch")[0] prediction_batch = tf.get_collection("predictions")[0] label_batch = tf.get_collection("labels")[0] loss = tf.get_collection("loss")[0] summary_op = tf.get_collection("summary_op")[0] saver = tf.train.Saver(tf.global_variables()) summary_writer = tf.summary.FileWriter( FLAGS.train_dir, graph=tf.get_default_graph()) evl_metrics = eval_util.EvaluationMetrics(FLAGS.num_classes, FLAGS.top_k) last_global_step_val = -1 last_global_step_val = evaluation_loop(video_id_batch, prediction_batch, label_batch, loss, summary_op, saver, summary_writer, evl_metrics, last_global_step_val) def main(unused_argv): logging.set_verbosity(tf.logging.INFO) print("tensorflow version: %s" % tf.__version__) evaluate() if __name__ == "__main__": app.run() ================================================ FILE: youtube-8m-ensemble/eval_util.py ================================================ # Copyright 2016 Google Inc. All Rights Reserved. # # Licensed under the Apache License, Version 2.0 (the "License"); # you may not use this file except in compliance with the License. # You may obtain a copy of the License at # # http://www.apache.org/licenses/LICENSE-2.0 # # Unless required by applicable law or agreed to in writing, software # distributed under the License is distributed on an "AS-IS" BASIS, # WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied. # See the License for the specific language governing permissions and # limitations under the License. """Provides functions to help with evaluating models.""" import datetime import numpy from tensorflow.python.platform import gfile import mean_average_precision_calculator as map_calculator import average_precision_calculator as ap_calculator def flatten(l): """ Merges a list of lists into a single list. """ return [item for sublist in l for item in sublist] def calculate_hit_at_one(predictions, actuals): """Performs a local (numpy) calculation of the hit at one. Args: predictions: Matrix containing the outputs of the model. Dimensions are 'batch' x 'num_classes'. actuals: Matrix containing the ground truth labels. Dimensions are 'batch' x 'num_classes'. Returns: float: The average hit at one across the entire batch. """ top_prediction = numpy.argmax(predictions, 1) hits = actuals[numpy.arange(actuals.shape[0]), top_prediction] return numpy.average(hits) def calculate_recall_at_n(predictions, actuals, n): """Performs a local (numpy) calculation of the recall@n Args: predictions: Matrix containing the outputs of the model. Dimensions are 'batch' x 'num_classes'. actuals: Matrix containing the ground truth labels. Dimensions are 'batch' x 'num_classes'. n: scalar of n Returns: float: The recall at n across the entire batch. """ aggregated_recall = 0.0 num_videos = actuals.shape[0] for row in numpy.arange(num_videos): num_labels = int(numpy.sum(actuals[row])) top_indices = numpy.argpartition(predictions[row], -n)[-n:] item_recall = 0.0 for label_index in top_indices: if predictions[row][label_index] > 0: item_recall += actuals[row][label_index] item_recall /= num_labels aggregated_recall += item_recall aggregated_recall /= num_videos return aggregated_recall def calculate_precision_at_equal_recall_rate(predictions, actuals): """Performs a local (numpy) calculation of the PERR. Args: predictions: Matrix containing the outputs of the model. Dimensions are 'batch' x 'num_classes'. actuals: Matrix containing the ground truth labels. Dimensions are 'batch' x 'num_classes'. Returns: float: The average precision at equal recall rate across the entire batch. """ aggregated_precision = 0.0 num_videos = actuals.shape[0] for row in numpy.arange(num_videos): num_labels = int(numpy.sum(actuals[row])) top_indices = numpy.argpartition(predictions[row], -num_labels)[-num_labels:] item_precision = 0.0 for label_index in top_indices: if predictions[row][label_index] > 0: item_precision += actuals[row][label_index] item_precision /= top_indices.size aggregated_precision += item_precision aggregated_precision /= num_videos return aggregated_precision def calculate_gap(predictions, actuals, top_k=20): """Performs a local (numpy) calculation of the global average precision. Only the top_k predictions are taken for each of the videos. Args: predictions: Matrix containing the outputs of the model. Dimensions are 'batch' x 'num_classes'. actuals: Matrix containing the ground truth labels. Dimensions are 'batch' x 'num_classes'. top_k: How many predictions to use per video. Returns: float: The global average precision. """ gap_calculator = ap_calculator.AveragePrecisionCalculator() sparse_predictions, sparse_labels, num_positives = top_k_by_class(predictions, actuals, top_k) gap_calculator.accumulate(flatten(sparse_predictions), flatten(sparse_labels), sum(num_positives)) return gap_calculator.peek_ap_at_n() def top_k_by_class(predictions, labels, k=20): """Extracts the top k predictions for each video, sorted by class. Args: predictions: A numpy matrix containing the outputs of the model. Dimensions are 'batch' x 'num_classes'. k: the top k non-zero entries to preserve in each prediction. Returns: A tuple (predictions,labels, true_positives). 'predictions' and 'labels' are lists of lists of floats. 'true_positives' is a list of scalars. The length of the lists are equal to the number of classes. The entries in the predictions variable are probability predictions, and the corresponding entries in the labels variable are the ground truth for those predictions. The entries in 'true_positives' are the number of true positives for each class in the ground truth. Raises: ValueError: An error occurred when the k is not a positive integer. """ if k <= 0: raise ValueError("k must be a positive integer.") k = min(k, predictions.shape[1]) num_classes = predictions.shape[1] prediction_triplets= [] for video_index in range(predictions.shape[0]): prediction_triplets.extend(top_k_triplets(predictions[video_index],labels[video_index], k)) out_predictions = [[] for v in range(num_classes)] out_labels = [[] for v in range(num_classes)] for triplet in prediction_triplets: out_predictions[triplet[0]].append(triplet[1]) out_labels[triplet[0]].append(triplet[2]) out_true_positives = [numpy.sum(labels[:,i]) for i in range(num_classes)] return out_predictions, out_labels, out_true_positives def top_k_triplets(predictions, labels, k=20): """Get the top_k for a 1-d numpy array. Returns a sparse list of tuples in (prediction, class) format""" m = len(predictions) k = min(k, m) indices = numpy.argpartition(predictions, -k)[-k:] return [(index, predictions[index], labels[index]) for index in indices] class EvaluationMetrics(object): """A class to store the evaluation metrics.""" def __init__(self, num_class, top_k): """Construct an EvaluationMetrics object to store the evaluation metrics. Args: num_class: A positive integer specifying the number of classes. top_k: A positive integer specifying how many predictions are considered per video. Raises: ValueError: An error occurred when MeanAveragePrecisionCalculator cannot not be constructed. """ self.sum_hit_at_one = 0.0 self.sum_perr = 0.0 self.sum_loss = 0.0 self.map_calculator = map_calculator.MeanAveragePrecisionCalculator(num_class) self.global_ap_calculator = ap_calculator.AveragePrecisionCalculator() self.top_k = top_k self.num_examples = 0 def accumulate(self, predictions, labels, loss): """Accumulate the metrics calculated locally for this mini-batch. Args: predictions: A numpy matrix containing the outputs of the model. Dimensions are 'batch' x 'num_classes'. labels: A numpy matrix containing the ground truth labels. Dimensions are 'batch' x 'num_classes'. loss: A numpy array containing the loss for each sample. Returns: dictionary: A dictionary storing the metrics for the mini-batch. Raises: ValueError: An error occurred when the shape of predictions and actuals does not match. """ batch_size = labels.shape[0] mean_hit_at_one = calculate_hit_at_one(predictions, labels) mean_perr = calculate_precision_at_equal_recall_rate(predictions, labels) mean_loss = numpy.mean(loss) # Take the top 20 predictions. sparse_predictions, sparse_labels, num_positives = top_k_by_class(predictions, labels, self.top_k) self.map_calculator.accumulate(sparse_predictions, sparse_labels, num_positives) self.global_ap_calculator.accumulate(flatten(sparse_predictions), flatten(sparse_labels), sum(num_positives)) self.num_examples += batch_size self.sum_hit_at_one += mean_hit_at_one * batch_size self.sum_perr += mean_perr * batch_size self.sum_loss += mean_loss * batch_size return {"hit_at_one": mean_hit_at_one, "perr": mean_perr, "loss": mean_loss} def get(self): """Calculate the evaluation metrics for the whole epoch. Raises: ValueError: If no examples were accumulated. Returns: dictionary: a dictionary storing the evaluation metrics for the epoch. The dictionary has the fields: avg_hit_at_one, avg_perr, avg_loss, and aps (default nan). """ if self.num_examples <= 0: raise ValueError("total_sample must be positive.") avg_hit_at_one = self.sum_hit_at_one / self.num_examples avg_perr = self.sum_perr / self.num_examples avg_loss = self.sum_loss / self.num_examples aps = self.map_calculator.peek_map_at_n() gap = self.global_ap_calculator.peek_ap_at_n() epoch_info_dict = {} return {"avg_hit_at_one": avg_hit_at_one, "avg_perr": avg_perr, "avg_loss": avg_loss, "aps": aps, "gap": gap} def clear(self): """Clear the evaluation metrics and reset the EvaluationMetrics object.""" self.sum_hit_at_one = 0.0 self.sum_perr = 0.0 self.sum_loss = 0.0 self.map_calculator.clear() self.global_ap_calculator.clear() self.num_examples = 0 ================================================ FILE: youtube-8m-ensemble/feature_transform.py ================================================ from tensorflow import flags flags.DEFINE_string("feature_transformer", "DefaultTransformer", "how to preprocess feature, defaults to identical, which means no transform") flags.DEFINE_string("engineer_types", "identical,avg,std,diff", "how to preprocess feature, defaults to identical, which means no transform") flags.DEFINE_integer("time_resolution", 8, "how many frames to mean pooling at a time") from all_feature_transform import * ================================================ FILE: youtube-8m-ensemble/inference-combine-tfrecords-frame.py ================================================ # Copyright 2016 Google Inc. All Rights Reserved. # # Licensed under the Apache License, Version 2.0 (the "License"); # you may not use this file except in compliance with the License. # You may obtain a copy of the License at # # http://www.apache.org/licenses/LICENSE-2.0 # # Unless required by applicable law or agreed to in writing, software # distributed under the License is distributed on an "AS-IS" BASIS, # WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied. # See the License for the specific language governing permissions and # limitations under the License. """Binary for combine model output and model input into one set of files.""" import os import sys import time import numpy import numpy as np import tensorflow as tf from tensorflow import app from tensorflow import flags from tensorflow import gfile from tensorflow import logging import utils import eval_util import losses import readers import ensemble_level_models FLAGS = flags.FLAGS if __name__ == '__main__': flags.DEFINE_string("output_dir", "", "The file to save the predictions to.") flags.DEFINE_string( "input_data_pattern", "", "File globs defining the input dataset in tensorflow.SequenceExample format.") flags.DEFINE_string( "prediction_data_pattern", "", "File globs defining the output dataset in tensorflow.SequenceExample format.") flags.DEFINE_string("input_feature_names", "mean_rgb,mean_audio", "Name of the feature " "to use for training.") flags.DEFINE_string("input_feature_sizes", "1024,128", "Length of the feature vectors.") flags.DEFINE_string("prediction_feature_names", "predictions", "Name of the feature " "to use for training.") flags.DEFINE_string("prediction_feature_sizes", "4716", "Length of the feature vectors.") flags.DEFINE_integer("batch_size", 256, "How many examples to process per batch.") flags.DEFINE_integer("file_size", 4096, "Number of frames per batch for DBoF.") flags.DEFINE_integer("file_num_mod", None, "file_num % 3 == file_num_mod will be output.") def find_class_by_name(name, modules): """Searches the provided modules for the named class and returns it.""" modules = [getattr(module, name, None) for module in modules] return next(a for a in modules if a) def get_input_data_tensors(reader, data_pattern, batch_size=256): logging.info("Using batch size of " + str(batch_size) + " for evaluation.") with tf.name_scope("eval_input"): files = gfile.Glob(data_pattern) if not files: raise IOError("Unable to find the evaluation files.") logging.info("number of evaluation files: " + str(len(files))) files.sort() filename_queue = tf.train.string_input_producer( files, shuffle=False, num_epochs=1) eval_data = reader.prepare_reader(filename_queue) return tf.train.batch( eval_data, batch_size=batch_size, capacity=4 * batch_size, allow_smaller_final_batch=True, enqueue_many=True) def build_graph(input_reader, input_data_pattern, prediction_reader, prediction_data_pattern, batch_size=256): """Creates the Tensorflow graph for evaluation. Args: all_readers: The data file reader. It should inherit from BaseReader. model: The core model (e.g. logistic or neural net). It should inherit from BaseModel. all_data_patterns: glob path to the evaluation data files. label_loss_fn: What kind of loss to apply to the model. It should inherit from BaseLoss. batch_size: How many examples to process at a time. """ video_ids_batch, model_rgbs_batch, model_audios_batch, labels_batch, num_frames_batch = ( get_input_data_tensors( input_reader, input_data_pattern, batch_size=batch_size)) video_ids_batch2, model_predictions_batch, labels_batch2, unused_num_frames2 = ( get_input_data_tensors( prediction_reader, prediction_data_pattern, batch_size=batch_size)) tf.add_to_collection("video_ids_batch", video_ids_batch) tf.add_to_collection("labels_batch", tf.cast(labels_batch, tf.float32)) tf.add_to_collection("rgbs_batch", model_rgbs_batch) tf.add_to_collection("audios_batch", model_audios_batch) tf.add_to_collection("predictions_batch", model_predictions_batch) tf.add_to_collection("num_frames_batch", num_frames_batch) def inference_loop(video_ids_batch, labels_batch, rgbs_batch, audios_batch, predictions_batch, num_frames_batch, output_dir, batch_size): with tf.Session() as sess: sess.run([tf.local_variables_initializer()]) # Start the queue runners. fetches = [video_ids_batch, labels_batch, rgbs_batch, audios_batch, predictions_batch, num_frames_batch] coord = tf.train.Coordinator() start_time = time.time() video_ids = [] video_labels = [] video_rgbs = [] video_audios = [] video_predictions = [] video_num_frames = [] filenum = 0 num_examples_processed = 0 total_num_examples_processed = 0 directory = FLAGS.output_dir if not os.path.exists(directory): os.makedirs(directory) try: threads = [] for qr in tf.get_collection(tf.GraphKeys.QUEUE_RUNNERS): threads.extend(qr.create_threads( sess, coord=coord, daemon=True, start=True)) while not coord.should_stop(): ids_val = None ids_val, labels_val, rgbs_val, audios_val, predictions_val, num_frames_val = sess.run(fetches) if filenum % 3 == FLAGS.file_num_mod or FLAGS.file_num_mod is None: video_ids.append(ids_val) video_labels.append(labels_val) video_rgbs.append(rgbs_val) video_audios.append(audios_val) video_predictions.append(predictions_val) video_num_frames.append(num_frames_val) num_examples_processed += len(ids_val) ids_shape = ids_val.shape[0] rgbs_shape = rgbs_val.shape[0] audios_shape = audios_val.shape[0] predictions_shape = predictions_val.shape[0] assert ids_shape == rgbs_shape == predictions_shape, "tensor ids(%d), rgbs(%d) and predictions(%d) should have equal rows" % (ids_shape, rgbs_shape, predictions_shape) ids_val = None now = time.time() logging.info("num examples processed: " + str(num_examples_processed) + " elapsed seconds: " + "{0:.2f}".format(now-start_time)) if num_examples_processed >= FLAGS.file_size: assert num_examples_processed==FLAGS.file_size, "num_examples_processed should be equal to %d"%FLAGS.file_size if filenum % 3 == FLAGS.file_num_mod or FLAGS.file_num_mod is None: video_ids = np.concatenate(video_ids, axis=0) video_labels = np.concatenate(video_labels, axis=0) video_rgbs = np.concatenate(video_rgbs, axis=0) video_audios = np.concatenate(video_audios, axis=0) video_predictions = np.concatenate(video_predictions, axis=0) video_num_frames = np.concatenate(video_num_frames, axis=0) write_to_record(video_ids, video_labels, video_rgbs, video_audios, video_predictions, video_num_frames, filenum, num_examples_processed) video_ids = [] video_labels = [] video_rgbs = [] video_audios = [] video_predictions = [] video_num_frames = [] filenum += 1 total_num_examples_processed += num_examples_processed now = time.time() logging.info(str(num_examples_processed) + " examples written to file elapsed seconds: " + "{0:.2f}".format(now-start_time)) num_examples_processed = 0 except tf.errors.OutOfRangeError as e: if ids_val is not None: video_ids.append(ids_val) video_labels.append(labels_val) video_rgbs.append(rgbs_val) video_audios.append(audios_val) video_predictions.append(predictions_val) video_num_frames.append(num_frames_val) num_examples_processed += len(ids_val) if 0 < num_examples_processed <= FLAGS.file_size: if filenum % 3 == FLAGS.file_num_mod or FLAGS.file_num_mod is None: video_ids = np.concatenate(video_ids, axis=0) video_labels = np.concatenate(video_labels, axis=0) video_rgbs = np.concatenate(video_rgbs, axis=0) video_audios = np.concatenate(video_audios, axis=0) video_predictions = np.concatenate(video_predictions, axis=0) video_num_frames = np.concatenate(video_num_frames, axis=0) write_to_record(video_ids, video_labels, video_rgbs, video_audios, video_predictions, video_num_frames, filenum, num_examples_processed) total_num_examples_processed += num_examples_processed now = time.time() logging.info("num examples processed: " + str(num_examples_processed) + " elapsed seconds: " + "{0:.2f}".format(now-start_time)) num_examples_processed = 0 logging.info("Done with inference. %d samples was written to %s" % (total_num_examples_processed, FLAGS.output_dir)) except Exception as e: # pylint: disable=broad-except logging.info("Unexpected exception: " + str(e)) finally: coord.request_stop() coord.join(threads, stop_grace_period_secs=10) def write_to_record(video_ids, video_labels, video_rgbs, video_audios, video_predictions, video_num_frames, filenum, num_examples_processed): writer = tf.python_io.TFRecordWriter(FLAGS.output_dir + '/' + 'predictions-%04d.tfrecord' % filenum) for i in range(num_examples_processed): video_id = video_ids[i] video_label = np.nonzero(video_labels[i,:])[0] video_rgb = video_rgbs[i,:] video_audio = video_audios[i,:] video_prediction = video_predictions[i,:] video_num_frame = video_num_frames[i] example = get_output_feature(video_id, video_label, video_rgb, video_audio, video_prediction, video_num_frame) serialized = example.SerializeToString() writer.write(serialized) writer.close() def get_output_feature(video_id, video_label, video_rgb, video_audio, video_prediction, video_num_frame): _bytes_feature_list = lambda x: tf.train.Feature(bytes_list=tf.train.BytesList(value=[x])) example = tf.train.SequenceExample( context = tf.train.Features(feature={ "video_id": tf.train.Feature(bytes_list=tf.train.BytesList(value=[video_id])), "labels": tf.train.Feature(int64_list=tf.train.Int64List(value=video_label)), "predictions": tf.train.Feature(float_list=tf.train.FloatList(value=video_prediction)) }), feature_lists = tf.train.FeatureLists(feature_list={ "rgb": tf.train.FeatureList(feature=map(_bytes_feature_list, video_rgb[:video_num_frame])), "audio": tf.train.FeatureList(feature=map(_bytes_feature_list, video_audio[:video_num_frame])), }) ) return example def main(unused_argv): logging.set_verbosity(tf.logging.INFO) with tf.Graph().as_default(): if FLAGS.input_data_pattern is "": raise IOError("'input_data_pattern' was not specified. " + "Nothing to evaluate.") if FLAGS.prediction_data_pattern is "": raise IOError("'prediction_data_pattern' was not specified. " + "Nothing to evaluate.") # convert feature_names and feature_sizes to lists of values input_feature_names, input_feature_sizes = utils.GetListOfFeatureNamesAndSizes( FLAGS.input_feature_names, FLAGS.input_feature_sizes) input_reader = readers.EnsembleFrameReader( feature_names=input_feature_names, feature_sizes=input_feature_sizes) prediction_feature_names, prediction_feature_sizes = utils.GetListOfFeatureNamesAndSizes( FLAGS.prediction_feature_names, FLAGS.prediction_feature_sizes) prediction_reader = readers.EnsembleReader( feature_names=prediction_feature_names, feature_sizes=prediction_feature_sizes) build_graph( input_reader=input_reader, prediction_reader=prediction_reader, input_data_pattern=FLAGS.input_data_pattern, prediction_data_pattern=FLAGS.prediction_data_pattern, batch_size=FLAGS.batch_size) logging.info("built evaluation graph") video_ids_batch = tf.get_collection("video_ids_batch")[0] labels_batch = tf.get_collection("labels_batch")[0] rgbs_batch = tf.get_collection("rgbs_batch")[0] audios_batch = tf.get_collection("audios_batch")[0] predictions_batch = tf.get_collection("predictions_batch")[0] num_frames_batch = tf.get_collection("num_frames_batch")[0] inference_loop(video_ids_batch, labels_batch, rgbs_batch, audios_batch, predictions_batch, num_frames_batch, FLAGS.output_dir, FLAGS.batch_size) if __name__ == "__main__": app.run() ================================================ FILE: youtube-8m-ensemble/inference-combine-tfrecords-video.py ================================================ # Copyright 2016 Google Inc. All Rights Reserved. # # Licensed under the Apache License, Version 2.0 (the "License"); # you may not use this file except in compliance with the License. # You may obtain a copy of the License at # # http://www.apache.org/licenses/LICENSE-2.0 # # Unless required by applicable law or agreed to in writing, software # distributed under the License is distributed on an "AS-IS" BASIS, # WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied. # See the License for the specific language governing permissions and # limitations under the License. """Binary for combine model output and model input into one set of files.""" import os import time import numpy import numpy as np import tensorflow as tf from tensorflow import app from tensorflow import flags from tensorflow import gfile from tensorflow import logging import utils import eval_util import losses import readers import ensemble_level_models FLAGS = flags.FLAGS if __name__ == '__main__': flags.DEFINE_string("output_dir", "", "The file to save the predictions to.") flags.DEFINE_string( "input_data_pattern", "", "File globs defining the input dataset in tensorflow.SequenceExample format.") flags.DEFINE_string( "prediction_data_pattern", "", "File globs defining the output dataset in tensorflow.SequenceExample format.") flags.DEFINE_string("input_feature_names", "mean_rgb,mean_audio", "Name of the feature " "to use for training.") flags.DEFINE_string("input_feature_sizes", "1024,128", "Length of the feature vectors.") flags.DEFINE_string("prediction_feature_names", "predictions", "Name of the feature " "to use for training.") flags.DEFINE_string("prediction_feature_sizes", "4716", "Length of the feature vectors.") flags.DEFINE_integer("batch_size", 256, "How many examples to process per batch.") flags.DEFINE_integer("file_size", 4096, "Number of frames per batch for DBoF.") def find_class_by_name(name, modules): """Searches the provided modules for the named class and returns it.""" modules = [getattr(module, name, None) for module in modules] return next(a for a in modules if a) def get_input_data_tensors(reader, data_pattern, batch_size=256): logging.info("Using batch size of " + str(batch_size) + " for evaluation.") with tf.name_scope("eval_input"): files = gfile.Glob(data_pattern) if not files: raise IOError("Unable to find the evaluation files.") logging.info("number of evaluation files: " + str(len(files))) files.sort() filename_queue = tf.train.string_input_producer( files, shuffle=False, num_epochs=1) eval_data = reader.prepare_reader(filename_queue) return tf.train.batch( eval_data, batch_size=batch_size, capacity=batch_size, allow_smaller_final_batch=True, enqueue_many=True) def build_graph(input_reader, input_data_pattern, prediction_reader, prediction_data_pattern, batch_size=256): """Creates the Tensorflow graph for evaluation. Args: all_readers: The data file reader. It should inherit from BaseReader. model: The core model (e.g. logistic or neural net). It should inherit from BaseModel. all_data_patterns: glob path to the evaluation data files. label_loss_fn: What kind of loss to apply to the model. It should inherit from BaseLoss. batch_size: How many examples to process at a time. """ video_ids_batch, model_inputs_batch, labels_batch, unused_num_frames = ( get_input_data_tensors( input_reader, input_data_pattern, batch_size=batch_size)) video_ids_batch2, model_predictions_batch, labels_batch2, unused_num_frames2 = ( get_input_data_tensors( prediction_reader, prediction_data_pattern, batch_size=batch_size)) video_ids_equal = tf.reduce_mean(tf.cast(tf.equal(video_ids_batch, video_ids_batch2), tf.float32)) labels_equal = tf.reduce_mean(tf.reduce_sum(tf.cast(tf.equal(labels_batch, labels_batch2), tf.float32), axis=1)) tf.add_to_collection("video_ids_equal", video_ids_equal) tf.add_to_collection("labels_equal", labels_equal) tf.add_to_collection("video_ids_batch", video_ids_batch) tf.add_to_collection("labels_batch", tf.cast(labels_batch, tf.float32)) tf.add_to_collection("inputs_batch", model_inputs_batch) tf.add_to_collection("predictions_batch", model_predictions_batch) def inference_loop(video_ids_batch, labels_batch, inputs_batch, predictions_batch, video_ids_equal, labels_equal, output_dir, batch_size): with tf.Session() as sess: sess.run([tf.local_variables_initializer()]) # Start the queue runners. fetches = [video_ids_batch, labels_batch, inputs_batch, predictions_batch, video_ids_equal, labels_equal] coord = tf.train.Coordinator() start_time = time.time() video_ids = [] video_labels = [] video_inputs = [] video_predictions = [] filenum = 0 num_examples_processed = 0 total_num_examples_processed = 0 directory = FLAGS.output_dir if not os.path.exists(directory): os.makedirs(directory) else: raise IOError("Output path exists! path='" + directory + "'") try: threads = [] for qr in tf.get_collection(tf.GraphKeys.QUEUE_RUNNERS): threads.extend(qr.create_threads( sess, coord=coord, daemon=True, start=True)) while not coord.should_stop(): ids_val = None ids_val, labels_val, inputs_val, predictions_val, ids_equal_val, labels_equal_val = sess.run(fetches) print "ids equal = %f" % (ids_equal_val) print "labels equal = %f" % (labels_equal_val) video_ids.append(ids_val) video_labels.append(labels_val) video_inputs.append(inputs_val) video_predictions.append(predictions_val) num_examples_processed += len(ids_val) ids_shape = ids_val.shape[0] inputs_shape = inputs_val.shape[0] predictions_shape = predictions_val.shape[0] assert ids_shape == inputs_shape == predictions_shape, "tensor ids(%d), inputs(%d) and predictions(%d) should have equal rows" % (ids_shape, inputs_shape, predictions_shape) ids_val = None if num_examples_processed >= FLAGS.file_size: assert num_examples_processed==FLAGS.file_size, "num_examples_processed should be equal to %d"%FLAGS.file_size video_ids = np.concatenate(video_ids, axis=0) video_labels = np.concatenate(video_labels, axis=0) video_inputs = np.concatenate(video_inputs, axis=0) video_predictions = np.concatenate(video_predictions, axis=0) write_to_record(video_ids, video_labels, video_inputs, video_predictions, filenum, num_examples_processed) video_ids = [] video_labels = [] video_inputs = [] video_predictions = [] filenum += 1 total_num_examples_processed += num_examples_processed now = time.time() logging.info("num examples processed: " + str(num_examples_processed) + " elapsed seconds: " + "{0:.2f}".format(now-start_time)) num_examples_processed = 0 except tf.errors.OutOfRangeError as e: if ids_val is not None: video_ids.append(ids_val) video_labels.append(labels_val) video_inputs.append(inputs_val) video_predictions.append(predictions_val) num_examples_processed += len(ids_val) if 0 < num_examples_processed <= FLAGS.file_size: video_ids = np.concatenate(video_ids, axis=0) video_labels = np.concatenate(video_labels, axis=0) video_inputs = np.concatenate(video_inputs, axis=0) video_predictions = np.concatenate(video_predictions, axis=0) write_to_record(video_ids, video_labels, video_inputs, video_predictions, filenum, num_examples_processed) total_num_examples_processed += num_examples_processed now = time.time() logging.info("num examples processed: " + str(num_examples_processed) + " elapsed seconds: " + "{0:.2f}".format(now-start_time)) num_examples_processed = 0 logging.info("Done with inference. %d samples was written to %s" % (total_num_examples_processed, FLAGS.output_dir)) except Exception as e: # pylint: disable=broad-except logging.info("Unexpected exception: " + str(e)) finally: coord.request_stop() coord.join(threads, stop_grace_period_secs=10) def write_to_record(video_ids, video_labels, video_inputs, video_predictions, filenum, num_examples_processed): writer = tf.python_io.TFRecordWriter(FLAGS.output_dir + '/' + 'predictions-%04d.tfrecord' % filenum) for i in range(num_examples_processed): video_id = video_ids[i] video_label = np.nonzero(video_labels[i,:])[0] video_input = video_inputs[i,:] video_prediction = video_predictions[i,:] example = get_output_feature(video_id, video_label, video_input, video_prediction) serialized = example.SerializeToString() writer.write(serialized) writer.close() def get_output_feature(video_id, video_label, video_input, video_prediction): feature_maps = {'video_id': tf.train.Feature(bytes_list=tf.train.BytesList(value=[video_id])), 'labels': tf.train.Feature(int64_list=tf.train.Int64List(value=video_label))} input_feature_names = FLAGS.input_feature_names.split(",") input_feature_sizes = map(int, FLAGS.input_feature_sizes.split(",")) input_feature_start = 0 for i in range(len(input_feature_names)): feature_maps[input_feature_names[i]] = tf.train.Feature( float_list=tf.train.FloatList( value=video_input[ input_feature_start : input_feature_start + input_feature_sizes[i]])) input_feature_start += input_feature_sizes[i] prediction_feature_names = FLAGS.prediction_feature_names.split(",") prediction_feature_sizes = map(int, FLAGS.prediction_feature_sizes.split(",")) prediction_feature_start = 0 for i in range(len(prediction_feature_names)): feature_maps[prediction_feature_names[i]] = tf.train.Feature( float_list=tf.train.FloatList( value=video_prediction[ prediction_feature_start : prediction_feature_start + prediction_feature_sizes[i]])) prediction_feature_start += prediction_feature_sizes[i] example = tf.train.Example(features=tf.train.Features(feature=feature_maps)) return example def main(unused_argv): logging.set_verbosity(tf.logging.INFO) with tf.Graph().as_default(): if FLAGS.input_data_pattern is "": raise IOError("'input_data_pattern' was not specified. " + "Nothing to evaluate.") if FLAGS.prediction_data_pattern is "": raise IOError("'prediction_data_pattern' was not specified. " + "Nothing to evaluate.") # convert feature_names and feature_sizes to lists of values input_feature_names, input_feature_sizes = utils.GetListOfFeatureNamesAndSizes( FLAGS.input_feature_names, FLAGS.input_feature_sizes) input_reader = readers.EnsembleReader( feature_names=input_feature_names, feature_sizes=input_feature_sizes) prediction_feature_names, prediction_feature_sizes = utils.GetListOfFeatureNamesAndSizes( FLAGS.prediction_feature_names, FLAGS.prediction_feature_sizes) prediction_reader = readers.EnsembleReader( feature_names=prediction_feature_names, feature_sizes=prediction_feature_sizes) build_graph( input_reader=input_reader, prediction_reader=prediction_reader, input_data_pattern=FLAGS.input_data_pattern, prediction_data_pattern=FLAGS.prediction_data_pattern, batch_size=FLAGS.batch_size) logging.info("built evaluation graph") video_ids_equal = tf.get_collection("video_ids_equal")[0] labels_equal = tf.get_collection("labels_equal")[0] video_ids_batch = tf.get_collection("video_ids_batch")[0] labels_batch = tf.get_collection("labels_batch")[0] inputs_batch = tf.get_collection("inputs_batch")[0] predictions_batch = tf.get_collection("predictions_batch")[0] inference_loop(video_ids_batch, labels_batch, inputs_batch, predictions_batch, video_ids_equal, labels_equal, FLAGS.output_dir, FLAGS.batch_size) if __name__ == "__main__": app.run() ================================================ FILE: youtube-8m-ensemble/inference-pre-ensemble.py ================================================ # Copyright 2016 Google Inc. All Rights Reserved. # # Licensed under the Apache License, Version 2.0 (the "License"); # you may not use this file except in compliance with the License. # You may obtain a copy of the License at # # http://www.apache.org/licenses/LICENSE-2.0 # # Unless required by applicable law or agreed to in writing, software # distributed under the License is distributed on an "AS-IS" BASIS, # WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied. # See the License for the specific language governing permissions and # limitations under the License. """Binary for generating predictions over a set of videos.""" import os import time import numpy import numpy as np import tensorflow as tf from tensorflow import app from tensorflow import flags from tensorflow import gfile from tensorflow import logging import utils import eval_util import losses import readers import ensemble_level_models FLAGS = flags.FLAGS if __name__ == '__main__': flags.DEFINE_string("model_checkpoint_path", None, "The file path to load the model from.") flags.DEFINE_string("train_dir", "", "The directory to load the model from.") flags.DEFINE_string("output_dir", "", "The file to save the predictions to.") flags.DEFINE_string( "input_data_patterns", "", "File globs defining the evaluation dataset in tensorflow.SequenceExample format.") flags.DEFINE_string( "input_data_pattern", None, "File globs for original model input.") flags.DEFINE_string("feature_names", "predictions", "Name of the feature " "to use for training.") flags.DEFINE_string("feature_sizes", "4716", "Length of the feature vectors.") # Model flags. flags.DEFINE_string( "model", "MeanModel", "Which architecture to use for the model.") flags.DEFINE_integer("batch_size", 256, "How many examples to process per batch.") flags.DEFINE_string("label_loss", "CrossEntropyLoss", "Loss computed on validation data") flags.DEFINE_integer("file_size", 4096, "Number of frames per batch for DBoF.") def find_class_by_name(name, modules): """Searches the provided modules for the named class and returns it.""" modules = [getattr(module, name, None) for module in modules] return next(a for a in modules if a) def get_input_data_tensors(reader, data_pattern, batch_size=256): logging.info("Using batch size of " + str(batch_size) + " for evaluation.") with tf.name_scope("eval_input"): files = gfile.Glob(data_pattern) if not files: raise IOError("Unable to find the evaluation files.") logging.info("number of evaluation files: " + str(len(files))) files.sort() filename_queue = tf.train.string_input_producer( files, shuffle=False, num_epochs=1) eval_data = reader.prepare_reader(filename_queue) return tf.train.batch( eval_data, batch_size=batch_size, capacity=4 * batch_size, allow_smaller_final_batch=True, enqueue_many=True) def build_graph(all_readers, all_data_patterns, input_reader, input_data_pattern, model, label_loss_fn, batch_size=256): """Creates the Tensorflow graph for evaluation. Args: all_readers: The data file reader. It should inherit from BaseReader. model: The core model (e.g. logistic or neural net). It should inherit from BaseModel. all_data_patterns: glob path to the evaluation data files. label_loss_fn: What kind of loss to apply to the model. It should inherit from BaseLoss. batch_size: How many examples to process at a time. """ global_step = tf.Variable(0, trainable=False, name="global_step") model_input_raw_tensors = [] labels_batch_tensor = None video_id_batch = None for reader, data_pattern in zip(all_readers, all_data_patterns): unused_video_id, model_input_raw, labels_batch, unused_num_frames = ( get_input_data_tensors( reader, data_pattern, batch_size=batch_size)) if labels_batch_tensor is None: labels_batch_tensor = labels_batch if video_id_batch is None: video_id_batch = unused_video_id model_input_raw_tensors.append(tf.expand_dims(model_input_raw, axis=2)) original_input = None if input_data_pattern is not None: unused_video_id, original_input, unused_labels_batch, unused_num_frames = ( get_input_data_tensors( input_reader, input_data_pattern, batch_size=batch_size)) model_input = tf.concat(model_input_raw_tensors, axis=2) labels_batch = labels_batch_tensor with tf.name_scope("model"): result = model.create_model(model_input, labels=labels_batch, vocab_size=reader.num_classes, original_input=original_input, is_training=False) predictions = result["predictions"] if "loss" in result.keys(): label_loss = result["loss"] else: label_loss = label_loss_fn.calculate_loss(predictions, labels_batch) tf.add_to_collection("global_step", global_step) tf.add_to_collection("loss", label_loss) tf.add_to_collection("predictions", predictions) tf.add_to_collection("input_batch", model_input) tf.add_to_collection("video_id_batch", video_id_batch) tf.add_to_collection("labels", tf.cast(labels_batch, tf.float32)) def inference_loop(video_id_batch, prediction_batch, label_batch, saver, output_dir, batch_size): with tf.Session() as sess: checkpoint = FLAGS.model_checkpoint_path if checkpoint is None: checkpoint = tf.train.latest_checkpoint(FLAGS.train_dir) if checkpoint: logging.info("Loading checkpoint for eval: " + checkpoint) saver.restore(sess, checkpoint) global_step_val = checkpoint.split("/")[-1].split("-")[-1] else: logging.info("No checkpoint file found.") return global_step_val sess.run([tf.local_variables_initializer()]) # Start the queue runners. fetches = [video_id_batch, prediction_batch, label_batch] coord = tf.train.Coordinator() start_time = time.time() video_ids = [] video_labels = [] video_features = [] filenum = 0 num_examples_processed = 0 total_num_examples_processed = 0 directory = FLAGS.output_dir if not os.path.exists(directory): os.makedirs(directory) else: raise IOError("Output path exists! path='" + directory + "'") try: threads = [] for qr in tf.get_collection(tf.GraphKeys.QUEUE_RUNNERS): threads.extend(qr.create_threads( sess, coord=coord, daemon=True, start=True)) logging.info("enter eval_once loop global_step_val = %s. ", global_step_val) while not coord.should_stop(): ids_val, predictions_val, labels_val = None, None, None ids_val, predictions_val, labels_val = sess.run(fetches) video_ids.append(ids_val) video_labels.append(labels_val) video_features.append(predictions_val) num_examples_processed += len(ids_val) ids_val, predictions_val, labels_val = None, None, None if num_examples_processed >= FLAGS.file_size: assert num_examples_processed==FLAGS.file_size, "num_examples_processed should be equal to %d"%FLAGS.file_size video_ids = np.concatenate(video_ids, axis=0) video_labels = np.concatenate(video_labels, axis=0) video_features = np.concatenate(video_features, axis=0) write_to_record(video_ids, video_labels, video_features, filenum, num_examples_processed) video_ids = [] video_labels = [] video_features = [] filenum += 1 total_num_examples_processed += num_examples_processed now = time.time() logging.info("num examples processed: " + str(num_examples_processed) + " elapsed seconds: " + "{0:.2f}".format(now-start_time)) num_examples_processed = 0 except tf.errors.OutOfRangeError as e: if ids_val is not None: video_ids.append(ids_val) video_labels.append(labels_val) video_features.append(predictions_val) num_examples_processed += len(ids_val) if 0 < num_examples_processed <= FLAGS.file_size: video_ids = np.concatenate(video_ids, axis=0) video_labels = np.concatenate(video_labels, axis=0) video_features = np.concatenate(video_features, axis=0) write_to_record(video_ids, video_labels, video_features, filenum, num_examples_processed) total_num_examples_processed += num_examples_processed now = time.time() logging.info("num examples processed: " + str(num_examples_processed) + " elapsed seconds: " + "{0:.2f}".format(now-start_time)) num_examples_processed = 0 logging.info("Done with inference. %d samples was written to %s" % (total_num_examples_processed, FLAGS.output_dir)) except Exception as e: # pylint: disable=broad-except logging.info("Unexpected exception: " + str(e)) finally: coord.request_stop() coord.join(threads, stop_grace_period_secs=10) def write_to_record(video_ids, video_labels, video_features, filenum, num_examples_processed): writer = tf.python_io.TFRecordWriter(FLAGS.output_dir + '/' + 'predictions-%04d.tfrecord' % filenum) for i in range(num_examples_processed): video_id = video_ids[i] video_label = np.nonzero(video_labels[i,:])[0] example = get_output_feature(video_id, video_label, [video_features[i,:]], ['predictions']) serialized = example.SerializeToString() writer.write(serialized) writer.close() def get_output_feature(video_id, video_label, video_feature, feature_names): feature_maps = {'video_id': tf.train.Feature(bytes_list=tf.train.BytesList(value=[video_id])), 'labels': tf.train.Feature(int64_list=tf.train.Int64List(value=video_label))} for feature_index in range(len(feature_names)): feature_maps[feature_names[feature_index]] = tf.train.Feature( float_list=tf.train.FloatList(value=video_feature[feature_index])) example = tf.train.Example(features=tf.train.Features(feature=feature_maps)) return example def main(unused_argv): logging.set_verbosity(tf.logging.INFO) with tf.Graph().as_default(): # convert feature_names and feature_sizes to lists of values feature_names, feature_sizes = utils.GetListOfFeatureNamesAndSizes( FLAGS.feature_names, FLAGS.feature_sizes) # prepare a reader for each single model prediction result all_readers = [] all_patterns = FLAGS.input_data_patterns all_patterns = map(lambda x: x.strip(), all_patterns.strip().strip(",").split(",")) for i in xrange(len(all_patterns)): reader = readers.EnsembleReader( feature_names=feature_names, feature_sizes=feature_sizes) all_readers.append(reader) input_reader = None input_data_pattern = None if FLAGS.input_data_pattern is not None: input_reader = readers.EnsembleReader( feature_names=["input"], feature_sizes=[1024+128]) input_data_pattern = FLAGS.input_data_pattern model = find_class_by_name(FLAGS.model, [ensemble_level_models])() label_loss_fn = find_class_by_name(FLAGS.label_loss, [losses])() if FLAGS.input_data_patterns is "": raise IOError("'input_data_patterns' was not specified. " + "Nothing to evaluate.") build_graph( all_readers=all_readers, input_reader=input_reader, all_data_patterns=all_patterns, input_data_pattern=input_data_pattern, model=model, label_loss_fn=label_loss_fn, batch_size=FLAGS.batch_size) logging.info("built evaluation graph") video_id_batch = tf.get_collection("video_id_batch")[0] prediction_batch = tf.get_collection("predictions")[0] label_batch = tf.get_collection("labels")[0] saver = tf.train.Saver(tf.global_variables()) inference_loop(video_id_batch, prediction_batch, label_batch, saver, FLAGS.output_dir, FLAGS.batch_size) if __name__ == "__main__": app.run() ================================================ FILE: youtube-8m-ensemble/inference.py ================================================ # # Licensed under the Apache License, Version 2.0 (the "License"); # you may not use this file except in compliance with the License. # You may obtain a copy of the License at # # http://www.apache.org/licenses/LICENSE-2.0 # # Unless required by applicable law or agreed to in writing, software # distributed under the License is distributed on an "AS-IS" BASIS, # WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied. # See the License for the specific language governing permissions and # limitations under the License. """Binary for evaluating Tensorflow models on the YouTube-8M dataset.""" import time import numpy import eval_util import losses import ensemble_level_models import readers import tensorflow as tf from tensorflow import app from tensorflow import flags from tensorflow import gfile from tensorflow import logging import utils FLAGS = flags.FLAGS if __name__ == "__main__": # Dataset flags. flags.DEFINE_string("model_checkpoint_path", None, "The file to load the model files from. ") flags.DEFINE_string("output_file", "", "The file to save the predictions to.") flags.DEFINE_string( "input_data_patterns", "", "File globs defining the evaluation dataset in tensorflow.SequenceExample format.") flags.DEFINE_string( "input_data_pattern", None, "File globs for original model input.") flags.DEFINE_string("feature_names", "predictions", "Name of the feature " "to use for training.") flags.DEFINE_string("feature_sizes", "4716", "Length of the feature vectors.") # Model flags. flags.DEFINE_string( "model", "LogisticModel", "Which architecture to use for the model.") flags.DEFINE_integer("batch_size", 256, "How many examples to process per batch.") # Other flags. flags.DEFINE_boolean("run_once", True, "Whether to run eval only once.") flags.DEFINE_integer("top_k", 20, "How many predictions to output per video.") def format_lines(video_ids, predictions, top_k): batch_size = len(video_ids) for video_index in range(batch_size): top_indices = numpy.argpartition(predictions[video_index], -top_k)[-top_k:] line = [(class_index, predictions[video_index][class_index]) for class_index in top_indices] line = sorted(line, key=lambda p: -p[1]) yield video_ids[video_index].decode('utf-8') + "," + " ".join("%i %f" % pair for pair in line) + "\n" def find_class_by_name(name, modules): """Searches the provided modules for the named class and returns it.""" modules = [getattr(module, name, None) for module in modules] return next(a for a in modules if a) def get_input_data_tensors(reader, data_pattern, batch_size=256): logging.info("Using batch size of " + str(batch_size) + " for evaluation.") with tf.name_scope("eval_input"): files = gfile.Glob(data_pattern) if not files: raise IOError("Unable to find the evaluation files.") logging.info("number of evaluation files: " + str(len(files))) files.sort() filename_queue = tf.train.string_input_producer( files, shuffle=False, num_epochs=1) eval_data = reader.prepare_reader(filename_queue) return tf.train.batch( eval_data, batch_size=batch_size, capacity=3 * batch_size, allow_smaller_final_batch=True, enqueue_many=True) def build_graph(all_readers, all_data_patterns, input_reader, input_data_pattern, model, batch_size=256): """Creates the Tensorflow graph for evaluation. Args: all_readers: The data file reader. It should inherit from BaseReader. model: The core model (e.g. logistic or neural net). It should inherit from BaseModel. all_data_patterns: glob path to the evaluation data files. batch_size: How many examples to process at a time. """ global_step = tf.Variable(0, trainable=False, name="global_step") model_input_raw_tensors = [] labels_batch_tensor = None video_id_batch = None for reader, data_pattern in zip(all_readers, all_data_patterns): unused_video_id, model_input_raw, labels_batch, unused_num_frames = ( get_input_data_tensors( reader, data_pattern, batch_size=batch_size)) if labels_batch_tensor is None: labels_batch_tensor = labels_batch if video_id_batch is None: video_id_batch = unused_video_id model_input_raw_tensors.append(tf.expand_dims(model_input_raw, axis=2)) original_input = None if input_data_pattern is not None: unused_video_id, original_input, unused_labels_batch, unused_num_frames = ( get_input_data_tensors( input_reader, input_data_pattern, batch_size=batch_size)) model_input = tf.concat(model_input_raw_tensors, axis=2) labels_batch = labels_batch_tensor with tf.name_scope("model"): result = model.create_model(model_input, labels=labels_batch, vocab_size=reader.num_classes, original_input=original_input, is_training=False) predictions = result["predictions"] tf.add_to_collection("global_step", global_step) tf.add_to_collection("predictions", predictions) tf.add_to_collection("input_batch", model_input) tf.add_to_collection("video_id_batch", video_id_batch) tf.add_to_collection("labels", tf.cast(labels_batch, tf.float32)) def inference_loop(video_id_batch, prediction_batch, label_batch, saver, out_file_location): top_k = FLAGS.top_k with tf.Session() as sess, gfile.Open(out_file_location, "w+") as out_file: checkpoint = FLAGS.model_checkpoint_path if checkpoint: logging.info("Loading checkpoint for eval: " + checkpoint) saver.restore(sess, checkpoint) global_step_val = checkpoint.split("/")[-1].split("-")[-1] else: logging.info("No checkpoint file found.") return global_step_val sess.run([tf.local_variables_initializer()]) # Start the queue runners. fetches = [video_id_batch, prediction_batch] coord = tf.train.Coordinator() try: threads = [] for qr in tf.get_collection(tf.GraphKeys.QUEUE_RUNNERS): threads.extend(qr.create_threads( sess, coord=coord, daemon=True, start=True)) logging.info("enter eval_once loop global_step_val = %s. ", global_step_val) num_examples_processed = 0 start_time = time.time() out_file.write("VideoId,LabelConfidencePairs\n") while not coord.should_stop(): batch_start_time = time.time() video_id_val, predictions_val = sess.run(fetches) now = time.time() num_examples_processed += len(video_id_val) logging.info("num examples processed: " + str(num_examples_processed) + " elapsed seconds: " + "{0:.2f}".format(now-start_time)) for line in format_lines(video_id_val, predictions_val, top_k): out_file.write(line) out_file.flush() except tf.errors.OutOfRangeError as e: logging.info('Done with inference. The output file was written to ' + out_file_location) except Exception as e: # pylint: disable=broad-except logging.info("Unexpected exception: " + str(e)) coord.request_stop(e) coord.request_stop() coord.join(threads, stop_grace_period_secs=10) def inference(): tf.set_random_seed(0) # for reproducibility with tf.Graph().as_default(): # convert feature_names and feature_sizes to lists of values feature_names, feature_sizes = utils.GetListOfFeatureNamesAndSizes( FLAGS.feature_names, FLAGS.feature_sizes) # prepare a reader for each single model prediction result all_readers = [] all_patterns = FLAGS.input_data_patterns all_patterns = map(lambda x: x.strip(), all_patterns.strip().strip(",").split(",")) for i in xrange(len(all_patterns)): reader = readers.EnsembleReader( feature_names=feature_names, feature_sizes=feature_sizes) all_readers.append(reader) input_reader = None input_data_pattern = None if FLAGS.input_data_pattern is not None: input_reader = readers.EnsembleReader( feature_names=["input"], feature_sizes=[1024+128]) input_data_pattern = FLAGS.input_data_pattern model = find_class_by_name(FLAGS.model, [ensemble_level_models])() if FLAGS.input_data_patterns is "": raise IOError("'input_data_patterns' was not specified. " + "Nothing to evaluate.") build_graph( all_readers=all_readers, all_data_patterns=all_patterns, input_reader=input_reader, input_data_pattern=input_data_pattern, model=model, batch_size=FLAGS.batch_size) logging.info("built evaluation graph") video_id_batch = tf.get_collection("video_id_batch")[0] prediction_batch = tf.get_collection("predictions")[0] label_batch = tf.get_collection("labels")[0] saver = tf.train.Saver(tf.global_variables()) inference_loop(video_id_batch, prediction_batch, label_batch, saver, FLAGS.output_file) def main(unused_argv): logging.set_verbosity(tf.logging.INFO) print("tensorflow version: %s" % tf.__version__) inference() if __name__ == "__main__": app.run() ================================================ FILE: youtube-8m-ensemble/losses.py ================================================ # Copyright 2016 Google Inc. All Rights Reserved. # # Licensed under the Apache License, Version 2.0 (the "License"); # you may not use this file except in compliance with the License. # You may obtain a copy of the License at # # http://www.apache.org/licenses/LICENSE-2.0 # # Unless required by applicable law or agreed to in writing, software # distributed under the License is distributed on an "AS-IS" BASIS, # WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied. # See the License for the specific language governing permissions and # limitations under the License. """Provides definitions for non-regularized training or test losses.""" import numpy as np import tensorflow as tf from tensorflow import flags FLAGS = flags.FLAGS flags.DEFINE_float("false_negative_punishment", 1.0, "punishment constant to 1 classified to 0") flags.DEFINE_float("false_positive_punishment", 1.0, "punishment constant to 0 classified to 1") flags.DEFINE_integer("num_classes", 4716, "number of classes") flags.DEFINE_float("support_loss_percent", 0.1, "the part that support loss (in multi-task scenario) take in the whole loss function.") flags.DEFINE_string("support_type", "vertical", "type of support label, vertical or frequent or vertical,frequent.") flags.DEFINE_integer("num_supports", 25, "Number of total support categories.") flags.DEFINE_integer("num_verticals", 25, "Number of total vertical categories.") flags.DEFINE_integer("num_frequents", 200, "Number of total frequent categories.") flags.DEFINE_string("vertical_file", "resources/vertical.tsv", "Location of label-vertical mapping file.") flags.DEFINE_float("batch_agreement", 0.1, "the batch_agreement parameter") flags.DEFINE_bool("label_smoothing", False, "whether do label smoothing") flags.DEFINE_float("label_smoothing_epsilon", 0.1, "whether do label smoothing") def smoothing(labels): print "label smoothing for", labels epsilon = FLAGS.label_smoothing_epsilon float_labels = tf.cast(labels, tf.float32) num_labels = tf.reduce_sum(float_labels, axis=1, keep_dims=True) K = float_labels.get_shape().as_list()[1] prior = num_labels / K smooth_labels = float_labels * (1.0 - epsilon) + prior * epsilon return smooth_labels class BaseLoss(object): """Inherit from this class when implementing new losses.""" def calculate_loss(self, unused_predictions, unused_labels, **unused_params): """Calculates the average loss of the examples in a mini-batch. Args: unused_predictions: a 2-d tensor storing the prediction scores, in which each row represents a sample in the mini-batch and each column represents a class. unused_labels: a 2-d tensor storing the labels, which has the same shape as the unused_predictions. The labels must be in the range of 0 and 1. unused_params: loss specific parameters. Returns: A scalar loss tensor. """ raise NotImplementedError() class CrossEntropyLoss(BaseLoss): """Calculate the cross entropy loss between the predictions and labels. """ def calculate_loss(self, predictions, labels, weights=None, **unused_params): with tf.name_scope("loss_xent"): epsilon = 10e-6 if FLAGS.label_smoothing: float_labels = smoothing(labels) else: float_labels = tf.cast(labels, tf.float32) cross_entropy_loss = float_labels * tf.log(predictions + epsilon) + ( 1 - float_labels) * tf.log(1 - predictions + epsilon) cross_entropy_loss = tf.negative(cross_entropy_loss) if weights is not None: print cross_entropy_loss, weights weighted_loss = tf.einsum("ij,i->ij", cross_entropy_loss, weights) print "create weighted_loss", weighted_loss return tf.reduce_mean(tf.reduce_sum(weighted_loss, 1)) else: return tf.reduce_mean(tf.reduce_sum(cross_entropy_loss, 1)) class HingeLoss(BaseLoss): """Calculate the hinge loss between the predictions and labels. Note the subgradient is used in the backpropagation, and thus the optimization may converge slower. The predictions trained by the hinge loss are between -1 and +1. """ def calculate_loss(self, predictions, labels, b=1.0, **unused_params): with tf.name_scope("loss_hinge"): float_labels = tf.cast(labels, tf.float32) all_zeros = tf.zeros(tf.shape(float_labels), dtype=tf.float32) all_ones = tf.ones(tf.shape(float_labels), dtype=tf.float32) sign_labels = tf.subtract(tf.scalar_mul(2, float_labels), all_ones) hinge_loss = tf.maximum( all_zeros, tf.scalar_mul(b, all_ones) - sign_labels * predictions) return tf.reduce_mean(tf.reduce_sum(hinge_loss, 1)) class SoftmaxLoss(BaseLoss): """Calculate the softmax loss between the predictions and labels. The function calculates the loss in the following way: first we feed the predictions to the softmax activation function and then we calculate the minus linear dot product between the logged softmax activations and the normalized ground truth label. It is an extension to the one-hot label. It allows for more than one positive labels for each sample. """ def calculate_loss(self, predictions, labels, **unused_params): with tf.name_scope("loss_softmax"): epsilon = 10e-8 float_labels = tf.cast(labels, tf.float32) # l1 normalization (labels are no less than 0) label_rowsum = tf.maximum( tf.reduce_sum(float_labels, 1, keep_dims=True), epsilon) norm_float_labels = tf.div(float_labels, label_rowsum) softmax_outputs = tf.nn.softmax(predictions) softmax_loss = tf.negative(tf.reduce_sum( tf.multiply(norm_float_labels, tf.log(softmax_outputs)), 1)) return tf.reduce_mean(softmax_loss) class MultiTaskLoss(BaseLoss): """This is a vitural loss """ def calculate_loss(self, unused_predictions, unused_labels, **unused_params): raise NotImplementedError() def get_support(self, labels, support_type=None): if support_type == None: support_type = FLAGS.support_type if "," in support_type: new_labels = [] for st in support_type.split(","): new_labels.append(tf.cast(self.get_support(labels, st), dtype=tf.float32)) support_labels = tf.concat(new_labels, axis=1) return support_labels elif support_type == "vertical": num_classes = FLAGS.num_classes num_verticals = FLAGS.num_verticals vertical_file = FLAGS.vertical_file vertical_mapping = np.zeros([num_classes, num_verticals], dtype=np.float32) float_labels = tf.cast(labels, dtype=tf.float32) with open(vertical_file) as F: for line in F: group = map(int, line.strip().split()) if len(group) == 2: x, y = group vertical_mapping[x, y] = 1 vm_init = tf.constant_initializer(vertical_mapping) vm = tf.get_variable("vm", shape = [num_classes, num_verticals], trainable=False, initializer=vm_init) vertical_labels = tf.matmul(float_labels, vm) return tf.cast(vertical_labels > 0.2, tf.float32) elif support_type == "frequent": num_frequents = FLAGS.num_frequents frequent_labels = tf.slice(labels, begin=[0, 0], size=[-1, num_frequents]) frequent_labels = tf.cast(frequent_labels, dtype=tf.float32) return frequent_labels elif support_type == "label": float_labels = tf.cast(labels, dtype=tf.float32) return float_labels else: raise NotImplementedError() class MultiTaskCrossEntropyLoss(MultiTaskLoss): """Calculate the loss between the predictions and labels. """ def calculate_loss(self, predictions, support_predictions, labels, **unused_params): support_labels = self.get_support(labels) ce_loss_fn = CrossEntropyLoss() cross_entropy_loss = ce_loss_fn.calculate_loss(predictions, labels, **unused_params) cross_entropy_loss2 = ce_loss_fn.calculate_loss(support_predictions, support_labels, **unused_params) return cross_entropy_loss * (1.0 - FLAGS.support_loss_percent) + cross_entropy_loss2 * FLAGS.support_loss_percent ================================================ FILE: youtube-8m-ensemble/mean_average_precision_calculator.py ================================================ # Copyright 2016 Google Inc. All Rights Reserved. # # Licensed under the Apache License, Version 2.0 (the "License"); # you may not use this file except in compliance with the License. # You may obtain a copy of the License at # # http://www.apache.org/licenses/LICENSE-2.0 # # Unless required by applicable law or agreed to in writing, software # distributed under the License is distributed on an "AS-IS" BASIS, # WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied. # See the License for the specific language governing permissions and # limitations under the License. """Calculate the mean average precision. It provides an interface for calculating mean average precision for an entire list or the top-n ranked items. Example usages: We first call the function accumulate many times to process parts of the ranked list. After processing all the parts, we call peek_map_at_n to calculate the mean average precision. ``` import random p = np.array([[random.random() for _ in xrange(50)] for _ in xrange(1000)]) a = np.array([[random.choice([0, 1]) for _ in xrange(50)] for _ in xrange(1000)]) # mean average precision for 50 classes. calculator = mean_average_precision_calculator.MeanAveragePrecisionCalculator( num_class=50) calculator.accumulate(p, a) aps = calculator.peek_map_at_n() ``` """ import numpy import average_precision_calculator class MeanAveragePrecisionCalculator(object): """This class is to calculate mean average precision. """ def __init__(self, num_class): """Construct a calculator to calculate the (macro) average precision. Args: num_class: A positive Integer specifying the number of classes. top_n_array: A list of positive integers specifying the top n for each class. The top n in each class will be used to calculate its average precision at n. The size of the array must be num_class. Raises: ValueError: An error occurred when num_class is not a positive integer; or the top_n_array is not a list of positive integers. """ if not isinstance(num_class, int) or num_class <= 1: raise ValueError("num_class must be a positive integer.") self._ap_calculators = [] # member of AveragePrecisionCalculator self._num_class = num_class # total number of classes for i in range(num_class): self._ap_calculators.append( average_precision_calculator.AveragePrecisionCalculator()) def accumulate(self, predictions, actuals, num_positives=None): """Accumulate the predictions and their ground truth labels. Args: predictions: A list of lists storing the prediction scores. The outer dimension corresponds to classes. actuals: A list of lists storing the ground truth labels. The dimensions should correspond to the predictions input. Any value larger than 0 will be treated as positives, otherwise as negatives. num_positives: If provided, it is a list of numbers representing the number of true positives for each class. If not provided, the number of true positives will be inferred from the 'actuals' array. Raises: ValueError: An error occurred when the shape of predictions and actuals does not match. """ if not num_positives: num_positives = [None for i in predictions.shape[1]] calculators = self._ap_calculators for i in range(len(predictions)): calculators[i].accumulate(predictions[i], actuals[i], num_positives[i]) def clear(self): for calculator in self._ap_calculators: calculator.clear() def is_empty(self): return ([calculator.heap_size for calculator in self._ap_calculators] == [0 for _ in range(self._num_class)]) def peek_map_at_n(self): """Peek the non-interpolated mean average precision at n. Returns: An array of non-interpolated average precision at n (default 0) for each class. """ aps = [self._ap_calculators[i].peek_ap_at_n() for i in range(self._num_class)] return aps ================================================ FILE: youtube-8m-ensemble/model_selection_scripts/.vimrc ================================================ set tabstop=2 set shiftwidth=2 set expandtab set autoindent ================================================ FILE: youtube-8m-ensemble/model_selection_scripts/extend-step-mean_model.sh ================================================ #!/bin/bash DIR="$( cd "$( dirname "${BASH_SOURCE[0]}" )" && pwd )" model_name="$1" candidates_conf="$2" train_path=/Youtube-8M/model_predictions_for_selection/ensemble_train model_path="${DIR}/../../model/${model_name}" all_models_conf="${model_path}/all_models.conf" for candidates in $(cat $candidates_conf); do echo "$candidates" train_data_patterns=$(python ${DIR}/get_patterns.py --train_path="$train_path" --candidates="$candidates") CUDA_VISIBLE_DEVICES=1 python ${DIR}/../eval.py \ --model_checkpoint_path="${model_path}/model.ckpt-0" \ --train_dir="${model_path}" \ --model="MeanModel" \ --echo_gap=True \ --batch_size=1024 \ --eval_data_patterns="$train_data_patterns" | tail -n 1 done ================================================ FILE: youtube-8m-ensemble/model_selection_scripts/get_extend_candidates.py ================================================ import sys from tensorflow import flags FLAGS = flags.FLAGS if __name__=="__main__": flags.DEFINE_string("top_k_file", "", "The file that contains the top-k ensemble models.") flags.DEFINE_string("all_models_conf", "", "The file that contains all available single models.") if __name__=="__main__": all_models = [line.strip() for line in open(FLAGS.all_models_conf) if len(line.strip()) > 0] extend_candidates = set() with open(FLAGS.top_k_file) as F: ensemble_models = [line.strip().split(",") for line in F.readlines() if len(line.strip()) > 0] for em in ensemble_models: for model in all_models: if model not in em: new_combination = ",".join(sorted(em + [model])) if new_combination not in extend_candidates: extend_candidates.add(new_combination) for candidate in extend_candidates: print candidate ================================================ FILE: youtube-8m-ensemble/model_selection_scripts/get_patterns.py ================================================ import sys from tensorflow import flags FLAGS = flags.FLAGS if __name__=="__main__": flags.DEFINE_string("train_path", "", "The directory where training files locates.") flags.DEFINE_string("candidates", "", "The candidate methods.") if __name__=="__main__": candidate_methods = map(lambda x: x.strip(), FLAGS.candidates.strip().split(",")) train_path = FLAGS.train_path output_path = ",".join(map(lambda x: "%s/%s/*.tfrecord"%(train_path, x), candidate_methods)) sys.stdout.write(output_path) sys.stdout.flush() ================================================ FILE: youtube-8m-ensemble/model_selection_scripts/get_top_k.py ================================================ import sys from tensorflow import flags FLAGS = flags.FLAGS if __name__=="__main__": flags.DEFINE_string("log_file", "", "The file that log models performances.") flags.DEFINE_string("sorted_log_file", "", "The file that log models performances (sorted by GAP).") flags.DEFINE_integer("top_k", 10, "The number of top models reserved.") if __name__=="__main__": log_file = FLAGS.log_file with open(log_file) as F: lines = F.readlines() models = map(lambda x: x.strip(), lines[::2]) perfs = map(lambda x: float(x.strip().split("=")[-1]), lines[1::2]) perfs = perfs[:len(models)] model_perfs = sorted(zip(perfs, models), reverse=True) with open(FLAGS.sorted_log_file, "w") as Fo: Fo.writelines(["%f\t%s\n"%(x,y) for x,y in model_perfs]) for perf, model in model_perfs[:FLAGS.top_k]: print model ================================================ FILE: youtube-8m-ensemble/model_selection_scripts/greedy-selection-mean_model.sh ================================================ #!/bin/bash DIR="$( cd "$( dirname "${BASH_SOURCE[0]}" )" && pwd )" model_name=$1 extend_step_script="${DIR}/extend-step-mean_model.sh" train_path=/Youtube-8M/model_predictions_for_selection/ensemble_train model_path="${DIR}/../../model/${model_name}" all_models_conf="${model_path}/all_models.conf" if [ -f $all_models_conf ]; then top_models_conf="${all_models_conf}" len_k_models_conf="${all_models_conf}" for step in {14..30}; do if [ $step -gt 1 ]; then top_models_conf="${model_path}/top_$((${step}-1))_models.conf" len_k_models_conf="${model_path}/len_${step}_models.conf" echo $top_models_conf $len_k_models_conf python ${DIR}/get_extend_candidates.py --top_k_file="$top_models_conf" --all_models_conf="${all_models_conf}" > $len_k_models_conf fi bash $extend_step_script $model_name ${len_k_models_conf} > ${model_path}/len_${step}_models.log python ${DIR}/get_top_k.py --top_k=2 --log_file="${model_path}/len_${step}_models.log" --sorted_log_file="${model_path}/len_${step}_models.sorted.log" > ${model_path}/top_${step}_models.conf top_models_conf="${model_path}/top_${step}_models.conf" done else echo $all_models_conf not found, did nothing fi ================================================ FILE: youtube-8m-ensemble/model_utils.py ================================================ # Copyright 2016 Google Inc. All Rights Reserved. # # Licensed under the Apache License, Version 2.0 (the "License"); # you may not use this file except in compliance with the License. # You may obtain a copy of the License at # # http://www.apache.org/licenses/LICENSE-2.0 # # Unless required by applicable law or agreed to in writing, software # distributed under the License is distributed on an "AS-IS" BASIS, # WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied. # See the License for the specific language governing permissions and # limitations under the License. """Contains a collection of util functions for model construction. """ import numpy import tensorflow as tf from tensorflow import logging from tensorflow import flags import tensorflow.contrib.slim as slim def SampleRandomSequence(model_input, num_frames, num_samples): """Samples a random sequence of frames of size num_samples. Args: model_input: A tensor of size batch_size x max_frames x feature_size num_frames: A tensor of size batch_size x 1 num_samples: A scalar Returns: `model_input`: A tensor of size batch_size x num_samples x feature_size """ batch_size = tf.shape(model_input)[0] frame_index_offset = tf.tile( tf.expand_dims(tf.range(num_samples), 0), [batch_size, 1]) max_start_frame_index = tf.maximum(num_frames - num_samples, 0) start_frame_index = tf.cast( tf.multiply( tf.random_uniform([batch_size, 1]), tf.cast(max_start_frame_index + 1, tf.float32)), tf.int32) frame_index = tf.minimum(start_frame_index + frame_index_offset, tf.cast(num_frames - 1, tf.int32)) batch_index = tf.tile( tf.expand_dims(tf.range(batch_size), 1), [1, num_samples]) index = tf.stack([batch_index, frame_index], 2) return tf.gather_nd(model_input, index) def SampleRandomFrames(model_input, num_frames, num_samples): """Samples a random set of frames of size num_samples. Args: model_input: A tensor of size batch_size x max_frames x feature_size num_frames: A tensor of size batch_size x 1 num_samples: A scalar Returns: `model_input`: A tensor of size batch_size x num_samples x feature_size """ batch_size = tf.shape(model_input)[0] frame_index = tf.cast( tf.multiply( tf.random_uniform([batch_size, num_samples]), tf.tile(tf.cast(num_frames, tf.float32), [1, num_samples])), tf.int32) batch_index = tf.tile( tf.expand_dims(tf.range(batch_size), 1), [1, num_samples]) index = tf.stack([batch_index, frame_index], 2) return tf.gather_nd(model_input, index) def FramePooling(frames, method, **unused_params): """Pools over the frames of a video. Args: frames: A tensor with shape [batch_size, num_frames, feature_size]. method: "average", "max", "attention", or "none". Returns: A tensor with shape [batch_size, feature_size] for average, max, or attention pooling. A tensor with shape [batch_size*num_frames, feature_size] for none pooling. Raises: ValueError: if method is other than "average", "max", "attention", or "none". """ if method == "average": return tf.reduce_mean(frames, 1) elif method == "max": return tf.reduce_max(frames, 1) elif method == "none": feature_size = frames.shape_as_list()[2] return tf.reshape(frames, [-1, feature_size]) else: raise ValueError("Unrecognized pooling method: %s" % method) ================================================ FILE: youtube-8m-ensemble/models.py ================================================ # Copyright 2016 Google Inc. All Rights Reserved. # # Licensed under the Apache License, Version 2.0 (the "License"); # you may not use this file except in compliance with the License. # You may obtain a copy of the License at # # http://www.apache.org/licenses/LICENSE-2.0 # # Unless required by applicable law or agreed to in writing, software # distributed under the License is distributed on an "AS-IS" BASIS, # WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied. # See the License for the specific language governing permissions and # limitations under the License. """Contains the base class for models.""" class BaseModel(object): """Inherit from this class when implementing new models.""" def create_model(self, unused_model_input, **unused_params): raise NotImplementedError() ================================================ FILE: youtube-8m-ensemble/readers.py ================================================ # Copyright 2016 Google Inc. All Rights Reserved. # # Licensed under the Apache License, Version 2.0 (the "License"); # you may not use this file except in compliance with the License. # You may obtain a copy of the License at # # http://www.apache.org/licenses/LICENSE-2.0 # # Unless required by applicable law or agreed to in writing, software # distributed under the License is distributed on an "AS-IS" BASIS, # WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied. # See the License for the specific language governing permissions and # limitations under the License. """Provides readers configured for different datasets.""" import sys import tensorflow as tf import utils from tensorflow import logging def resize_axis(tensor, axis, new_size, fill_value=0): tensor = tf.convert_to_tensor(tensor) shape = tf.unstack(tf.shape(tensor)) pad_shape = shape[:] pad_shape[axis] = tf.maximum(0, new_size - shape[axis]) shape[axis] = tf.minimum(shape[axis], new_size) shape = tf.stack(shape) resized = tf.concat([ tf.slice(tensor, tf.zeros_like(shape), shape), tf.fill(tf.stack(pad_shape), tf.cast(fill_value, tensor.dtype)) ], axis) # Update shape. new_shape = tensor.get_shape().as_list() # A copy is being made. new_shape[axis] = new_size resized.set_shape(new_shape) return resized class BaseReader(object): """Inherit from this class when implementing new readers.""" def prepare_reader(self, unused_filename_queue): """Create a thread for generating prediction and label tensors.""" raise NotImplementedError() class EnsembleReader(BaseReader): def __init__(self, num_classes=4716, feature_sizes=[1024], feature_names=["mean_inc3"]): assert len(feature_names) == len(feature_sizes), \ "length of feature_names (={}) != length of feature_sizes (={})".format( \ len(feature_names), len(feature_sizes)) self.num_classes = num_classes self.feature_sizes = feature_sizes self.feature_names = feature_names def prepare_reader(self, filename_queue, batch_size=1024): reader = tf.TFRecordReader() _, serialized_examples = reader.read_up_to(filename_queue, batch_size) # set the mapping from the fields to data types in the proto num_features = len(self.feature_names) assert num_features > 0, "self.feature_names is empty!" assert len(self.feature_names) == len(self.feature_sizes), \ "length of feature_names (={}) != length of feature_sizes (={})".format( \ len(self.feature_names), len(self.feature_sizes)) feature_map = {"video_id": tf.FixedLenFeature([], tf.string), "labels": tf.VarLenFeature(tf.int64)} for feature_index in range(num_features): feature_map[self.feature_names[feature_index]] = tf.FixedLenFeature( [self.feature_sizes[feature_index]], tf.float32) features = tf.parse_example(serialized_examples, features=feature_map) labels = tf.sparse_to_indicator(features["labels"], self.num_classes) labels.set_shape([None, self.num_classes]) concatenated_features = tf.concat([ features[feature_name] for feature_name in self.feature_names], 1) return features["video_id"], concatenated_features, labels, tf.ones([tf.shape(serialized_examples)[0]]) class EnsembleFrameReader(BaseReader): def __init__(self, num_classes=4716, feature_sizes=[1024], feature_names=["mean_inc3"], max_frames=300): assert len(feature_names) == len(feature_sizes), \ "length of feature_names (={}) != length of feature_sizes (={})".format( \ len(feature_names), len(feature_sizes)) self.num_classes = num_classes self.feature_sizes = feature_sizes self.feature_names = feature_names self.max_frames = max_frames def get_video_matrix(self, features, feature_size, max_frames): reshaped_features = tf.reshape(features, [-1]) num_frames = tf.minimum(tf.shape(reshaped_features)[0], max_frames) feature_matrix = resize_axis(reshaped_features, 0, max_frames, fill_value="") return feature_matrix, num_frames def prepare_reader(self, filename_queue): reader = tf.TFRecordReader() _, serialized_example = reader.read(filename_queue) contexts, features = tf.parse_single_sequence_example( serialized_example, context_features={ "video_id": tf.FixedLenFeature([], tf.string), "labels": tf.VarLenFeature(tf.int64)}, sequence_features={ "rgb": tf.FixedLenSequenceFeature([], dtype=tf.string), "audio": tf.FixedLenSequenceFeature([], dtype=tf.string), }) # read ground truth labels labels = (tf.cast( tf.sparse_to_dense(contexts["labels"].values, (self.num_classes,), 1, validate_indices=False), tf.bool)) rgbs, num_frames = self.get_video_matrix(features["rgb"], 1024, self.max_frames) audios, num_frames = self.get_video_matrix(features["audio"], 1024, self.max_frames) batch_video_ids = tf.expand_dims(contexts["video_id"], 0) batch_rgbs = tf.expand_dims(rgbs, 0) batch_audios = tf.expand_dims(audios, 0) batch_labels = tf.expand_dims(labels, 0) batch_frames = tf.expand_dims(num_frames, 0) return batch_video_ids, batch_rgbs, batch_audios, batch_labels, batch_frames ================================================ FILE: youtube-8m-ensemble/top_k_scripts/eval-attention_matrix_model.sh ================================================ model=$1 conf=$2 moe=$3 att=$4 DEFAULT_GPU_ID=0 if [ -z ${CUDA_VISIBLE_DEVICES+x} ]; then GPU_ID=$DEFAULT_GPU_ID echo "set CUDA_VISIBLE_DEVICES to default('$GPU_ID')" else GPU_ID=$CUDA_VISIBLE_DEVICES echo "set CUDA_VISIBLE_DEVICES to external('$GPU_ID')" fi validate_path=/Youtube-8M/model_predictions/ensemble_validate validate_data_patterns="" for d in $(cat $conf); do validate_data_patterns="${validate_path}/${d}/*.tfrecord${validate_data_patterns:+,$validate_data_patterns}" done echo "$validate_data_patterns" input_data_pattern="${validate_path}/model_input/*.tfrecord" start=0 EVERY=300 DIR="$(pwd)" MODEL_DIR="${DIR}/../model/${model}" \ for checkpoint in $(cd $MODEL_DIR && python ${DIR}/training_utils/select.py $EVERY); do echo $checkpoint; if [ $checkpoint -gt $start ]; then echo $checkpoint; CUDA_VISIBLE_DEVICES="$GPU_ID" python eval.py \ --model_checkpoint_path="../model/${model}/model.ckpt-${checkpoint}" \ --train_dir="../model/${model}" \ --model="AttentionMatrixModel" \ --moe_num_mixtures=$moe \ --attention_matrix_rank=$att \ --eval_data_patterns="$validate_data_patterns" \ --input_data_pattern="$input_data_pattern" fi done ================================================ FILE: youtube-8m-ensemble/top_k_scripts/infer-attention_matrix_model.sh ================================================ model=$1 conf=$2 checkpoint=$3 test_path=/Youtube-8M/model_predictions/test test_data_patterns="" for d in $(cat $conf); do test_data_patterns="${test_path}/${d}/*.tfrecord${test_data_patterns:+,$test_data_patterns}" done echo "$test_data_patterns" input_data_pattern="${test_path}/model_input/*.tfrecord" #CUDA_VISIBLE_DEVICES=0 python inference.py \ python inference.py \ --model_checkpoint_path="../model/${model}/model.ckpt-${checkpoint}" \ --output_file="../model/${model}/predictions.${model}.csv" \ --model="AttentionMatrixModel" \ --moe_num_mixtures=4 \ --attention_matrix_rank=8 \ --batch_size=1024 \ --input_data_pattern="$input_data_pattern" \ --input_data_patterns="$test_data_patterns" ================================================ FILE: youtube-8m-ensemble/top_k_scripts/preensemble-attention_matrix_model.sh ================================================ model=$1 conf=$2 part=$3 checkpoint=$4 DEFAULT_GPU_ID=0 if [ -z ${CUDA_VISIBLE_DEVICES+x} ]; then GPU_ID=$DEFAULT_GPU_ID echo "set CUDA_VISIBLE_DEVICES to default('$GPU_ID')" else GPU_ID=$CUDA_VISIBLE_DEVICES echo "set CUDA_VISIBLE_DEVICES to external('$GPU_ID')" fi test_path=/Youtube-8M/model_predictions/${part} test_data_patterns="" for d in $(cat $conf); do test_data_patterns="${test_path}/${d}/*.tfrecord${test_data_patterns:+,$test_data_patterns}" done echo "$test_data_patterns" input_data_pattern="${test_path}/model_input/*.tfrecord" CUDA_VISIBLE_DEVICES="$GPU_ID" python inference-pre-ensemble.py \ --output_dir="/Youtube-8M/model_predictions/${part}/${model}" \ --model_checkpoint_path="../model/${model}/model.ckpt-${checkpoint}" \ --input_data_patterns="$test_data_patterns" \ --input_data_pattern="$input_data_pattern" \ --model="AttentionMatrixModel" \ --moe_num_mixtures=4 \ --attention_matrix_rank=8 \ --batch_size=1024 \ --file_size=4096 ================================================ FILE: youtube-8m-ensemble/top_k_scripts/run_top_k.sh ================================================ #!/bin/bash for i in 8 12 16 20; do bash top_k_scripts/train-attention_matrix_model.sh model_selection/top_${i}_model ../model/model_selection/top_${i}_model.conf 4 $(($i/4)) bash top_k_scripts/eval-attention_matrix_model.sh model_selection/top_${i}_model ../model/model_selection/top_${i}_model.conf 4 $(($i/4)) done ================================================ FILE: youtube-8m-ensemble/top_k_scripts/train-attention_matrix_model.sh ================================================ model=$1 conf=$2 moe=$3 att=$4 DEFAULT_GPU_ID=0 if [ -z ${CUDA_VISIBLE_DEVICES+x} ]; then GPU_ID=$DEFAULT_GPU_ID echo "set CUDA_VISIBLE_DEVICES to default('$GPU_ID')" else GPU_ID=$CUDA_VISIBLE_DEVICES echo "set CUDA_VISIBLE_DEVICES to external('$GPU_ID')" fi train_path=/Youtube-8M/model_predictions/ensemble_train train_data_patterns="" for d in $(cat $conf); do train_data_patterns="${train_path}/${d}/*.tfrecord${train_data_patterns:+,$train_data_patterns}" done echo $train_data_patterns input_data_pattern="${train_path}/model_input/*.tfrecord" CUDA_VISIBLE_DEVICES="$GPU_ID" python train.py \ --train_dir="../model/${model}" \ --train_data_patterns="$train_data_patterns" \ --input_data_pattern="$input_data_pattern" \ --model=AttentionMatrixModel \ --base_learning_rate=0.004 \ --moe_num_mixtures=${moe} \ --attention_matrix_rank=${att} \ --keep_checkpoint_every_n_hours=0.1 \ --batch_size=1024 \ --num_epochs=4 ================================================ FILE: youtube-8m-ensemble/train.py ================================================ # Copyright 2016 Google Inc. All Rights Reserved. # # Licensed under the Apache License, Version 2.0 (the "License"); # you may not use this file except in compliance with the License. # You may obtain a copy of the License at # # http://www.apache.org/licenses/LICENSE-2.0 # # Unless required by applicable law or agreed to in writing, software # distributed under the License is distributed on an "AS-IS" BASIS, # WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied. # See the License for the specific language governing permissions and # limitations under the License. """Binary for training Tensorflow models on the YouTube-8M dataset.""" import json import os import time import numpy import numpy as np import eval_util import losses import ensemble_level_models import readers import tensorflow as tf import tensorflow.contrib.slim as slim from tensorflow import app from tensorflow import flags from tensorflow import gfile from tensorflow import logging import utils FLAGS = flags.FLAGS if __name__ == "__main__": # Dataset flags. flags.DEFINE_string("train_dir", "/tmp/yt8m_model/", "The directory to save the model files in.") flags.DEFINE_string( "train_data_patterns", "", "Multiple file globs for the training dataset.") flags.DEFINE_string( "input_data_pattern", None, "File globs for original model input.") flags.DEFINE_string("feature_names", "predictions", "Name of the feature " "to use for training.") flags.DEFINE_string("feature_sizes", "4716", "Length of the feature vectors.") # Model flags. flags.DEFINE_string( "model", "LogisticModel", "Which architecture to use for the model. Models are defined " "in models.py.") flags.DEFINE_bool( "multitask", False, "Whether to consider support_predictions") flags.DEFINE_bool( "start_new_model", False, "If set, this will not resume from a checkpoint and will instead create a" " new model instance.") # Training flags. flags.DEFINE_integer("batch_size", 256, "How many examples to process per batch for training.") flags.DEFINE_string("label_loss", "CrossEntropyLoss", "Which loss function to use for training the model.") flags.DEFINE_float( "regularization_penalty", 1, "How much weight to give to the regularization loss (the label loss has " "a weight of 1).") flags.DEFINE_float("base_learning_rate", 0.01, "Which learning rate to start with.") flags.DEFINE_float("learning_rate_decay", 0.95, "Learning rate decay factor to be applied every " "learning_rate_decay_examples.") flags.DEFINE_float("learning_rate_decay_examples", 1000000, "Multiply current learning rate by learning_rate_decay " "every learning_rate_decay_examples.") flags.DEFINE_integer("num_epochs", 10, "How many passes to make over the dataset before " "halting training.") flags.DEFINE_float("keep_checkpoint_every_n_hours", 0.1, "How many hours before saving a new checkpoint") flags.DEFINE_bool("reweight", False, "Whether to load model weight from file.") flags.DEFINE_string("sample_vocab_file", "", "Where to load video_id vocabulary.") flags.DEFINE_string("sample_freq_file", "", "Where to load sample frequency.") # Other flags. flags.DEFINE_string("optimizer", "AdamOptimizer", "What optimizer class to use.") flags.DEFINE_float("clip_gradient_norm", 1.0, "Norm to clip gradients to.") flags.DEFINE_bool( "log_device_placement", False, "Whether to write the device on which every op will run into the " "logs on startup.") flags.DEFINE_integer("recall_at_n", 100, "N in recall@N.") flags.DEFINE_bool("training", True, "Whether to train") flags.DEFINE_bool( "dropout", False, "Whether to consider dropout") flags.DEFINE_float("keep_prob", 1.0, "probability to keep output (used in dropout, keep it unchanged in validationg and test)") flags.DEFINE_float("noise_level", 0.0, "standard deviation of noise (added to hidden nodes)") def get_input_data_tensors(reader, data_pattern, batch_size=256, num_epochs=None): logging.info("Using batch size of " + str(batch_size) + " for training.") with tf.name_scope("train_input"): files = gfile.Glob(data_pattern) if not files: raise IOError("Unable to find training files. data_pattern='" + data_pattern + "'.") logging.info("Number of training files: %s.", str(len(files))) files.sort() filename_queue = tf.train.string_input_producer( files, num_epochs=num_epochs, shuffle=False) training_data = reader.prepare_reader(filename_queue) return tf.train.batch( training_data, batch_size=batch_size, capacity=FLAGS.batch_size * 4, allow_smaller_final_batch=True, enqueue_many=True) def find_class_by_name(name, modules): """Searches the provided modules for the named class and returns it.""" modules = [getattr(module, name, None) for module in modules] return next(a for a in modules if a) def get_video_weights_array(): weight_lines = open(FLAGS.sample_freq_file).readlines() weights = numpy.array(map(float, weight_lines)) weights = weights.reshape([len(weight_lines)]) return weights, len(weight_lines) def optional_assign_weights(sess, weights_input, weights_assignment): if weights_input is not None: weights, length = get_video_weights_array() _ = sess.run(weights_assignment, feed_dict={weights_input: weights}) print "Assigned weights from %s" % FLAGS.sample_freq_file else: print "Collection weights_input not found" def get_video_weights(video_id_batch): video_id_to_index = tf.contrib.lookup.string_to_index_table_from_file( vocabulary_file=FLAGS.sample_vocab_file, default_value=0) indexes = video_id_to_index.lookup(video_id_batch) weights, length = get_video_weights_array() weights_input = tf.placeholder(tf.float32, shape=[length], name="sample_weights_input") weights_tensor = tf.get_variable("sample_weights", shape=[length], trainable=False, dtype=tf.float32, initializer=tf.constant_initializer(weights)) weights_assignment = tf.assign(weights_tensor, weights_input) tf.add_to_collection("weights_input", weights_input) tf.add_to_collection("weights_assignment", weights_assignment) video_weight_batch = tf.nn.embedding_lookup(weights_tensor, indexes) return video_weight_batch def build_graph(all_readers, all_train_data_patterns, input_reader, input_data_pattern, model, label_loss_fn=losses.CrossEntropyLoss(), batch_size=256, base_learning_rate=0.01, learning_rate_decay_examples=1000000, learning_rate_decay=0.95, optimizer_class=tf.train.AdamOptimizer, clip_gradient_norm=1.0, regularization_penalty=1, num_epochs=None): """Creates the Tensorflow graph. This will only be called once in the life of a training model, because after the graph is created the model will be restored from a meta graph file rather than being recreated. Args: all_readers: The data file readers. Every element in it should inherit from BaseReader. model: The core model (e.g. logistic or neural net). It should inherit from BaseModel. train_data_patterns: glob paths to the training data files. label_loss_fn: What kind of loss to apply to the model. It should inherit from BaseLoss. batch_size: How many examples to process at a time. base_learning_rate: What learning rate to initialize the optimizer with. optimizer_class: Which optimization algorithm to use. clip_gradient_norm: Magnitude of the gradient to clip to. regularization_penalty: How much weight to give the regularization loss compared to the label loss. num_epochs: How many passes to make over the data. 'None' means an unlimited number of passes. """ global_step = tf.Variable(0, trainable=False, name="global_step") learning_rate = tf.train.exponential_decay( base_learning_rate, global_step * batch_size, learning_rate_decay_examples, learning_rate_decay, staircase=True) tf.summary.scalar('learning_rate', learning_rate) original_input = None if input_data_pattern is not None: original_video_id, original_input, unused_labels_batch, unused_num_frames = ( get_input_data_tensors( input_reader, input_data_pattern, batch_size=batch_size, num_epochs=num_epochs)) optimizer = optimizer_class(learning_rate) model_input_raw_tensors = [] labels_batch_tensor = None for reader, data_pattern in zip(all_readers, all_train_data_patterns): video_id, model_input_raw, labels_batch, unused_num_frames = ( get_input_data_tensors( reader, data_pattern, batch_size=batch_size, num_epochs=num_epochs)) if labels_batch_tensor is None: labels_batch_tensor = labels_batch model_input_raw_tensors.append(tf.expand_dims(model_input_raw, axis=2)) if original_input is not None: id_match = tf.ones_like(original_video_id, dtype=tf.float32) id_match = id_match * tf.cast(tf.equal(original_video_id, video_id), dtype=tf.float32) tf.summary.scalar("model/id_match", tf.reduce_mean(id_match)) model_input = tf.concat(model_input_raw_tensors, axis=2) labels_batch = labels_batch_tensor tf.summary.histogram("model/input", model_input) with tf.name_scope("model"): if FLAGS.noise_level > 0: noise_level_tensor = tf.placeholder_with_default(0.0, shape=[], name="noise_level") else: noise_level_tensor = None if FLAGS.dropout: keep_prob_tensor = tf.placeholder_with_default(1.0, shape=[], name="keep_prob") result = model.create_model( model_input, labels=labels_batch, vocab_size=reader.num_classes, original_input=original_input, dropout=FLAGS.dropout, keep_prob=keep_prob_tensor, noise_level=noise_level_tensor) else: result = model.create_model( model_input, labels=labels_batch, vocab_size=reader.num_classes, original_input=original_input, noise_level=noise_level_tensor) for variable in slim.get_model_variables(): tf.summary.histogram(variable.op.name, variable) predictions = result["predictions"] if "loss" in result.keys(): label_loss = result["loss"] else: video_weights_batch = None if FLAGS.reweight: video_weights_batch = get_video_weights(video_id) else: video_weights_batch = None if FLAGS.multitask: print "using multitask loss" support_predictions = result["support_predictions"] tf.summary.histogram("model/support_predictions", support_predictions) print "support_predictions", support_predictions label_loss = label_loss_fn.calculate_loss(predictions, support_predictions, labels_batch, weights=video_weights_batch) else: print "using original loss" label_loss = label_loss_fn.calculate_loss(predictions, labels_batch, weights=video_weights_batch) tf.summary.histogram("model/predictions", predictions) tf.summary.scalar("label_loss", label_loss) if "regularization_loss" in result.keys(): reg_loss = result["regularization_loss"] else: reg_loss = tf.constant(0.0) reg_losses = tf.losses.get_regularization_losses() if reg_losses: reg_loss += tf.add_n(reg_losses) if regularization_penalty != 0: tf.summary.scalar("reg_loss", reg_loss) # Adds update_ops (e.g., moving average updates in batch normalization) as # a dependency to the train_op. update_ops = tf.get_collection(tf.GraphKeys.UPDATE_OPS) if "update_ops" in result.keys(): update_ops += result["update_ops"] if update_ops: with tf.control_dependencies(update_ops): barrier = tf.no_op(name="gradient_barrier") with tf.control_dependencies([barrier]): label_loss = tf.identity(label_loss) # Incorporate the L2 weight penalties etc. final_loss = regularization_penalty * reg_loss + label_loss if FLAGS.training: gradients = optimizer.compute_gradients(final_loss, colocate_gradients_with_ops=False) if clip_gradient_norm > 0: with tf.name_scope('clip_grads'): gradients = utils.clip_gradient_norms(gradients , clip_gradient_norm) train_op = optimizer.apply_gradients(gradients, global_step=global_step) else: train_op = tf.no_op() tf.add_to_collection("global_step", global_step) tf.add_to_collection("loss", label_loss) tf.add_to_collection("predictions", predictions) tf.add_to_collection("input_batch_raw", model_input) tf.add_to_collection("input_batch", model_input) tf.add_to_collection("labels", tf.cast(labels_batch, tf.float32)) tf.add_to_collection("train_op", train_op) if FLAGS.dropout: tf.add_to_collection("keep_prob", keep_prob_tensor) if FLAGS.noise_level > 0: tf.add_to_collection("noise_level", noise_level_tensor) class Trainer(object): """A Trainer to train a Tensorflow graph.""" def __init__(self, cluster, task, train_dir, log_device_placement=True): """"Creates a Trainer. Args: cluster: A tf.train.ClusterSpec if the execution is distributed. None otherwise. task: A TaskSpec describing the job type and the task index. """ self.cluster = cluster self.task = task self.is_master = (task.type == "master" and task.index == 0) self.train_dir = train_dir self.config = tf.ConfigProto(log_device_placement=log_device_placement) if self.is_master and self.task.index > 0: raise StandardError("%s: Only one replica of master expected", task_as_string(self.task)) def run(self, start_new_model=False): """Performs training on the currently defined Tensorflow graph. Returns: A tuple of the training Hit@1 and the training PERR. """ if self.is_master and start_new_model: self.remove_training_directory(self.train_dir) target, device_fn = self.start_server_if_distributed() meta_filename = self.get_meta_filename(start_new_model, self.train_dir) with tf.Graph().as_default() as graph: if meta_filename: saver = self.recover_model(meta_filename) with tf.device(device_fn): if not meta_filename: saver = self.build_model() global_step = tf.get_collection("global_step")[0] loss = tf.get_collection("loss")[0] predictions = tf.get_collection("predictions")[0] labels = tf.get_collection("labels")[0] train_op = tf.get_collection("train_op")[0] init_op = tf.global_variables_initializer() if FLAGS.dropout: keep_prob_tensor = tf.get_collection("keep_prob")[0] if FLAGS.noise_level > 0: noise_level_tensor = tf.get_collection("noise_level")[0] if FLAGS.reweight: weights_input, weights_assignment = None, None if len(tf.get_collection("weights_input")) > 0: weights_input = tf.get_collection("weights_input")[0] weights_assignment = tf.get_collection("weights_assignment")[0] sv = tf.train.Supervisor( graph, logdir=self.train_dir, init_op=init_op, is_chief=self.is_master, global_step=global_step, save_model_secs=6 * 60, save_summaries_secs=120, saver=saver) logging.info("%s: Starting managed session.", task_as_string(self.task)) with sv.managed_session(target, config=self.config) as sess: # re-assign weights if FLAGS.reweight: optional_assign_weights(sess, weights_input, weights_assignment) try: logging.info("%s: Entering training loop.", task_as_string(self.task)) while not sv.should_stop(): batch_start_time = time.time() custom_feed = {} if FLAGS.dropout: custom_feed[keep_prob_tensor] = FLAGS.keep_prob if FLAGS.noise_level > 0: custom_feed[noise_level_tensor] = FLAGS.noise_level _, global_step_val, loss_val, predictions_val, labels_val = sess.run( [train_op, global_step, loss, predictions, labels], feed_dict=custom_feed) seconds_per_batch = time.time() - batch_start_time if self.is_master: examples_per_second = labels_val.shape[0] / seconds_per_batch hit_at_one = eval_util.calculate_hit_at_one(predictions_val, labels_val) perr = eval_util.calculate_precision_at_equal_recall_rate( predictions_val, labels_val) recall = "N/A" if False: recall = eval_util.calculate_recall_at_n( predictions_val, labels_val, FLAGS.recall_at_n) sv.summary_writer.add_summary( utils.MakeSummary("model/Training_Recall@%d" % FLAGS.recall_at_n, recall), global_step_val) recall = "%.2f" % recall gap = eval_util.calculate_gap(predictions_val, labels_val) logging.info( "%s: training step " + str(global_step_val) + "| Hit@1: " + ("%.2f" % hit_at_one) + " PERR: " + ("%.2f" % perr) + " GAP: " + ("%.2f" % gap) + " Recall@%d: " % FLAGS.recall_at_n + recall + " Loss: " + str(loss_val), task_as_string(self.task)) sv.summary_writer.add_summary( utils.MakeSummary("model/Training_Hit@1", hit_at_one), global_step_val) sv.summary_writer.add_summary( utils.MakeSummary("model/Training_Perr", perr), global_step_val) sv.summary_writer.add_summary( utils.MakeSummary("model/Training_GAP", gap), global_step_val) sv.summary_writer.add_summary( utils.MakeSummary("global_step/Examples/Second", examples_per_second), global_step_val) sv.summary_writer.flush() except tf.errors.OutOfRangeError: logging.info("%s: Done training -- epoch limit reached.", task_as_string(self.task)) logging.info("%s: Exited training loop.", task_as_string(self.task)) sv.Stop() def start_server_if_distributed(self): """Starts a server if the execution is distributed.""" if self.cluster: logging.info("%s: Starting trainer within cluster %s.", task_as_string(self.task), self.cluster.as_dict()) server = start_server(self.cluster, self.task) target = server.target device_fn = tf.train.replica_device_setter( ps_device="/job:ps", worker_device="/job:%s/task:%d" % (self.task.type, self.task.index), cluster=self.cluster) else: target = "" device_fn = "" return (target, device_fn) def remove_training_directory(self, train_dir): """Removes the training directory.""" try: logging.info( "%s: Removing existing train directory.", task_as_string(self.task)) gfile.DeleteRecursively(train_dir) except: logging.error( "%s: Failed to delete directory " + train_dir + " when starting a new model. Please delete it manually and" + " try again.", task_as_string(self.task)) def get_meta_filename(self, start_new_model, train_dir): if start_new_model: logging.info("%s: Flag 'start_new_model' is set. Building a new model.", task_as_string(self.task)) return None latest_checkpoint = tf.train.latest_checkpoint(train_dir) if not latest_checkpoint: logging.info("%s: No checkpoint file found. Building a new model.", task_as_string(self.task)) return None meta_filename = latest_checkpoint + ".meta" if not gfile.Exists(meta_filename): logging.info("%s: No meta graph file found. Building a new model.", task_as_string(self.task)) return None else: return meta_filename def recover_model(self, meta_filename): logging.info("%s: Restoring from meta graph file %s", task_as_string(self.task), meta_filename) return tf.train.import_meta_graph(meta_filename) def build_model(self): """Find the model and build the graph.""" # Convert feature_names and feature_sizes to lists of values. feature_names, feature_sizes = utils.GetListOfFeatureNamesAndSizes( FLAGS.feature_names, FLAGS.feature_sizes) # prepare a reader for each single model prediction result all_readers = [] all_patterns = FLAGS.train_data_patterns all_patterns = map(lambda x: x.strip(), all_patterns.strip().strip(",").split(",")) for i in xrange(len(all_patterns)): all_readers.append(readers.EnsembleReader( feature_names=feature_names, feature_sizes=feature_sizes)) input_reader = None input_data_pattern = None if FLAGS.input_data_pattern is not None: input_reader = readers.EnsembleReader( feature_names=["input"], feature_sizes=[1024+128]) input_data_pattern = FLAGS.input_data_pattern # Find the model. model = find_class_by_name(FLAGS.model, [ensemble_level_models])() label_loss_fn = find_class_by_name(FLAGS.label_loss, [losses])() optimizer_class = find_class_by_name(FLAGS.optimizer, [tf.train]) build_graph(all_readers=all_readers, all_train_data_patterns=all_patterns, input_reader=input_reader, input_data_pattern=input_data_pattern, model=model, optimizer_class=optimizer_class, clip_gradient_norm=FLAGS.clip_gradient_norm, label_loss_fn=label_loss_fn, base_learning_rate=FLAGS.base_learning_rate, learning_rate_decay=FLAGS.learning_rate_decay, learning_rate_decay_examples=FLAGS.learning_rate_decay_examples, regularization_penalty=FLAGS.regularization_penalty, batch_size=FLAGS.batch_size, num_epochs=FLAGS.num_epochs) logging.info("%s: Built graph.", task_as_string(self.task)) return tf.train.Saver(max_to_keep=3, keep_checkpoint_every_n_hours=FLAGS.keep_checkpoint_every_n_hours) class ParameterServer(object): """A parameter server to serve variables in a distributed execution.""" def __init__(self, cluster, task): """Creates a ParameterServer. Args: cluster: A tf.train.ClusterSpec if the execution is distributed. None otherwise. task: A TaskSpec describing the job type and the task index. """ self.cluster = cluster self.task = task def run(self): """Starts the parameter server.""" logging.info("%s: Starting parameter server within cluster %s.", task_as_string(self.task), self.cluster.as_dict()) server = start_server(self.cluster, self.task) server.join() def start_server(cluster, task): """Creates a Server. Args: cluster: A tf.train.ClusterSpec if the execution is distributed. None otherwise. task: A TaskSpec describing the job type and the task index. """ if not task.type: raise ValueError("%s: The task type must be specified." % task_as_string(task)) if task.index is None: raise ValueError("%s: The task index must be specified." % task_as_string(task)) # Create and start a server. return tf.train.Server( tf.train.ClusterSpec(cluster), protocol="grpc", job_name=task.type, task_index=task.index) def task_as_string(task): return "/job:%s/task:%s" % (task.type, task.index) def main(unused_argv): # Load the environment. env = json.loads(os.environ.get("TF_CONFIG", "{}")) # Load the cluster data from the environment. cluster_data = env.get("cluster", None) cluster = tf.train.ClusterSpec(cluster_data) if cluster_data else None # Load the task data from the environment. task_data = env.get("task", None) or {"type": "master", "index": 0} task = type("TaskSpec", (object,), task_data) # Logging the version. logging.set_verbosity(tf.logging.INFO) logging.info("%s: Tensorflow version: %s.", task_as_string(task), tf.__version__) # Dispatch to a master, a worker, or a parameter server. if not cluster or task.type == "master" or task.type == "worker": Trainer(cluster, task, FLAGS.train_dir, FLAGS.log_device_placement).run( start_new_model=FLAGS.start_new_model) elif task.type == "ps": ParameterServer(cluster, task).run() else: raise ValueError("%s: Invalid task_type: %s." % (task_as_string(task), task.type)) if __name__ == "__main__": app.run() ================================================ FILE: youtube-8m-ensemble/training_utils/del.py ================================================ import os check = {} check_list = [] for filename in os.listdir("."): if filename.endswith("meta"): checkpoint = int(filename.split("-")[1].split(".")[0]) check_list.append(checkpoint) check_list.sort() for checkpoint in check_list: if check.has_key(checkpoint / 4000): print checkpoint else: check[checkpoint / 4000] = True ================================================ FILE: youtube-8m-ensemble/training_utils/sample_conf.py ================================================ import random from datetime import datetime import tensorflow as tf from tensorflow import flags FLAGS = flags.FLAGS if __name__=="__main__": flags.DEFINE_string("main_conf_file", "", "The conf file to sample from.") flags.DEFINE_string("sub_conf_file", "", "The conf file to randomly generate.") if __name__=="__main__": with open(FLAGS.main_conf_file) as F: methods = [] sample_methods = [] # methods for line in F: m = line.strip() if m: methods.append(m) # random sample random.seed(datetime.now()) for i in xrange(len(methods)): index = random.randint(0,len(methods)-1) m = methods[index] sample_methods.append(m) # output weight with open(FLAGS.sub_conf_file, "w") as out: out.writelines([m+"\n" for m in sample_methods]) ================================================ FILE: youtube-8m-ensemble/training_utils/sample_freq.py ================================================ import random from datetime import datetime import tensorflow as tf from tensorflow import flags FLAGS = flags.FLAGS if __name__=="__main__": flags.DEFINE_string("video_id_file", "", "The file in which every line is a video_id.") flags.DEFINE_string("output_freq_file", "", "Output the corresponding freq of video_ids.") if __name__=="__main__": with open(FLAGS.video_id_file) as F: word_list = [] freq_dict = {} # OOV line = F.next() for line in F: word = line.strip() if word: word_list.append(word) freq_dict[word] = 0 # random sample random.seed(datetime.now()) for i in xrange(len(word_list)): index = random.randint(0,len(word_list)-1) word = word_list[index] freq_dict[word] += 1 # get weight word_weights = [] word_weights.append(1) for word in word_list: word_weights.append(freq_dict[word]) # output weight with open(FLAGS.output_freq_file, "w") as out: out.writelines([str(item)+"\n" for item in word_weights]) ================================================ FILE: youtube-8m-ensemble/training_utils/select.py ================================================ import os import sys every = int(sys.argv[1]) check = {} check_list = [] for filename in os.listdir("."): if filename.endswith("meta"): checkpoint = int(filename.split("-")[1].split(".")[0]) check_list.append(checkpoint) check_list.sort() for checkpoint in check_list: if check.has_key(checkpoint / every): pass else: check[checkpoint / every] = True print checkpoint ================================================ FILE: youtube-8m-ensemble/utils.py ================================================ # Copyright 2016 Google Inc. All Rights Reserved. # # Licensed under the Apache License, Version 2.0 (the "License"); # you may not use this file except in compliance with the License. # You may obtain a copy of the License at # # http://www.apache.org/licenses/LICENSE-2.0 # # Unless required by applicable law or agreed to in writing, software # distributed under the License is distributed on an "AS-IS" BASIS, # WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied. # See the License for the specific language governing permissions and # limitations under the License. """Contains a collection of util functions for training and evaluating. """ import numpy import tensorflow as tf from tensorflow import logging def Dequantize(feat_vector, max_quantized_value=2, min_quantized_value=-2): """Dequantize the feature from the byte format to the float format. Args: feat_vector: the input 1-d vector. max_quantized_value: the maximum of the quantized value. min_quantized_value: the minimum of the quantized value. Returns: A float vector which has the same shape as feat_vector. """ assert max_quantized_value > min_quantized_value quantized_range = max_quantized_value - min_quantized_value scalar = quantized_range / 255.0 bias = (quantized_range / 512.0) + min_quantized_value return feat_vector * scalar + bias def MakeSummary(name, value): """Creates a tf.Summary proto with the given name and value.""" summary = tf.Summary() val = summary.value.add() val.tag = str(name) val.simple_value = float(value) return summary def AddGlobalStepSummary(summary_writer, global_step_val, global_step_info_dict, summary_scope="Eval"): """Add the global_step summary to the Tensorboard. Args: summary_writer: Tensorflow summary_writer. global_step_val: a int value of the global step. global_step_info_dict: a dictionary of the evaluation metrics calculated for a mini-batch. summary_scope: Train or Eval. Returns: A string of this global_step summary """ this_hit_at_one = global_step_info_dict["hit_at_one"] this_perr = global_step_info_dict["perr"] this_loss = global_step_info_dict["loss"] examples_per_second = global_step_info_dict.get("examples_per_second", -1) summary_writer.add_summary( MakeSummary("GlobalStep/" + summary_scope + "_Hit@1", this_hit_at_one), global_step_val) summary_writer.add_summary( MakeSummary("GlobalStep/" + summary_scope + "_Perr", this_perr), global_step_val) summary_writer.add_summary( MakeSummary("GlobalStep/" + summary_scope + "_Loss", this_loss), global_step_val) if examples_per_second != -1: summary_writer.add_summary( MakeSummary("GlobalStep/" + summary_scope + "_Example_Second", examples_per_second), global_step_val) summary_writer.flush() info = ("global_step {0} | Batch Hit@1: {1:.3f} | Batch PERR: {2:.3f} | Batch Loss: {3:.3f} " "| Examples_per_sec: {4:.3f}").format( global_step_val, this_hit_at_one, this_perr, this_loss, examples_per_second) return info def AddEpochSummary(summary_writer, global_step_val, epoch_info_dict, summary_scope="Eval"): """Add the epoch summary to the Tensorboard. Args: summary_writer: Tensorflow summary_writer. global_step_val: a int value of the global step. epoch_info_dict: a dictionary of the evaluation metrics calculated for the whole epoch. summary_scope: Train or Eval. Returns: A string of this global_step summary """ epoch_id = epoch_info_dict["epoch_id"] avg_hit_at_one = epoch_info_dict["avg_hit_at_one"] avg_perr = epoch_info_dict["avg_perr"] avg_loss = epoch_info_dict["avg_loss"] aps = epoch_info_dict["aps"] gap = epoch_info_dict["gap"] mean_ap = numpy.mean(aps) summary_writer.add_summary( MakeSummary("Epoch/" + summary_scope + "_Avg_Hit@1", avg_hit_at_one), global_step_val) summary_writer.add_summary( MakeSummary("Epoch/" + summary_scope + "_Avg_Perr", avg_perr), global_step_val) summary_writer.add_summary( MakeSummary("Epoch/" + summary_scope + "_Avg_Loss", avg_loss), global_step_val) summary_writer.add_summary( MakeSummary("Epoch/" + summary_scope + "_MAP", mean_ap), global_step_val) summary_writer.add_summary( MakeSummary("Epoch/" + summary_scope + "_GAP", gap), global_step_val) summary_writer.flush() info = ("epoch/eval number {0} | Avg_Hit@1: {1:.3f} | Avg_PERR: {2:.3f} " "| MAP: {3:.3f} | GAP: {4:.3f} | Avg_Loss: {5:3f}").format( epoch_id, avg_hit_at_one, avg_perr, mean_ap, gap, avg_loss) return info def GetListOfFeatureNamesAndSizes(feature_names, feature_sizes): """Extract the list of feature names and the dimensionality of each feature from string of comma separated values. Args: feature_names: string containing comma separated list of feature names feature_sizes: string containing comma separated list of feature sizes Returns: List of the feature names and list of the dimensionality of each feature. Elements in the first/second list are strings/integers. """ list_of_feature_names = [ feature_names.strip() for feature_names in feature_names.split(',')] list_of_feature_sizes = [ int(feature_sizes) for feature_sizes in feature_sizes.split(',')] if len(list_of_feature_names) != len(list_of_feature_sizes): logging.error("length of the feature names (=" + str(len(list_of_feature_names)) + ") != length of feature " "sizes (=" + str(len(list_of_feature_sizes)) + ")") return list_of_feature_names, list_of_feature_sizes def clip_gradient_norms(gradients_to_variables, max_norm): clipped_grads_and_vars = [] for grad, var in gradients_to_variables: if grad is not None: if isinstance(grad, tf.IndexedSlices): tmp = tf.clip_by_norm(grad.values, max_norm) grad = tf.IndexedSlices(tmp, grad.indices, grad.dense_shape) else: grad = tf.clip_by_norm(grad, max_norm) clipped_grads_and_vars.append((grad, var)) return clipped_grads_and_vars ================================================ FILE: youtube-8m-wangheda/.vimrc ================================================ set tabstop=2 set shiftwidth=2 set expandtab set autoindent color desert ================================================ FILE: youtube-8m-wangheda/CONTRIBUTING.md ================================================ # How to contribute We are accepting patches and contributions to this project. To set expectations, this project is primarily intended to be a flexible starting point for researchers working with the YouTube-8M dataset. As such, we would like to keep it simple. We are more likely to accept small bug fixes and optimizations, and less likely to accept patches which add significant complexity. For the latter type of contribution, we recommend creating a Github fork of the project instead. If you would like to contribute, there are a few small guidelines you need to follow. ## Contributor License Agreement Contributions to any Google project must be accompanied by a Contributor License Agreement. This is necessary because you own the copyright to your changes, even after your contribution becomes part of this project. So this agreement simply gives us permission to use and redistribute your contributions as part of the project. Head over to to see your current agreements on file or to sign a new one. You generally only need to submit a CLA once, so if you've already submitted one (even if it was for a different project), you probably don't need to do it again. ## Code reviews All submissions, including submissions by project members, require review. We use GitHub pull requests for this purpose. Consult [GitHub Help] for more information on using pull requests. [GitHub Help]: https://help.github.com/articles/about-pull-requests/ ================================================ FILE: youtube-8m-wangheda/LICENSE ================================================ Apache License Version 2.0, January 2004 http://www.apache.org/licenses/ TERMS AND CONDITIONS FOR USE, REPRODUCTION, AND DISTRIBUTION 1. Definitions. "License" shall mean the terms and conditions for use, reproduction, and distribution as defined by Sections 1 through 9 of this document. "Licensor" shall mean the copyright owner or entity authorized by the copyright owner that is granting the License. "Legal Entity" shall mean the union of the acting entity and all other entities that control, are controlled by, or are under common control with that entity. For the purposes of this definition, "control" means (i) the power, direct or indirect, to cause the direction or management of such entity, whether by contract or otherwise, or (ii) ownership of fifty percent (50%) or more of the outstanding shares, or (iii) beneficial ownership of such entity. "You" (or "Your") shall mean an individual or Legal Entity exercising permissions granted by this License. "Source" form shall mean the preferred form for making modifications, including but not limited to software source code, documentation source, and configuration files. "Object" form shall mean any form resulting from mechanical transformation or translation of a Source form, including but not limited to compiled object code, generated documentation, and conversions to other media types. "Work" shall mean the work of authorship, whether in Source or Object form, made available under the License, as indicated by a copyright notice that is included in or attached to the work (an example is provided in the Appendix below). "Derivative Works" shall mean any work, whether in Source or Object form, that is based on (or derived from) the Work and for which the editorial revisions, annotations, elaborations, or other modifications represent, as a whole, an original work of authorship. For the purposes of this License, Derivative Works shall not include works that remain separable from, or merely link (or bind by name) to the interfaces of, the Work and Derivative Works thereof. "Contribution" shall mean any work of authorship, including the original version of the Work and any modifications or additions to that Work or Derivative Works thereof, that is intentionally submitted to Licensor for inclusion in the Work by the copyright owner or by an individual or Legal Entity authorized to submit on behalf of the copyright owner. For the purposes of this definition, "submitted" means any form of electronic, verbal, or written communication sent to the Licensor or its representatives, including but not limited to communication on electronic mailing lists, source code control systems, and issue tracking systems that are managed by, or on behalf of, the Licensor for the purpose of discussing and improving the Work, but excluding communication that is conspicuously marked or otherwise designated in writing by the copyright owner as "Not a Contribution." "Contributor" shall mean Licensor and any individual or Legal Entity on behalf of whom a Contribution has been received by Licensor and subsequently incorporated within the Work. 2. Grant of Copyright License. Subject to the terms and conditions of this License, each Contributor hereby grants to You a perpetual, worldwide, non-exclusive, no-charge, royalty-free, irrevocable copyright license to reproduce, prepare Derivative Works of, publicly display, publicly perform, sublicense, and distribute the Work and such Derivative Works in Source or Object form. 3. Grant of Patent License. Subject to the terms and conditions of this License, each Contributor hereby grants to You a perpetual, worldwide, non-exclusive, no-charge, royalty-free, irrevocable (except as stated in this section) patent license to make, have made, use, offer to sell, sell, import, and otherwise transfer the Work, where such license applies only to those patent claims licensable by such Contributor that are necessarily infringed by their Contribution(s) alone or by combination of their Contribution(s) with the Work to which such Contribution(s) was submitted. If You institute patent litigation against any entity (including a cross-claim or counterclaim in a lawsuit) alleging that the Work or a Contribution incorporated within the Work constitutes direct or contributory patent infringement, then any patent licenses granted to You under this License for that Work shall terminate as of the date such litigation is filed. 4. Redistribution. You may reproduce and distribute copies of the Work or Derivative Works thereof in any medium, with or without modifications, and in Source or Object form, provided that You meet the following conditions: (a) You must give any other recipients of the Work or Derivative Works a copy of this License; and (b) You must cause any modified files to carry prominent notices stating that You changed the files; and (c) You must retain, in the Source form of any Derivative Works that You distribute, all copyright, patent, trademark, and attribution notices from the Source form of the Work, excluding those notices that do not pertain to any part of the Derivative Works; and (d) If the Work includes a "NOTICE" text file as part of its distribution, then any Derivative Works that You distribute must include a readable copy of the attribution notices contained within such NOTICE file, excluding those notices that do not pertain to any part of the Derivative Works, in at least one of the following places: within a NOTICE text file distributed as part of the Derivative Works; within the Source form or documentation, if provided along with the Derivative Works; or, within a display generated by the Derivative Works, if and wherever such third-party notices normally appear. The contents of the NOTICE file are for informational purposes only and do not modify the License. You may add Your own attribution notices within Derivative Works that You distribute, alongside or as an addendum to the NOTICE text from the Work, provided that such additional attribution notices cannot be construed as modifying the License. You may add Your own copyright statement to Your modifications and may provide additional or different license terms and conditions for use, reproduction, or distribution of Your modifications, or for any such Derivative Works as a whole, provided Your use, reproduction, and distribution of the Work otherwise complies with the conditions stated in this License. 5. Submission of Contributions. Unless You explicitly state otherwise, any Contribution intentionally submitted for inclusion in the Work by You to the Licensor shall be under the terms and conditions of this License, without any additional terms or conditions. Notwithstanding the above, nothing herein shall supersede or modify the terms of any separate license agreement you may have executed with Licensor regarding such Contributions. 6. Trademarks. This License does not grant permission to use the trade names, trademarks, service marks, or product names of the Licensor, except as required for reasonable and customary use in describing the origin of the Work and reproducing the content of the NOTICE file. 7. Disclaimer of Warranty. Unless required by applicable law or agreed to in writing, Licensor provides the Work (and each Contributor provides its Contributions) on an "AS IS" BASIS, WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied, including, without limitation, any warranties or conditions of TITLE, NON-INFRINGEMENT, MERCHANTABILITY, or FITNESS FOR A PARTICULAR PURPOSE. You are solely responsible for determining the appropriateness of using or redistributing the Work and assume any risks associated with Your exercise of permissions under this License. 8. Limitation of Liability. In no event and under no legal theory, whether in tort (including negligence), contract, or otherwise, unless required by applicable law (such as deliberate and grossly negligent acts) or agreed to in writing, shall any Contributor be liable to You for damages, including any direct, indirect, special, incidental, or consequential damages of any character arising as a result of this License or out of the use or inability to use the Work (including but not limited to damages for loss of goodwill, work stoppage, computer failure or malfunction, or any and all other commercial damages or losses), even if such Contributor has been advised of the possibility of such damages. 9. Accepting Warranty or Additional Liability. While redistributing the Work or Derivative Works thereof, You may choose to offer, and charge a fee for, acceptance of support, warranty, indemnity, or other liability obligations and/or rights consistent with this License. However, in accepting such obligations, You may act only on Your own behalf and on Your sole responsibility, not on behalf of any other Contributor, and only if You agree to indemnify, defend, and hold each Contributor harmless for any liability incurred by, or claims asserted against, such Contributor by reason of your accepting any such warranty or additional liability. END OF TERMS AND CONDITIONS APPENDIX: How to apply the Apache License to your work. To apply the Apache License to your work, attach the following boilerplate notice, with the fields enclosed by brackets "[]" replaced with your own identifying information. (Don't include the brackets!) The text should be enclosed in the appropriate comment syntax for the file format. We also recommend that a file or class name and description of purpose be included on the same "printed page" as the copyright notice for easier identification within third-party archives. Copyright [yyyy] [name of copyright owner] Licensed under the Apache License, Version 2.0 (the "License"); you may not use this file except in compliance with the License. You may obtain a copy of the License at http://www.apache.org/licenses/LICENSE-2.0 Unless required by applicable law or agreed to in writing, software distributed under the License is distributed on an "AS IS" BASIS, WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied. See the License for the specific language governing permissions and limitations under the License. ================================================ FILE: youtube-8m-wangheda/README.md ================================================ # YouTube-8M Tensorflow Starter Code This repo contains starter code for training and evaluating machine learning models over the [YouTube-8M](https://research.google.com/youtube8m/) dataset. The code gives an end-to-end working example for reading the dataset, training a TensorFlow model, and evaluating the performance of the model. Out of the box, you can train several [model architectures](#overview-of-models) over either frame-level or video-level features. The code can easily be extended to train your own custom-defined models. It is possible to train and evaluate on YouTube-8M in two ways: on Google Cloud or on your own machine. This README provides instructions for both. ## Table of Contents * [Running on Google's Cloud Machine Learning Platform](#running-on-googles-cloud-machine-learning-platform) * [Requirements](#requirements) * [Testing Locally](#testing-locally) * [Training on the Cloud over Video-Level Features](#training-on-video-level-features) * [Evaluation and Inference](#evaluation-and-inference) * [Accessing Files on Google Cloud](#accessing-files-on-google-cloud) * [Using Frame-Level Features](#using-frame-level-features) * [Using Audio Features](#using-audio-features) * [Running on Your Own Machine](#running-on-your-own-machine) * [Requirements](#requirements-1) * [Training on Video-Level Features](#training-on-video-level-features-1) * [Evaluation and Inference](#evaluation-and-inference-1) * [Using Frame-Level Features](#using-frame-level-features-1) * [Using Audio Features](#using-audio-features-1) * [Ground-Truth Label Files](#ground-truth-label-files) * [Overview of Models](#overview-of-models) * [Video-Level Models](#video-level-models) * [Frame-Level Models](#frame-level-models) * [Overview of Files](#overview-of-files) * [Training](#training) * [Evaluation](#evaluation) * [Inference](#inference) * [Misc](#misc) * [About This Project](#about-this-project) ## Running on Google's Cloud Machine Learning Platform ### Requirements This option requires you to have an appropriately configured Google Cloud Platform account. To create and configure your account, please make sure you follow the instructions [here](https://cloud.google.com/ml/docs/how-tos/getting-set-up). If you are participating in the Google Cloud & YouTube-8M Video Understanding Challenge hosted on [kaggle](https://www.kaggle.com/c/youtube8m), see [these instructions](https://www.kaggle.com/c/youtube8m#getting-started-with-google-cloud) instead. Please also verify that you have Python 2.7+ and Tensorflow 1.0.0 or higher installed by running the following commands: ```sh python --version python -c 'import tensorflow as tf; print(tf.__version__)' ``` ### Testing Locally All gcloud commands should be done from the directory *immediately above* the source code. You should be able to see the source code directory if you run 'ls'. As you are developing your own models, you will want to test them quickly to flush out simple problems without having to submit them to the cloud. You can use the `gcloud beta ml local` set of commands for that. Here is an example command line for video-level training: ```sh gcloud beta ml local train \ --package-path=youtube-8m --module-name=youtube-8m.train -- \ --train_data_pattern='gs://youtube8m-ml/1/video_level/train/train*.tfrecord' \ --train_dir=/tmp/yt8m_train --start_new_model ``` You might want to download some training shards locally to speed things up and allow you to work offline. The command below will copy 10 out of the 4096 training data files to the current directory. ```sh # Downloads 55MB of data. gsutil cp gs://us.data.yt8m.org/1/video_level/train/traina[0-9].tfrecord . ``` Once you download the files, you can point the job to them using the 'train_data_pattern' argument (i.e. instead of pointing to the "gs://..." files, you point to the local files). Once your model is working locally, you can scale up on the Cloud which is described below. ### Training on the Cloud over Video-Level Features The following commands will train a model on Google Cloud over video-level features. ```sh BUCKET_NAME=gs://${USER}_yt8m_train_bucket # (One Time) Create a storage bucket to store training logs and checkpoints. gsutil mb -l us-east1 $BUCKET_NAME # Submit the training job. JOB_NAME=yt8m_train_$(date +%Y%m%d_%H%M%S); gcloud --verbosity=debug beta ml jobs \ submit training $JOB_NAME \ --package-path=youtube-8m --module-name=youtube-8m.train \ --staging-bucket=$BUCKET_NAME --region=us-east1 \ --config=youtube-8m/cloudml-gpu.yaml \ -- --train_data_pattern='gs://youtube8m-ml-us-east1/1/video_level/train/train*.tfrecord' \ --train_dir=$BUCKET_NAME/yt8m_train_video_level_logistic_model ``` In the 'gsutil' command above, the 'package-path' flag refers to the directory containing the 'train.py' script and more generally the python package which should be deployed to the cloud worker. The module-name refers to the specific python script which should be executed (in this case the train module). It may take several minutes before the job starts running on Google Cloud. When it starts you will see outputs like the following: ``` training step 270| Hit@1: 0.68 PERR: 0.52 Loss: 638.453 training step 271| Hit@1: 0.66 PERR: 0.49 Loss: 635.537 training step 272| Hit@1: 0.70 PERR: 0.52 Loss: 637.564 ``` At this point you can disconnect your console by pressing "ctrl-c". The model will continue to train indefinitely in the Cloud. Later, you can check on its progress or halt the job by visiting the [Google Cloud ML Jobs console](https://console.cloud.google.com/ml/jobs). You can train many jobs at once and use tensorboard to compare their performance visually. ```sh tensorboard --logdir=$BUCKET_NAME --port=8080 ``` Once tensorboard is running, you can access it at the following url: [http://localhost:8080](http://localhost:8080). If you are using Google Cloud Shell, you can instead click the Web Preview button on the upper left corner of the Cloud Shell window and select "Preview on port 8080". This will bring up a new browser tab with the Tensorboard view. ### Evaluation and Inference Here's how to evaluate a model on the validation dataset: ```sh JOB_TO_EVAL=yt8m_train_video_level_logistic_model JOB_NAME=yt8m_eval_$(date +%Y%m%d_%H%M%S); gcloud --verbosity=debug beta ml jobs \ submit training $JOB_NAME \ --package-path=youtube-8m --module-name=youtube-8m.eval \ --staging-bucket=$BUCKET_NAME --region=us-east1 \ --config=youtube-8m/cloudml-gpu.yaml \ -- --eval_data_pattern='gs://youtube8m-ml-us-east1/1/video_level/validate/validate*.tfrecord' \ --train_dir=$BUCKET_NAME/${JOB_TO_EVAL} --run_once=True ``` And here's how to perform inference with a model on the test set: ```sh JOB_TO_EVAL=yt8m_train_video_level_logistic_model JOB_NAME=yt8m_inference_$(date +%Y%m%d_%H%M%S); gcloud --verbosity=debug beta ml jobs \ submit training $JOB_NAME \ --package-path=youtube-8m --module-name=youtube-8m.inference \ --staging-bucket=$BUCKET_NAME --region=us-east1 \ --config=youtube-8m/cloudml-gpu.yaml \ -- --input_data_pattern='gs://youtube8m-ml/1/video_level/test/test*.tfrecord' \ --train_dir=$BUCKET_NAME/${JOB_TO_EVAL} \ --output_file=$BUCKET_NAME/${JOB_TO_EVAL}/predictions.csv ``` Note the confusing use of 'training' in the above gcloud commands. Despite the name, the 'training' argument really just offers a cloud hosted python/tensorflow service. From the point of view of the Cloud Platform, there is no distinction between our training and inference jobs. The Cloud ML platform also offers specialized functionality for prediction with Tensorflow models, but discussing that is beyond the scope of this readme. Once these job starts executing you will see outputs similar to the following for the evaluation code: ``` examples_processed: 1024 | global_step 447044 | Batch Hit@1: 0.782 | Batch PERR: 0.637 | Batch Loss: 7.821 | Examples_per_sec: 834.658 ``` and the following for the inference code: ``` num examples processed: 8192 elapsed seconds: 14.85 ``` ### Accessing Files on Google Cloud You can browse the storage buckets you created on Google Cloud, for example, to access the trained models, prediction CSV files, etc. by visiting the [Google Cloud storage browser](https://console.cloud.google.com/storage/browser). Alternatively, you can use the 'gsutil' command to download the files directly. For example, to download the output of the inference code from the previous section to your local machine, run: ``` gsutil cp $BUCKET_NAME/${JOB_TO_EVAL}/predictions.csv . ``` ### Using Frame-Level Features Append ```sh --frame_features=True --model=FrameLevelLogisticModel --feature_names="rgb" \ --feature_sizes="1024" --batch_size=128 \ --train_dir=$BUCKET_NAME/yt8m_train_frame_level_logistic_model ``` to the 'gcloud' commands given above, and change 'video_level' in paths to 'frame_level'. Here is a sample command to kick-off a frame-level job: ```sh JOB_NAME=yt8m_train_$(date +%Y%m%d_%H%M%S); gcloud --verbosity=debug beta ml jobs \ submit training $JOB_NAME \ --package-path=youtube-8m --module-name=youtube-8m.train \ --staging-bucket=$BUCKET_NAME --region=us-east1 \ --config=youtube-8m/cloudml-gpu.yaml \ -- --train_data_pattern='gs://youtube8m-ml-us-east1/1/frame_level/train/train*.tfrecord' \ --frame_features=True --model=FrameLevelLogisticModel --feature_names="rgb" \ --feature_sizes="1024" --batch_size=128 \ --train_dir=$BUCKET_NAME/yt8m_train_frame_level_logistic_model ``` The 'FrameLevelLogisticModel' is designed to provide equivalent results to a logistic model trained over the video-level features. Please look at the 'video_level_models.py' or 'frame_level_models.py' files to see how to implement your own models. ### Using Audio Features The feature files (both Frame-Level and Video-Level) contain two sets of features: 1) visual and 2) audio. The code defaults to using the visual features only, but it is possible to use audio features instead of (or besides) visual features. To specify the (combination of) features to use you must set `--feature_names` and `--feature_sizes` flags. The visual and audio features are called 'rgb' and 'audio' and have 1024 and 128 dimensions, respectively. The two flags take a comma-separated list of values in string. For example, to use audio-visual Video-Level features the flags must be set as follows: ``` --feature_names="mean_rgb, mean_audio" --feature_sizes="1024, 128" ``` Similarly, to use audio-visual Frame-Level features use: ``` --feature_names="rgb, audio" --feature_sizes="1024, 128" ``` **NOTE:** Make sure the set of features and the order in which the appear in the lists provided to the two flags above match. Also, the order must match when running training, evaluation, or inference. ## Running on Your Own Machine ### Requirements The starter code requires Tensorflow. If you haven't installed it yet, follow the instructions on [tensorflow.org](https://www.tensorflow.org/install/). This code has been tested with Tensorflow 1.0.0. Going forward, we will continue to target the latest released version of Tensorflow. Please verify that you have Python 2.7+ and Tensorflow 1.0.0 or higher installed by running the following commands: ```sh python --version python -c 'import tensorflow as tf; print(tf.__version__)' ``` You can find complete instructions for downloading the dataset on the [YouTube-8M website](https://research.google.com/youtube8m/download.html). We recommend downloading the smaller video-level features dataset first when getting started. To do that, run: ``` mkdir -p features; cd features curl data.yt8m.org/download.py | partition=1/video_level/train mirror=us python ``` This will download the full set of video level features, which takes up 31GB of space. If you are located outside of North America, you should change the flag 'mirror' to 'eu' for Europe or 'asia' for Asia to speed up the transfer of the files. Change 'train' to 'validate'/'test' and re-run the command to download the other splits of the dataset. Change 'video_level' to 'frame_level' to download the frame-level features. The complete frame-level features take about 1.71TB of space. You can set the environment variable 'shard' to 'm,n' to download only m/n-th of the data. For example, to download 1/100-th of the frame-level features from the training set, run: ``` curl data.yt8m.org/download.py | shard=1,100 partition=1/frame_level/train mirror=us python ``` ### Training on Video-Level Features To start training a logistic model on the video-level features, run ```sh MODEL_DIR=/tmp/yt8m python train.py --train_data_pattern='/path/to/features/train*.tfrecord' --train_dir=$MODEL_DIR/video_level_logistic_model ``` Since the dataset is sharded into 4096 individual files, we use a wildcard (\*) to represent all of those files. By default, the training code will frequently write _checkpoint_ files (i.e. values of all trainable parameters, at the current training iteration). These will be written to the `--train_dir`. If you re-use a `--train_dir`, the trainer will first restore the latest checkpoint written in that directory. This only works if the architecture of the checkpoint matches the graph created by the training code. If you are in active development/debugging phase, consider adding `--start_new_model` flag to your run configuration. ### Evaluation and Inference To evaluate the model, run ```sh python eval.py --eval_data_pattern='/path/to/features/validate*.tfrecord' --train_dir=$MODEL_DIR/video_level_logistic_model --run_once=True ``` As the model is training or evaluating, you can view the results on tensorboard by running ```sh tensorboard --logdir=$MODEL_DIR ``` and navigating to http://localhost:6006 in your web browser. When you are happy with your model, you can generate a csv file of predictions from it by running ```sh python inference.py --output_file=$MODEL_DIR/video_level_logistic_model/predictions.csv --input_data_pattern='/path/to/features/test*.tfrecord' --train_dir=$MODEL_DIR/video_level_logistic_model ``` This will output the top 20 predicted labels from the model for every example to 'predictions.csv'. ### Using Frame-Level Features Follow the same instructions as above, appending `--frame_features=True --model=FrameLevelLogisticModel --feature_names="rgb" --feature_sizes="1024" --train_dir=$MODEL_DIR/frame_level_logistic_model` for the 'train.py', 'eval.py', and 'inference.py' scripts. The 'FrameLevelLogisticModel' is designed to provide equivalent results to a logistic model trained over the video-level features. Please look at the 'models.py' file to see how to implement your own models. ### Using Audio Features See [Using Audio Features](#using-audio-features) section above. ### Ground-Truth Label Files We also provide CSV files containing the ground-truth label information of the 'train' and 'validation' partitions of the dataset. These files can be downloaded using 'gsutil' command: ``` gsutil cp gs://us.data.yt8m.org/1/ground_truth_labels/train_labels.csv /destination/folder/ gsutil cp gs://us.data.yt8m.org/1/ground_truth_labels/validate_labels.csv /destination/folder/ ``` or directly using the following links: * [http://us.data.yt8m.org/1/ground_truth_labels/train_labels.csv](http://us.data.yt8m.org/1/ground_truth_labels/train_labels.csv) * [http://us.data.yt8m.org/1/ground_truth_labels/validate_labels.csv](http://us.data.yt8m.org/1/ground_truth_labels/validate_labels.csv) Each line in the files starts with the video id and is followed by the list of ground-truth labels corresponding to that video. For example, for a video with id 'VIDEO_ID' and two lables 'LABLE1' and 'LABEL2' we store the following line: ``` VIDEO_ID,LABEL1 LABEL2 ``` ## Overview of Models This sample code contains implementations of the models given in the [YouTube-8M technical report](https://arxiv.org/abs/1609.08675). ### Video-Level Models * `LogisticModel`: Linear projection of the output features into the label space, followed by a sigmoid function to convert logit values to probabilities. * `MoeModel`: A per-class softmax distribution over a configurable number of logistic classifiers. One of the classifiers in the mixture is not trained, and always predicts 0. ### Frame-Level Models * `LstmModel`: Processes the features for each frame using a multi-layered LSTM neural net. The final internal state of the LSTM is input to a video-level model for classification. Note that you will need to change the learning rate to 0.001 when using this model. * `DbofModel`: Projects the features for each frame into a higher dimensional 'clustering' space, pools across frames in that space, and then uses a video-level model to classify the now aggregated features. * `FrameLevelLogisticModel`: Equivalent to 'LogisticModel', but performs average-pooling on the fly over frame-level features rather than using pre-aggregated features. ## Overview of Files ### Training * `train.py`: The primary script for training models. * `losses.py`: Contains definitions for loss functions. * `models.py`: Contains the base class for defining a model. * `video_level_models.py`: Contains definitions for models that take aggregated features as input. * `frame_level_models.py`: Contains definitions for models that take frame- level features as input. * `model_util.py`: Contains functions that are of general utility for implementing models. * `readers.py`: Contains definitions for the Video dataset and Frame dataset readers. ### Evaluation * `eval.py`: The primary script for evaluating models. * `eval_util.py`: Provides a class that calculates all evaluation metrics. * `average_precision_calculator.py`: Functions for calculating average precision. * `mean_average_precision_calculator.py`: Functions for calculating mean average precision. ### Inference * `inference.py`: Generates an output file containing predictions of the model over a set of videos. ### Misc * `README.md`: This documentation. * `utils.py`: Common functions. ## About This Project This project is meant help people quickly get started working with the [YouTube-8M](https://research.google.com/youtube8m/) dataset. This is not an official Google product. ================================================ FILE: youtube-8m-wangheda/__init__.py ================================================ # Copyright 2016 Google Inc. All Rights Reserved. # # Licensed under the Apache License, Version 2.0 (the "License"); # you may not use this file except in compliance with the License. # You may obtain a copy of the License at # # http://www.apache.org/licenses/LICENSE-2.0 # # Unless required by applicable law or agreed to in writing, software # distributed under the License is distributed on an "AS-IS" BASIS, # WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied. # See the License for the specific language governing permissions and # limitations under the License. ================================================ FILE: youtube-8m-wangheda/all_data_augmentation/__init__.py ================================================ from default_augmenter import * from half_augmenter import * from half_video_augmenter import * from noise_augmenter import * from clipping_augmenter import * ================================================ FILE: youtube-8m-wangheda/all_data_augmentation/clipping_augmenter.py ================================================ import tensorflow as tf from tensorflow import flags FLAGS = flags.FLAGS class ClippingAugmenter: """This only works with frame data""" def augment(self, model_input_raw, num_frames, labels_batch, **unused_params): assert(FLAGS.frame_feature, "AugmentationTransformer only works with frame feature") feature_dim = len(model_input_raw.get_shape()) - 1 frame_dim = len(model_input_raw.get_shape()) - 2 max_frame = model_input_raw.get_shape().as_list()[frame_dim] limit = tf.cast(tf.reduce_min(num_frames) / 4.0, tf.int32) offset = tf.random_uniform(shape=[], dtype=tf.int32) % limit input_trans1 = tf.pad(model_input_raw[:,offset:,:], paddings=[0,offset,0]) num_frames_trans1 = num_frames - offset num_frames_trans1 = tf.cast( tf.random_uniform(shape=num_frames.shape, minval=0.75, maxval=1.0, dtype=tf.float32) * num_frames_trans1, tf.int32) model_input = tf.concat([model_input_raw, input_trans1], axis=0) labels_batch = tf.concat([labels_batch, labels_batch], axis=0) num_frames = tf.concat([num_frames, num_frames_trans1], axis=0) return model_input, labels_batch, num_frames_new ================================================ FILE: youtube-8m-wangheda/all_data_augmentation/default_augmenter.py ================================================ import tensorflow as tf from tensorflow import flags FLAGS = flags.FLAGS class DefaultAugmenter: """This only works with frame data""" def augment(self, model_input_raw, num_frames, labels_batch, **unused_params): print "DefaultAugmenter" return model_input_raw, labels_batch, num_frames ================================================ FILE: youtube-8m-wangheda/all_data_augmentation/half_augmenter.py ================================================ import tensorflow as tf from tensorflow import flags FLAGS = flags.FLAGS class HalfAugmenter: """This only works with frame data""" def augment(self, model_input_raw, num_frames, labels_batch, **unused_params): assert(FLAGS.frame_features, "HalfAugmenter only works with frame feature") print "using HalfAugmeter" feature_dim = len(model_input_raw.get_shape()) - 1 frame_dim = len(model_input_raw.get_shape()) - 2 max_frame = model_input_raw.get_shape().as_list()[frame_dim] seg_length = max(int(max_frame / 2), 1) seg_num_frames = tf.maximum(num_frames / 2, 1) seg_inputs = [] seg_frames = [] seg_labels = [] seg_inputs.append(model_input_raw) seg_frames.append(num_frames) seg_labels.append(labels_batch) for i in xrange(2): begin_frames = tf.reshape(seg_num_frames*i, [-1,1]) frames_index = tf.reshape(tf.range(seg_length), [1,seg_length]) frames_index = begin_frames + frames_index batch_size = tf.shape(model_input_raw)[0] batch_index = tf.tile(tf.expand_dims(tf.range(batch_size), 1), [1, seg_length]) index = tf.stack([batch_index, tf.cast(frames_index,dtype=tf.int32)], 2) seg_input = tf.gather_nd(model_input_raw, index) seg_input = tf.pad(seg_input, paddings=[[0,0],[0, max_frame-seg_length],[0,0]]) seg_input = seg_input * tf.expand_dims(tf.sequence_mask(seg_num_frames, maxlen=max_frame, dtype=tf.float32), axis=2) seg_inputs.append(seg_input) seg_frames.append(seg_num_frames) seg_labels.append(labels_batch) new_input_raw = tf.concat(seg_inputs, axis=0) new_num_frames = tf.concat(seg_frames, axis=0) new_labels_batch = tf.concat(seg_labels, axis=0) return new_input_raw, new_labels_batch, new_num_frames ================================================ FILE: youtube-8m-wangheda/all_data_augmentation/half_video_augmenter.py ================================================ import tensorflow as tf from tensorflow import flags FLAGS = flags.FLAGS class HalfVideoAugmenter: """This only works with frame data""" def augment(self, model_input_raw, num_frames, labels_batch, **unused_params): assert(FLAGS.frame_features, "HalfVideoAugmenter only works with frame feature") print "using HalfVideoAugmeter" new_input_raw, new_labels_batch, new_num_frames = self.frame_augment(model_input_raw, num_frames, labels_batch) new_float_frames = tf.expand_dims(tf.cast(new_num_frames, tf.float32), axis=1) aggregated_inputs = tf.reduce_sum(new_input_raw, axis=1) / new_float_frames return aggregated_inputs, new_labels_batch, new_num_frames def frame_augment(self, model_input_raw, num_frames, labels_batch, **unused_params): feature_dim = len(model_input_raw.get_shape()) - 1 frame_dim = len(model_input_raw.get_shape()) - 2 max_frame = model_input_raw.get_shape().as_list()[frame_dim] seg_length = max(int(max_frame / 2), 1) seg_num_frames = tf.maximum(num_frames / 2, 1) seg_inputs = [] seg_frames = [] seg_labels = [] seg_inputs.append(model_input_raw) seg_frames.append(num_frames) seg_labels.append(labels_batch) for i in xrange(2): begin_frames = tf.reshape(seg_num_frames*i, [-1,1]) frames_index = tf.reshape(tf.range(seg_length), [1,seg_length]) frames_index = begin_frames + frames_index batch_size = tf.shape(model_input_raw)[0] batch_index = tf.tile(tf.expand_dims(tf.range(batch_size), 1), [1, seg_length]) index = tf.stack([batch_index, tf.cast(frames_index,dtype=tf.int32)], 2) seg_input = tf.gather_nd(model_input_raw, index) seg_input = tf.pad(seg_input, paddings=[[0,0],[0, max_frame-seg_length],[0,0]]) seg_input = seg_input * tf.expand_dims(tf.sequence_mask(seg_num_frames, maxlen=max_frame, dtype=tf.float32), axis=2) seg_inputs.append(seg_input) seg_frames.append(seg_num_frames) seg_labels.append(labels_batch) new_input_raw = tf.concat(seg_inputs, axis=0) new_num_frames = tf.concat(seg_frames, axis=0) new_labels_batch = tf.concat(seg_labels, axis=0) return new_input_raw, new_labels_batch, new_num_frames ================================================ FILE: youtube-8m-wangheda/all_data_augmentation/noise_augmenter.py ================================================ import tensorflow as tf from tensorflow import flags FLAGS = flags.FLAGS class NoiseAugmenter: """This only works with frame data""" def augment(self, model_input_raw, num_frames, labels_batch, **unused_params): print "NoiseAugmenter", model_input_raw.shape, "noise =", FLAGS.input_noise_level noise_input = tf.random_normal(tf.shape(model_input_raw), mean=0.0, stddev=FLAGS.input_noise_level) model_input = model_input_raw + noise_input return model_input, labels_batch, num_frames ================================================ FILE: youtube-8m-wangheda/all_feature_transform/__init__.py ================================================ from default_transformer import * from identical_transformer import * from engineer_transformer import * from avg_transformer import * from resolution_transformer import * ================================================ FILE: youtube-8m-wangheda/all_feature_transform/avg_transformer.py ================================================ import tensorflow as tf class AvgTransformer: def transform(self, model_input_raw, num_frames, **unused_params): float_num_frames = tf.cast(tf.expand_dims(num_frames, 1), tf.float32) feature_size = model_input_raw.get_shape().as_list()[2] denominators = tf.reshape( tf.tile(float_num_frames, [1, feature_size]), [-1, feature_size]) avg_pooled = tf.reduce_sum(model_input_raw, axis=[1]) / denominators feature_dim = len(avg_pooled.get_shape()) - 1 model_input = tf.nn.l2_normalize(avg_pooled, feature_dim) return model_input, num_frames ================================================ FILE: youtube-8m-wangheda/all_feature_transform/default_transformer.py ================================================ import tensorflow as tf class DefaultTransformer: def transform(self, model_input_raw, num_frames, **unused_params): feature_dim = len(model_input_raw.get_shape()) - 1 model_input = tf.nn.l2_normalize(model_input_raw, feature_dim) return model_input, num_frames ================================================ FILE: youtube-8m-wangheda/all_feature_transform/engineer_transformer.py ================================================ import tensorflow as tf import numpy as np from tensorflow import flags FLAGS = flags.FLAGS class EngineerTransformer: """feature transform by feature engineering""" def transform(self, model_input_raw, num_frames, **unused_params): feature_dim = len(model_input_raw.get_shape()) - 1 engineer_types = map(lambda x: x.strip(), FLAGS.engineer_types.split(",")) feature_list = [] mask = self.mask(model_input_raw, num_frames) for etype in engineer_types: if etype == "avg": feature_list.append(self.avg(model_input_raw, num_frames, mask)) elif etype == "std": feature_list.append(self.std(model_input_raw, num_frames, mask)) elif etype == "diff": feature_list.append(self.diff(model_input_raw, num_frames, mask)) else: feature_list.append(model_input_raw) model_input = tf.nn.l2_normalize(tf.concat(model_input_raw, axis=feature_dim), feature_dim) return model_input, num_frames def mask(self, model_input_raw, num_frames): max_frames = model_input_raw.get_shape().as_list()[1] mask_array = [] for i in xrange(max_frames + 1): tmp = [0.0] * max_frames for j in xrange(i): tmp[j] = 1.0 mask_array.append(tmp) mask_array = np.array(mask_array) mask_init = tf.constant_initializer(mask_array) mask_emb = tf.get_variable("mask_emb", shape = [max_frames + 1, max_frames], dtype = tf.float32, trainable = False, initializer = mask_init) mask = tf.nn.embedding_lookup(mask_emb, num_frames) return mask def avg(self, model_input_raw, num_frames, mask): max_frames = model_input_raw.get_shape().as_list()[1] num_frames_matrix = tf.maximum(tf.cast( tf.expand_dims(num_frames, axis=1), dtype=tf.float32), 1.0) mean_matrix = mask / num_frames_matrix mean_input = tf.einsum("ijk,ij->ik", model_input_raw, mean_matrix) mean_input_tile = tf.tile(tf.expand_dims(mean_input, axis=1), multiples=[1,max_frames,1]) return mean_input_tile def std(self, model_input_raw, num_frames, mask): mean_input = self.avg(model_input_raw, num_frames, mask) error = tf.einsum("ijk,ij->ijk", model_input_raw - mean_input, mask) return error def diff(self, model_input_raw, num_frames, mask): max_frames = model_input_raw.get_shape().as_list()[1] shift_input1 = tf.pad(model_input_raw, paddings=[[0,0], [0,1], [0,0]]) shift_input2 = tf.pad(model_input_raw, paddings=[[0,0], [1,0], [0,0]]) diff_input = shift_input1 - shift_input2 difference = tf.einsum("ijk,ij->ijk", diff_input[:,:max_frames,:], mask) return difference ================================================ FILE: youtube-8m-wangheda/all_feature_transform/identical_transformer.py ================================================ import tensorflow as tf class IdenticalTransformer: def transform(self, model_input_raw, num_frames, **unused_params): return model_input_raw, num_frames ================================================ FILE: youtube-8m-wangheda/all_feature_transform/resolution_transformer.py ================================================ import tensorflow as tf from tensorflow import flags FLAGS = flags.FLAGS class ResolutionTransformer: def resolution(self, model_input_raw, num_frames): resolution = FLAGS.time_resolution frame_dim = len(model_input_raw.get_shape()) - 2 feature_dim = len(model_input_raw.get_shape()) - 1 max_frames = model_input_raw.get_shape().as_list()[frame_dim] num_features = model_input_raw.get_shape().as_list()[feature_dim] new_max_frames = max_frames / resolution cut_frames = new_max_frames * resolution model_input_raw = model_input_raw[:, :cut_frames, :] model_input_raw = tf.reshape(model_input_raw, shape=[-1,new_max_frames,resolution,num_features]) model_input_raw = tf.reduce_mean(model_input_raw, axis=2) num_frames = num_frames / resolution return model_input_raw, num_frames def transform(self, model_input_raw, num_frames, **unused_params): model_input_raw, num_frames = self.resolution(model_input_raw, num_frames) feature_dim = len(model_input_raw.get_shape()) - 1 model_input = tf.nn.l2_normalize(model_input_raw, feature_dim) return model_input, num_frames ================================================ FILE: youtube-8m-wangheda/all_frame_models/.vimrc ================================================ syntax on set tabstop=2 set shiftwidth=2 set expandtab set autoindent ================================================ FILE: youtube-8m-wangheda/all_frame_models/__init__.py ================================================ from bilstm_model import * from biunilstm_model import * from cnn_kmax_model import * from dbof_model import * from frame_seg_model import * from deep_lstm_model import * from gru_pooling_model import * from gru_with_pooling_model import * from logistic_model import * from lstm_advanced_model import * from lstm_attention_lstm_model import * from lstm_attention_model import * from lstm_attention_max_pooling_model import * from distillchain_lstm_attention_max_pooling_model import * from lstm_positional_attention_max_pooling_model import * from lstm_divided_model import * from lstm_memory_model import * from layernorm_lstm_memory_model import * from lstm_parallel_memory_model import * from lstm_parallel_finaloutput_model import * from distillchain_lstm_parallel_finaloutput_model import * from cnn_model import * from cnn_deep_combine_chain_model import * from distillchain_cnn_deep_combine_chain_model import * from positional_cnn_deep_combine_chain_model import * from multi_view_cnn_deep_combine_chain_model import * from lstm_cnn_deep_combine_chain_model import * from distillchain_lstm_cnn_deep_combine_chain_model import * from deep_cnn_deep_combine_chain_model import * from cnn_lstm_memory_model import * from cnn_lstm_memory_multitask_model import * from cnn_lstm_memory_normalization_model import * from mm_lstm_memory_model import * from lstm_memory_normalization_model import * from lstm_model import * from lstm_multi_attention_model import * from lstm_multi_pooling_model import * from lstm_parallel_model import * from lstm_pooling_model import * from lstm_with_mean_input_model import * from lstm_with_pooling_model import * from progressive_attention_lstm_model import * from lstm_memory_multitask_model import * from lstm_memory_chain_model import * from lstm_memory_deep_chain_model import * from distillchain_lstm_memory_deep_combine_chain_model import * from lstm_memory_input_chain_model import * from lstm_memory_parallel_chain_model import * from wide_and_deep_model import * from multires_lstm_memory_deep_combine_chain_model import * from framehop_lstm_memory_deep_combine_chain_model import * from framehop_lstm_memory_model import * from lstm_look_back_model import * from lstm_auxloss_deep_combine_chain_model import * from multiscale_cnn_lstm_model import * from distillchain_multiscale_cnn_lstm_model import * ================================================ FILE: youtube-8m-wangheda/all_frame_models/bilstm_model.py ================================================ import sys import models import model_utils import math import numpy as np import video_level_models import tensorflow as tf import utils import tensorflow.contrib.slim as slim from tensorflow import flags FLAGS = flags.FLAGS class BiLstmModel(models.BaseModel): def create_model(self, model_input, vocab_size, num_frames, **unused_params): """Creates a model which uses a stack of Bi LSTMs to represent the video. Args: model_input: A 'batch_size' x 'max_frames' x 'num_features' matrix of input features. vocab_size: The number of classes in the dataset. num_frames: A vector of length 'batch' which indicates the number of frames for each video (before padding). Returns: A dictionary with a tensor containing the probability predictions of the model in the 'predictions' key. The dimensions of the tensor are 'batch_size' x 'num_classes'. """ lstm_size = int(FLAGS.lstm_cells) number_of_layers = FLAGS.lstm_layers ## Batch normalize the input fw_stacked_lstm = tf.contrib.rnn.MultiRNNCell( [ tf.contrib.rnn.BasicLSTMCell( lstm_size, forget_bias=1.0, state_is_tuple=False) for _ in range(number_of_layers) ], state_is_tuple=False) bw_stacked_lstm = tf.contrib.rnn.MultiRNNCell( [ tf.contrib.rnn.BasicLSTMCell( lstm_size, forget_bias=1.0, state_is_tuple=False) for _ in range(number_of_layers) ], state_is_tuple=False) loss = 0.0 with tf.variable_scope("RNN"): outputs, states = tf.nn.bidirectional_dynamic_rnn(cell_fw = fw_stacked_lstm, cell_bw = bw_stacked_lstm, inputs = model_input, sequence_length=num_frames, swap_memory=FLAGS.rnn_swap_memory, dtype=tf.float32) state_fw, state_bw = states state = tf.concat([state_fw, state_bw], axis = 1) aggregated_model = getattr(video_level_models, FLAGS.video_level_classifier_model) return aggregated_model().create_model( model_input=state, original_input=model_input, vocab_size=vocab_size, **unused_params) ================================================ FILE: youtube-8m-wangheda/all_frame_models/biunilstm_model.py ================================================ import sys import models import model_utils import math import numpy as np import video_level_models import tensorflow as tf import utils import tensorflow.contrib.slim as slim from tensorflow import flags FLAGS = flags.FLAGS class BiUniLstmModel(models.BaseModel): def create_model(self, model_input, vocab_size, num_frames, **unused_params): """Creates a model which uses a stack of Bi-Uni LSTMs to represent the video. Args: model_input: A 'batch_size' x 'max_frames' x 'num_features' matrix of input features. vocab_size: The number of classes in the dataset. num_frames: A vector of length 'batch' which indicates the number of frames for each video (before padding). Returns: A dictionary with a tensor containing the probability predictions of the model in the 'predictions' key. The dimensions of the tensor are 'batch_size' x 'num_classes'. """ lstm_size = int(FLAGS.lstm_cells) ## Batch normalize the input fw_cell = tf.contrib.rnn.BasicLSTMCell( lstm_size, forget_bias=1.0, state_is_tuple=False) bw_cell = tf.contrib.rnn.BasicLSTMCell( lstm_size, forget_bias=1.0, state_is_tuple=False) cell = tf.contrib.rnn.BasicLSTMCell( lstm_size, forget_bias=1.0, state_is_tuple=False) loss = 0.0 with tf.variable_scope("RNN"): l1_outputs, l1_states = tf.nn.bidirectional_dynamic_rnn(cell_fw = fw_cell, cell_bw = bw_cell, inputs = model_input, sequence_length=num_frames, swap_memory=FLAGS.rnn_swap_memory, dtype=tf.float32) l1_outputs = tf.concat(l1_outputs, axis = 2) l2_outputs, l2_states = tf.nn.dynamic_rnn(cell=cell, inputs=l1_outputs, sequence_length=num_frames, swap_memory=FLAGS.rnn_swap_memory, dtype=tf.float32) state_fw, state_bw = l1_states state = tf.concat([state_fw, state_bw, l2_states], axis = 1) aggregated_model = getattr(video_level_models, FLAGS.video_level_classifier_model) return aggregated_model().create_model( model_input=state, original_input=model_input, vocab_size=vocab_size, **unused_params) ================================================ FILE: youtube-8m-wangheda/all_frame_models/cnn_deep_combine_chain_model.py ================================================ import math import models import tensorflow as tf import numpy as np import utils from tensorflow import flags import tensorflow.contrib.slim as slim FLAGS = flags.FLAGS class CnnDeepCombineChainModel(models.BaseModel): """A softmax over a mixture of logistic models (with L2 regularization).""" def cnn(self, model_input, l2_penalty=1e-8, num_filters = [1024, 1024, 1024], filter_sizes = [1,2,3], sub_scope="", **unused_params): max_frames = model_input.get_shape().as_list()[1] num_features = model_input.get_shape().as_list()[2] shift_inputs = [] for i in xrange(max(filter_sizes)): if i == 0: shift_inputs.append(model_input) else: shift_inputs.append(tf.pad(model_input, paddings=[[0,0],[i,0],[0,0]])[:,:max_frames,:]) cnn_outputs = [] for nf, fs in zip(num_filters, filter_sizes): sub_input = tf.concat(shift_inputs[:fs], axis=2) sub_filter = tf.get_variable(sub_scope+"cnn-filter-len%d"%fs, shape=[num_features*fs, nf], dtype=tf.float32, initializer=tf.truncated_normal_initializer(mean=0.0, stddev=0.1), regularizer=tf.contrib.layers.l2_regularizer(l2_penalty)) cnn_outputs.append(tf.einsum("ijk,kl->ijl", sub_input, sub_filter)) cnn_output = tf.concat(cnn_outputs, axis=2) return cnn_output def create_model(self, model_input, vocab_size, num_frames, num_mixtures=None, l2_penalty=1e-8, sub_scope="", original_input=None, **unused_params): num_supports = FLAGS.num_supports num_layers = FLAGS.deep_chain_layers relu_cells = FLAGS.deep_chain_relu_cells max_frames = model_input.get_shape().as_list()[1] relu_layers = [] support_predictions = [] mask = self.get_mask(max_frames, num_frames) mean_input = tf.einsum("ijk,ij->ik", model_input, mask) \ / tf.expand_dims(tf.cast(num_frames, dtype=tf.float32), dim=1) mean_relu = slim.fully_connected( mean_input, relu_cells, activation_fn=tf.nn.relu, weights_regularizer=slim.l2_regularizer(l2_penalty), scope=sub_scope+"mean-relu") mean_relu_norm = tf.nn.l2_normalize(mean_relu, dim=1) relu_layers.append(mean_relu_norm) cnn_output = self.cnn(model_input, num_filters=[relu_cells,relu_cells,relu_cells*2], filter_sizes=[1,2,3], sub_scope=sub_scope+"cnn0") max_cnn_output = tf.reduce_max(cnn_output, axis=1) normalized_cnn_output = tf.nn.l2_normalize(max_cnn_output, dim=1) next_input = normalized_cnn_output for layer in xrange(num_layers): sub_prediction = self.sub_model(next_input, vocab_size, sub_scope=sub_scope+"prediction-%d"%layer) support_predictions.append(sub_prediction) sub_relu = slim.fully_connected( sub_prediction, relu_cells, activation_fn=tf.nn.relu, weights_regularizer=slim.l2_regularizer(l2_penalty), scope=sub_scope+"relu-%d"%layer) relu_norm = tf.nn.l2_normalize(sub_relu, dim=1) relu_layers.append(relu_norm) cnn_output = self.cnn(model_input, num_filters=[relu_cells,relu_cells,relu_cells*2], filter_sizes=[1,2,3], sub_scope=sub_scope+"cnn%d"%(layer+1)) max_cnn_output = tf.reduce_max(cnn_output, axis=1) normalized_cnn_output = tf.nn.l2_normalize(max_cnn_output, dim=1) next_input = tf.concat([mean_input, normalized_cnn_output] + relu_layers, axis=1) main_predictions = self.sub_model(next_input, vocab_size, sub_scope=sub_scope+"-main") support_predictions = tf.concat(support_predictions, axis=1) return {"predictions": main_predictions, "support_predictions": support_predictions} def sub_model(self, model_input, vocab_size, num_mixtures=None, l2_penalty=1e-8, sub_scope="", **unused_params): num_mixtures = num_mixtures or FLAGS.moe_num_mixtures gate_activations = slim.fully_connected( model_input, vocab_size * (num_mixtures + 1), activation_fn=None, biases_initializer=None, weights_regularizer=slim.l2_regularizer(l2_penalty), scope="gates-"+sub_scope) expert_activations = slim.fully_connected( model_input, vocab_size * num_mixtures, activation_fn=None, weights_regularizer=slim.l2_regularizer(l2_penalty), scope="experts-"+sub_scope) gating_distribution = tf.nn.softmax(tf.reshape( gate_activations, [-1, num_mixtures + 1])) # (Batch * #Labels) x (num_mixtures + 1) expert_distribution = tf.nn.sigmoid(tf.reshape( expert_activations, [-1, num_mixtures])) # (Batch * #Labels) x num_mixtures final_probabilities_by_class_and_batch = tf.reduce_sum( gating_distribution[:, :num_mixtures] * expert_distribution, 1) final_probabilities = tf.reshape(final_probabilities_by_class_and_batch, [-1, vocab_size]) return final_probabilities def get_mask(self, max_frames, num_frames): mask_array = [] for i in xrange(max_frames + 1): tmp = [0.0] * max_frames for j in xrange(i): tmp[j] = 1.0 mask_array.append(tmp) mask_array = np.array(mask_array) mask_init = tf.constant_initializer(mask_array) mask_emb = tf.get_variable("mask_emb", shape = [max_frames + 1, max_frames], dtype = tf.float32, trainable = False, initializer = mask_init) mask = tf.nn.embedding_lookup(mask_emb, num_frames) return mask ================================================ FILE: youtube-8m-wangheda/all_frame_models/cnn_kmax_model.py ================================================ import sys import models import model_utils import math import numpy as np import video_level_models import tensorflow as tf import utils import tensorflow.contrib.slim as slim from tensorflow import flags FLAGS = flags.FLAGS class CnnKmaxModel(models.BaseModel): def create_model(self, model_input, vocab_size, num_frames, l2_penalty=1e-8, **unused_params): """Creates a model which uses a stack of LSTMs to represent the video. Args: model_input: A 'batch_size' x 'max_frames' x 'num_features' matrix of input features. vocab_size: The number of classes in the dataset. num_frames: A vector of length 'batch' which indicates the number of frames for each video (before padding). Returns: A dictionary with a tensor containing the probability predictions of the model in the 'predictions' key. The dimensions of the tensor are 'batch_size' x 'num_classes'. """ # Create a convolution + maxpool layer for each filter size filter_sizes = map(int, FLAGS.cnn_filter_sizes.split(",")) filter_nums = map(int, FLAGS.cnn_filter_nums.split(",")) pooling_k = FLAGS.cnn_pooling_k assert len(filter_sizes) == len(filter_nums), \ "length of filter_sizes (={}) != length of filter_nums (={})".format( \ len(filter_sizes), len(filter_nums)) batch_size, max_frames, num_features = model_input.get_shape().as_list() with tf.variable_scope("CNN"): # set channel dimension to 1 # cnn_input.shape = [batch, in_height, in_width, in_channels] cnn_input = tf.expand_dims(model_input, axis = 3) cnn_output = [] for filter_size, filter_num in zip(filter_sizes, filter_nums): with tf.name_scope("conv-maxpool-%s" % filter_size): # Convolution Layer # filter.shape = [filter_height, filter_width, in_channels, out_channels] filter_shape = [filter_size, num_features, 1, filter_num] W = tf.Variable(tf.truncated_normal(filter_shape, stddev=0.1), name="W") b = tf.Variable(tf.constant(0.1, shape=[filter_num]), name="b") tf.add_to_collection(name=tf.GraphKeys.REGULARIZATION_LOSSES, value=l2_penalty*tf.nn.l2_loss(W)) tf.add_to_collection(name=tf.GraphKeys.REGULARIZATION_LOSSES, value=l2_penalty*tf.nn.l2_loss(b)) # conv.shape = [batch, -1, 1, out_channels] conv = tf.nn.conv2d(cnn_input, W, strides=[1, 1, 1, 1], padding="VALID",name="conv") # add bias conv = tf.nn.bias_add(conv, b) if pooling_k == 1: _, conv_len, _, _= conv.shape.as_list() conv_kmax = tf.reduce_max(conv, axis = 1) conv_flat = tf.reshape(conv_kmax, [-1, filter_num]) else: conv_kmax = tf.transpose(tf.squeeze(conv, axis = 2), perm = [0, 2, 1]) conv_kmax, _ = tf.nn.top_k(conv_kmax, k = pooling_k, sorted = True) conv_flat = tf.reshape(conv_kmax, [-1, filter_num * pooling_k]) cnn_output.append(conv_flat) cnn_output = tf.concat(cnn_output, axis = 1) aggregated_model = getattr(video_level_models, FLAGS.video_level_classifier_model) return aggregated_model().create_model( model_input=conv_flat, original_input=model_input, vocab_size=vocab_size, **unused_params) ================================================ FILE: youtube-8m-wangheda/all_frame_models/cnn_lstm_memory_model.py ================================================ import sys import models import model_utils import math import numpy as np import video_level_models import tensorflow as tf import utils import tensorflow.contrib.slim as slim from tensorflow import flags FLAGS = flags.FLAGS class CnnLstmMemoryModel(models.BaseModel): def cnn(self, model_input, l2_penalty=1e-8, num_filters = [1024, 1024, 1024], filter_sizes = [1,2,3], **unused_params): max_frames = model_input.get_shape().as_list()[1] num_features = model_input.get_shape().as_list()[2] shift_inputs = [] for i in xrange(max(filter_sizes)): if i == 0: shift_inputs.append(model_input) else: shift_inputs.append(tf.pad(model_input, paddings=[[0,0],[i,0],[0,0]])[:,:max_frames,:]) cnn_outputs = [] for nf, fs in zip(num_filters, filter_sizes): sub_input = tf.concat(shift_inputs[:fs], axis=2) sub_filter = tf.get_variable("cnn-filter-len%d"%fs, shape=[num_features*fs, nf], dtype=tf.float32, initializer=tf.truncated_normal_initializer(mean=0.0, stddev=0.1), regularizer=tf.contrib.layers.l2_regularizer(l2_penalty)) cnn_outputs.append(tf.einsum("ijk,kl->ijl", sub_input, sub_filter)) cnn_output = tf.concat(cnn_outputs, axis=2) return cnn_output def create_model(self, model_input, vocab_size, num_frames, **unused_params): """Creates a model which uses a stack of LSTMs to represent the video. Args: model_input: A 'batch_size' x 'max_frames' x 'num_features' matrix of input features. vocab_size: The number of classes in the dataset. num_frames: A vector of length 'batch' which indicates the number of frames for each video (before padding). Returns: A dictionary with a tensor containing the probability predictions of the model in the 'predictions' key. The dimensions of the tensor are 'batch_size' x 'num_classes'. """ lstm_size = int(FLAGS.lstm_cells) number_of_layers = FLAGS.lstm_layers cnn_output = self.cnn(model_input, num_filters=[1024,1024,1024], filter_sizes=[1,2,3]) normalized_cnn_output = tf.nn.l2_normalize(cnn_output, dim=2) ## Batch normalize the input stacked_lstm = tf.contrib.rnn.MultiRNNCell( [ tf.contrib.rnn.BasicLSTMCell( lstm_size, forget_bias=1.0, state_is_tuple=True) for _ in range(number_of_layers) ], state_is_tuple=True) loss = 0.0 with tf.variable_scope("RNN"): outputs, state = tf.nn.dynamic_rnn(stacked_lstm, normalized_cnn_output, sequence_length=num_frames, swap_memory=FLAGS.rnn_swap_memory, dtype=tf.float32) final_state = tf.concat(map(lambda x: x.c, state), axis = 1) aggregated_model = getattr(video_level_models, FLAGS.video_level_classifier_model) return aggregated_model().create_model( model_input=final_state, original_input=model_input, vocab_size=vocab_size, **unused_params) ================================================ FILE: youtube-8m-wangheda/all_frame_models/cnn_lstm_memory_multitask_model.py ================================================ import sys import models import model_utils import math import numpy as np import video_level_models import tensorflow as tf import utils import tensorflow.contrib.slim as slim from tensorflow import flags FLAGS = flags.FLAGS class CnnLstmMemoryMultiTaskModel(models.BaseModel): def cnn(self, model_input, l2_penalty=1e-8, num_filters = [1024, 1024, 1024], filter_sizes = [1,2,3], **unused_params): max_frames = model_input.get_shape().as_list()[1] num_features = model_input.get_shape().as_list()[2] shift_inputs = [] for i in xrange(max(filter_sizes)): if i == 0: shift_inputs.append(model_input) else: shift_inputs.append(tf.pad(model_input, paddings=[[0,0],[i,0],[0,0]])[:,:max_frames,:]) cnn_outputs = [] for nf, fs in zip(num_filters, filter_sizes): sub_input = tf.concat(shift_inputs[:fs], axis=2) sub_filter = tf.get_variable("cnn-filter-len%d"%fs, shape=[num_features*fs, nf], dtype=tf.float32, initializer=tf.truncated_normal_initializer(mean=0.0, stddev=0.1), regularizer=tf.contrib.layers.l2_regularizer(l2_penalty)) cnn_outputs.append(tf.einsum("ijk,kl->ijl", sub_input, sub_filter)) cnn_output = tf.concat(cnn_outputs, axis=2) return cnn_output def create_model(self, model_input, vocab_size, num_frames, **unused_params): """Creates a model which uses a stack of LSTMs to represent the video. Args: model_input: A 'batch_size' x 'max_frames' x 'num_features' matrix of input features. vocab_size: The number of classes in the dataset. num_frames: A vector of length 'batch' which indicates the number of frames for each video (before padding). Returns: A dictionary with a tensor containing the probability predictions of the model in the 'predictions' key. The dimensions of the tensor are 'batch_size' x 'num_classes'. """ lstm_size = int(FLAGS.lstm_cells) number_of_layers = FLAGS.lstm_layers max_frames = model_input.get_shape().as_list()[1] cnn_output = self.cnn(model_input, num_filters=[1024,1024,1024], filter_sizes=[1,2,3]) normalized_cnn_output = tf.nn.l2_normalize(cnn_output, dim=2) ## Batch normalize the input stacked_lstm = tf.contrib.rnn.MultiRNNCell( [ tf.contrib.rnn.BasicLSTMCell( lstm_size, forget_bias=1.0, state_is_tuple=True) for _ in range(number_of_layers) ], state_is_tuple=True) loss = 0.0 with tf.variable_scope("RNN"): outputs, state = tf.nn.dynamic_rnn(stacked_lstm, normalized_cnn_output, sequence_length=num_frames, swap_memory=FLAGS.rnn_swap_memory, dtype=tf.float32) final_state = tf.concat(map(lambda x: x.c, state), axis = 1) mask = self.get_mask(max_frames, num_frames) mean_cnn_output = tf.einsum("ijk,ij->ik", normalized_cnn_output, mask) \ / tf.expand_dims(tf.cast(num_frames, dtype=tf.float32), dim=1) support_model = getattr(video_level_models, FLAGS.video_level_classifier_support_model) support_predictions = support_model().create_model( model_input=mean_cnn_output, original_input=model_input, vocab_size=vocab_size, num_mixtures=2, sub_scope="support", **unused_params) aggregated_model = getattr(video_level_models, FLAGS.video_level_classifier_model) predictions = aggregated_model().create_model( model_input=final_state, original_input=model_input, vocab_size=vocab_size, sub_scope="main", **unused_params) return {"predictions": predictions["predictions"], "support_predictions": support_predictions["predictions"]} def get_mask(self, max_frames, num_frames): mask_array = [] for i in xrange(max_frames + 1): tmp = [0.0] * max_frames for j in xrange(i): tmp[j] = 1.0 mask_array.append(tmp) mask_array = np.array(mask_array) mask_init = tf.constant_initializer(mask_array) mask_emb = tf.get_variable("mask_emb", shape = [max_frames + 1, max_frames], dtype = tf.float32, trainable = False, initializer = mask_init) mask = tf.nn.embedding_lookup(mask_emb, num_frames) return mask ================================================ FILE: youtube-8m-wangheda/all_frame_models/cnn_lstm_memory_normalization_model.py ================================================ import sys import models import model_utils import math import numpy as np import video_level_models import tensorflow as tf import utils import tensorflow.contrib.slim as slim from tensorflow import flags FLAGS = flags.FLAGS class CnnLstmMemoryNormalizationModel(models.BaseModel): def cnn(self, model_input, l2_penalty=1e-8, num_filters = [1024, 1024, 1024], filter_sizes = [1,2,3], **unused_params): max_frames = model_input.get_shape().as_list()[1] num_features = model_input.get_shape().as_list()[2] normalize_class = getattr(self, FLAGS.lstm_normalization, self.identical) shift_inputs = [] for i in xrange(max(filter_sizes)): if i == 0: shift_inputs.append(model_input) else: shift_inputs.append(tf.pad(model_input, paddings=[[0,0],[i,0],[0,0]])[:,:max_frames,:]) cnn_outputs = [] for nf, fs in zip(num_filters, filter_sizes): sub_input = tf.concat(shift_inputs[:fs], axis=2) sub_filter = tf.get_variable("cnn-filter-len%d"%fs, shape=[num_features*fs, nf], dtype=tf.float32, initializer=tf.truncated_normal_initializer(mean=0.0, stddev=0.1), regularizer=tf.contrib.layers.l2_regularizer(l2_penalty)) cnn_outputs.append(tf.einsum("ijk,kl->ijl", sub_input, sub_filter)) cnn_output = tf.concat(cnn_outputs, axis=2) return cnn_output def create_model(self, model_input, vocab_size, num_frames, **unused_params): """Creates a model which uses a stack of LSTMs to represent the video. Args: model_input: A 'batch_size' x 'max_frames' x 'num_features' matrix of input features. vocab_size: The number of classes in the dataset. num_frames: A vector of length 'batch' which indicates the number of frames for each video (before padding). Returns: A dictionary with a tensor containing the probability predictions of the model in the 'predictions' key. The dimensions of the tensor are 'batch_size' x 'num_classes'. """ lstm_size = int(FLAGS.lstm_cells) number_of_layers = FLAGS.lstm_layers cnn_output = self.cnn(model_input, num_filters=[1024,1024,1024], filter_sizes=[1,2,3]) outputs, state = cnn_output, [] for layer in xrange(number_of_layers): with tf.variable_scope("RNN-layer%d" % layer): outputs = normalize_class(outputs) cell = tf.contrib.rnn.BasicLSTMCell( lstm_size, forget_bias=1.0, state_is_tuple=True) layer_outputs, layer_state = tf.nn.dynamic_rnn(cell, outputs, sequence_length=num_frames, swap_memory=FLAGS.rnn_swap_memory, dtype=tf.float32) state.append(layer_state) outputs = layer_outputs final_state = tf.concat(map(lambda x: x.c, state), axis = 1) aggregated_model = getattr(video_level_models, FLAGS.video_level_classifier_model) return aggregated_model().create_model( model_input=final_state, original_input=model_input, vocab_size=vocab_size, **unused_params) def layer_normalize(self, input_raw, epsilon=1e-8): feature_dim = len(input_raw.get_shape()) - 1 mean_input = tf.reduce_mean(input_raw, axis=feature_dim, keep_dims=True) std_input = tf.sqrt(tf.reduce_mean(tf.square(input_raw-mean_input), axis=feature_dim, keep_dims=True)) std_input = tf.maximum(std_input, epsilon) output = (input_raw - mean_input) / std_input return output def l2_normalize(self, input_raw, epsilon=1e-8): feature_dim = len(input_raw.get_shape()) - 1 output = tf.nn.l2_normalize(input_raw, dim=feature_dim) return output def identical(self, input_raw, epsilon=1e-8): return input_raw ================================================ FILE: youtube-8m-wangheda/all_frame_models/cnn_model.py ================================================ import math import models import tensorflow as tf import numpy as np import utils from tensorflow import flags import tensorflow.contrib.slim as slim FLAGS = flags.FLAGS class CnnModel(models.BaseModel): """A softmax over a mixture of logistic models (with L2 regularization).""" def cnn(self, model_input, l2_penalty=1e-8, num_filters = [1024, 1024, 1024], filter_sizes = [1,2,3], sub_scope="", **unused_params): max_frames = model_input.get_shape().as_list()[1] num_features = model_input.get_shape().as_list()[2] shift_inputs = [] for i in xrange(max(filter_sizes)): if i == 0: shift_inputs.append(model_input) else: shift_inputs.append(tf.pad(model_input, paddings=[[0,0],[i,0],[0,0]])[:,:max_frames,:]) cnn_outputs = [] for nf, fs in zip(num_filters, filter_sizes): sub_input = tf.concat(shift_inputs[:fs], axis=2) sub_filter = tf.get_variable(sub_scope+"cnn-filter-len%d"%fs, shape=[num_features*fs, nf], dtype=tf.float32, initializer=tf.truncated_normal_initializer(mean=0.0, stddev=0.1), regularizer=tf.contrib.layers.l2_regularizer(l2_penalty)) cnn_outputs.append(tf.einsum("ijk,kl->ijl", sub_input, sub_filter)) cnn_output = tf.concat(cnn_outputs, axis=2) return cnn_output def create_model(self, model_input, vocab_size, num_frames, num_mixtures=None, l2_penalty=1e-8, sub_scope="", original_input=None, **unused_params): num_supports = FLAGS.num_supports num_filters = FLAGS.cnn_num_filters max_frames = model_input.get_shape().as_list()[1] cnn_output = self.cnn(model_input, num_filters=[num_filters,num_filters,num_filters*2], filter_sizes=[1,2,3], sub_scope=sub_scope+"cnn") max_cnn_output = tf.reduce_max(cnn_output, axis=1) normalized_cnn_output = tf.nn.l2_normalize(max_cnn_output, dim=1) predictions = self.sub_model(normalized_cnn_output, vocab_size, sub_scope=sub_scope+"main") return {"predictions": predictions} def sub_model(self, model_input, vocab_size, num_mixtures=None, l2_penalty=1e-8, sub_scope="", **unused_params): num_mixtures = num_mixtures or FLAGS.moe_num_mixtures gate_activations = slim.fully_connected( model_input, vocab_size * (num_mixtures + 1), activation_fn=None, biases_initializer=None, weights_regularizer=slim.l2_regularizer(l2_penalty), scope="gates-"+sub_scope) expert_activations = slim.fully_connected( model_input, vocab_size * num_mixtures, activation_fn=None, weights_regularizer=slim.l2_regularizer(l2_penalty), scope="experts-"+sub_scope) gating_distribution = tf.nn.softmax(tf.reshape( gate_activations, [-1, num_mixtures + 1])) # (Batch * #Labels) x (num_mixtures + 1) expert_distribution = tf.nn.sigmoid(tf.reshape( expert_activations, [-1, num_mixtures])) # (Batch * #Labels) x num_mixtures final_probabilities_by_class_and_batch = tf.reduce_sum( gating_distribution[:, :num_mixtures] * expert_distribution, 1) final_probabilities = tf.reshape(final_probabilities_by_class_and_batch, [-1, vocab_size]) return final_probabilities ================================================ FILE: youtube-8m-wangheda/all_frame_models/dbof_model.py ================================================ import sys import models import model_utils import math import numpy as np import video_level_models import tensorflow as tf import utils import tensorflow.contrib.slim as slim from tensorflow import flags FLAGS = flags.FLAGS class DbofModel(models.BaseModel): """Creates a Deep Bag of Frames model. The model projects the features for each frame into a higher dimensional 'clustering' space, pools across frames in that space, and then uses a configurable video-level model to classify the now aggregated features. The model will randomly sample either frames or sequences of frames during training to speed up convergence. Args: model_input: A 'batch_size' x 'max_frames' x 'num_features' matrix of input features. vocab_size: The number of classes in the dataset. num_frames: A vector of length 'batch' which indicates the number of frames for each video (before padding). Returns: A dictionary with a tensor containing the probability predictions of the model in the 'predictions' key. The dimensions of the tensor are 'batch_size' x 'num_classes'. """ def create_model(self, model_input, vocab_size, num_frames, iterations=None, add_batch_norm=None, sample_random_frames=None, cluster_size=None, hidden_size=None, is_training=True, **unused_params): iterations = iterations or FLAGS.iterations add_batch_norm = add_batch_norm or FLAGS.dbof_add_batch_norm random_frames = sample_random_frames or FLAGS.sample_random_frames cluster_size = cluster_size or FLAGS.dbof_cluster_size hidden1_size = hidden_size or FLAGS.dbof_hidden_size num_frames = tf.cast(tf.expand_dims(num_frames, 1), tf.float32) if random_frames: model_input = model_utils.SampleRandomFrames(model_input, num_frames, iterations) else: model_input = model_utils.SampleRandomSequence(model_input, num_frames, iterations) max_frames = model_input.get_shape().as_list()[1] feature_size = model_input.get_shape().as_list()[2] reshaped_input = tf.reshape(model_input, [-1, feature_size]) tf.summary.histogram("input_hist", reshaped_input) if add_batch_norm: reshaped_input = slim.batch_norm( reshaped_input, center=True, scale=True, is_training=is_training, scope="input_bn") cluster_weights = tf.Variable(tf.random_normal( [feature_size, cluster_size], stddev=1 / math.sqrt(feature_size))) tf.summary.histogram("cluster_weights", cluster_weights) activation = tf.matmul(reshaped_input, cluster_weights) if add_batch_norm: activation = slim.batch_norm( activation, center=True, scale=True, is_training=is_training, scope="cluster_bn") else: cluster_biases = tf.Variable( tf.random_normal( [cluster_size], stddev=1 / math.sqrt(feature_size))) tf.summary.histogram("cluster_biases", cluster_biases) activation += cluster_biases activation = tf.nn.relu6(activation) tf.summary.histogram("cluster_output", activation) activation = tf.reshape(activation, [-1, max_frames, cluster_size]) activation = model_utils.FramePooling(activation, FLAGS.dbof_pooling_method) hidden1_weights = tf.Variable(tf.random_normal( [cluster_size, hidden1_size], stddev=1 / math.sqrt(cluster_size))) tf.summary.histogram("hidden1_weights", hidden1_weights) activation = tf.matmul(activation, hidden1_weights) if add_batch_norm: activation = slim.batch_norm( activation, center=True, scale=True, is_training=is_training, scope="hidden1_bn") else: hidden1_biases = tf.Variable( tf.random_normal( [hidden1_size], stddev=0.01)) tf.summary.histogram("hidden1_biases", hidden1_biases) activation += hidden1_biases activation = tf.nn.relu6(activation) tf.summary.histogram("hidden1_output", activation) aggregated_model = getattr(video_level_models, FLAGS.video_level_classifier_model) return aggregated_model().create_model( model_input=activation, original_input=model_input, vocab_size=vocab_size, **unused_params) ================================================ FILE: youtube-8m-wangheda/all_frame_models/deep_cnn_deep_combine_chain_model.py ================================================ import math import models import tensorflow as tf import numpy as np import utils from tensorflow import flags import tensorflow.contrib.slim as slim FLAGS = flags.FLAGS class DeepCnnDeepCombineChainModel(models.BaseModel): """A softmax over a mixture of logistic models (with L2 regularization).""" def cnn(self, model_input, l2_penalty=1e-8, num_filters = [128,128,256], filter_sizes = [1,2,3], sub_scope="", **unused_params): max_frames = model_input.get_shape().as_list()[1] num_features = model_input.get_shape().as_list()[2] shift_inputs = [] for i in xrange(max(filter_sizes)): if i == 0: shift_inputs.append(model_input) else: shift_inputs.append(tf.pad(model_input, paddings=[[0,0],[i,0],[0,0]])[:,:max_frames,:]) cnn_outputs = [] for nf, fs in zip(num_filters, filter_sizes): sub_input = tf.concat(shift_inputs[:fs], axis=2) sub_filter = tf.get_variable(sub_scope+"filter-len%d"%fs, shape=[num_features*fs, nf], dtype=tf.float32, initializer=tf.truncated_normal_initializer(mean=0.0, stddev=0.1), regularizer=tf.contrib.layers.l2_regularizer(l2_penalty)) cnn_outputs.append(tf.einsum("ijk,kl->ijl", sub_input, sub_filter)) cnn_output = tf.concat(cnn_outputs, axis=2) return cnn_output def deep_cnn(self, model_input, l2_penalty=1e-8, num_layers=3, num_filters = [[128,128,256],[128,128],[128,128]], filter_sizes = [[1,2,3],[2,3],[2,3]], sub_scope="", **unused_params): max_frames = model_input.get_shape().as_list()[1] num_features = model_input.get_shape().as_list()[2] cnn_outputs = [] time_pooled = model_input for i in xrange(num_layers): cnn_output = self.cnn(time_pooled, num_filters=num_filters[i], filter_sizes=filter_sizes[i], sub_scope=sub_scope+"cnn%d"%i) cnn_outputs.append(cnn_output) if i+1 < num_layers: time_pooled = tf.nn.pool(cnn_output, window_shape=[2], pooling_type="MAX", padding="VALID", strides=[2]) return cnn_outputs def create_model(self, model_input, vocab_size, num_frames, num_mixtures=None, dropout=False, keep_prob=None, noise_level=None, l2_penalty=1e-8, sub_scope="", original_input=None, **unused_params): num_supports = FLAGS.num_supports num_layers = FLAGS.deep_chain_layers relu_cells = FLAGS.deep_chain_relu_cells cnn_size = FLAGS.deep_cnn_base_size max_frames = model_input.get_shape().as_list()[1] relu_layers = [] support_predictions = [] mask = self.get_mask(max_frames, num_frames) mean_input = tf.einsum("ijk,ij->ik", model_input, mask) \ / tf.expand_dims(tf.cast(num_frames, dtype=tf.float32), dim=1) mean_relu = slim.fully_connected( mean_input, relu_cells, activation_fn=tf.nn.relu, weights_regularizer=slim.l2_regularizer(l2_penalty), scope=sub_scope+"mean-relu") mean_relu_norm = tf.nn.l2_normalize(mean_relu, dim=1) relu_layers.append(mean_relu_norm) cnn_outputs = self.deep_cnn(model_input, sub_scope=sub_scope+"deepcnn0", num_layers=3, num_filters = [[cnn_size,cnn_size,cnn_size*2],[cnn_size,cnn_size],[cnn_size,cnn_size]], filter_sizes = [[1,2,3],[2,3],[2,3]]) max_cnn_output = tf.concat(map(lambda x: tf.reduce_max(x, axis=1), cnn_outputs), axis=1) normalized_cnn_output = tf.nn.l2_normalize(max_cnn_output, dim=1) next_input = normalized_cnn_output for layer in xrange(num_layers): sub_prediction = self.sub_model(next_input, vocab_size, sub_scope=sub_scope+"prediction-%d"%layer, dropout=dropout, keep_prob=keep_prob, noise_level=noise_level) support_predictions.append(sub_prediction) sub_relu = slim.fully_connected( sub_prediction, relu_cells, activation_fn=tf.nn.relu, weights_regularizer=slim.l2_regularizer(l2_penalty), scope=sub_scope+"relu-%d"%layer) relu_norm = tf.nn.l2_normalize(sub_relu, dim=1) relu_layers.append(relu_norm) cnn_outputs = self.deep_cnn(model_input, sub_scope=sub_scope+"deepcnn%d"%(layer+1), num_layers=3, num_filters = [[cnn_size,cnn_size,cnn_size*2],[cnn_size,cnn_size],[cnn_size,cnn_size]], filter_sizes = [[1,2,3],[2,3],[2,3]]) max_cnn_output = tf.concat(map(lambda x: tf.reduce_max(x, axis=1), cnn_outputs), axis=1) normalized_cnn_output = tf.nn.l2_normalize(max_cnn_output, dim=1) next_input = tf.concat([mean_input, normalized_cnn_output] + relu_layers, axis=1) main_predictions = self.sub_model(next_input, vocab_size, sub_scope=sub_scope+"-main", dropout=dropout, keep_prob=keep_prob, noise_level=noise_level) support_predictions = tf.concat(support_predictions, axis=1) return {"predictions": main_predictions, "support_predictions": support_predictions} def sub_model(self, model_input, vocab_size, num_mixtures=None, dropout=False, keep_prob=None, noise_level=None, l2_penalty=1e-8, sub_scope="", **unused_params): num_mixtures = num_mixtures or FLAGS.moe_num_mixtures if dropout: print "adding dropout to", model_input model_input = tf.nn.dropout(model_input, keep_prob=keep_prob) gate_activations = slim.fully_connected( model_input, vocab_size * (num_mixtures + 1), activation_fn=None, biases_initializer=None, weights_regularizer=slim.l2_regularizer(l2_penalty), scope="gates-"+sub_scope) expert_activations = slim.fully_connected( model_input, vocab_size * num_mixtures, activation_fn=None, weights_regularizer=slim.l2_regularizer(l2_penalty), scope="experts-"+sub_scope) gating_distribution = tf.nn.softmax(tf.reshape( gate_activations, [-1, num_mixtures + 1])) # (Batch * #Labels) x (num_mixtures + 1) expert_distribution = tf.nn.sigmoid(tf.reshape( expert_activations, [-1, num_mixtures])) # (Batch * #Labels) x num_mixtures final_probabilities_by_class_and_batch = tf.reduce_sum( gating_distribution[:, :num_mixtures] * expert_distribution, 1) final_probabilities = tf.reshape(final_probabilities_by_class_and_batch, [-1, vocab_size]) return final_probabilities def get_mask(self, max_frames, num_frames): mask_array = [] for i in xrange(max_frames + 1): tmp = [0.0] * max_frames for j in xrange(i): tmp[j] = 1.0 mask_array.append(tmp) mask_array = np.array(mask_array) mask_init = tf.constant_initializer(mask_array) mask_emb = tf.get_variable("mask_emb", shape = [max_frames + 1, max_frames], dtype = tf.float32, trainable = False, initializer = mask_init) mask = tf.nn.embedding_lookup(mask_emb, num_frames) return mask ================================================ FILE: youtube-8m-wangheda/all_frame_models/deep_lstm_model.py ================================================ import sys import models import model_utils import math import numpy as np import video_level_models import tensorflow as tf import utils import tensorflow.contrib.slim as slim from tensorflow import flags FLAGS = flags.FLAGS class DeepLstmModel(models.BaseModel): def create_model(self, model_input, vocab_size, num_frames, **unused_params): """Creates a model which uses a stack of LSTMs to represent the video. Args: model_input: A 'batch_size' x 'max_frames' x 'num_features' matrix of input features. vocab_size: The number of classes in the dataset. num_frames: A vector of length 'batch' which indicates the number of frames for each video (before padding). Returns: A dictionary with a tensor containing the probability predictions of the model in the 'predictions' key. The dimensions of the tensor are 'batch_size' x 'num_classes'. """ lstm_size = int(FLAGS.lstm_cells) number_of_layers = FLAGS.lstm_layers ## Batch normalize the input stacked_lstm = tf.contrib.rnn.MultiRNNCell( [ tf.contrib.rnn.BasicLSTMCell( lstm_size / (2 ** layer_ind), forget_bias=1.0, state_is_tuple=False) for layer_ind in range(number_of_layers) ], state_is_tuple=False) loss = 0.0 with tf.variable_scope("RNN"): outputs, state = tf.nn.dynamic_rnn(stacked_lstm, model_input, sequence_length=num_frames, time_major=False, swap_memory=FLAGS.rnn_swap_memory, dtype=tf.float32) concat_state = tf.concat(state, axis = 1) num_frames_matrix = tf.maximum(tf.cast(tf.expand_dims(num_frames, axis=1), dtype=tf.float32), 1.0) pooling_output = tf.reduce_sum(outputs, axis = 1) / num_frames_matrix final_output = tf.concat([pooling_output, concat_state], axis = 1) aggregated_model = getattr(video_level_models, FLAGS.video_level_classifier_model) predictions = aggregated_model().create_model( model_input=final_output, original_input=model_input, vocab_size=vocab_size, **unused_params) return predictions ================================================ FILE: youtube-8m-wangheda/all_frame_models/distillchain_cnn_deep_combine_chain_model.py ================================================ import math import models import tensorflow as tf import numpy as np import utils from tensorflow import flags import tensorflow.contrib.slim as slim FLAGS = flags.FLAGS class DistillchainCnnDeepCombineChainModel(models.BaseModel): """A softmax over a mixture of logistic models (with L2 regularization).""" def cnn(self, model_input, l2_penalty=1e-8, num_filters = [1024, 1024, 1024], filter_sizes = [1,2,3], sub_scope="", **unused_params): max_frames = model_input.get_shape().as_list()[1] num_features = model_input.get_shape().as_list()[2] shift_inputs = [] for i in xrange(max(filter_sizes)): if i == 0: shift_inputs.append(model_input) else: shift_inputs.append(tf.pad(model_input, paddings=[[0,0],[i,0],[0,0]])[:,:max_frames,:]) cnn_outputs = [] for nf, fs in zip(num_filters, filter_sizes): sub_input = tf.concat(shift_inputs[:fs], axis=2) sub_filter = tf.get_variable(sub_scope+"cnn-filter-len%d"%fs, shape=[num_features*fs, nf], dtype=tf.float32, initializer=tf.truncated_normal_initializer(mean=0.0, stddev=0.1), regularizer=tf.contrib.layers.l2_regularizer(l2_penalty)) cnn_outputs.append(tf.einsum("ijk,kl->ijl", sub_input, sub_filter)) cnn_output = tf.concat(cnn_outputs, axis=2) return cnn_output def create_model(self, model_input, vocab_size, num_frames, num_mixtures=None, l2_penalty=1e-8, sub_scope="", original_input=None, distillation_predictions=None, **unused_params): assert distillation_predictions is not None, "distillation feature must be used" distillchain_relu_cells = FLAGS.distillchain_relu_cells num_supports = FLAGS.num_supports num_layers = FLAGS.deep_chain_layers relu_cells = FLAGS.deep_chain_relu_cells max_frames = model_input.get_shape().as_list()[1] relu_layers = [] support_predictions = [] distill_relu = slim.fully_connected( distillation_predictions, distillchain_relu_cells, activation_fn=tf.nn.relu, weights_regularizer=slim.l2_regularizer(l2_penalty), scope="distillrelu") distill_norm = tf.nn.l2_normalize(distill_relu, dim=1) relu_layers.append(distill_norm) mask = self.get_mask(max_frames, num_frames) mean_input = tf.einsum("ijk,ij->ik", model_input, mask) \ / tf.expand_dims(tf.cast(num_frames, dtype=tf.float32), dim=1) mean_relu = slim.fully_connected( mean_input, relu_cells, activation_fn=tf.nn.relu, weights_regularizer=slim.l2_regularizer(l2_penalty), scope=sub_scope+"mean-relu") mean_relu_norm = tf.nn.l2_normalize(mean_relu, dim=1) relu_layers.append(mean_relu_norm) cnn_output = self.cnn(model_input, num_filters=[relu_cells,relu_cells,relu_cells*2], filter_sizes=[1,2,3], sub_scope=sub_scope+"cnn0") max_cnn_output = tf.reduce_max(cnn_output, axis=1) normalized_cnn_output = tf.nn.l2_normalize(max_cnn_output, dim=1) next_input = tf.concat([normalized_cnn_output] + relu_layers, axis=1) for layer in xrange(num_layers): sub_prediction = self.sub_model(next_input, vocab_size, sub_scope=sub_scope+"prediction-%d"%layer) support_predictions.append(sub_prediction) sub_relu = slim.fully_connected( sub_prediction, relu_cells, activation_fn=tf.nn.relu, weights_regularizer=slim.l2_regularizer(l2_penalty), scope=sub_scope+"relu-%d"%layer) relu_norm = tf.nn.l2_normalize(sub_relu, dim=1) relu_layers.append(relu_norm) cnn_output = self.cnn(model_input, num_filters=[relu_cells,relu_cells,relu_cells*2], filter_sizes=[1,2,3], sub_scope=sub_scope+"cnn%d"%(layer+1)) max_cnn_output = tf.reduce_max(cnn_output, axis=1) normalized_cnn_output = tf.nn.l2_normalize(max_cnn_output, dim=1) next_input = tf.concat([normalized_cnn_output] + relu_layers, axis=1) main_predictions = self.sub_model(next_input, vocab_size, sub_scope=sub_scope+"-main") support_predictions = tf.concat(support_predictions, axis=1) return {"predictions": main_predictions, "support_predictions": support_predictions} def sub_model(self, model_input, vocab_size, num_mixtures=None, l2_penalty=1e-8, sub_scope="", **unused_params): num_mixtures = num_mixtures or FLAGS.moe_num_mixtures gate_activations = slim.fully_connected( model_input, vocab_size * (num_mixtures + 1), activation_fn=None, biases_initializer=None, weights_regularizer=slim.l2_regularizer(l2_penalty), scope="gates-"+sub_scope) expert_activations = slim.fully_connected( model_input, vocab_size * num_mixtures, activation_fn=None, weights_regularizer=slim.l2_regularizer(l2_penalty), scope="experts-"+sub_scope) gating_distribution = tf.nn.softmax(tf.reshape( gate_activations, [-1, num_mixtures + 1])) # (Batch * #Labels) x (num_mixtures + 1) expert_distribution = tf.nn.sigmoid(tf.reshape( expert_activations, [-1, num_mixtures])) # (Batch * #Labels) x num_mixtures final_probabilities_by_class_and_batch = tf.reduce_sum( gating_distribution[:, :num_mixtures] * expert_distribution, 1) final_probabilities = tf.reshape(final_probabilities_by_class_and_batch, [-1, vocab_size]) return final_probabilities def get_mask(self, max_frames, num_frames): mask_array = [] for i in xrange(max_frames + 1): tmp = [0.0] * max_frames for j in xrange(i): tmp[j] = 1.0 mask_array.append(tmp) mask_array = np.array(mask_array) mask_init = tf.constant_initializer(mask_array) mask_emb = tf.get_variable("mask_emb", shape = [max_frames + 1, max_frames], dtype = tf.float32, trainable = False, initializer = mask_init) mask = tf.nn.embedding_lookup(mask_emb, num_frames) return mask ================================================ FILE: youtube-8m-wangheda/all_frame_models/distillchain_lstm_attention_max_pooling_model.py ================================================ import math import models import tensorflow as tf import numpy as np import utils from tensorflow import flags import tensorflow.contrib.slim as slim FLAGS = flags.FLAGS class DistillchainLstmAttentionMaxPoolingModel(models.BaseModel): """Max pooling over temporal weighted sums (attention) of lstm outputs.""" def create_model(self, model_input, vocab_size, num_frames, num_mixtures=None, l2_penalty=1e-8, sub_scope="", distillation_predictions=None, original_input=None, **unused_params): """Creates a model which uses a stack of LSTMs to represent the video. Args: model_input: A 'batch_size' x 'max_frames' x 'num_features' matrix of input features. vocab_size: The number of classes in the dataset. num_frames: A vector of length 'batch' which indicates the number of frames for each video (before padding). Returns: A dictionary with a tensor containing the probability predictions of the model in the 'predictions' key. The dimensions of the tensor are 'batch_size' x 'num_classes'. """ assert distillation_predictions is not None, "distillation feature must be used" distillchain_relu_cells = FLAGS.distillchain_relu_cells lstm_size = int(FLAGS.lstm_cells) number_of_layers = FLAGS.lstm_layers num_attentions = FLAGS.lstm_attentions batch_size, max_frames, num_features = model_input.get_shape().as_list() mask = tf.sequence_mask(lengths=num_frames, maxlen=max_frames, dtype=tf.float32) distill_relu = slim.fully_connected( distillation_predictions, distillchain_relu_cells, activation_fn=tf.nn.relu, weights_regularizer=slim.l2_regularizer(l2_penalty), scope="distillrelu") distill_norm = tf.nn.l2_normalize(distill_relu, dim=1) tiled_distill_norm = tf.tile( input=tf.expand_dims(distill_norm, axis=1), multiples=[1,num_attentions,1]) ## Batch normalize the input stacked_lstm = tf.contrib.rnn.MultiRNNCell( [ tf.contrib.rnn.BasicLSTMCell( lstm_size, forget_bias=1.0, state_is_tuple=True) for _ in range(number_of_layers) ], state_is_tuple=True) with tf.variable_scope("RNN"): outputs, state = tf.nn.dynamic_rnn(stacked_lstm, model_input, sequence_length=num_frames, swap_memory=FLAGS.rnn_swap_memory, dtype=tf.float32) attention_activations = slim.fully_connected( tf.concat([model_input, outputs], axis=2), num_attentions, activation_fn=None, weights_regularizer=slim.l2_regularizer(l2_penalty), scope="attention-"+sub_scope) # Batch x #Attentions x #Frames attention_weights = tf.einsum("ijk,ij->ikj", tf.nn.softmax(attention_activations, dim=1), mask) attention_weights = attention_weights / tf.reduce_sum(attention_weights, axis=2, keep_dims=True) # Batch x #Attentions x #Features attention_outputs = tf.einsum("ijk,ilj->ilk", outputs, attention_weights) attention_outputs = tf.concat([attention_outputs, tiled_distill_norm], axis=2) moe_predictions = self.sub_moe(attention_outputs, vocab_size, sub_scope="sub-moe") predictions = tf.reshape(moe_predictions, [-1, num_attentions, vocab_size]) max_predictions = tf.reduce_max(predictions, axis=1) return {"predictions": max_predictions} def sub_moe(self, model_input, vocab_size, num_mixtures=None, l2_penalty=1e-8, sub_scope="", **unused_params): num_mixtures = num_mixtures or FLAGS.moe_num_mixtures gate_activations = slim.fully_connected( model_input, vocab_size * (num_mixtures + 1), activation_fn=None, biases_initializer=None, weights_regularizer=slim.l2_regularizer(l2_penalty), scope="gates-"+sub_scope) expert_activations = slim.fully_connected( model_input, vocab_size * num_mixtures, activation_fn=None, weights_regularizer=slim.l2_regularizer(l2_penalty), scope="experts-"+sub_scope) gating_distribution = tf.nn.softmax(tf.reshape( gate_activations, [-1, num_mixtures + 1])) # (Batch * #Labels) x (num_mixtures + 1) expert_distribution = tf.nn.sigmoid(tf.reshape( expert_activations, [-1, num_mixtures])) # (Batch * #Labels) x num_mixtures final_probabilities_by_class_and_batch = tf.reduce_sum( gating_distribution[:, :num_mixtures] * expert_distribution, 1) final_probabilities = tf.reshape(final_probabilities_by_class_and_batch, [-1, vocab_size]) return final_probabilities ================================================ FILE: youtube-8m-wangheda/all_frame_models/distillchain_lstm_cnn_deep_combine_chain_model.py ================================================ import math import models import tensorflow as tf import numpy as np import utils from tensorflow import flags import tensorflow.contrib.slim as slim FLAGS = flags.FLAGS class DistillchainLstmCnnDeepCombineChainModel(models.BaseModel): """A softmax over a mixture of logistic models (with L2 regularization).""" def cnn(self, model_input, l2_penalty=1e-8, num_filters = [1024, 1024, 1024], filter_sizes = [1,2,3], sub_scope="", **unused_params): max_frames = model_input.get_shape().as_list()[1] num_features = model_input.get_shape().as_list()[2] shift_inputs = [] for i in xrange(max(filter_sizes)): if i == 0: shift_inputs.append(model_input) else: shift_inputs.append(tf.pad(model_input, paddings=[[0,0],[i,0],[0,0]])[:,:max_frames,:]) cnn_outputs = [] for nf, fs in zip(num_filters, filter_sizes): sub_input = tf.concat(shift_inputs[:fs], axis=2) sub_filter = tf.get_variable(sub_scope+"cnn-filter-len%d"%fs, shape=[num_features*fs, nf], dtype=tf.float32, initializer=tf.truncated_normal_initializer(mean=0.0, stddev=0.1), regularizer=tf.contrib.layers.l2_regularizer(l2_penalty)) cnn_outputs.append(tf.einsum("ijk,kl->ijl", sub_input, sub_filter)) cnn_output = tf.concat(cnn_outputs, axis=2) return cnn_output def create_model(self, model_input, vocab_size, num_frames, num_mixtures=None, l2_penalty=1e-8, sub_scope="", original_input=None, distillation_predictions=None, **unused_params): assert distillation_predictions is not None, "distillation feature must be used" distillchain_relu_cells = FLAGS.distillchain_relu_cells num_supports = FLAGS.num_supports num_layers = FLAGS.deep_chain_layers relu_cells = FLAGS.deep_chain_relu_cells max_frames = model_input.get_shape().as_list()[1] relu_layers = [] support_predictions = [] # distill distill_relu = slim.fully_connected( distillation_predictions, distillchain_relu_cells, activation_fn=tf.nn.relu, weights_regularizer=slim.l2_regularizer(l2_penalty), scope="distillrelu") distill_norm = tf.nn.l2_normalize(distill_relu, dim=1) relu_layers.append(distill_norm) # mean mask = tf.sequence_mask(num_frames, maxlen=max_frames, dtype=tf.float32) mean_input = tf.einsum("ijk,ij->ik", model_input, mask) \ / tf.expand_dims(tf.cast(num_frames, dtype=tf.float32), dim=1) mean_relu = slim.fully_connected( mean_input, relu_cells, activation_fn=tf.nn.relu, weights_regularizer=slim.l2_regularizer(l2_penalty), scope=sub_scope+"mean-relu") mean_relu_norm = tf.nn.l2_normalize(mean_relu, dim=1) relu_layers.append(mean_relu_norm) # LSTM lstm_output = self.lstmoutput(model_input, vocab_size, num_frames) # CNN cnn_output = self.cnn(lstm_output, num_filters=[relu_cells,2*relu_cells,relu_cells], filter_sizes=[1,2,3], sub_scope=sub_scope+"cnn0") max_cnn_output = tf.reduce_max(cnn_output, axis=1) normalized_cnn_output = tf.nn.l2_normalize(max_cnn_output, dim=1) next_input = tf.concat([normalized_cnn_output] + relu_layers, axis=1) for layer in xrange(num_layers): sub_prediction = self.sub_model(next_input, vocab_size, sub_scope=sub_scope+"prediction-%d"%layer) support_predictions.append(sub_prediction) sub_relu = slim.fully_connected( sub_prediction, relu_cells, activation_fn=tf.nn.relu, weights_regularizer=slim.l2_regularizer(l2_penalty), scope=sub_scope+"relu-%d"%layer) relu_norm = tf.nn.l2_normalize(sub_relu, dim=1) relu_layers.append(relu_norm) cnn_output = self.cnn(lstm_output, num_filters=[relu_cells,relu_cells,relu_cells*2], filter_sizes=[1,2,3], sub_scope=sub_scope+"cnn%d"%(layer+1)) max_cnn_output = tf.reduce_max(cnn_output, axis=1) normalized_cnn_output = tf.nn.l2_normalize(max_cnn_output, dim=1) next_input = tf.concat([normalized_cnn_output] + relu_layers, axis=1) main_predictions = self.sub_model(next_input, vocab_size, sub_scope=sub_scope+"-main") support_predictions = tf.concat(support_predictions, axis=1) return {"predictions": main_predictions, "support_predictions": support_predictions} def sub_model(self, model_input, vocab_size, num_mixtures=None, l2_penalty=1e-8, sub_scope="", **unused_params): num_mixtures = num_mixtures or FLAGS.moe_num_mixtures gate_activations = slim.fully_connected( model_input, vocab_size * (num_mixtures + 1), activation_fn=None, biases_initializer=None, weights_regularizer=slim.l2_regularizer(l2_penalty), scope="gates-"+sub_scope) expert_activations = slim.fully_connected( model_input, vocab_size * num_mixtures, activation_fn=None, weights_regularizer=slim.l2_regularizer(l2_penalty), scope="experts-"+sub_scope) gating_distribution = tf.nn.softmax(tf.reshape( gate_activations, [-1, num_mixtures + 1])) # (Batch * #Labels) x (num_mixtures + 1) expert_distribution = tf.nn.sigmoid(tf.reshape( expert_activations, [-1, num_mixtures])) # (Batch * #Labels) x num_mixtures final_probabilities_by_class_and_batch = tf.reduce_sum( gating_distribution[:, :num_mixtures] * expert_distribution, 1) final_probabilities = tf.reshape(final_probabilities_by_class_and_batch, [-1, vocab_size]) return final_probabilities def get_mask(self, max_frames, num_frames): mask_array = [] for i in xrange(max_frames + 1): tmp = [0.0] * max_frames for j in xrange(i): tmp[j] = 1.0 mask_array.append(tmp) mask_array = np.array(mask_array) mask_init = tf.constant_initializer(mask_array) mask_emb = tf.get_variable("mask_emb", shape = [max_frames + 1, max_frames], dtype = tf.float32, trainable = False, initializer = mask_init) mask = tf.nn.embedding_lookup(mask_emb, num_frames) return mask def lstmoutput(self, model_input, vocab_size, num_frames): number_of_layers = FLAGS.lstm_layers lstm_sizes = map(int, FLAGS.lstm_cells.split(",")) feature_names, feature_sizes = utils.GetListOfFeatureNamesAndSizes( FLAGS.feature_names, FLAGS.feature_sizes) sub_inputs = [tf.nn.l2_normalize(x, dim=2) for x in tf.split(model_input, feature_sizes, axis = 2)] assert len(lstm_sizes) == len(feature_sizes), \ "length of lstm_sizes (={}) != length of feature_sizes (={})".format( \ len(lstm_sizes), len(feature_sizes)) outputs = [] for i in xrange(len(feature_sizes)): with tf.variable_scope("RNN%d" % i): sub_input = sub_inputs[i] lstm_size = lstm_sizes[i] ## Batch normalize the input stacked_lstm = tf.contrib.rnn.MultiRNNCell( [ tf.contrib.rnn.BasicLSTMCell( lstm_size, forget_bias=1.0, state_is_tuple=True) for _ in range(number_of_layers) ], state_is_tuple=True) output, state = tf.nn.dynamic_rnn(stacked_lstm, sub_input, sequence_length=num_frames, swap_memory=FLAGS.rnn_swap_memory, dtype=tf.float32) outputs.append(output) # concat final_output = tf.concat(outputs, axis=2) return final_output ================================================ FILE: youtube-8m-wangheda/all_frame_models/distillchain_lstm_memory_deep_combine_chain_model.py ================================================ import sys import models import model_utils import math import numpy as np import video_level_models import tensorflow as tf import utils import tensorflow.contrib.slim as slim from tensorflow import flags FLAGS = flags.FLAGS class DistillchainLstmMemoryDeepCombineChainModel(models.BaseModel): """Classifier chain model of lstm memory""" def create_model(self, model_input, vocab_size, num_frames, l2_penalty=1e-8, sub_scope="", original_input=None, distillation_predictions=None, **unused_params): assert distillation_predictions is not None, "distillation feature must be used" distillchain_relu_cells = FLAGS.distillchain_relu_cells lstm_size = int(FLAGS.lstm_cells) number_of_layers = FLAGS.lstm_layers num_supports = FLAGS.num_supports num_layers = FLAGS.deep_chain_layers relu_cells = FLAGS.deep_chain_relu_cells max_frames = model_input.get_shape().as_list()[1] relu_layers = [] support_predictions = [] # distill predictions distill_relu = slim.fully_connected( distillation_predictions, distillchain_relu_cells, activation_fn=tf.nn.relu, weights_regularizer=slim.l2_regularizer(l2_penalty), scope="distill-relu") distill_norm = tf.nn.l2_normalize(distill_relu, dim=1) relu_layers.append(distill_norm) # mean input mask = tf.sequence_mask(num_frames, maxlen=max_frames, dtype=tf.float32) mean_input = tf.einsum("ijk,ij->ik", model_input, mask) \ / tf.expand_dims(tf.cast(num_frames, dtype=tf.float32), dim=1) mean_relu = slim.fully_connected( mean_input, relu_cells, activation_fn=tf.nn.relu, weights_regularizer=slim.l2_regularizer(l2_penalty), scope=sub_scope+"mean-relu") mean_relu_norm = tf.nn.l2_normalize(mean_relu, dim=1) relu_layers.append(mean_relu_norm) lstm_output = self.sub_lstm(model_input, num_frames, lstm_size, number_of_layers, sub_scope="lstm-%d"%0) normalized_lstm_output = tf.nn.l2_normalize(lstm_output, dim=1) next_input = tf.concat([normalized_lstm_output] + relu_layers, axis=1) for layer in xrange(num_layers): sub_prediction = self.sub_moe(next_input, vocab_size, sub_scope=sub_scope+"prediction-%d"%layer) support_predictions.append(sub_prediction) sub_relu = slim.fully_connected( sub_prediction, relu_cells, activation_fn=tf.nn.relu, weights_regularizer=slim.l2_regularizer(l2_penalty), scope=sub_scope+"relu-%d"%layer) relu_norm = tf.nn.l2_normalize(sub_relu, dim=1) relu_layers.append(relu_norm) lstm_output = self.sub_lstm(model_input, num_frames, lstm_size, number_of_layers, sub_scope="lstm-%d"%(layer+1)) normalized_lstm_output = tf.nn.l2_normalize(lstm_output, dim=1) next_input = tf.concat([normalized_lstm_output] + relu_layers, axis=1) main_predictions = self.sub_moe(next_input, vocab_size, sub_scope=sub_scope+"-main") support_predictions = tf.concat(support_predictions, axis=1) return {"predictions": main_predictions, "support_predictions": support_predictions} def sub_lstm(self, model_input, num_frames, lstm_size, number_of_layers, sub_scope=""): stacked_lstm = tf.contrib.rnn.MultiRNNCell( [ tf.contrib.rnn.BasicLSTMCell( lstm_size, forget_bias=1.0, state_is_tuple=True) for _ in range(number_of_layers) ], state_is_tuple=True) loss = 0.0 with tf.variable_scope(sub_scope+"-RNN"): outputs, state = tf.nn.dynamic_rnn(stacked_lstm, model_input, sequence_length=num_frames, swap_memory=FLAGS.rnn_swap_memory, dtype=tf.float32) final_state = tf.concat(map(lambda x: x.c, state), axis = 1) return final_state def sub_moe(self, model_input, vocab_size, num_mixtures=None, l2_penalty=1e-8, sub_scope="", **unused_params): num_mixtures = num_mixtures or FLAGS.moe_num_mixtures gate_activations = slim.fully_connected( model_input, vocab_size * (num_mixtures + 1), activation_fn=None, biases_initializer=None, weights_regularizer=slim.l2_regularizer(l2_penalty), scope="gates-"+sub_scope) expert_activations = slim.fully_connected( model_input, vocab_size * num_mixtures, activation_fn=None, weights_regularizer=slim.l2_regularizer(l2_penalty), scope="experts-"+sub_scope) gating_distribution = tf.nn.softmax(tf.reshape( gate_activations, [-1, num_mixtures + 1])) # (Batch * #Labels) x (num_mixtures + 1) expert_distribution = tf.nn.sigmoid(tf.reshape( expert_activations, [-1, num_mixtures])) # (Batch * #Labels) x num_mixtures final_probabilities_by_class_and_batch = tf.reduce_sum( gating_distribution[:, :num_mixtures] * expert_distribution, 1) final_probabilities = tf.reshape(final_probabilities_by_class_and_batch, [-1, vocab_size]) return final_probabilities ================================================ FILE: youtube-8m-wangheda/all_frame_models/distillchain_lstm_parallel_finaloutput_model.py ================================================ import sys import models import model_utils import math import numpy as np import video_level_models import tensorflow as tf import utils import tensorflow.contrib.slim as slim from tensorflow import flags FLAGS = flags.FLAGS class DistillchainLstmParallelFinaloutputModel(models.BaseModel): def create_model(self, model_input, vocab_size, num_frames, distillation_predictions=None, l2_penalty=1e-8, **unused_params): """Creates a model which uses a stack of LSTMs to represent the video. Args: model_input: A 'batch_size' x 'max_frames' x 'num_features' matrix of input features. vocab_size: The number of classes in the dataset. num_frames: A vector of length 'batch' which indicates the number of frames for each video (before padding). Returns: A dictionary with a tensor containing the probability predictions of the model in the 'predictions' key. The dimensions of the tensor are 'batch_size' x 'num_classes'. """ assert distillation_predictions is not None, "distillation feature must be used" relu_cells = FLAGS.distillchain_relu_cells number_of_layers = FLAGS.lstm_layers distill_relu = slim.fully_connected( distillation_predictions, relu_cells, activation_fn=tf.nn.relu, weights_regularizer=slim.l2_regularizer(l2_penalty), scope="distillrelu") distill_norm = tf.nn.l2_normalize(distill_relu, dim=1) lstm_sizes = map(int, FLAGS.lstm_cells.split(",")) feature_names, feature_sizes = utils.GetListOfFeatureNamesAndSizes( FLAGS.feature_names, FLAGS.feature_sizes) sub_inputs = [tf.nn.l2_normalize(x, dim=2) for x in tf.split(model_input, feature_sizes, axis = 2)] assert len(lstm_sizes) == len(feature_sizes), \ "length of lstm_sizes (={}) != length of feature_sizes (={})".format( \ len(lstm_sizes), len(feature_sizes)) outputs = [] states = [] for i in xrange(len(feature_sizes)): with tf.variable_scope("RNN%d" % i): sub_input = sub_inputs[i] lstm_size = lstm_sizes[i] ## Batch normalize the input stacked_lstm = tf.contrib.rnn.MultiRNNCell( [ tf.contrib.rnn.BasicLSTMCell( lstm_size, forget_bias=1.0, state_is_tuple=True) for _ in range(number_of_layers) ], state_is_tuple=True) output, state = tf.nn.dynamic_rnn(stacked_lstm, sub_input, sequence_length=num_frames, swap_memory=FLAGS.rnn_swap_memory, dtype=tf.float32) outputs.append(output) states.extend(map(lambda x: x.h, state)) # concat final_state = tf.concat(states + [distill_norm], axis = 1) aggregated_model = getattr(video_level_models, FLAGS.video_level_classifier_model) return aggregated_model().create_model( model_input=final_state, original_input=model_input, vocab_size=vocab_size, **unused_params) ================================================ FILE: youtube-8m-wangheda/all_frame_models/distillchain_multiscale_cnn_lstm_model.py ================================================ import math import models import tensorflow as tf import numpy as np import utils from tensorflow import flags import tensorflow.contrib.slim as slim FLAGS = flags.FLAGS class DistillchainMultiscaleCnnLstmModel(models.BaseModel): def cnn(self, model_input, l2_penalty=1e-8, num_filters = [1024, 1024, 1024], filter_sizes = [1,2,3], sub_scope="", **unused_params): max_frames = model_input.get_shape().as_list()[1] num_features = model_input.get_shape().as_list()[2] shift_inputs = [] for i in xrange(max(filter_sizes)): if i == 0: shift_inputs.append(model_input) else: shift_inputs.append(tf.pad(model_input, paddings=[[0,0],[i,0],[0,0]])[:,:max_frames,:]) cnn_outputs = [] for nf, fs in zip(num_filters, filter_sizes): sub_input = tf.concat(shift_inputs[:fs], axis=2) sub_filter = tf.get_variable(sub_scope+"cnn-filter-len%d"%fs, shape=[num_features*fs, nf], dtype=tf.float32, initializer=tf.truncated_normal_initializer(mean=0.0, stddev=0.1), regularizer=tf.contrib.layers.l2_regularizer(l2_penalty)) cnn_outputs.append(tf.einsum("ijk,kl->ijl", sub_input, sub_filter)) cnn_output = tf.concat(cnn_outputs, axis=2) cnn_output = slim.batch_norm( cnn_output, center=True, scale=True, is_training=FLAGS.is_training, scope=sub_scope+"cluster_bn") return cnn_output def moe(self,model_input, vocab_size, num_mixtures=None, l2_penalty=1e-8, scopename="", **unused_params): num_mixtures = num_mixtures or FLAGS.moe_num_mixtures gate_activations = slim.fully_connected( model_input, vocab_size * (num_mixtures + 1), activation_fn=None, biases_initializer=None, weights_regularizer=slim.l2_regularizer(l2_penalty), scope="gates"+scopename) expert_activations = slim.fully_connected( model_input, vocab_size * num_mixtures, activation_fn=None, weights_regularizer=slim.l2_regularizer(l2_penalty), scope="experts"+scopename) gating_distribution = tf.nn.softmax(tf.reshape( gate_activations, [-1, num_mixtures + 1])) # (Batch * #Labels) x (num_mixtures + 1) expert_distribution = tf.nn.sigmoid(tf.reshape( expert_activations, [-1, num_mixtures])) # (Batch * #Labels) x num_mixtures final_probabilities_by_class_and_batch = tf.reduce_sum( gating_distribution[:, :num_mixtures] * expert_distribution, 1) final_probabilities = tf.reshape(final_probabilities_by_class_and_batch, [-1, vocab_size]) return final_probabilities def rnn(self, model_input, lstm_size, num_frames, sub_scope="", **unused_params): cell = tf.contrib.rnn.BasicLSTMCell(lstm_size, forget_bias=1.0, state_is_tuple=True) with tf.variable_scope("RNN-"+sub_scope): outputs, state = tf.nn.dynamic_rnn(cell, model_input, sequence_length=num_frames, swap_memory=True, dtype=tf.float32) # return final memory return state.c def create_model(self, model_input, vocab_size, num_frames, distillation_predictions=None, l2_penalty=1e-8, **unused_params): assert distillation_predictions is not None, "distillation feature must be used" distillchain_relu_cells = FLAGS.distillchain_relu_cells num_layers = FLAGS.multiscale_cnn_lstm_layers lstm_size = int(FLAGS.lstm_cells) pool_size=2 num_filters=[256,256,512] filter_sizes=[1,2,3] features_size = sum(num_filters) sub_predictions = [] cnn_input = model_input distill_relu = slim.fully_connected( distillation_predictions, distillchain_relu_cells, activation_fn=tf.nn.relu, weights_regularizer=slim.l2_regularizer(l2_penalty), scope="distillrelu") distill_norm = tf.nn.l2_normalize(distill_relu, dim=1) cnn_max_frames = model_input.get_shape().as_list()[1] for layer in range(num_layers): cnn_output = self.cnn(cnn_input, num_filters=num_filters, filter_sizes=filter_sizes, sub_scope="cnn%d"%(layer+1)) cnn_output_relu = tf.nn.relu(cnn_output) lstm_memory = self.rnn(cnn_output_relu, lstm_size, num_frames, sub_scope="rnn%d"%(layer+1)) memory_and_predictions = tf.concat([lstm_memory, distill_norm], axis=1) sub_prediction = self.moe(memory_and_predictions, vocab_size, scopename="moe%d"%(layer+1)) sub_predictions.append(sub_prediction) cnn_max_frames /= pool_size max_pooled_cnn_output = tf.reduce_max( tf.reshape( cnn_output_relu[:, :cnn_max_frames*2, :], [-1, cnn_max_frames, pool_size, features_size] ), axis=2) # for the next cnn layer cnn_input = max_pooled_cnn_output num_frames = tf.maximum(num_frames/pool_size, 1) support_predictions = tf.concat(sub_predictions, axis=1) predictions = tf.add_n(sub_predictions) / len(sub_predictions) return {"predictions": predictions, "support_predictions": support_predictions} ================================================ FILE: youtube-8m-wangheda/all_frame_models/frame_seg_model.py ================================================ import sys import models import model_utils import math import numpy as np import video_level_models import tensorflow as tf import utils import tensorflow.contrib.slim as slim from tensorflow import flags FLAGS = flags.FLAGS class FrameSegModel(models.BaseModel): def create_model(self, model_input, vocab_size, num_frames, l2_penalty=1e-8, **unused_params): relu_cells = FLAGS.frame_seg_relu_cells float_num_frames = tf.cast(num_frames, tf.float32) divisions = [0.0, 0.1, 0.5, 0.9, 1.0] relu_layers = [] for i in xrange(4): frame_start = float_num_frames * divisions[i] frame_end = float_num_frames * divisions[i+1] mean_frame = self.frame_mean(model_input, frame_start, frame_end) mean_frame = tf.nn.l2_normalize(mean_frame, dim=1) mean_relu = slim.fully_connected( mean_frame, relu_cells, activation_fn=tf.nn.relu, weights_regularizer=slim.l2_regularizer(l2_penalty), scope="relu%d"%i) relu_layers.append(tf.nn.l2_normalize(mean_relu, dim=1)) relu_layer = tf.concat(relu_layers, axis=1) aggregated_model = getattr(video_level_models, FLAGS.video_level_classifier_model) return aggregated_model().create_model( model_input=relu_layer, original_input=model_input, vocab_size=vocab_size, **unused_params) def frame_mean(self, model_input, frame_start, frame_end, **unused_params): max_frames = model_input.shape.as_list()[-2] frame_start = tf.cast(frame_start, tf.int32) frame_end = tf.cast(frame_end, tf.int32) frame_length = tf.expand_dims(tf.cast(frame_end - frame_start, tf.float32), axis=1) frame_mask = tf.sequence_mask(frame_end, maxlen=max_frames, dtype=tf.float32) \ - tf.sequence_mask(frame_start, maxlen=max_frames, dtype=tf.float32) mean_frame = tf.einsum("ijk,ij->ik", model_input, frame_mask) / (0.1 + frame_length) return mean_frame ================================================ FILE: youtube-8m-wangheda/all_frame_models/framehop_lstm_memory_deep_combine_chain_model.py ================================================ import sys import models import model_utils import math import numpy as np import video_level_models import tensorflow as tf import utils import tensorflow.contrib.slim as slim from tensorflow import flags FLAGS = flags.FLAGS class FramehopLstmMemoryDeepCombineChainModel(models.BaseModel): """Classifier chain model of lstm memory""" def lstm(self, model_input, vocab_size, num_frames, sub_scope="", feature_names=None, feature_sizes=None, **unused_params): number_of_layers = FLAGS.lstm_layers lstm_sizes = map(int, FLAGS.lstm_cells.split(",")) if feature_names is None: feature_names = FLAGS.feature_names if feature_sizes is None: feature_sizes = FLAGS.feature_sizes feature_names, feature_sizes = utils.GetListOfFeatureNamesAndSizes( feature_names, feature_sizes) sub_inputs = [tf.nn.l2_normalize(x, dim=2) for x in tf.split(model_input, feature_sizes, axis = 2)] print model_input assert len(lstm_sizes) == len(feature_sizes), \ "length of lstm_sizes (={}) != length of feature_sizes (={})".format( \ len(lstm_sizes), len(feature_sizes)) states = [] outputs = [] for i in xrange(len(feature_sizes)): with tf.variable_scope(sub_scope+"RNN%d" % i): print sub_scope + "RNN%d"%i, lstm_sizes print sub_inputs sub_input = sub_inputs[i] lstm_size = lstm_sizes[i] ## Batch normalize the input stacked_lstm = tf.contrib.rnn.MultiRNNCell( [ tf.contrib.rnn.BasicLSTMCell( lstm_size, forget_bias=1.0, state_is_tuple=True) for _ in range(number_of_layers) ], state_is_tuple=True) output, state = tf.nn.dynamic_rnn(stacked_lstm, sub_input, sequence_length=num_frames, swap_memory=FLAGS.rnn_swap_memory, dtype=tf.float32) states.extend(map(lambda x: x.c, state)) outputs.append(output) final_state = tf.concat(states, axis = 1) final_output = tf.concat(outputs, axis = 2) return final_state, final_output def create_model(self, model_input, vocab_size, num_mixtures=None, l2_penalty=1e-8, sub_scope="", original_input=None, dropout=False, keep_prob=None, noise_level=None, num_frames=None, **unused_params): num_supports = FLAGS.num_supports num_layers = FLAGS.deep_chain_layers relu_cells = FLAGS.deep_chain_relu_cells relu_type = FLAGS.deep_chain_relu_type use_length = FLAGS.deep_chain_use_length additional_features = [] if use_length: print "using length as feature" additional_features.append(self.get_length_code(num_frames)) lstm_input, lstm_frames = self.resolution(model_input, num_frames, 1) lstm_state, lstm_output = self.lstm(lstm_input, vocab_size, num_frames=lstm_frames, sub_scope="lstm%d"%0) next_input = tf.concat([lstm_state] + additional_features, axis=1) support_predictions = [] for layer in xrange(num_layers): sub_prediction = self.sub_model(next_input, vocab_size, sub_scope=sub_scope+"prediction-%d"%layer, dropout=dropout, keep_prob=keep_prob, noise_level=noise_level) support_predictions.append(sub_prediction) sub_activation = slim.fully_connected( sub_prediction, relu_cells, activation_fn=None, weights_regularizer=slim.l2_regularizer(l2_penalty), scope=sub_scope+"relu-%d"%layer) if relu_type == "elu": sub_relu = tf.nn.elu(sub_activation) else: # default: relu sub_relu = tf.nn.relu(sub_activation) if noise_level is not None: print "adding noise to sub_relu, level = ", noise_level sub_relu = sub_relu + tf.random_normal(tf.shape(sub_relu), mean=0.0, stddev=noise_level) relu_norm = tf.nn.l2_normalize(sub_relu, dim=1) additional_features.append(relu_norm) print "before resolution", lstm_output lstm_input, lstm_frames = self.resolution(lstm_output, num_frames, 2) print "before resolution", lstm_input lstm_state, lstm_output = self.lstm(lstm_input, vocab_size, num_frames=lstm_frames, feature_sizes=FLAGS.lstm_cells, sub_scope="lstm%d"%(layer+1)) next_input = tf.concat([lstm_state] + additional_features, axis=1) main_predictions = self.sub_model(next_input, vocab_size, sub_scope=sub_scope+"-main") support_predictions = tf.concat(support_predictions, axis=1) return {"predictions": main_predictions, "support_predictions": support_predictions} def sub_model(self, model_input, vocab_size, num_mixtures=None, l2_penalty=1e-8, sub_scope="", dropout=False, keep_prob=None, noise_level=None, **unused_params): num_mixtures = num_mixtures or FLAGS.moe_num_mixtures if dropout: print "adding dropout to moe, keep_prob = ", keep_prob model_input = tf.nn.dropout(model_input, keep_prob=keep_prob) gate_activations = slim.fully_connected( model_input, vocab_size * (num_mixtures + 1), activation_fn=None, biases_initializer=None, weights_regularizer=slim.l2_regularizer(l2_penalty), scope="gates-"+sub_scope) expert_activations = slim.fully_connected( model_input, vocab_size * num_mixtures, activation_fn=None, weights_regularizer=slim.l2_regularizer(l2_penalty), scope="experts-"+sub_scope) gating_distribution = tf.nn.softmax(tf.reshape( gate_activations, [-1, num_mixtures + 1])) # (Batch * #Labels) x (num_mixtures + 1) expert_distribution = tf.nn.sigmoid(tf.reshape( expert_activations, [-1, num_mixtures])) # (Batch * #Labels) x num_mixtures final_probabilities_by_class_and_batch = tf.reduce_sum( gating_distribution[:, :num_mixtures] * expert_distribution, 1) final_probabilities = tf.reshape(final_probabilities_by_class_and_batch, [-1, vocab_size]) return final_probabilities def get_length_code(self, num_frames): code_0 = tf.cast(num_frames <= 60, dtype=tf.int32) code_1 = tf.cast(num_frames > 60, dtype=tf.int32) * tf.cast(num_frames <= 120, dtype=tf.int32) code_2 = tf.cast(num_frames > 120, dtype=tf.int32) * tf.cast(num_frames <= 180, dtype=tf.int32) code_3 = tf.cast(num_frames > 180, dtype=tf.int32) * tf.cast(num_frames <= 240, dtype=tf.int32) code_4 = tf.cast(num_frames > 240, dtype=tf.int32) codes = map(lambda x: tf.expand_dims(x, dim=1), [code_0, code_1, code_2, code_3, code_4]) length_code = tf.cast(tf.concat(codes, axis=1), dtype=tf.float32) return length_code def resolution(self, model_input_raw, num_frames, resolution, method="SELECT"): frame_dim = len(model_input_raw.get_shape()) - 2 feature_dim = len(model_input_raw.get_shape()) - 1 max_frames = model_input_raw.get_shape().as_list()[frame_dim] num_features = model_input_raw.get_shape().as_list()[feature_dim] if resolution > 1: new_max_frames = max_frames / resolution cut_frames = new_max_frames * resolution model_input_raw = model_input_raw[:, :cut_frames, :] model_input_raw = tf.reshape(model_input_raw, shape=[-1,new_max_frames,resolution,num_features]) if method == "MEAN": model_input_raw = tf.reduce_mean(model_input_raw, axis=2) elif method == "MAX": model_input_raw = tf.reduce_max(model_input_raw, axis=2) elif method == "SELECT": model_input_raw = model_input_raw[:,:,resolution-1,:] model_input = tf.nn.l2_normalize(model_input_raw, feature_dim) num_frames = num_frames / resolution else: model_input = tf.nn.l2_normalize(model_input_raw, feature_dim) return model_input, num_frames ================================================ FILE: youtube-8m-wangheda/all_frame_models/framehop_lstm_memory_model.py ================================================ import sys import models import model_utils import math import numpy as np import video_level_models import tensorflow as tf import utils import tensorflow.contrib.slim as slim from tensorflow import flags FLAGS = flags.FLAGS class FramehopLstmMemoryModel(models.BaseModel): """Classifier chain model of lstm memory""" def lstm(self, model_input, vocab_size, num_frames, sub_scope="", feature_names=None, feature_sizes=None, **unused_params): number_of_layers = FLAGS.lstm_layers lstm_sizes = map(int, FLAGS.lstm_cells.split(",")) if feature_names is None: feature_names = FLAGS.feature_names if feature_sizes is None: feature_sizes = FLAGS.feature_sizes feature_names, feature_sizes = utils.GetListOfFeatureNamesAndSizes( feature_names, feature_sizes) sub_inputs = [tf.nn.l2_normalize(x, dim=2) for x in tf.split(model_input, feature_sizes, axis = 2)] print model_input assert len(lstm_sizes) == len(feature_sizes), \ "length of lstm_sizes (={}) != length of feature_sizes (={})".format( \ len(lstm_sizes), len(feature_sizes)) states = [] outputs = [] for i in xrange(len(feature_sizes)): with tf.variable_scope(sub_scope+"RNN%d" % i): print sub_scope + "RNN%d"%i, lstm_sizes print sub_inputs sub_input = sub_inputs[i] lstm_size = lstm_sizes[i] ## Batch normalize the input stacked_lstm = tf.contrib.rnn.MultiRNNCell( [ tf.contrib.rnn.BasicLSTMCell( lstm_size, forget_bias=1.0, state_is_tuple=True) for _ in range(number_of_layers) ], state_is_tuple=True) output, state = tf.nn.dynamic_rnn(stacked_lstm, sub_input, sequence_length=num_frames, swap_memory=FLAGS.rnn_swap_memory, dtype=tf.float32) states.extend(map(lambda x: x.c, state)) outputs.append(output) final_state = tf.concat(states, axis = 1) final_output = tf.concat(outputs, axis = 2) return final_state, final_output def create_model(self, model_input, vocab_size, num_mixtures=None, l2_penalty=1e-8, sub_scope="", original_input=None, dropout=False, keep_prob=None, noise_level=None, num_frames=None, **unused_params): num_supports = FLAGS.num_supports num_layers = FLAGS.deep_chain_layers relu_cells = FLAGS.deep_chain_relu_cells relu_type = FLAGS.deep_chain_relu_type use_length = FLAGS.deep_chain_use_length lstm_states = [] if use_length: print "using length as feature" additional_features.append(self.get_length_code(num_frames)) lstm_input, lstm_frames = self.resolution(model_input, num_frames, 1) lstm_state, lstm_output = self.lstm(lstm_input, vocab_size, num_frames=lstm_frames, sub_scope="lstm%d"%0) lstm_states.append(lstm_state) support_predictions = [] for layer in xrange(num_layers): lstm_input, lstm_frames = self.resolution(lstm_output, num_frames, 2) lstm_state, lstm_output = self.lstm(lstm_input, vocab_size, num_frames=lstm_frames, feature_sizes=FLAGS.lstm_cells, sub_scope="lstm%d"%(layer+1)) lstm_states.append(lstm_state) final_states = tf.concat(lstm_states, axis=1) aggregated_model = getattr(video_level_models, FLAGS.video_level_classifier_model) return aggregated_model().create_model( model_input=final_states, original_input=model_input, vocab_size=vocab_size, num_frames=num_frames, **unused_params) def get_length_code(self, num_frames): code_0 = tf.cast(num_frames <= 60, dtype=tf.int32) code_1 = tf.cast(num_frames > 60, dtype=tf.int32) * tf.cast(num_frames <= 120, dtype=tf.int32) code_2 = tf.cast(num_frames > 120, dtype=tf.int32) * tf.cast(num_frames <= 180, dtype=tf.int32) code_3 = tf.cast(num_frames > 180, dtype=tf.int32) * tf.cast(num_frames <= 240, dtype=tf.int32) code_4 = tf.cast(num_frames > 240, dtype=tf.int32) codes = map(lambda x: tf.expand_dims(x, dim=1), [code_0, code_1, code_2, code_3, code_4]) length_code = tf.cast(tf.concat(codes, axis=1), dtype=tf.float32) return length_code def resolution(self, model_input_raw, num_frames, resolution, method="SELECT"): frame_dim = len(model_input_raw.get_shape()) - 2 feature_dim = len(model_input_raw.get_shape()) - 1 max_frames = model_input_raw.get_shape().as_list()[frame_dim] num_features = model_input_raw.get_shape().as_list()[feature_dim] if resolution > 1: new_max_frames = max_frames / resolution cut_frames = new_max_frames * resolution model_input_raw = model_input_raw[:, :cut_frames, :] model_input_raw = tf.reshape(model_input_raw, shape=[-1,new_max_frames,resolution,num_features]) if method == "MEAN": model_input_raw = tf.reduce_mean(model_input_raw, axis=2) elif method == "MAX": model_input_raw = tf.reduce_max(model_input_raw, axis=2) elif method == "SELECT": model_input_raw = model_input_raw[:,:,resolution-1,:] model_input = tf.nn.l2_normalize(model_input_raw, feature_dim) num_frames = num_frames / resolution else: model_input = tf.nn.l2_normalize(model_input_raw, feature_dim) return model_input, num_frames ================================================ FILE: youtube-8m-wangheda/all_frame_models/gru_pooling_model.py ================================================ import sys import models import model_utils import math import numpy as np import video_level_models import tensorflow as tf import utils import tensorflow.contrib.slim as slim from tensorflow import flags FLAGS = flags.FLAGS class GruPoolingModel(models.BaseModel): def create_model(self, model_input, vocab_size, num_frames, **unused_params): """Creates a model which uses a stack of LSTMs to represent the video. Args: model_input: A 'batch_size' x 'max_frames' x 'num_features' matrix of input features. vocab_size: The number of classes in the dataset. num_frames: A vector of length 'batch' which indicates the number of frames for each video (before padding). Returns: A dictionary with a tensor containing the probability predictions of the model in the 'predictions' key. The dimensions of the tensor are 'batch_size' x 'num_classes'. """ gru_size = FLAGS.gru_cells number_of_layers = FLAGS.gru_layers ## Batch normalize the input stacked_gru = tf.contrib.rnn.MultiRNNCell( [ tf.contrib.rnn.GRUCell(gru_size) for _ in range(number_of_layers) ], state_is_tuple=False) loss = 0.0 with tf.variable_scope("RNN"): outputs, state = tf.nn.dynamic_rnn(stacked_gru, model_input, sequence_length=num_frames, time_major=False, swap_memory=FLAGS.rnn_swap_memory, dtype=tf.float32) num_frames_matrix = tf.maximum(tf.cast(tf.expand_dims(num_frames, axis=1), dtype=tf.float32), tf.ones([batch_size, 1])) pooling_output = tf.reduce_sum(outputs, axis = 1) / num_frames_matrix aggregated_model = getattr(video_level_models, FLAGS.video_level_classifier_model) return aggregated_model().create_model( model_input=pooling_output, original_input=model_input, vocab_size=vocab_size, **unused_params) ================================================ FILE: youtube-8m-wangheda/all_frame_models/gru_with_pooling_model.py ================================================ import sys import models import model_utils import math import numpy as np import video_level_models import tensorflow as tf import utils import tensorflow.contrib.slim as slim from tensorflow import flags FLAGS = flags.FLAGS class GruWithPoolingModel(models.BaseModel): def create_model(self, model_input, vocab_size, num_frames, **unused_params): """Creates a model which uses a stack of LSTMs to represent the video. Args: model_input: A 'batch_size' x 'max_frames' x 'num_features' matrix of input features. vocab_size: The number of classes in the dataset. num_frames: A vector of length 'batch' which indicates the number of frames for each video (before padding). Returns: A dictionary with a tensor containing the probability predictions of the model in the 'predictions' key. The dimensions of the tensor are 'batch_size' x 'num_classes'. """ gru_size = FLAGS.gru_cells number_of_layers = FLAGS.gru_layers ## Batch normalize the input stacked_gru = tf.contrib.rnn.MultiRNNCell( [ tf.contrib.rnn.GRUCell(gru_size) for _ in range(number_of_layers) ], state_is_tuple=False) loss = 0.0 with tf.variable_scope("RNN"): outputs, state = tf.nn.dynamic_rnn(stacked_gru, model_input, sequence_length=num_frames, time_major=False, swap_memory=FLAGS.rnn_swap_memory, dtype=tf.float32) num_frames_matrix = tf.maximum(tf.cast(tf.expand_dims(num_frames, axis=1), dtype=tf.float32), tf.ones([batch_size, 1])) pooling_output = tf.reduce_sum(outputs, axis = 1) / num_frames_matrix final_output = tf.concat([pooling_output, state], axis = 1) aggregated_model = getattr(video_level_models, FLAGS.video_level_classifier_model) return aggregated_model().create_model( model_input=final_output, original_input=model_input, vocab_size=vocab_size, **unused_params) ================================================ FILE: youtube-8m-wangheda/all_frame_models/layernorm_lstm_memory_model.py ================================================ import sys import models import model_utils import math import numpy as np import video_level_models import tensorflow as tf import utils import tensorflow.contrib.slim as slim from tensorflow import flags FLAGS = flags.FLAGS class LayerNormLstmMemoryModel(models.BaseModel): def create_model(self, model_input, vocab_size, num_frames, dropout=False, keep_prob=None, noise_level=None, **unused_params): """Creates a model which uses a stack of LSTMs to represent the video. Args: model_input: A 'batch_size' x 'max_frames' x 'num_features' matrix of input features. vocab_size: The number of classes in the dataset. num_frames: A vector of length 'batch' which indicates the number of frames for each video (before padding). Returns: A dictionary with a tensor containing the probability predictions of the model in the 'predictions' key. The dimensions of the tensor are 'batch_size' x 'num_classes'. """ lstm_size = int(FLAGS.lstm_cells) number_of_layers = FLAGS.lstm_layers ## Batch normalize the input if dropout: stacked_lstm = tf.contrib.rnn.MultiRNNCell( [ tf.contrib.rnn.LayerNormBasicLSTMCell( lstm_size, dropout_keep_prob=keep_prob) for _ in range(number_of_layers) ], state_is_tuple=True) else: stacked_lstm = tf.contrib.rnn.MultiRNNCell( [ tf.contrib.rnn.LayerNormBasicLSTMCell( lstm_size) for _ in range(number_of_layers) ], state_is_tuple=True) loss = 0.0 with tf.variable_scope("RNN"): outputs, state = tf.nn.dynamic_rnn(stacked_lstm, model_input, sequence_length=num_frames, swap_memory=FLAGS.rnn_swap_memory, dtype=tf.float32) final_state = tf.concat(map(lambda x: x.c, state), axis = 1) if noise_level is not None: final_state = final_state + tf.random_normal(tf.shape(final_state), mean=0.0, stddev=noise_level) aggregated_model = getattr(video_level_models, FLAGS.video_level_classifier_model) return aggregated_model().create_model( model_input=final_state, original_input=model_input, vocab_size=vocab_size, **unused_params) ================================================ FILE: youtube-8m-wangheda/all_frame_models/logistic_model.py ================================================ import sys import models import model_utils import math import numpy as np import video_level_models import tensorflow as tf import utils import tensorflow.contrib.slim as slim from tensorflow import flags FLAGS = flags.FLAGS class FrameLevelLogisticModel(models.BaseModel): def create_model(self, model_input, vocab_size, num_frames, **unused_params): """Creates a model which uses a logistic classifier over the average of the frame-level features. This class is intended to be an example for implementors of frame level models. If you want to train a model over averaged features it is more efficient to average them beforehand rather than on the fly. Args: model_input: A 'batch_size' x 'max_frames' x 'num_features' matrix of input features. vocab_size: The number of classes in the dataset. num_frames: A vector of length 'batch' which indicates the number of frames for each video (before padding). Returns: A dictionary with a tensor containing the probability predictions of the model in the 'predictions' key. The dimensions of the tensor are 'batch_size' x 'num_classes'. """ num_frames = tf.cast(tf.expand_dims(num_frames, 1), tf.float32) feature_size = model_input.get_shape().as_list()[2] denominators = tf.reshape( tf.tile(num_frames, [1, feature_size]), [-1, feature_size]) avg_pooled = tf.reduce_sum(model_input, axis=[1]) / denominators output = slim.fully_connected( avg_pooled, vocab_size, activation_fn=tf.nn.sigmoid, weights_regularizer=slim.l2_regularizer(1e-8)) return {"predictions": output} ================================================ FILE: youtube-8m-wangheda/all_frame_models/lstm_advanced_model.py ================================================ import sys import models import model_utils import math import numpy as np import video_level_models import tensorflow as tf import utils import tensorflow.contrib.slim as slim from tensorflow import flags FLAGS = flags.FLAGS class LstmAdvancedModel(models.BaseModel): def create_model(self, model_input, vocab_size, num_frames, **unused_params): """Creates a model which uses a stack of LSTMs to represent the video. Args: model_input: A 'batch_size' x 'max_frames' x 'num_features' matrix of input features. vocab_size: The number of classes in the dataset. num_frames: A vector of length 'batch' which indicates the number of frames for each video (before padding). Returns: A dictionary with a tensor containing the probability predictions of the model in the 'predictions' key. The dimensions of the tensor are 'batch_size' x 'num_classes'. """ lstm_size = int(FLAGS.lstm_cells) number_of_layers = FLAGS.lstm_layers ## Batch normalize the input stacked_lstm = tf.contrib.rnn.MultiRNNCell( [ tf.contrib.rnn.LSTMCell( lstm_size, forget_bias=1.0, state_is_tuple=False, use_peepholes=False) for _ in range(number_of_layers) ], state_is_tuple=False) loss = 0.0 with tf.variable_scope("RNN"): outputs, state = tf.nn.dynamic_rnn(stacked_lstm, model_input, sequence_length=num_frames, swap_memory=FLAGS.rnn_swap_memory, dtype=tf.float32) aggregated_model = getattr(video_level_models, FLAGS.video_level_classifier_model) return aggregated_model().create_model( model_input=state, original_input=model_input, vocab_size=vocab_size, **unused_params) ================================================ FILE: youtube-8m-wangheda/all_frame_models/lstm_attention_lstm_model.py ================================================ import sys import models import model_utils import math import numpy as np import video_level_models import tensorflow as tf import utils import tensorflow.contrib.slim as slim from tensorflow import flags FLAGS = flags.FLAGS class LstmAttentionLstmModel(models.BaseModel): def create_model(self, model_input, vocab_size, num_frames, **unused_params): """Creates a model which uses a stack of LSTMs to represent the video. Args: model_input: A 'batch_size' x 'max_frames' x 'num_features' matrix of input features. vocab_size: The number of classes in the dataset. num_frames: A vector of length 'batch' which indicates the number of frames for each video (before padding). Returns: A dictionary with a tensor containing the probability predictions of the model in the 'predictions' key. The dimensions of the tensor are 'batch_size' x 'num_classes'. """ lstm_size = int(FLAGS.lstm_cells) number_of_layers = FLAGS.lstm_layers l2_penalty = unused_params.get("l2_penalty", 1e-8) max_frames = model_input.get_shape().as_list()[1] mask_array = [] for i in xrange(max_frames + 1): tmp = [0.0] * max_frames for j in xrange(i): tmp[j] = 1.0 mask_array.append(tmp) mask_array = np.array(mask_array) mask_init = tf.constant_initializer(mask_array) mask_emb = tf.get_variable("mask_emb", shape = [max_frames + 1, max_frames], dtype = tf.float32, trainable = False, initializer = mask_init) ## Batch normalize the input stacked_cell = tf.contrib.rnn.MultiRNNCell( [ tf.contrib.rnn.BasicLSTMCell( lstm_size, forget_bias=1.0, state_is_tuple=True) for _ in range(number_of_layers) ], state_is_tuple=True) stacked_att_cell = tf.contrib.rnn.MultiRNNCell( [ tf.contrib.rnn.BasicLSTMCell( lstm_size, forget_bias=1.0, state_is_tuple=True) for _ in range(number_of_layers) ], state_is_tuple=True) loss = 0.0 with tf.variable_scope("ATT"): att_outputs, att_state = tf.nn.dynamic_rnn(stacked_att_cell, model_input, sequence_length=num_frames, swap_memory=FLAGS.rnn_swap_memory, dtype=tf.float32) att_final_memory = tf.concat(map(lambda x: x.c, att_state), axis = 1) with tf.variable_scope("RNN"): outputs, state = tf.nn.dynamic_rnn(stacked_cell, model_input, sequence_length=num_frames, swap_memory=FLAGS.rnn_swap_memory, dtype=tf.float32) final_memory = tf.concat(map(lambda x: x.c, state), axis = 1) num_frames_matrix = tf.maximum(tf.cast( tf.expand_dims(tf.expand_dims(num_frames, axis=1), axis=2), dtype=tf.float32), 1.0) mask = tf.expand_dims(tf.nn.embedding_lookup(mask_emb, num_frames), axis = 2) attention_fc = slim.fully_connected( att_outputs, 1, activation_fn=tf.nn.sigmoid, weights_regularizer=slim.l2_regularizer(l2_penalty)) print attention_fc attention = attention_fc * mask attention_sum = tf.reduce_sum(attention, axis = 1, keep_dims = True) + 1e-8 attention = attention / attention_sum print attention attended_output = tf.reduce_sum(attention * outputs, axis = 1) final_state = tf.concat([attended_output, att_final_memory, final_memory], axis = 1) print "final_state", final_state aggregated_model = getattr(video_level_models, FLAGS.video_level_classifier_model) return aggregated_model().create_model( model_input=final_state, original_input=model_input, vocab_size=vocab_size, **unused_params) ================================================ FILE: youtube-8m-wangheda/all_frame_models/lstm_attention_max_pooling_model.py ================================================ import math import models import tensorflow as tf import numpy as np import utils from tensorflow import flags import tensorflow.contrib.slim as slim FLAGS = flags.FLAGS class LstmAttentionMaxPoolingModel(models.BaseModel): """Max pooling over temporal weighted sums (attention) of lstm outputs.""" def create_model(self, model_input, vocab_size, num_frames, num_mixtures=None, l2_penalty=1e-8, sub_scope="", original_input=None, **unused_params): """Creates a model which uses a stack of LSTMs to represent the video. Args: model_input: A 'batch_size' x 'max_frames' x 'num_features' matrix of input features. vocab_size: The number of classes in the dataset. num_frames: A vector of length 'batch' which indicates the number of frames for each video (before padding). Returns: A dictionary with a tensor containing the probability predictions of the model in the 'predictions' key. The dimensions of the tensor are 'batch_size' x 'num_classes'. """ lstm_size = int(FLAGS.lstm_cells) number_of_layers = FLAGS.lstm_layers num_attentions = FLAGS.lstm_attentions batch_size, max_frames, num_features = model_input.get_shape().as_list() mask = tf.sequence_mask(lengths=num_frames, maxlen=max_frames, dtype=tf.float32) ## Batch normalize the input stacked_lstm = tf.contrib.rnn.MultiRNNCell( [ tf.contrib.rnn.BasicLSTMCell( lstm_size, forget_bias=1.0, state_is_tuple=True) for _ in range(number_of_layers) ], state_is_tuple=True) with tf.variable_scope("RNN"): outputs, state = tf.nn.dynamic_rnn(stacked_lstm, model_input, sequence_length=num_frames, swap_memory=FLAGS.rnn_swap_memory, dtype=tf.float32) attention_activations = slim.fully_connected( tf.concat([model_input, outputs], axis=2), num_attentions, activation_fn=None, weights_regularizer=slim.l2_regularizer(l2_penalty), scope="attention-"+sub_scope) # Batch x #Attentions x #Frames attention_weights = tf.einsum("ijk,ij->ikj", tf.nn.softmax(attention_activations, dim=1), mask) attention_weights = attention_weights / tf.reduce_sum(attention_weights, axis=2, keep_dims=True) # Batch x #Attentions x #Features attention_outputs = tf.einsum("ijk,ilj->ilk", outputs, attention_weights) moe_predictions = self.sub_moe(attention_outputs, vocab_size, sub_scope="sub-moe") predictions = tf.reshape(moe_predictions, [-1, num_attentions, vocab_size]) max_predictions = tf.reduce_max(predictions, axis=1) return {"predictions": max_predictions} def sub_moe(self, model_input, vocab_size, num_mixtures=None, l2_penalty=1e-8, sub_scope="", **unused_params): num_mixtures = num_mixtures or FLAGS.moe_num_mixtures gate_activations = slim.fully_connected( model_input, vocab_size * (num_mixtures + 1), activation_fn=None, biases_initializer=None, weights_regularizer=slim.l2_regularizer(l2_penalty), scope="gates-"+sub_scope) expert_activations = slim.fully_connected( model_input, vocab_size * num_mixtures, activation_fn=None, weights_regularizer=slim.l2_regularizer(l2_penalty), scope="experts-"+sub_scope) gating_distribution = tf.nn.softmax(tf.reshape( gate_activations, [-1, num_mixtures + 1])) # (Batch * #Labels) x (num_mixtures + 1) expert_distribution = tf.nn.sigmoid(tf.reshape( expert_activations, [-1, num_mixtures])) # (Batch * #Labels) x num_mixtures final_probabilities_by_class_and_batch = tf.reduce_sum( gating_distribution[:, :num_mixtures] * expert_distribution, 1) final_probabilities = tf.reshape(final_probabilities_by_class_and_batch, [-1, vocab_size]) return final_probabilities ================================================ FILE: youtube-8m-wangheda/all_frame_models/lstm_attention_model.py ================================================ import sys import models import model_utils import math import numpy as np import video_level_models import tensorflow as tf import utils import tensorflow.contrib.slim as slim from tensorflow import flags FLAGS = flags.FLAGS class LstmAttentionModel(models.BaseModel): def create_model(self, model_input, vocab_size, num_frames, **unused_params): """Creates a model which uses a stack of LSTMs to represent the video. Args: model_input: A 'batch_size' x 'max_frames' x 'num_features' matrix of input features. vocab_size: The number of classes in the dataset. num_frames: A vector of length 'batch' which indicates the number of frames for each video (before padding). Returns: A dictionary with a tensor containing the probability predictions of the model in the 'predictions' key. The dimensions of the tensor are 'batch_size' x 'num_classes'. """ lstm_size = int(FLAGS.lstm_cells) number_of_layers = FLAGS.lstm_layers l2_penalty = unused_params.get("l2_penalty", 1e-8) max_frames = model_input.get_shape().as_list()[1] mask_array = [] for i in xrange(max_frames + 1): tmp = [0.0] * max_frames for j in xrange(i): tmp[j] = 1.0 mask_array.append(tmp) mask_array = np.array(mask_array) mask_init = tf.constant_initializer(mask_array) mask_emb = tf.get_variable("mask_emb", shape = [max_frames + 1, max_frames], dtype = tf.float32, trainable = False, initializer = mask_init) ## Batch normalize the input stacked_lstm = tf.contrib.rnn.MultiRNNCell( [ tf.contrib.rnn.BasicLSTMCell( lstm_size, forget_bias=1.0, state_is_tuple=False) for _ in range(number_of_layers) ], state_is_tuple=False) loss = 0.0 with tf.variable_scope("RNN"): outputs, state = tf.nn.dynamic_rnn(stacked_lstm, model_input, sequence_length=num_frames, swap_memory=FLAGS.rnn_swap_memory, dtype=tf.float32) num_frames_matrix = tf.maximum(tf.cast( tf.expand_dims(tf.expand_dims(num_frames, axis=1), axis=2), dtype=tf.float32), 1.0) mask = tf.expand_dims(tf.nn.embedding_lookup(mask_emb, num_frames), axis = 2) attention_fc = slim.fully_connected( outputs, 1, activation_fn=tf.nn.sigmoid, weights_regularizer=slim.l2_regularizer(l2_penalty)) print attention_fc attention = attention_fc * mask attention_sum = tf.reduce_sum(attention, axis = 1, keep_dims = True) + 1e-8 attention = attention / attention_sum print attention mean_input = tf.reduce_sum(attention * model_input, axis = 1) final_state = tf.concat([mean_input, state], axis = 1) print mean_input aggregated_model = getattr(video_level_models, FLAGS.video_level_classifier_model) return aggregated_model().create_model( model_input=final_state, original_input=model_input, vocab_size=vocab_size, **unused_params) ================================================ FILE: youtube-8m-wangheda/all_frame_models/lstm_auxloss_deep_combine_chain_model.py ================================================ import sys import models import model_utils import math import numpy as np import video_level_models import tensorflow as tf import utils import tensorflow.contrib.slim as slim from tensorflow import flags FLAGS = flags.FLAGS class LstmAuxlossDeepCombineChainModel(models.BaseModel): def create_model(self, model_input, vocab_size, num_frames, num_mixtures=None, l2_penalty=1e-8, sub_scope="", original_input=None, **unused_params): """Creates a model that use different times of output of lstm model_input: A 'batch_size' x 'max_frames' x 'num_features' matrix of input features. """ number_of_layers = FLAGS.lstm_layers deep_chain_layers = FLAGS.deep_chain_layers relu_cells = FLAGS.deep_chain_relu_cells batch_size = tf.shape(model_input)[0] lstm_sizes = map(int, FLAGS.lstm_cells.split(",")) feature_names, feature_sizes = utils.GetListOfFeatureNamesAndSizes( FLAGS.feature_names, FLAGS.feature_sizes) sub_inputs = [tf.nn.l2_normalize(x, dim=2) for x in tf.split(model_input, feature_sizes, axis = 2)] assert len(lstm_sizes) == len(feature_sizes), \ "length of lstm_sizes (={}) != length of feature_sizes (={})".format( \ len(lstm_sizes), len(feature_sizes)) outputs = [] for i in xrange(len(feature_sizes)): with tf.variable_scope("RNN%d" % i): sub_input = sub_inputs[i] lstm_size = lstm_sizes[i] ## Batch normalize the input stacked_lstm = tf.contrib.rnn.MultiRNNCell( [ tf.contrib.rnn.BasicLSTMCell( lstm_size, forget_bias=1.0, state_is_tuple=True) for _ in range(number_of_layers) ], state_is_tuple=True) output, state = tf.nn.dynamic_rnn(stacked_lstm, sub_input, sequence_length=num_frames, swap_memory=FLAGS.rnn_swap_memory, time_major=False, dtype=tf.float32) outputs.append(output) # generate time to get frame output num_layers = deep_chain_layers + 1 timed_num_frames = [] print "num_layers =", num_layers unit_num_frames = num_frames / num_layers for i in xrange(deep_chain_layers): timed_num_frames.append(unit_num_frames * (i+1)) timed_num_frames.append(num_frames) # generate frame outputs from certain time points (num_frames - 1 to turn length into index (base 0)) timed_outputs = [] for sub_num_frames in timed_num_frames: frame_outputs = [] for output in outputs: frame_index = tf.stack([tf.range(batch_size), tf.maximum(sub_num_frames-1,0)], axis=1) frame_output = tf.gather_nd(output, frame_index) frame_outputs.append(frame_output) timed_outputs.append(tf.concat(frame_outputs, axis=1)) # deep combine model predictions = [] relu_layers = [] for layer in xrange(num_layers): if relu_layers: next_input = tf.concat([timed_outputs[layer]] + relu_layers, axis=1) else: next_input = timed_outputs[layer] sub_prediction = self.sub_model(next_input, vocab_size, sub_scope=sub_scope+"prediction-%d"%layer) predictions.append(sub_prediction) sub_relu = slim.fully_connected( sub_prediction, relu_cells, activation_fn=tf.nn.relu, weights_regularizer=slim.l2_regularizer(l2_penalty), scope=sub_scope+"relu-%d"%layer) if layer + 1 < num_layers: relu_norm = tf.nn.l2_normalize(sub_relu, dim=1) relu_layers.append(relu_norm) main_predictions = predictions[-1] support_predictions = tf.concat(predictions[:-1], axis=1) return {"predictions": main_predictions, "support_predictions": support_predictions} def sub_model(self, model_input, vocab_size, num_mixtures=None, l2_penalty=1e-8, sub_scope="", **unused_params): num_mixtures = num_mixtures or FLAGS.moe_num_mixtures gate_activations = slim.fully_connected( model_input, vocab_size * (num_mixtures + 1), activation_fn=None, biases_initializer=None, weights_regularizer=slim.l2_regularizer(l2_penalty), scope="gates-"+sub_scope) expert_activations = slim.fully_connected( model_input, vocab_size * num_mixtures, activation_fn=None, weights_regularizer=slim.l2_regularizer(l2_penalty), scope="experts-"+sub_scope) gating_distribution = tf.nn.softmax(tf.reshape( gate_activations, [-1, num_mixtures + 1])) # (Batch * #Labels) x (num_mixtures + 1) expert_distribution = tf.nn.sigmoid(tf.reshape( expert_activations, [-1, num_mixtures])) # (Batch * #Labels) x num_mixtures final_probabilities_by_class_and_batch = tf.reduce_sum( gating_distribution[:, :num_mixtures] * expert_distribution, 1) final_probabilities = tf.reshape(final_probabilities_by_class_and_batch, [-1, vocab_size]) return final_probabilities ================================================ FILE: youtube-8m-wangheda/all_frame_models/lstm_cnn_deep_combine_chain_model.py ================================================ import math import models import tensorflow as tf import numpy as np import utils from tensorflow import flags import tensorflow.contrib.slim as slim FLAGS = flags.FLAGS class LstmCnnDeepCombineChainModel(models.BaseModel): """A softmax over a mixture of logistic models (with L2 regularization).""" def cnn(self, model_input, l2_penalty=1e-8, num_filters = [1024, 1024, 1024], filter_sizes = [1,2,3], sub_scope="", **unused_params): max_frames = model_input.get_shape().as_list()[1] num_features = model_input.get_shape().as_list()[2] shift_inputs = [] for i in xrange(max(filter_sizes)): if i == 0: shift_inputs.append(model_input) else: shift_inputs.append(tf.pad(model_input, paddings=[[0,0],[i,0],[0,0]])[:,:max_frames,:]) cnn_outputs = [] for nf, fs in zip(num_filters, filter_sizes): sub_input = tf.concat(shift_inputs[:fs], axis=2) sub_filter = tf.get_variable(sub_scope+"cnn-filter-len%d"%fs, shape=[num_features*fs, nf], dtype=tf.float32, initializer=tf.truncated_normal_initializer(mean=0.0, stddev=0.1), regularizer=tf.contrib.layers.l2_regularizer(l2_penalty)) cnn_outputs.append(tf.einsum("ijk,kl->ijl", sub_input, sub_filter)) cnn_output = tf.concat(cnn_outputs, axis=2) return cnn_output def create_model(self, model_input, vocab_size, num_frames, num_mixtures=None, l2_penalty=1e-8, sub_scope="", original_input=None, **unused_params): num_supports = FLAGS.num_supports num_layers = FLAGS.deep_chain_layers relu_cells = FLAGS.deep_chain_relu_cells max_frames = model_input.get_shape().as_list()[1] relu_layers = [] support_predictions = [] mask = self.get_mask(max_frames, num_frames) mean_input = tf.einsum("ijk,ij->ik", model_input, mask) \ / tf.expand_dims(tf.cast(num_frames, dtype=tf.float32), dim=1) # LSTM lstm_output = self.lstmoutput(model_input, vocab_size, num_frames) mean_relu = slim.fully_connected( mean_input, relu_cells, activation_fn=tf.nn.relu, weights_regularizer=slim.l2_regularizer(l2_penalty), scope=sub_scope+"mean-relu") mean_relu_norm = tf.nn.l2_normalize(mean_relu, dim=1) relu_layers.append(mean_relu_norm) cnn_output = self.cnn(lstm_output, num_filters=[relu_cells,2*relu_cells,relu_cells], filter_sizes=[1,2,3], sub_scope=sub_scope+"cnn0") max_cnn_output = tf.reduce_max(cnn_output, axis=1) normalized_cnn_output = tf.nn.l2_normalize(max_cnn_output, dim=1) next_input = normalized_cnn_output for layer in xrange(num_layers): sub_prediction = self.sub_model(next_input, vocab_size, sub_scope=sub_scope+"prediction-%d"%layer) support_predictions.append(sub_prediction) sub_relu = slim.fully_connected( sub_prediction, relu_cells, activation_fn=tf.nn.relu, weights_regularizer=slim.l2_regularizer(l2_penalty), scope=sub_scope+"relu-%d"%layer) relu_norm = tf.nn.l2_normalize(sub_relu, dim=1) relu_layers.append(relu_norm) cnn_output = self.cnn(lstm_output, num_filters=[relu_cells,relu_cells,relu_cells*2], filter_sizes=[1,2,3], sub_scope=sub_scope+"cnn%d"%(layer+1)) max_cnn_output = tf.reduce_max(cnn_output, axis=1) normalized_cnn_output = tf.nn.l2_normalize(max_cnn_output, dim=1) next_input = tf.concat([normalized_cnn_output] + relu_layers, axis=1) main_predictions = self.sub_model(next_input, vocab_size, sub_scope=sub_scope+"-main") support_predictions = tf.concat(support_predictions, axis=1) return {"predictions": main_predictions, "support_predictions": support_predictions} def sub_model(self, model_input, vocab_size, num_mixtures=None, l2_penalty=1e-8, sub_scope="", **unused_params): num_mixtures = num_mixtures or FLAGS.moe_num_mixtures gate_activations = slim.fully_connected( model_input, vocab_size * (num_mixtures + 1), activation_fn=None, biases_initializer=None, weights_regularizer=slim.l2_regularizer(l2_penalty), scope="gates-"+sub_scope) expert_activations = slim.fully_connected( model_input, vocab_size * num_mixtures, activation_fn=None, weights_regularizer=slim.l2_regularizer(l2_penalty), scope="experts-"+sub_scope) gating_distribution = tf.nn.softmax(tf.reshape( gate_activations, [-1, num_mixtures + 1])) # (Batch * #Labels) x (num_mixtures + 1) expert_distribution = tf.nn.sigmoid(tf.reshape( expert_activations, [-1, num_mixtures])) # (Batch * #Labels) x num_mixtures final_probabilities_by_class_and_batch = tf.reduce_sum( gating_distribution[:, :num_mixtures] * expert_distribution, 1) final_probabilities = tf.reshape(final_probabilities_by_class_and_batch, [-1, vocab_size]) return final_probabilities def get_mask(self, max_frames, num_frames): mask_array = [] for i in xrange(max_frames + 1): tmp = [0.0] * max_frames for j in xrange(i): tmp[j] = 1.0 mask_array.append(tmp) mask_array = np.array(mask_array) mask_init = tf.constant_initializer(mask_array) mask_emb = tf.get_variable("mask_emb", shape = [max_frames + 1, max_frames], dtype = tf.float32, trainable = False, initializer = mask_init) mask = tf.nn.embedding_lookup(mask_emb, num_frames) return mask def lstmoutput(self, model_input, vocab_size, num_frames): number_of_layers = FLAGS.lstm_layers lstm_sizes = map(int, FLAGS.lstm_cells.split(",")) feature_names, feature_sizes = utils.GetListOfFeatureNamesAndSizes( FLAGS.feature_names, FLAGS.feature_sizes) sub_inputs = [tf.nn.l2_normalize(x, dim=2) for x in tf.split(model_input, feature_sizes, axis = 2)] assert len(lstm_sizes) == len(feature_sizes), \ "length of lstm_sizes (={}) != length of feature_sizes (={})".format( \ len(lstm_sizes), len(feature_sizes)) outputs = [] for i in xrange(len(feature_sizes)): with tf.variable_scope("RNN%d" % i): sub_input = sub_inputs[i] lstm_size = lstm_sizes[i] ## Batch normalize the input stacked_lstm = tf.contrib.rnn.MultiRNNCell( [ tf.contrib.rnn.BasicLSTMCell( lstm_size, forget_bias=1.0, state_is_tuple=True) for _ in range(number_of_layers) ], state_is_tuple=True) output, state = tf.nn.dynamic_rnn(stacked_lstm, sub_input, sequence_length=num_frames, swap_memory=FLAGS.rnn_swap_memory, dtype=tf.float32) outputs.append(output) # concat final_output = tf.concat(outputs, axis=2) return final_output ================================================ FILE: youtube-8m-wangheda/all_frame_models/lstm_divided_model.py ================================================ import sys import models import model_utils import math import numpy as np import video_level_models import tensorflow as tf import utils import tensorflow.contrib.slim as slim from tensorflow import flags FLAGS = flags.FLAGS class LstmDividedModel(models.BaseModel): def create_model(self, model_input, vocab_size, num_frames, **unused_params): """Creates a model which uses a stack of LSTMs to represent the video. Args: model_input: A 'batch_size' x 'max_frames' x 'num_features' matrix of input features. vocab_size: The number of classes in the dataset. num_frames: A vector of length 'batch' which indicates the number of frames for each video (before padding). Returns: A dictionary with a tensor containing the probability predictions of the model in the 'predictions' key. The dimensions of the tensor are 'batch_size' x 'num_classes'. """ lstm_size = int(FLAGS.lstm_cells) number_of_layers = FLAGS.lstm_layers with tf.device("/gpu:1"): ## Batch normalize the input stacked_lstm = tf.contrib.rnn.MultiRNNCell( [ tf.contrib.rnn.BasicLSTMCell( lstm_size, forget_bias=1.0, state_is_tuple=False) for _ in range(number_of_layers) ], state_is_tuple=False) loss = 0.0 with tf.variable_scope("RNN"): outputs, state = tf.nn.dynamic_rnn(stacked_lstm, model_input, sequence_length=num_frames, swap_memory=FLAGS.rnn_swap_memory, dtype=tf.float32) with tf.device("/gpu:0"): aggregated_model = getattr(video_level_models, FLAGS.video_level_classifier_model) predictions = aggregated_model().create_model( model_input=state, original_input=model_input, vocab_size=vocab_size, **unused_params) return predictions ================================================ FILE: youtube-8m-wangheda/all_frame_models/lstm_look_back_model.py ================================================ import sys import models import model_utils import math import numpy as np import video_level_models import tensorflow as tf import utils import tensorflow.contrib.slim as slim from tensorflow import flags FLAGS = flags.FLAGS class LstmLookBackModel(models.BaseModel): def shift(self, model_input, shift_width, **unused_params): max_frames = model_input.get_shape().as_list()[1] num_features = model_input.get_shape().as_list()[2] shift_inputs = [] for i in xrange(shift_width): if i == 0: shift_inputs.append(model_input) else: shift_inputs.append(tf.pad(model_input, paddings=[[0,0],[i,0],[0,0]])[:,:max_frames,:]) shift_output = tf.concat(shift_inputs, axis=2) return shift_output def create_model(self, model_input, vocab_size, num_frames, **unused_params): """Creates a model which uses a stack of LSTMs to represent the video. Args: model_input: A 'batch_size' x 'max_frames' x 'num_features' matrix of input features. vocab_size: The number of classes in the dataset. num_frames: A vector of length 'batch' which indicates the number of frames for each video (before padding). Returns: A dictionary with a tensor containing the probability predictions of the model in the 'predictions' key. The dimensions of the tensor are 'batch_size' x 'num_classes'. """ number_of_layers = FLAGS.lstm_layers shift_width = FLAGS.lstm_look_back lstm_sizes = map(int, FLAGS.lstm_cells.split(",")) feature_names, feature_sizes = utils.GetListOfFeatureNamesAndSizes( FLAGS.feature_names, FLAGS.feature_sizes) sub_inputs = [tf.nn.l2_normalize(x, dim=2) for x in tf.split(model_input, feature_sizes, axis = 2)] assert len(lstm_sizes) == len(feature_sizes), \ "length of lstm_sizes (={}) != length of feature_sizes (={})".format( \ len(lstm_sizes), len(feature_sizes)) if shift_width < 1: shift_width = 1 outputs = [] states = [] for i in xrange(len(feature_sizes)): with tf.variable_scope("RNN%d" % i): sub_input = sub_inputs[i] lstm_size = lstm_sizes[i] look_back = self.shift(sub_input, shift_width=shift_width) ## Batch normalize the input stacked_lstm = tf.contrib.rnn.MultiRNNCell( [ tf.contrib.rnn.BasicLSTMCell( lstm_size, forget_bias=1.0, state_is_tuple=True) for _ in range(number_of_layers) ], state_is_tuple=True) output, state = tf.nn.dynamic_rnn(stacked_lstm, look_back, sequence_length=num_frames, swap_memory=FLAGS.rnn_swap_memory, dtype=tf.float32) outputs.append(output) states.extend(map(lambda x: x.h, state)) # concat final_state = tf.concat(states, axis = 1) aggregated_model = getattr(video_level_models, FLAGS.video_level_classifier_model) return aggregated_model().create_model( model_input=final_state, original_input=model_input, vocab_size=vocab_size, **unused_params) ================================================ FILE: youtube-8m-wangheda/all_frame_models/lstm_memory_chain_model.py ================================================ import sys import models import model_utils import math import numpy as np import video_level_models import tensorflow as tf import utils import tensorflow.contrib.slim as slim from tensorflow import flags FLAGS = flags.FLAGS class LstmMemoryChainModel(models.BaseModel): """Classifier chain model of lstm memory""" def create_model(self, model_input, vocab_size, num_frames, **unused_params): """Creates a model which uses a stack of LSTMs to represent the video. Args: model_input: A 'batch_size' x 'max_frames' x 'num_features' matrix of input features. vocab_size: The number of classes in the dataset. num_frames: A vector of length 'batch' which indicates the number of frames for each video (before padding). Returns: A dictionary with a tensor containing the probability predictions of the model in the 'predictions' key. The dimensions of the tensor are 'batch_size' x 'num_classes'. """ lstm_size = int(FLAGS.lstm_cells) number_of_layers = FLAGS.lstm_layers num_supports = FLAGS.num_supports ## Batch normalize the input stacked_lstm = tf.contrib.rnn.MultiRNNCell( [ tf.contrib.rnn.BasicLSTMCell( lstm_size, forget_bias=1.0, state_is_tuple=True) for _ in range(number_of_layers) ], state_is_tuple=True) loss = 0.0 with tf.variable_scope("RNN"): outputs, state = tf.nn.dynamic_rnn(stacked_lstm, model_input, sequence_length=num_frames, swap_memory=FLAGS.rnn_swap_memory, dtype=tf.float32) final_state = tf.concat(map(lambda x: x.c, state), axis = 1) aggregated_model = getattr(video_level_models, FLAGS.video_level_classifier_model) support_predictions = aggregated_model().create_model( model_input=final_state, original_input=model_input, vocab_size=num_supports, sub_scope="support", **unused_params) support_predictions = support_predictions["predictions"] main_state = tf.concat([final_state, support_predictions], axis=1) predictions = aggregated_model().create_model( model_input=main_state, original_input=model_input, vocab_size=vocab_size, sub_scope="main", **unused_params) predictions["support_predictions"] = support_predictions return predictions ================================================ FILE: youtube-8m-wangheda/all_frame_models/lstm_memory_deep_chain_model.py ================================================ import sys import models import model_utils import math import numpy as np import video_level_models import tensorflow as tf import utils import tensorflow.contrib.slim as slim from tensorflow import flags FLAGS = flags.FLAGS class LstmMemoryDeepChainModel(models.BaseModel): """Classifier chain model of lstm memory""" def create_model(self, model_input, vocab_size, num_frames, l2_penalty=1e-8, sub_scope="", original_input=None, **unused_params): """Creates a model which uses a stack of LSTMs to represent the video. Args: model_input: A 'batch_size' x 'max_frames' x 'num_features' matrix of input features. vocab_size: The number of classes in the dataset. num_frames: A vector of length 'batch' which indicates the number of frames for each video (before padding). Returns: A dictionary with a tensor containing the probability predictions of the model in the 'predictions' key. The dimensions of the tensor are 'batch_size' x 'num_classes'. """ lstm_size = int(FLAGS.lstm_cells) number_of_layers = FLAGS.lstm_layers num_supports = FLAGS.num_supports num_layers = FLAGS.deep_chain_layers relu_cells = FLAGS.deep_chain_relu_cells lstm_output = self.sub_lstm(model_input, num_frames, lstm_size, number_of_layers, sub_scope="lstm-%d"%0) next_input = lstm_output support_predictions = [] for layer in xrange(num_layers): sub_prediction = self.sub_moe(next_input, vocab_size, sub_scope=sub_scope+"prediction-%d"%layer) sub_relu = slim.fully_connected( sub_prediction, relu_cells, activation_fn=tf.nn.relu, weights_regularizer=slim.l2_regularizer(l2_penalty), scope=sub_scope+"relu-%d"%layer) relu_norm = tf.nn.l2_normalize(sub_relu, dim=1) lstm_output = self.sub_lstm(model_input, num_frames, lstm_size, number_of_layers, sub_scope="lstm-%d"%(layer+1)) next_input = tf.concat([lstm_output, relu_norm], axis=1) support_predictions.append(sub_prediction) main_predictions = self.sub_moe(next_input, vocab_size, sub_scope=sub_scope+"-main") support_predictions = tf.concat(support_predictions, axis=1) return {"predictions": main_predictions, "support_predictions": support_predictions} def sub_lstm(self, model_input, num_frames, lstm_size, number_of_layers, sub_scope=""): stacked_lstm = tf.contrib.rnn.MultiRNNCell( [ tf.contrib.rnn.BasicLSTMCell( lstm_size, forget_bias=1.0, state_is_tuple=True) for _ in range(number_of_layers) ], state_is_tuple=True) loss = 0.0 with tf.variable_scope(sub_scope+"-RNN"): outputs, state = tf.nn.dynamic_rnn(stacked_lstm, model_input, sequence_length=num_frames, swap_memory=FLAGS.rnn_swap_memory, dtype=tf.float32) final_state = tf.concat(map(lambda x: x.c, state), axis = 1) return final_state def sub_moe(self, model_input, vocab_size, num_mixtures=None, l2_penalty=1e-8, sub_scope="", **unused_params): num_mixtures = num_mixtures or FLAGS.moe_num_mixtures gate_activations = slim.fully_connected( model_input, vocab_size * (num_mixtures + 1), activation_fn=None, biases_initializer=None, weights_regularizer=slim.l2_regularizer(l2_penalty), scope="gates-"+sub_scope) expert_activations = slim.fully_connected( model_input, vocab_size * num_mixtures, activation_fn=None, weights_regularizer=slim.l2_regularizer(l2_penalty), scope="experts-"+sub_scope) gating_distribution = tf.nn.softmax(tf.reshape( gate_activations, [-1, num_mixtures + 1])) # (Batch * #Labels) x (num_mixtures + 1) expert_distribution = tf.nn.sigmoid(tf.reshape( expert_activations, [-1, num_mixtures])) # (Batch * #Labels) x num_mixtures final_probabilities_by_class_and_batch = tf.reduce_sum( gating_distribution[:, :num_mixtures] * expert_distribution, 1) final_probabilities = tf.reshape(final_probabilities_by_class_and_batch, [-1, vocab_size]) return final_probabilities ================================================ FILE: youtube-8m-wangheda/all_frame_models/lstm_memory_input_chain_model.py ================================================ import sys import models import model_utils import math import numpy as np import video_level_models import tensorflow as tf import utils import tensorflow.contrib.slim as slim from tensorflow import flags FLAGS = flags.FLAGS class LstmMemoryInputChainModel(models.BaseModel): """Classifier chain model of lstm memory""" def create_model(self, model_input, vocab_size, num_frames, l2_penalty=1e-8, **unused_params): """Creates a model which uses a stack of LSTMs to represent the video. Args: model_input: A 'batch_size' x 'max_frames' x 'num_features' matrix of input features. vocab_size: The number of classes in the dataset. num_frames: A vector of length 'batch' which indicates the number of frames for each video (before padding). Returns: A dictionary with a tensor containing the probability predictions of the model in the 'predictions' key. The dimensions of the tensor are 'batch_size' x 'num_classes'. """ lstm_size = int(FLAGS.lstm_cells) number_of_layers = FLAGS.lstm_layers num_supports = FLAGS.num_supports num_support_embedding = 200 aggregated_model = getattr(video_level_models, FLAGS.video_level_classifier_model) mean_input = tf.reduce_mean(model_input, axis=1) support_predictions = aggregated_model().create_model( model_input=mean_input, original_input=model_input, vocab_size=num_supports, sub_scope="support", **unused_params) support_predictions = support_predictions["predictions"] support_activations = tf.stop_gradient(support_predictions * 0.06 - 0.03) max_frames = model_input.shape.as_list()[1] support_input = tf.tile( tf.expand_dims(support_activations, axis=1), multiples=[1, max_frames, 1]) actual_input = tf.concat([model_input, support_input], axis=2) normalized_input = tf.nn.l2_normalize(actual_input, dim=2) ## Batch normalize the input stacked_lstm = tf.contrib.rnn.MultiRNNCell( [ tf.contrib.rnn.BasicLSTMCell( lstm_size, forget_bias=1.0, state_is_tuple=True) for _ in range(number_of_layers) ], state_is_tuple=True) loss = 0.0 with tf.variable_scope("RNN"): outputs, state = tf.nn.dynamic_rnn(stacked_lstm, normalized_input, sequence_length=num_frames, swap_memory=FLAGS.rnn_swap_memory, dtype=tf.float32) final_state = tf.concat(map(lambda x: x.c, state), axis = 1) predictions = aggregated_model().create_model( model_input=final_state, original_input=model_input, vocab_size=vocab_size, sub_scope="main", **unused_params) predictions["support_predictions"] = support_predictions return predictions ================================================ FILE: youtube-8m-wangheda/all_frame_models/lstm_memory_model.py ================================================ import sys import models import model_utils import math import numpy as np import video_level_models import tensorflow as tf import utils import tensorflow.contrib.slim as slim from tensorflow import flags FLAGS = flags.FLAGS class LstmMemoryModel(models.BaseModel): def create_model(self, model_input, vocab_size, num_frames, dropout=False, keep_prob=None, noise_level=None, **unused_params): """Creates a model which uses a stack of LSTMs to represent the video. Args: model_input: A 'batch_size' x 'max_frames' x 'num_features' matrix of input features. vocab_size: The number of classes in the dataset. num_frames: A vector of length 'batch' which indicates the number of frames for each video (before padding). Returns: A dictionary with a tensor containing the probability predictions of the model in the 'predictions' key. The dimensions of the tensor are 'batch_size' x 'num_classes'. """ lstm_size = int(FLAGS.lstm_cells) number_of_layers = FLAGS.lstm_layers ## Batch normalize the input if dropout: stacked_lstm = tf.contrib.rnn.MultiRNNCell( [ tf.contrib.rnn.DropoutWrapper( tf.contrib.rnn.BasicLSTMCell( lstm_size, forget_bias=1.0, state_is_tuple=True), input_keep_prob=keep_prob) for _ in range(number_of_layers) ], state_is_tuple=True) else: stacked_lstm = tf.contrib.rnn.MultiRNNCell( [ tf.contrib.rnn.BasicLSTMCell( lstm_size, forget_bias=1.0, state_is_tuple=True) for _ in range(number_of_layers) ], state_is_tuple=True) loss = 0.0 with tf.variable_scope("RNN"): outputs, state = tf.nn.dynamic_rnn(stacked_lstm, model_input, sequence_length=num_frames, swap_memory=FLAGS.rnn_swap_memory, dtype=tf.float32) final_state = tf.concat(map(lambda x: x.c, state), axis = 1) if noise_level is not None: final_state = final_state + tf.random_normal(tf.shape(final_state), mean=0.0, stddev=noise_level) aggregated_model = getattr(video_level_models, FLAGS.video_level_classifier_model) return aggregated_model().create_model( model_input=final_state, original_input=model_input, vocab_size=vocab_size, num_frames=num_frames, **unused_params) ================================================ FILE: youtube-8m-wangheda/all_frame_models/lstm_memory_multitask_model.py ================================================ import sys import models import model_utils import math import numpy as np import video_level_models import tensorflow as tf import utils import tensorflow.contrib.slim as slim from tensorflow import flags FLAGS = flags.FLAGS class LstmMemoryMultitaskModel(models.BaseModel): def create_model(self, model_input, vocab_size, num_frames, **unused_params): """Creates a model which uses a stack of LSTMs to represent the video. Args: model_input: A 'batch_size' x 'max_frames' x 'num_features' matrix of input features. vocab_size: The number of classes in the dataset. num_frames: A vector of length 'batch' which indicates the number of frames for each video (before padding). Returns: A dictionary with a tensor containing the probability predictions of the model in the 'predictions' key. The dimensions of the tensor are 'batch_size' x 'num_classes'. """ lstm_size = int(FLAGS.lstm_cells) number_of_layers = FLAGS.lstm_layers num_supports = FLAGS.num_supports ## Batch normalize the input stacked_lstm = tf.contrib.rnn.MultiRNNCell( [ tf.contrib.rnn.BasicLSTMCell( lstm_size, forget_bias=1.0, state_is_tuple=True) for _ in range(number_of_layers) ], state_is_tuple=True) loss = 0.0 with tf.variable_scope("RNN"): outputs, state = tf.nn.dynamic_rnn(stacked_lstm, model_input, sequence_length=num_frames, swap_memory=FLAGS.rnn_swap_memory, dtype=tf.float32) final_state = tf.concat(map(lambda x: x.c, state), axis = 1) aggregated_model = getattr(video_level_models, FLAGS.video_level_classifier_model) predictions = aggregated_model().create_model( model_input=final_state, original_input=model_input, vocab_size=vocab_size, sub_scope="main", **unused_params) support_model = getattr(video_level_models, FLAGS.video_level_classifier_support_model) support_predictions = support_model().create_model( model_input=final_state, original_input=model_input, vocab_size=num_supports, sub_scope="support", **unused_params) predictions["support_predictions"] = support_predictions["predictions"] return predictions ================================================ FILE: youtube-8m-wangheda/all_frame_models/lstm_memory_normalization_model.py ================================================ import sys import models import model_utils import math import numpy as np import video_level_models import tensorflow as tf import utils import tensorflow.contrib.slim as slim from tensorflow import flags FLAGS = flags.FLAGS class LstmMemoryNormalizationModel(models.BaseModel): """layer normalization (not the Hinton one)""" def create_model(self, model_input, vocab_size, num_frames, **unused_params): """Creates a model which uses a stack of LSTMs to represent the video. Args: model_input: A 'batch_size' x 'max_frames' x 'num_features' matrix of input features. vocab_size: The number of classes in the dataset. num_frames: A vector of length 'batch' which indicates the number of frames for each video (before padding). Returns: A dictionary with a tensor containing the probability predictions of the model in the 'predictions' key. The dimensions of the tensor are 'batch_size' x 'num_classes'. """ lstm_size = int(FLAGS.lstm_cells) number_of_layers = FLAGS.lstm_layers normalize_class = getattr(self, FLAGS.lstm_normalization, self.identical) outputs, state = model_input, [] for layer in xrange(number_of_layers): with tf.variable_scope("RNN-layer%d" % layer): outputs = normalize_class(outputs) cell = tf.contrib.rnn.BasicLSTMCell( lstm_size, forget_bias=1.0, state_is_tuple=True) layer_outputs, layer_state = tf.nn.dynamic_rnn(cell, outputs, sequence_length=num_frames, swap_memory=FLAGS.rnn_swap_memory, dtype=tf.float32) state.append(layer_state) outputs = layer_outputs final_state = tf.concat(map(lambda x: x.c, state), axis = 1) aggregated_model = getattr(video_level_models, FLAGS.video_level_classifier_model) return aggregated_model().create_model( model_input=final_state, original_input=model_input, vocab_size=vocab_size, **unused_params) def layer_normalize(self, input_raw, epsilon=1e-8): feature_dim = len(input_raw.get_shape()) - 1 mean_input = tf.reduce_mean(input_raw, axis=feature_dim, keep_dims=True) std_input = tf.sqrt(tf.reduce_mean(tf.square(input_raw-mean_input), axis=feature_dim, keep_dims=True)) std_input = tf.maximum(std_input, epsilon) output = (input_raw - mean_input) / std_input return output def l2_normalize(self, input_raw, epsilon=1e-8): feature_dim = len(input_raw.get_shape()) - 1 output = tf.nn.l2_normalize(input_raw, dim=feature_dim) return output def identical(self, input_raw, epsilon=1e-8): return input_raw ================================================ FILE: youtube-8m-wangheda/all_frame_models/lstm_memory_parallel_chain_model.py ================================================ import sys import models import model_utils import math import numpy as np import video_level_models import tensorflow as tf import utils import tensorflow.contrib.slim as slim from tensorflow import flags FLAGS = flags.FLAGS class LstmMemoryParallelChainModel(models.BaseModel): """Classifier chain model of lstm memory""" def create_model(self, model_input, vocab_size, num_frames, **unused_params): """Creates a model which uses a stack of LSTMs to represent the video. Args: model_input: A 'batch_size' x 'max_frames' x 'num_features' matrix of input features. vocab_size: The number of classes in the dataset. num_frames: A vector of length 'batch' which indicates the number of frames for each video (before padding). Returns: A dictionary with a tensor containing the probability predictions of the model in the 'predictions' key. The dimensions of the tensor are 'batch_size' x 'num_classes'. """ lstm_size = int(FLAGS.lstm_cells) support_lstm_size = lstm_size / 2 number_of_layers = FLAGS.lstm_layers num_supports = FLAGS.num_supports aggregated_model = getattr(video_level_models, FLAGS.video_level_classifier_model) stacked_lstm_support = tf.contrib.rnn.MultiRNNCell( [ tf.contrib.rnn.BasicLSTMCell( support_lstm_size, forget_bias=1.0, state_is_tuple=True) for _ in range(number_of_layers) ], state_is_tuple=True) with tf.variable_scope("RNN-support"): support_outputs, support_state = tf.nn.dynamic_rnn(stacked_lstm_support, model_input, sequence_length=num_frames, swap_memory=FLAGS.rnn_swap_memory, dtype=tf.float32) support_final_state = tf.concat(map(lambda x: x.c, support_state), axis = 1) support_predictions = aggregated_model().create_model( model_input=support_final_state, original_input=model_input, vocab_size=num_supports, sub_scope="support", **unused_params) support_predictions = support_predictions["predictions"] stacked_lstm = tf.contrib.rnn.MultiRNNCell( [ tf.contrib.rnn.BasicLSTMCell( lstm_size, forget_bias=1.0, state_is_tuple=True) for _ in range(number_of_layers) ], state_is_tuple=True) with tf.variable_scope("RNN-main"): outputs, state = tf.nn.dynamic_rnn(stacked_lstm, model_input, sequence_length=num_frames, swap_memory=FLAGS.rnn_swap_memory, dtype=tf.float32) final_state = tf.concat(map(lambda x: x.c, state), axis = 1) main_state = tf.concat([final_state, support_predictions], axis=1) predictions = aggregated_model().create_model( model_input=main_state, original_input=model_input, vocab_size=vocab_size, sub_scope="main", **unused_params) predictions["support_predictions"] = support_predictions return predictions ================================================ FILE: youtube-8m-wangheda/all_frame_models/lstm_model.py ================================================ import sys import models import model_utils import math import numpy as np import video_level_models import tensorflow as tf import utils import tensorflow.contrib.slim as slim from tensorflow import flags FLAGS = flags.FLAGS class LstmModel(models.BaseModel): def create_model(self, model_input, vocab_size, num_frames, **unused_params): """Creates a model which uses a stack of LSTMs to represent the video. Args: model_input: A 'batch_size' x 'max_frames' x 'num_features' matrix of input features. vocab_size: The number of classes in the dataset. num_frames: A vector of length 'batch' which indicates the number of frames for each video (before padding). Returns: A dictionary with a tensor containing the probability predictions of the model in the 'predictions' key. The dimensions of the tensor are 'batch_size' x 'num_classes'. """ lstm_size = int(FLAGS.lstm_cells) number_of_layers = FLAGS.lstm_layers ## Batch normalize the input stacked_lstm = tf.contrib.rnn.MultiRNNCell( [ tf.contrib.rnn.BasicLSTMCell( lstm_size, forget_bias=1.0, state_is_tuple=False) for _ in range(number_of_layers) ], state_is_tuple=False) loss = 0.0 with tf.variable_scope("RNN"): outputs, state = tf.nn.dynamic_rnn(stacked_lstm, model_input, sequence_length=num_frames, swap_memory=FLAGS.rnn_swap_memory, dtype=tf.float32) aggregated_model = getattr(video_level_models, FLAGS.video_level_classifier_model) predictions = aggregated_model().create_model( model_input=state, original_input=model_input, vocab_size=vocab_size, **unused_params) print predictions return predictions ================================================ FILE: youtube-8m-wangheda/all_frame_models/lstm_multi_attention_model.py ================================================ import sys import models import model_utils import math import numpy as np import video_level_models import tensorflow as tf import utils import tensorflow.contrib.slim as slim from tensorflow import flags FLAGS = flags.FLAGS class LstmMultiAttentionModel(models.BaseModel): def create_model(self, model_input, vocab_size, num_frames, **unused_params): """Creates a model which uses a stack of LSTMs to represent the video. Args: model_input: A 'batch_size' x 'max_frames' x 'num_features' matrix of input features. vocab_size: The number of classes in the dataset. num_frames: A vector of length 'batch' which indicates the number of frames for each video (before padding). Returns: A dictionary with a tensor containing the probability predictions of the model in the 'predictions' key. The dimensions of the tensor are 'batch_size' x 'num_classes'. """ lstm_size = int(FLAGS.lstm_cells) number_of_layers = FLAGS.lstm_layers attention_size = FLAGS.attention_size l2_penalty = unused_params.get("l2_penalty", 1e-8) max_frames = model_input.get_shape().as_list()[1] mask_array = [] for i in xrange(max_frames + 1): tmp = [0.0] * max_frames for j in xrange(i): tmp[j] = 1.0 mask_array.append(tmp) mask_array = np.array(mask_array) mask_init = tf.constant_initializer(mask_array) mask_emb = tf.get_variable("mask_emb", shape = [max_frames + 1, max_frames], dtype = tf.float32, trainable = False, initializer = mask_init) ## Batch normalize the input stacked_lstm = tf.contrib.rnn.MultiRNNCell( [ tf.contrib.rnn.BasicLSTMCell( lstm_size, forget_bias=1.0, state_is_tuple=True) for _ in range(number_of_layers) ], state_is_tuple=True) loss = 0.0 with tf.variable_scope("RNN"): outputs, state = tf.nn.dynamic_rnn(stacked_lstm, model_input, sequence_length=num_frames, swap_memory=FLAGS.rnn_swap_memory, dtype=tf.float32) num_frames_matrix = tf.maximum(tf.cast( tf.expand_dims(tf.expand_dims(num_frames, axis=1), axis=2), dtype=tf.float32), 1.0) mask = tf.expand_dims(tf.nn.embedding_lookup(mask_emb, num_frames), axis = 2) attention_fc = slim.fully_connected( outputs, attention_size, activation_fn=tf.nn.sigmoid, weights_regularizer=slim.l2_regularizer(l2_penalty)) print attention_fc attention = attention_fc * mask attention_sum = tf.reduce_sum(attention, axis = 1, keep_dims = True) + 1e-8 attention = attention / attention_sum print attention, model_input attention_input = tf.einsum("ijk,ijl->ikl", attention, model_input) print attention_input aggregated_model = getattr(video_level_models, FLAGS.video_level_classifier_model) attention_output = aggregated_model().create_model( model_input=attention_input, original_input=model_input, vocab_size=vocab_size, **unused_params)["predictions"] print attention_output final_output = tf.reduce_max(tf.reshape(attention_output, [-1, attention_size, vocab_size]), axis = 1) return {"predictions": final_output} ================================================ FILE: youtube-8m-wangheda/all_frame_models/lstm_multi_pooling_model.py ================================================ import sys import models import model_utils import math import numpy as np import video_level_models import tensorflow as tf import utils import tensorflow.contrib.slim as slim from tensorflow import flags FLAGS = flags.FLAGS class LstmMultiPoolingModel(models.BaseModel): def create_model(self, model_input, vocab_size, num_frames, **unused_params): """Creates a model which uses a stack of LSTMs to represent the video. Args: model_input: A 'batch_size' x 'max_frames' x 'num_features' matrix of input features. vocab_size: The number of classes in the dataset. num_frames: A vector of length 'batch' which indicates the number of frames for each video (before padding). Returns: A dictionary with a tensor containing the probability predictions of the model in the 'predictions' key. The dimensions of the tensor are 'batch_size' x 'num_classes'. """ lstm_size = int(FLAGS.lstm_cells) number_of_layers = FLAGS.lstm_layers batch_size = FLAGS.batch_size pooled_output = [] outputs = model_input num_frames_matrix = tf.maximum(tf.cast(tf.expand_dims(num_frames, axis=1), dtype=tf.float32), tf.ones([batch_size, 1])) pooled_output.append(tf.reduce_sum(outputs, axis = 1) / num_frames_matrix) for layer in xrange(number_of_layers): ## Batch normalize the input cell = tf.contrib.rnn.BasicLSTMCell( lstm_size, forget_bias=1.0, state_is_tuple=False) loss = 0.0 with tf.variable_scope("RNN-layer%d" % layer): outputs, state = tf.nn.dynamic_rnn(cell, outputs, sequence_length=num_frames, time_major=False, swap_memory=FLAGS.rnn_swap_memory, dtype=tf.float32) pooled_output.append(tf.reduce_sum(outputs, axis = 1) / num_frames_matrix) final_output = tf.concat(pooled_output, axis = 1) aggregated_model = getattr(video_level_models, FLAGS.video_level_classifier_model) return aggregated_model().create_model( model_input=final_output, original_input=model_input, vocab_size=vocab_size, **unused_params) ================================================ FILE: youtube-8m-wangheda/all_frame_models/lstm_parallel_finaloutput_model.py ================================================ import sys import models import model_utils import math import numpy as np import video_level_models import tensorflow as tf import utils import tensorflow.contrib.slim as slim from tensorflow import flags FLAGS = flags.FLAGS class LstmParallelFinaloutputModel(models.BaseModel): def create_model(self, model_input, vocab_size, num_frames, **unused_params): """Creates a model which uses a stack of LSTMs to represent the video. Args: model_input: A 'batch_size' x 'max_frames' x 'num_features' matrix of input features. vocab_size: The number of classes in the dataset. num_frames: A vector of length 'batch' which indicates the number of frames for each video (before padding). Returns: A dictionary with a tensor containing the probability predictions of the model in the 'predictions' key. The dimensions of the tensor are 'batch_size' x 'num_classes'. """ number_of_layers = FLAGS.lstm_layers lstm_sizes = map(int, FLAGS.lstm_cells.split(",")) feature_names, feature_sizes = utils.GetListOfFeatureNamesAndSizes( FLAGS.feature_names, FLAGS.feature_sizes) sub_inputs = [tf.nn.l2_normalize(x, dim=2) for x in tf.split(model_input, feature_sizes, axis = 2)] assert len(lstm_sizes) == len(feature_sizes), \ "length of lstm_sizes (={}) != length of feature_sizes (={})".format( \ len(lstm_sizes), len(feature_sizes)) outputs = [] states = [] for i in xrange(len(feature_sizes)): with tf.variable_scope("RNN%d" % i): sub_input = sub_inputs[i] lstm_size = lstm_sizes[i] ## Batch normalize the input stacked_lstm = tf.contrib.rnn.MultiRNNCell( [ tf.contrib.rnn.BasicLSTMCell( lstm_size, forget_bias=1.0, state_is_tuple=True) for _ in range(number_of_layers) ], state_is_tuple=True) output, state = tf.nn.dynamic_rnn(stacked_lstm, sub_input, sequence_length=num_frames, swap_memory=FLAGS.rnn_swap_memory, dtype=tf.float32) outputs.append(output) states.extend(map(lambda x: x.h, state)) # concat final_state = tf.concat(states, axis = 1) aggregated_model = getattr(video_level_models, FLAGS.video_level_classifier_model) return aggregated_model().create_model( model_input=final_state, original_input=model_input, vocab_size=vocab_size, **unused_params) ================================================ FILE: youtube-8m-wangheda/all_frame_models/lstm_parallel_memory_model.py ================================================ import sys import models import model_utils import math import numpy as np import video_level_models import tensorflow as tf import utils import tensorflow.contrib.slim as slim from tensorflow import flags FLAGS = flags.FLAGS class LstmParallelMemoryModel(models.BaseModel): def create_model(self, model_input, vocab_size, num_frames, **unused_params): """Creates a model which uses a stack of LSTMs to represent the video. Args: model_input: A 'batch_size' x 'max_frames' x 'num_features' matrix of input features. vocab_size: The number of classes in the dataset. num_frames: A vector of length 'batch' which indicates the number of frames for each video (before padding). Returns: A dictionary with a tensor containing the probability predictions of the model in the 'predictions' key. The dimensions of the tensor are 'batch_size' x 'num_classes'. """ number_of_layers = FLAGS.lstm_layers lstm_sizes = map(int, FLAGS.lstm_cells.split(",")) feature_names, feature_sizes = utils.GetListOfFeatureNamesAndSizes( FLAGS.feature_names, FLAGS.feature_sizes) sub_inputs = [tf.nn.l2_normalize(x, dim=2) for x in tf.split(model_input, feature_sizes, axis = 2)] assert len(lstm_sizes) == len(feature_sizes), \ "length of lstm_sizes (={}) != length of feature_sizes (={})".format( \ len(lstm_sizes), len(feature_sizes)) loss = 0.0 outputs = [] states = [] for i in xrange(len(feature_sizes)): with tf.variable_scope("RNN%d" % i): sub_input = sub_inputs[i] lstm_size = lstm_sizes[i] ## Batch normalize the input stacked_lstm = tf.contrib.rnn.MultiRNNCell( [ tf.contrib.rnn.BasicLSTMCell( lstm_size, forget_bias=1.0, state_is_tuple=True) for _ in range(number_of_layers) ], state_is_tuple=True) output, state = tf.nn.dynamic_rnn(stacked_lstm, sub_input, sequence_length=num_frames, swap_memory=FLAGS.rnn_swap_memory, dtype=tf.float32) outputs.append(output) states.extend(map(lambda x: x.c, state)) # concat final_state = tf.concat(states, axis = 1) aggregated_model = getattr(video_level_models, FLAGS.video_level_classifier_model) return aggregated_model().create_model( model_input=final_state, original_input=model_input, vocab_size=vocab_size, **unused_params) ================================================ FILE: youtube-8m-wangheda/all_frame_models/lstm_parallel_model.py ================================================ import sys import models import model_utils import math import numpy as np import video_level_models import tensorflow as tf import utils import tensorflow.contrib.slim as slim from tensorflow import flags FLAGS = flags.FLAGS class LstmParallelModel(models.BaseModel): def create_model(self, model_input, vocab_size, num_frames, **unused_params): """Creates a model which uses a stack of LSTMs to represent the video. Args: model_input: A 'batch_size' x 'max_frames' x 'num_features' matrix of input features. vocab_size: The number of classes in the dataset. num_frames: A vector of length 'batch' which indicates the number of frames for each video (before padding). Returns: A dictionary with a tensor containing the probability predictions of the model in the 'predictions' key. The dimensions of the tensor are 'batch_size' x 'num_classes'. """ number_of_layers = FLAGS.lstm_layers lstm_sizes = map(int, FLAGS.lstm_cells.split(",")) feature_names, feature_sizes = utils.GetListOfFeatureNamesAndSizes( FLAGS.feature_names, FLAGS.feature_sizes) sub_inputs = tf.split(model_input, feature_sizes, axis = 2) assert len(lstm_sizes) == len(feature_sizes), \ "length of lstm_sizes (={}) != length of feature_sizes (={})".format( \ len(lstm_sizes), len(feature_sizes)) loss = 0.0 outputs = [] states = [] for i in xrange(len(feature_sizes)): with tf.variable_scope("RNN%d" % i): sub_input = sub_inputs[i] lstm_size = lstm_sizes[i] ## Batch normalize the input stacked_lstm = tf.contrib.rnn.MultiRNNCell( [ tf.contrib.rnn.BasicLSTMCell( lstm_size, forget_bias=1.0, state_is_tuple=False) for _ in range(number_of_layers) ], state_is_tuple=False) output, state = tf.nn.dynamic_rnn(stacked_lstm, sub_input, sequence_length=num_frames, swap_memory=FLAGS.rnn_swap_memory, dtype=tf.float32) outputs.append(output) states.append(state) # concat final_state = tf.concat(states, axis = 1) aggregated_model = getattr(video_level_models, FLAGS.video_level_classifier_model) return aggregated_model().create_model( model_input=final_state, original_input=model_input, vocab_size=vocab_size, **unused_params) ================================================ FILE: youtube-8m-wangheda/all_frame_models/lstm_pooling_model.py ================================================ import sys import models import model_utils import math import numpy as np import video_level_models import tensorflow as tf import utils import tensorflow.contrib.slim as slim from tensorflow import flags FLAGS = flags.FLAGS class LstmPoolingModel(models.BaseModel): def create_model(self, model_input, vocab_size, num_frames, **unused_params): """Creates a model which uses a stack of LSTMs to represent the video. Args: model_input: A 'batch_size' x 'max_frames' x 'num_features' matrix of input features. vocab_size: The number of classes in the dataset. num_frames: A vector of length 'batch' which indicates the number of frames for each video (before padding). Returns: A dictionary with a tensor containing the probability predictions of the model in the 'predictions' key. The dimensions of the tensor are 'batch_size' x 'num_classes'. """ lstm_size = int(FLAGS.lstm_cells) number_of_layers = FLAGS.lstm_layers batch_size = FLAGS.batch_size ## Batch normalize the input stacked_lstm = tf.contrib.rnn.MultiRNNCell( [ tf.contrib.rnn.BasicLSTMCell( lstm_size, forget_bias=1.0, state_is_tuple=False) for _ in range(number_of_layers) ], state_is_tuple=False) loss = 0.0 with tf.variable_scope("RNN"): outputs, state = tf.nn.dynamic_rnn(stacked_lstm, model_input, sequence_length=num_frames, time_major=False, swap_memory=FLAGS.rnn_swap_memory, dtype=tf.float32) num_frames_matrix = tf.maximum(tf.cast(tf.expand_dims(num_frames, axis=1), dtype=tf.float32), tf.ones([batch_size, 1])) pooling_output = tf.reduce_sum(outputs, axis = 1) / num_frames_matrix aggregated_model = getattr(video_level_models, FLAGS.video_level_classifier_model) return aggregated_model().create_model( model_input=pooling_output, original_input=model_input, vocab_size=vocab_size, **unused_params) ================================================ FILE: youtube-8m-wangheda/all_frame_models/lstm_positional_attention_max_pooling_model.py ================================================ import math import models import tensorflow as tf import numpy as np import utils from tensorflow import flags import tensorflow.contrib.slim as slim FLAGS = flags.FLAGS class LstmPositionalAttentionMaxPoolingModel(models.BaseModel): """Max pooling over temporal weighted sums (attention) of lstm outputs.""" def create_model(self, model_input, vocab_size, num_frames, num_mixtures=None, l2_penalty=1e-8, sub_scope="", original_input=None, **unused_params): """Creates a model which uses a stack of LSTMs to represent the video. Args: model_input: A 'batch_size' x 'max_frames' x 'num_features' matrix of input features. vocab_size: The number of classes in the dataset. num_frames: A vector of length 'batch' which indicates the number of frames for each video (before padding). Returns: A dictionary with a tensor containing the probability predictions of the model in the 'predictions' key. The dimensions of the tensor are 'batch_size' x 'num_classes'. """ lstm_size = int(FLAGS.lstm_cells) number_of_layers = FLAGS.lstm_layers num_attentions = FLAGS.lstm_attentions batch_size, max_frames, num_features = model_input.get_shape().as_list() mask = tf.sequence_mask(lengths=num_frames, maxlen=max_frames, dtype=tf.float32) ## Batch normalize the input stacked_lstm = tf.contrib.rnn.MultiRNNCell( [ tf.contrib.rnn.BasicLSTMCell( lstm_size, forget_bias=1.0, state_is_tuple=True) for _ in range(number_of_layers) ], state_is_tuple=True) with tf.variable_scope("RNN"): outputs, state = tf.nn.dynamic_rnn(stacked_lstm, model_input, sequence_length=num_frames, swap_memory=FLAGS.rnn_swap_memory, dtype=tf.float32) positional_embedding = self.get_positional_embedding(model_input, num_frames) mean_input = self.get_mean_input(model_input, num_frames) attention_activations = slim.fully_connected( tf.concat([model_input, positional_embedding, mean_input, outputs], axis=2), num_attentions, activation_fn=None, weights_regularizer=slim.l2_regularizer(l2_penalty), scope="attention-"+sub_scope) # Batch x #Attentions x #Frames attention_weights = tf.einsum("ijk,ij->ikj", tf.nn.softmax(attention_activations, dim=1), mask) attention_weights = attention_weights / tf.reduce_sum(attention_weights, axis=2, keep_dims=True) # Batch x #Attentions x #Features attention_outputs = tf.einsum("ijk,ilj->ilk", outputs, attention_weights) moe_predictions = self.sub_moe(attention_outputs, vocab_size, sub_scope="sub-moe") predictions = tf.reshape(moe_predictions, [-1, num_attentions, vocab_size]) max_predictions = tf.reduce_max(predictions, axis=1) return {"predictions": max_predictions} def get_positional_embedding(self, model_input, num_frames, l2_penalty=1e-8): embedding_size = FLAGS.positional_embedding_size batch_size, max_frames, num_features = model_input.get_shape().as_list() positional_embedding = tf.get_variable("positional_embedding", dtype=tf.float32, shape=[1, max_frames, embedding_size], regularizer=tf.contrib.layers.l2_regularizer(l2_penalty)) tiled_embedding = tf.tile(positional_embedding, multiples=[tf.shape(model_input)[0],1,1]) return tiled_embedding def get_mean_input(self, model_input, num_frames): batch_size, max_frames, num_features = model_input.get_shape().as_list() mask = tf.sequence_mask(lengths=num_frames, maxlen=max_frames, dtype=tf.float32) mean_input = tf.einsum("ijk,ij->ik", model_input, mask) / tf.expand_dims(tf.cast(num_frames, dtype=tf.float32), dim=1) tiled_mean_input = tf.tile(tf.expand_dims(mean_input, dim=1), multiples=[1,max_frames,1]) return tiled_mean_input def sub_moe(self, model_input, vocab_size, num_mixtures=None, l2_penalty=1e-8, sub_scope="", **unused_params): num_mixtures = num_mixtures or FLAGS.moe_num_mixtures gate_activations = slim.fully_connected( model_input, vocab_size * (num_mixtures + 1), activation_fn=None, biases_initializer=None, weights_regularizer=slim.l2_regularizer(l2_penalty), scope="gates-"+sub_scope) expert_activations = slim.fully_connected( model_input, vocab_size * num_mixtures, activation_fn=None, weights_regularizer=slim.l2_regularizer(l2_penalty), scope="experts-"+sub_scope) gating_distribution = tf.nn.softmax(tf.reshape( gate_activations, [-1, num_mixtures + 1])) # (Batch * #Labels) x (num_mixtures + 1) expert_distribution = tf.nn.sigmoid(tf.reshape( expert_activations, [-1, num_mixtures])) # (Batch * #Labels) x num_mixtures final_probabilities_by_class_and_batch = tf.reduce_sum( gating_distribution[:, :num_mixtures] * expert_distribution, 1) final_probabilities = tf.reshape(final_probabilities_by_class_and_batch, [-1, vocab_size]) return final_probabilities ================================================ FILE: youtube-8m-wangheda/all_frame_models/lstm_with_mean_input_model.py ================================================ import sys import models import model_utils import math import numpy as np import video_level_models import tensorflow as tf import utils import tensorflow.contrib.slim as slim from tensorflow import flags FLAGS = flags.FLAGS class LstmWithMeanInputModel(models.BaseModel): def create_model(self, model_input, vocab_size, num_frames, **unused_params): """Creates a model which uses a stack of LSTMs to represent the video. Args: model_input: A 'batch_size' x 'max_frames' x 'num_features' matrix of input features. vocab_size: The number of classes in the dataset. num_frames: A vector of length 'batch' which indicates the number of frames for each video (before padding). Returns: A dictionary with a tensor containing the probability predictions of the model in the 'predictions' key. The dimensions of the tensor are 'batch_size' x 'num_classes'. """ lstm_size = int(FLAGS.lstm_cells) number_of_layers = FLAGS.lstm_layers batch_size = FLAGS.batch_size ## Batch normalize the input stacked_lstm = tf.contrib.rnn.MultiRNNCell( [ tf.contrib.rnn.BasicLSTMCell( lstm_size, forget_bias=1.0, state_is_tuple=False) for _ in range(number_of_layers) ], state_is_tuple=False) loss = 0.0 with tf.variable_scope("RNN"): outputs, state = tf.nn.dynamic_rnn(stacked_lstm, model_input, sequence_length=num_frames, time_major=False, swap_memory=FLAGS.rnn_swap_memory, dtype=tf.float32) with tf.device("/gpu:0"): num_frames_matrix = tf.maximum(tf.cast(tf.expand_dims(num_frames, axis=1), dtype=tf.float32), 1.0) mean_input = tf.reduce_sum(model_input, axis = 1) / num_frames_matrix final_output = tf.concat([mean_input, state], axis = 1) aggregated_model = getattr(video_level_models, FLAGS.video_level_classifier_model) predictions = aggregated_model().create_model( model_input=final_output, original_input=model_input, vocab_size=vocab_size, **unused_params) return predictions ================================================ FILE: youtube-8m-wangheda/all_frame_models/lstm_with_pooling_model.py ================================================ import sys import models import model_utils import math import numpy as np import video_level_models import tensorflow as tf import utils import tensorflow.contrib.slim as slim from tensorflow import flags FLAGS = flags.FLAGS class LstmWithPoolingModel(models.BaseModel): def create_model(self, model_input, vocab_size, num_frames, **unused_params): """Creates a model which uses a stack of LSTMs to represent the video. Args: model_input: A 'batch_size' x 'max_frames' x 'num_features' matrix of input features. vocab_size: The number of classes in the dataset. num_frames: A vector of length 'batch' which indicates the number of frames for each video (before padding). Returns: A dictionary with a tensor containing the probability predictions of the model in the 'predictions' key. The dimensions of the tensor are 'batch_size' x 'num_classes'. """ lstm_size = int(FLAGS.lstm_cells) number_of_layers = FLAGS.lstm_layers ## Batch normalize the input stacked_lstm = tf.contrib.rnn.MultiRNNCell( [ tf.contrib.rnn.BasicLSTMCell( lstm_size, forget_bias=1.0, state_is_tuple=False) for _ in range(number_of_layers) ], state_is_tuple=False) loss = 0.0 with tf.variable_scope("RNN"): outputs, state = tf.nn.dynamic_rnn(stacked_lstm, model_input, sequence_length=num_frames, time_major=False, swap_memory=FLAGS.rnn_swap_memory, dtype=tf.float32) num_frames_matrix = tf.maximum(tf.cast(tf.expand_dims(num_frames, axis=1), dtype=tf.float32), 1.0) pooling_output = tf.reduce_sum(outputs, axis = 1) / num_frames_matrix final_output = tf.concat([pooling_output, state], axis = 1) aggregated_model = getattr(video_level_models, FLAGS.video_level_classifier_model) return aggregated_model().create_model( model_input=final_output, original_input=model_input, vocab_size=vocab_size, **unused_params) ================================================ FILE: youtube-8m-wangheda/all_frame_models/mm_lstm_memory_model.py ================================================ import sys import models import model_utils import math import numpy as np import video_level_models import tensorflow as tf import utils import tensorflow.contrib.slim as slim from tensorflow import flags FLAGS = flags.FLAGS class MatchingMatrixLstmMemoryModel(models.BaseModel): def matching_matrix(self, model_input, vocab_size, l2_penalty=1e-8, **unused_params): max_frames = model_input.get_shape().as_list()[1] num_features = model_input.get_shape().as_list()[2] embedding_size = FLAGS.mm_label_embedding model_input = tf.reshape(model_input, [-1, num_features]) frame_relu = slim.fully_connected( model_input, embedding_size, activation_fn=tf.nn.relu, biases_initializer=None, weights_regularizer=slim.l2_regularizer(l2_penalty), scope="mm_relu") frame_activation = slim.fully_connected( frame_relu, embedding_size, activation_fn=tf.nn.tanh, biases_initializer=None, weights_regularizer=slim.l2_regularizer(l2_penalty), scope="mm_activation") label_embedding = tf.get_variable("label_embedding", shape=[vocab_size,embedding_size], dtype=tf.float32, initializer=tf.truncated_normal_initializer(mean=0.0, stddev=0.5), regularizer=slim.l2_regularizer(l2_penalty), trainable=True) mm_matrix = tf.einsum("ik,jk->ij", frame_activation, label_embedding) mm_output = tf.reshape(mm_matrix, [-1,max_frames,vocab_size]) return mm_output def create_model(self, model_input, vocab_size, num_frames, **unused_params): """Creates a model which uses a stack of LSTMs to represent the video. Args: model_input: A 'batch_size' x 'max_frames' x 'num_features' matrix of input features. vocab_size: The number of classes in the dataset. num_frames: A vector of length 'batch' which indicates the number of frames for each video (before padding). Returns: A dictionary with a tensor containing the probability predictions of the model in the 'predictions' key. The dimensions of the tensor are 'batch_size' x 'num_classes'. """ lstm_size = int(FLAGS.lstm_cells) number_of_layers = FLAGS.lstm_layers mm_output = self.matching_matrix(model_input, vocab_size) normalized_mm_output = tf.nn.l2_normalize(mm_output, dim=2) ## Batch normalize the input stacked_lstm = tf.contrib.rnn.MultiRNNCell( [ tf.contrib.rnn.BasicLSTMCell( lstm_size, forget_bias=1.0, state_is_tuple=True) for _ in range(number_of_layers) ], state_is_tuple=True) loss = 0.0 with tf.variable_scope("RNN"): outputs, state = tf.nn.dynamic_rnn(stacked_lstm, normalized_mm_output, sequence_length=num_frames, swap_memory=FLAGS.rnn_swap_memory, dtype=tf.float32) final_state = tf.concat(map(lambda x: x.c, state), axis = 1) aggregated_model = getattr(video_level_models, FLAGS.video_level_classifier_model) return aggregated_model().create_model( model_input=final_state, original_input=model_input, vocab_size=vocab_size, **unused_params) ================================================ FILE: youtube-8m-wangheda/all_frame_models/multi_view_cnn_deep_combine_chain_model.py ================================================ import math import models import tensorflow as tf import numpy as np import utils from tensorflow import flags import tensorflow.contrib.slim as slim FLAGS = flags.FLAGS class MultiViewCnnDeepCombineChainModel(models.BaseModel): def cnn(self, model_input, l2_penalty=1e-8, num_filters = [1024, 1024, 1024], filter_sizes = [1,2,3], sub_scope="", **unused_params): max_frames = model_input.get_shape().as_list()[1] num_features = model_input.get_shape().as_list()[2] shift_inputs = [] for i in xrange(max(filter_sizes)): if i == 0: shift_inputs.append(model_input) else: shift_inputs.append(tf.pad(model_input, paddings=[[0,0],[i,0],[0,0]])[:,:max_frames,:]) cnn_outputs = [] for nf, fs in zip(num_filters, filter_sizes): sub_input = tf.concat(shift_inputs[:fs], axis=2) sub_filter = tf.get_variable(sub_scope+"cnn-filter-len%d"%fs, shape=[num_features*fs, nf], dtype=tf.float32, initializer=tf.truncated_normal_initializer(mean=0.0, stddev=0.1), regularizer=tf.contrib.layers.l2_regularizer(l2_penalty)) cnn_outputs.append(tf.einsum("ijk,kl->ijl", sub_input, sub_filter)) cnn_output = tf.concat(cnn_outputs, axis=2) return cnn_output def create_model(self, model_input, vocab_size, num_frames, num_mixtures=None, l2_penalty=1e-8, sub_scope="", original_input=None, **unused_params): num_supports = FLAGS.num_supports num_layers = FLAGS.deep_chain_layers relu_cells = FLAGS.deep_chain_relu_cells max_frames = model_input.get_shape().as_list()[1] relu_layers = [] support_predictions = [] mask = self.get_mask(max_frames, num_frames) mean_input = tf.einsum("ijk,ij->ik", model_input, mask) \ / tf.expand_dims(tf.cast(num_frames, dtype=tf.float32), dim=1) mean_relu = slim.fully_connected( mean_input, relu_cells, activation_fn=tf.nn.relu, weights_regularizer=slim.l2_regularizer(l2_penalty), scope=sub_scope+"mean-relu") mean_relu_norm = tf.nn.l2_normalize(mean_relu, dim=1) relu_layers.append(mean_relu_norm) cnn_output = self.cnn(model_input, num_filters=[relu_cells,2*relu_cells,relu_cells], filter_sizes=[1,2,3], sub_scope=sub_scope+"cnn0") mv_cnn_output = self.multiview(cnn_output, axis=1) normalized_cnn_output = tf.nn.l2_normalize(mv_cnn_output, dim=1) next_input = normalized_cnn_output for layer in xrange(num_layers): sub_prediction = self.sub_model(next_input, vocab_size, sub_scope=sub_scope+"prediction-%d"%layer) support_predictions.append(sub_prediction) sub_relu = slim.fully_connected( sub_prediction, relu_cells, activation_fn=tf.nn.relu, weights_regularizer=slim.l2_regularizer(l2_penalty), scope=sub_scope+"relu-%d"%layer) relu_norm = tf.nn.l2_normalize(sub_relu, dim=1) relu_layers.append(relu_norm) cnn_output = self.cnn(model_input, num_filters=[relu_cells,relu_cells,relu_cells*2], filter_sizes=[1,2,3], sub_scope=sub_scope+"cnn%d"%(layer+1)) mv_cnn_output = self.multiview(cnn_output, axis=1) normalized_cnn_output = tf.nn.l2_normalize(mv_cnn_output, dim=1) next_input = tf.concat([mean_input, normalized_cnn_output] + relu_layers, axis=1) main_predictions = self.sub_model(next_input, vocab_size, sub_scope=sub_scope+"-main") support_predictions = tf.concat(support_predictions, axis=1) return {"predictions": main_predictions, "support_predictions": support_predictions} def multiview(self, cnn_output, axis=1): max_view = tf.reduce_max(cnn_output, axis=1) mean_view = tf.reduce_mean(cnn_output, axis=1) min_view = tf.reduce_min(cnn_output, axis=1) multi_view = tf.concat([max_view, mean_view, min_view], axis=1) return multi_view def sub_model(self, model_input, vocab_size, num_mixtures=None, l2_penalty=1e-8, sub_scope="", **unused_params): num_mixtures = num_mixtures or FLAGS.moe_num_mixtures gate_activations = slim.fully_connected( model_input, vocab_size * (num_mixtures + 1), activation_fn=None, biases_initializer=None, weights_regularizer=slim.l2_regularizer(l2_penalty), scope="gates-"+sub_scope) expert_activations = slim.fully_connected( model_input, vocab_size * num_mixtures, activation_fn=None, weights_regularizer=slim.l2_regularizer(l2_penalty), scope="experts-"+sub_scope) gating_distribution = tf.nn.softmax(tf.reshape( gate_activations, [-1, num_mixtures + 1])) # (Batch * #Labels) x (num_mixtures + 1) expert_distribution = tf.nn.sigmoid(tf.reshape( expert_activations, [-1, num_mixtures])) # (Batch * #Labels) x num_mixtures final_probabilities_by_class_and_batch = tf.reduce_sum( gating_distribution[:, :num_mixtures] * expert_distribution, 1) final_probabilities = tf.reshape(final_probabilities_by_class_and_batch, [-1, vocab_size]) return final_probabilities def get_mask(self, max_frames, num_frames): mask_array = [] for i in xrange(max_frames + 1): tmp = [0.0] * max_frames for j in xrange(i): tmp[j] = 1.0 mask_array.append(tmp) mask_array = np.array(mask_array) mask_init = tf.constant_initializer(mask_array) mask_emb = tf.get_variable("mask_emb", shape = [max_frames + 1, max_frames], dtype = tf.float32, trainable = False, initializer = mask_init) mask = tf.nn.embedding_lookup(mask_emb, num_frames) return mask ================================================ FILE: youtube-8m-wangheda/all_frame_models/multires_lstm_memory_deep_combine_chain_model.py ================================================ import sys import models import model_utils import math import numpy as np import video_level_models import tensorflow as tf import utils import tensorflow.contrib.slim as slim from tensorflow import flags FLAGS = flags.FLAGS class MultiresLstmMemoryDeepCombineChainModel(models.BaseModel): """Classifier chain model of lstm memory""" def lstm(self, model_input, vocab_size, num_frames, sub_scope="", **unused_params): number_of_layers = FLAGS.lstm_layers lstm_sizes = map(int, FLAGS.lstm_cells.split(",")) feature_names, feature_sizes = utils.GetListOfFeatureNamesAndSizes( FLAGS.feature_names, FLAGS.feature_sizes) sub_inputs = [tf.nn.l2_normalize(x, dim=2) for x in tf.split(model_input, feature_sizes, axis = 2)] assert len(lstm_sizes) == len(feature_sizes), \ "length of lstm_sizes (={}) != length of feature_sizes (={})".format( \ len(lstm_sizes), len(feature_sizes)) states = [] for i in xrange(len(feature_sizes)): with tf.variable_scope(sub_scope+"RNN%d" % i): sub_input = sub_inputs[i] lstm_size = lstm_sizes[i] ## Batch normalize the input stacked_lstm = tf.contrib.rnn.MultiRNNCell( [ tf.contrib.rnn.BasicLSTMCell( lstm_size, forget_bias=1.0, state_is_tuple=True) for _ in range(number_of_layers) ], state_is_tuple=True) output, state = tf.nn.dynamic_rnn(stacked_lstm, sub_input, sequence_length=num_frames, swap_memory=FLAGS.rnn_swap_memory, dtype=tf.float32) states.extend(map(lambda x: x.c, state)) final_state = tf.concat(states, axis = 1) return final_state def create_model(self, model_input, vocab_size, num_mixtures=None, l2_penalty=1e-8, sub_scope="", original_input=None, dropout=False, keep_prob=None, noise_level=None, num_frames=None, **unused_params): num_supports = FLAGS.num_supports num_layers = FLAGS.deep_chain_layers relu_cells = FLAGS.deep_chain_relu_cells relu_type = FLAGS.deep_chain_relu_type use_length = FLAGS.deep_chain_use_length num_resolutions = [int(2**(num_layers-i)) for i in xrange(num_layers+1)] additional_features = [] if use_length: print "using length as feature" additional_features.append(self.get_length_code(num_frames)) lstm_input, lstm_frames = self.resolution(model_input, num_frames, num_resolutions[0]) lstm_output = self.lstm(lstm_input, vocab_size, num_frames=lstm_frames, sub_scope="lstm%d"%0) next_input = tf.concat([lstm_output] + additional_features, axis=1) support_predictions = [] for layer in xrange(num_layers): sub_prediction = self.sub_model(next_input, vocab_size, sub_scope=sub_scope+"prediction-%d"%layer, dropout=dropout, keep_prob=keep_prob, noise_level=noise_level) support_predictions.append(sub_prediction) sub_activation = slim.fully_connected( sub_prediction, relu_cells, activation_fn=None, weights_regularizer=slim.l2_regularizer(l2_penalty), scope=sub_scope+"relu-%d"%layer) if relu_type == "elu": sub_relu = tf.nn.elu(sub_activation) else: # default: relu sub_relu = tf.nn.relu(sub_activation) if noise_level is not None: print "adding noise to sub_relu, level = ", noise_level sub_relu = sub_relu + tf.random_normal(tf.shape(sub_relu), mean=0.0, stddev=noise_level) relu_norm = tf.nn.l2_normalize(sub_relu, dim=1) additional_features.append(relu_norm) lstm_input, lstm_frames = self.resolution(model_input, num_frames, num_resolutions[layer+1]) lstm_output = self.lstm(lstm_input, vocab_size, num_frames=lstm_frames, sub_scope="lstm%d"%(layer+1)) next_input = tf.concat([lstm_output] + additional_features, axis=1) main_predictions = self.sub_model(next_input, vocab_size, sub_scope=sub_scope+"-main") support_predictions = tf.concat(support_predictions, axis=1) return {"predictions": main_predictions, "support_predictions": support_predictions} def sub_model(self, model_input, vocab_size, num_mixtures=None, l2_penalty=1e-8, sub_scope="", dropout=False, keep_prob=None, noise_level=None, **unused_params): num_mixtures = num_mixtures or FLAGS.moe_num_mixtures if dropout: print "adding dropout to moe, keep_prob = ", keep_prob model_input = tf.nn.dropout(model_input, keep_prob=keep_prob) gate_activations = slim.fully_connected( model_input, vocab_size * (num_mixtures + 1), activation_fn=None, biases_initializer=None, weights_regularizer=slim.l2_regularizer(l2_penalty), scope="gates-"+sub_scope) expert_activations = slim.fully_connected( model_input, vocab_size * num_mixtures, activation_fn=None, weights_regularizer=slim.l2_regularizer(l2_penalty), scope="experts-"+sub_scope) gating_distribution = tf.nn.softmax(tf.reshape( gate_activations, [-1, num_mixtures + 1])) # (Batch * #Labels) x (num_mixtures + 1) expert_distribution = tf.nn.sigmoid(tf.reshape( expert_activations, [-1, num_mixtures])) # (Batch * #Labels) x num_mixtures final_probabilities_by_class_and_batch = tf.reduce_sum( gating_distribution[:, :num_mixtures] * expert_distribution, 1) final_probabilities = tf.reshape(final_probabilities_by_class_and_batch, [-1, vocab_size]) return final_probabilities def get_length_code(self, num_frames): code_0 = tf.cast(num_frames <= 60, dtype=tf.int32) code_1 = tf.cast(num_frames > 60, dtype=tf.int32) * tf.cast(num_frames <= 120, dtype=tf.int32) code_2 = tf.cast(num_frames > 120, dtype=tf.int32) * tf.cast(num_frames <= 180, dtype=tf.int32) code_3 = tf.cast(num_frames > 180, dtype=tf.int32) * tf.cast(num_frames <= 240, dtype=tf.int32) code_4 = tf.cast(num_frames > 240, dtype=tf.int32) codes = map(lambda x: tf.expand_dims(x, dim=1), [code_0, code_1, code_2, code_3, code_4]) length_code = tf.cast(tf.concat(codes, axis=1), dtype=tf.float32) return length_code def resolution(self, model_input_raw, num_frames, resolution): frame_dim = len(model_input_raw.get_shape()) - 2 feature_dim = len(model_input_raw.get_shape()) - 1 max_frames = model_input_raw.get_shape().as_list()[frame_dim] num_features = model_input_raw.get_shape().as_list()[feature_dim] if resolution > 1: new_max_frames = max_frames / resolution cut_frames = new_max_frames * resolution model_input_raw = model_input_raw[:, :cut_frames, :] model_input_raw = tf.reshape(model_input_raw, shape=[-1,new_max_frames,resolution,num_features]) model_input_raw = tf.reduce_mean(model_input_raw, axis=2) model_input = tf.nn.l2_normalize(model_input_raw, feature_dim) num_frames = num_frames / resolution else: model_input = tf.nn.l2_normalize(model_input_raw, feature_dim) return model_input, num_frames ================================================ FILE: youtube-8m-wangheda/all_frame_models/multiscale_cnn_lstm_model.py ================================================ import math import models import tensorflow as tf import numpy as np import utils from tensorflow import flags import tensorflow.contrib.slim as slim FLAGS = flags.FLAGS class MultiscaleCnnLstmModel(models.BaseModel): def cnn(self, model_input, l2_penalty=1e-8, num_filters = [1024, 1024, 1024], filter_sizes = [1,2,3], sub_scope="", **unused_params): max_frames = model_input.get_shape().as_list()[1] num_features = model_input.get_shape().as_list()[2] shift_inputs = [] for i in xrange(max(filter_sizes)): if i == 0: shift_inputs.append(model_input) else: shift_inputs.append(tf.pad(model_input, paddings=[[0,0],[i,0],[0,0]])[:,:max_frames,:]) cnn_outputs = [] for nf, fs in zip(num_filters, filter_sizes): sub_input = tf.concat(shift_inputs[:fs], axis=2) sub_filter = tf.get_variable(sub_scope+"cnn-filter-len%d"%fs, shape=[num_features*fs, nf], dtype=tf.float32, initializer=tf.truncated_normal_initializer(mean=0.0, stddev=0.1), regularizer=tf.contrib.layers.l2_regularizer(l2_penalty)) cnn_outputs.append(tf.einsum("ijk,kl->ijl", sub_input, sub_filter)) cnn_output = tf.concat(cnn_outputs, axis=2) cnn_output = slim.batch_norm( cnn_output, center=True, scale=True, is_training=FLAGS.is_training, scope=sub_scope+"cluster_bn") return cnn_output def moe(self,model_input, vocab_size, num_mixtures=None, l2_penalty=1e-8, scopename="", **unused_params): num_mixtures = num_mixtures or FLAGS.moe_num_mixtures gate_activations = slim.fully_connected( model_input, vocab_size * (num_mixtures + 1), activation_fn=None, biases_initializer=None, weights_regularizer=slim.l2_regularizer(l2_penalty), scope="gates"+scopename) expert_activations = slim.fully_connected( model_input, vocab_size * num_mixtures, activation_fn=None, weights_regularizer=slim.l2_regularizer(l2_penalty), scope="experts"+scopename) gating_distribution = tf.nn.softmax(tf.reshape( gate_activations, [-1, num_mixtures + 1])) # (Batch * #Labels) x (num_mixtures + 1) expert_distribution = tf.nn.sigmoid(tf.reshape( expert_activations, [-1, num_mixtures])) # (Batch * #Labels) x num_mixtures final_probabilities_by_class_and_batch = tf.reduce_sum( gating_distribution[:, :num_mixtures] * expert_distribution, 1) final_probabilities = tf.reshape(final_probabilities_by_class_and_batch, [-1, vocab_size]) return final_probabilities def rnn(self, model_input, lstm_size, num_frames, sub_scope="", **unused_params): cell = tf.contrib.rnn.BasicLSTMCell(lstm_size, forget_bias=1.0, state_is_tuple=True) with tf.variable_scope("RNN-"+sub_scope): outputs, state = tf.nn.dynamic_rnn(cell, model_input, sequence_length=num_frames, swap_memory=True, dtype=tf.float32) # return final memory return state.c def create_model(self, model_input, vocab_size, num_frames, l2_penalty=1e-8, **unused_params): num_layers = FLAGS.multiscale_cnn_lstm_layers lstm_size = int(FLAGS.lstm_cells) pool_size=2 num_filters=[256,256,512] filter_sizes=[1,2,3] features_size = sum(num_filters) sub_predictions = [] cnn_input = model_input cnn_max_frames = model_input.get_shape().as_list()[1] for layer in range(num_layers): cnn_output = self.cnn(cnn_input, num_filters=num_filters, filter_sizes=filter_sizes, sub_scope="cnn%d"%(layer+1)) cnn_output_relu = tf.nn.relu(cnn_output) lstm_memory = self.rnn(cnn_output_relu, lstm_size, num_frames, sub_scope="rnn%d"%(layer+1)) sub_prediction = self.moe(lstm_memory, vocab_size, scopename="moe%d"%(layer+1)) sub_predictions.append(sub_prediction) cnn_max_frames /= pool_size max_pooled_cnn_output = tf.reduce_max( tf.reshape( cnn_output_relu[:, :cnn_max_frames*2, :], [-1, cnn_max_frames, pool_size, features_size] ), axis=2) # for the next cnn layer cnn_input = max_pooled_cnn_output num_frames = tf.maximum(num_frames/pool_size, 1) support_predictions = tf.concat(sub_predictions, axis=1) predictions = tf.add_n(sub_predictions) / len(sub_predictions) return {"predictions": predictions, "support_predictions": support_predictions} ================================================ FILE: youtube-8m-wangheda/all_frame_models/positional_cnn_deep_combine_chain_model.py ================================================ import math import models import tensorflow as tf import numpy as np import utils from tensorflow import flags import tensorflow.contrib.slim as slim FLAGS = flags.FLAGS class PositionalCnnDeepCombineChainModel(models.BaseModel): """Add positional embedding for cnn""" def cnn(self, model_input, l2_penalty=1e-8, num_filters = [1024, 1024, 1024], filter_sizes = [1,2,3], sub_scope="", **unused_params): max_frames = model_input.get_shape().as_list()[1] num_features = model_input.get_shape().as_list()[2] shift_inputs = [] for i in xrange(max(filter_sizes)): if i == 0: shift_inputs.append(model_input) else: shift_inputs.append(tf.pad(model_input, paddings=[[0,0],[i,0],[0,0]])[:,:max_frames,:]) cnn_outputs = [] for nf, fs in zip(num_filters, filter_sizes): sub_input = tf.concat(shift_inputs[:fs], axis=2) sub_filter = tf.get_variable(sub_scope+"cnn-filter-len%d"%fs, shape=[num_features*fs, nf], dtype=tf.float32, initializer=tf.truncated_normal_initializer(mean=0.0, stddev=0.1), regularizer=tf.contrib.layers.l2_regularizer(l2_penalty)) cnn_outputs.append(tf.einsum("ijk,kl->ijl", sub_input, sub_filter)) cnn_output = tf.concat(cnn_outputs, axis=2) return cnn_output def add_positional_embedding(self, model_input, num_frames, l2_penalty=1e-8): batch_size, max_frames, num_features = model_input.get_shape().as_list() positional_embedding = tf.get_variable("positional_embedding", dtype=tf.float32, shape=[1, max_frames, num_features], initializer=tf.zeros_initializer(), regularizer=tf.contrib.layers.l2_regularizer(l2_penalty)) mask = tf.sequence_mask(lengths=num_frames, maxlen=max_frames, dtype=tf.float32) model_input_with_positional_embedding = tf.einsum("ijk,ij->ijk", model_input + positional_embedding, mask) return model_input_with_positional_embedding def gated_linear_unit(self, input1, input2): return tf.nn.softmax(input1) * input2 def create_model(self, model_input, vocab_size, num_frames, num_mixtures=None, l2_penalty=1e-8, sub_scope="", original_input=None, **unused_params): num_supports = FLAGS.num_supports num_layers = FLAGS.deep_chain_layers relu_cells = FLAGS.deep_chain_relu_cells max_frames = model_input.get_shape().as_list()[1] relu_layers = [] support_predictions = [] # add positional embedding model_input = self.add_positional_embedding(model_input, num_frames, l2_penalty=l2_penalty) mask = tf.sequence_mask(lengths=num_frames, maxlen=max_frames, dtype=tf.float32) mean_input = tf.einsum("ijk,ij->ik", model_input, mask) \ / tf.expand_dims(tf.cast(num_frames, dtype=tf.float32), dim=1) mean_relu = slim.fully_connected( mean_input, relu_cells, activation_fn=tf.nn.relu, weights_regularizer=slim.l2_regularizer(l2_penalty), scope=sub_scope+"mean-relu") mean_relu_norm = tf.nn.l2_normalize(mean_relu, dim=1) relu_layers.append(mean_relu_norm) cnn_output1 = self.cnn(model_input, num_filters=[relu_cells,2*relu_cells,relu_cells], filter_sizes=[1,2,3], sub_scope=sub_scope+"cnn0-output1") cnn_output2 = self.cnn(model_input, num_filters=[relu_cells,2*relu_cells,relu_cells], filter_sizes=[1,2,3], sub_scope=sub_scope+"cnn0-output2") cnn_output = self.gated_linear_unit(cnn_output1, cnn_output2) max_cnn_output = tf.reduce_max(cnn_output, axis=1) normalized_cnn_output = tf.nn.l2_normalize(max_cnn_output, dim=1) next_input = normalized_cnn_output for layer in xrange(num_layers): sub_prediction = self.sub_model(next_input, vocab_size, sub_scope=sub_scope+"prediction-%d"%layer) support_predictions.append(sub_prediction) sub_relu = slim.fully_connected( sub_prediction, relu_cells, activation_fn=tf.nn.relu, weights_regularizer=slim.l2_regularizer(l2_penalty), scope=sub_scope+"relu-%d"%layer) relu_norm = tf.nn.l2_normalize(sub_relu, dim=1) relu_layers.append(relu_norm) cnn_output1 = self.cnn(model_input, num_filters=[relu_cells,2*relu_cells,relu_cells], filter_sizes=[1,2,3], sub_scope=sub_scope+"cnn%d-output1"%(layer+1)) cnn_output2 = self.cnn(model_input, num_filters=[relu_cells,2*relu_cells,relu_cells], filter_sizes=[1,2,3], sub_scope=sub_scope+"cnn%d-output2"%(layer+1)) cnn_output = self.gated_linear_unit(cnn_output1, cnn_output2) max_cnn_output = tf.reduce_max(cnn_output, axis=1) normalized_cnn_output = tf.nn.l2_normalize(max_cnn_output, dim=1) next_input = tf.concat([mean_input, normalized_cnn_output] + relu_layers, axis=1) main_predictions = self.sub_model(next_input, vocab_size, sub_scope=sub_scope+"-main") support_predictions = tf.concat(support_predictions, axis=1) return {"predictions": main_predictions, "support_predictions": support_predictions} def sub_model(self, model_input, vocab_size, num_mixtures=None, l2_penalty=1e-8, sub_scope="", **unused_params): num_mixtures = num_mixtures or FLAGS.moe_num_mixtures gate_activations = slim.fully_connected( model_input, vocab_size * (num_mixtures + 1), activation_fn=None, biases_initializer=None, weights_regularizer=slim.l2_regularizer(l2_penalty), scope="gates-"+sub_scope) expert_activations = slim.fully_connected( model_input, vocab_size * num_mixtures, activation_fn=None, weights_regularizer=slim.l2_regularizer(l2_penalty), scope="experts-"+sub_scope) gating_distribution = tf.nn.softmax(tf.reshape( gate_activations, [-1, num_mixtures + 1])) # (Batch * #Labels) x (num_mixtures + 1) expert_distribution = tf.nn.sigmoid(tf.reshape( expert_activations, [-1, num_mixtures])) # (Batch * #Labels) x num_mixtures final_probabilities_by_class_and_batch = tf.reduce_sum( gating_distribution[:, :num_mixtures] * expert_distribution, 1) final_probabilities = tf.reshape(final_probabilities_by_class_and_batch, [-1, vocab_size]) return final_probabilities def get_mask(self, max_frames, num_frames): mask_array = [] for i in xrange(max_frames + 1): tmp = [0.0] * max_frames for j in xrange(i): tmp[j] = 1.0 mask_array.append(tmp) mask_array = np.array(mask_array) mask_init = tf.constant_initializer(mask_array) mask_emb = tf.get_variable("mask_emb", shape = [max_frames + 1, max_frames], dtype = tf.float32, trainable = False, initializer = mask_init) mask = tf.nn.embedding_lookup(mask_emb, num_frames) return mask ================================================ FILE: youtube-8m-wangheda/all_frame_models/progressive_attention_lstm_model.py ================================================ import sys import models import model_utils import math import numpy as np import video_level_models import tensorflow as tf import utils import tensorflow.contrib.slim as slim from tensorflow import flags FLAGS = flags.FLAGS class ProgressiveAttentionLstmModel(models.BaseModel): def create_model(self, model_input, vocab_size, num_frames, **unused_params): """Creates a model which uses a stack of LSTMs to represent the video. Args: model_input: A 'batch_size' x 'max_frames' x 'num_features' matrix of input features. vocab_size: The number of classes in the dataset. num_frames: A vector of length 'batch' which indicates the number of frames for each video (before padding). Returns: A dictionary with a tensor containing the probability predictions of the model in the 'predictions' key. The dimensions of the tensor are 'batch_size' x 'num_classes'. """ lstm_size = int(FLAGS.lstm_cells) number_of_layers = FLAGS.lstm_layers num_attentions = FLAGS.num_attentions print model_input.get_shape().as_list() max_frames = model_input.get_shape().as_list()[1] ## Batch normalize the input stacked_cell = tf.contrib.rnn.MultiRNNCell( [ tf.contrib.rnn.BasicLSTMCell( lstm_size, forget_bias=1.0, state_is_tuple=True) for _ in range(number_of_layers) ], state_is_tuple=True) att_cell = tf.contrib.rnn.AttentionCellWrapper(cell = tf.contrib.rnn.BasicLSTMCell( lstm_size, forget_bias=1.0, state_is_tuple=True), attn_length=1, state_is_tuple=True) loss = 0.0 with tf.variable_scope("RNN"): outputs, state = tf.nn.dynamic_rnn(stacked_cell, model_input, sequence_length=num_frames, swap_memory=FLAGS.rnn_swap_memory, dtype=tf.float32) final_memory = tf.concat(map(lambda x: x.c, state), axis = 1) with tf.variable_scope("ATT"): att_outputs, att_state = tf.nn.dynamic_rnn(att_cell, outputs, sequence_length=tf.ones_like(num_frames, dtype=tf.int32)*num_attentions, swap_memory=FLAGS.rnn_swap_memory, dtype=tf.float32) print att_outputs print att_state att_state, _, _ = att_state print att_state att_final_memory = att_state.c final_state = tf.concat([att_final_memory, final_memory], axis = 1) print "final_state", final_state aggregated_model = getattr(video_level_models, FLAGS.video_level_classifier_model) predictions = aggregated_model().create_model( model_input=final_state, original_input=model_input, vocab_size=vocab_size, **unused_params) print predictions return predictions ================================================ FILE: youtube-8m-wangheda/all_frame_models/wide_and_deep_model.py ================================================ import sys import models import model_utils import math import numpy as np import video_level_models import frame_level_models import tensorflow as tf import utils import tensorflow.contrib.slim as slim from tensorflow import flags FLAGS = flags.FLAGS class WideAndDeepModel(models.BaseModel): def create_model(self, model_input, vocab_size, num_frames, l2_penalty=1e-8, **unused_params): """ A super model that combine one or more models """ models = FLAGS.wide_and_deep_models outputs = [] for model_name in map(lambda x: x.strip(), models.split(",")): model = getattr(frame_level_models, model_name, None)() output = model.create_model(model_input, vocab_size, num_frames, l2_penalty=l2_penalty, **unused_params)["predictions"] outputs.append(tf.expand_dims(output, axis=2)) num_models = len(outputs) model_outputs = tf.concat(outputs, axis=2) # linear_combination = tf.get_variable("combine", shape=[vocab_size,num_models], # dtype=tf.float32, initializer=tf.zeros_initializer(), # regularizer=slim.l2_regularizer(l2_penalty)) # combination = tf.nn.softmax(linear_combination) combination = tf.fill(dims=[vocab_size,num_models], value=1.0/num_models) output_sum = tf.einsum("ijk,jk->ij", model_outputs, combination) return {"predictions": output_sum} ================================================ FILE: youtube-8m-wangheda/all_video_models/.vimrc ================================================ set tabstop=2 set shiftwidth=2 set expandtab set autoindent ================================================ FILE: youtube-8m-wangheda/all_video_models/__init__.py ================================================ from logistic_model import * from moe_model import * from mlp_moe_model import * from chain_moe_model import * from deep_chain_model import * from deep_combine_chain_model import * from distillchain_deep_combine_chain_model import * from hidden_chain_model import * from hidden_combine_chain_model import * from chain_main_relu_moe_model import * from chain_support_relu_moe_model import * from shortcut_chain_support_relu_moe_model import * from multitask_moe_model import * from stage2_logistic_model import * from multitask_divergence_moe_model import * from multitask_divergence_deep_combine_chain_model import * ================================================ FILE: youtube-8m-wangheda/all_video_models/chain_main_relu_moe_model.py ================================================ import math import models import tensorflow as tf import utils from tensorflow import flags import tensorflow.contrib.slim as slim FLAGS = flags.FLAGS class ChainMainReluMoeModel(models.BaseModel): """A softmax over a mixture of logistic models (with L2 regularization).""" def create_model(self, model_input, vocab_size, num_mixtures=None, l2_penalty=1e-8, sub_scope="", original_input=None, **unused_params): num_supports = FLAGS.num_supports input_size = model_input.shape.as_list()[1] support_predictions = self.sub_model(model_input, num_supports, sub_scope=sub_scope+"-support") main_relu = slim.fully_connected( model_input, input_size, activation_fn=tf.nn.relu, weights_regularizer=slim.l2_regularizer(l2_penalty), scope="main-relu-"+sub_scope) main_input = tf.concat([main_relu, support_predictions], axis=1) main_predictions = self.sub_model(main_input, vocab_size, sub_scope=sub_scope+"-main") return {"predictions": main_predictions, "support_predictions": support_predictions} def sub_model(self, model_input, vocab_size, num_mixtures=None, l2_penalty=1e-8, sub_scope="", **unused_params): num_mixtures = num_mixtures or FLAGS.moe_num_mixtures gate_activations = slim.fully_connected( model_input, vocab_size * (num_mixtures + 1), activation_fn=None, biases_initializer=None, weights_regularizer=slim.l2_regularizer(l2_penalty), scope="gates-"+sub_scope) expert_activations = slim.fully_connected( model_input, vocab_size * num_mixtures, activation_fn=None, weights_regularizer=slim.l2_regularizer(l2_penalty), scope="experts-"+sub_scope) gating_distribution = tf.nn.softmax(tf.reshape( gate_activations, [-1, num_mixtures + 1])) # (Batch * #Labels) x (num_mixtures + 1) expert_distribution = tf.nn.sigmoid(tf.reshape( expert_activations, [-1, num_mixtures])) # (Batch * #Labels) x num_mixtures final_probabilities_by_class_and_batch = tf.reduce_sum( gating_distribution[:, :num_mixtures] * expert_distribution, 1) final_probabilities = tf.reshape(final_probabilities_by_class_and_batch, [-1, vocab_size]) return final_probabilities ================================================ FILE: youtube-8m-wangheda/all_video_models/chain_moe_model.py ================================================ import math import models import tensorflow as tf import utils from tensorflow import flags import tensorflow.contrib.slim as slim FLAGS = flags.FLAGS class ChainMoeModel(models.BaseModel): """A softmax over a mixture of logistic models (with L2 regularization).""" def create_model(self, model_input, vocab_size, num_mixtures=None, l2_penalty=1e-8, sub_scope="", original_input=None, **unused_params): num_supports = FLAGS.num_supports support_predictions = self.sub_model(model_input, num_supports, sub_scope=sub_scope+"-support") main_input = tf.concat([model_input, support_predictions], axis=1) main_predictions = self.sub_model(main_input, vocab_size, sub_scope=sub_scope+"-main") return {"predictions": main_predictions, "support_predictions": support_predictions} def sub_model(self, model_input, vocab_size, num_mixtures=None, l2_penalty=1e-8, sub_scope="", **unused_params): num_mixtures = num_mixtures or FLAGS.moe_num_mixtures gate_activations = slim.fully_connected( model_input, vocab_size * (num_mixtures + 1), activation_fn=None, biases_initializer=None, weights_regularizer=slim.l2_regularizer(l2_penalty), scope="gates"+sub_scope) expert_activations = slim.fully_connected( model_input, vocab_size * num_mixtures, activation_fn=None, weights_regularizer=slim.l2_regularizer(l2_penalty), scope="experts"+sub_scope) gating_distribution = tf.nn.softmax(tf.reshape( gate_activations, [-1, num_mixtures + 1])) # (Batch * #Labels) x (num_mixtures + 1) expert_distribution = tf.nn.sigmoid(tf.reshape( expert_activations, [-1, num_mixtures])) # (Batch * #Labels) x num_mixtures final_probabilities_by_class_and_batch = tf.reduce_sum( gating_distribution[:, :num_mixtures] * expert_distribution, 1) final_probabilities = tf.reshape(final_probabilities_by_class_and_batch, [-1, vocab_size]) return final_probabilities ================================================ FILE: youtube-8m-wangheda/all_video_models/chain_support_relu_moe_model.py ================================================ import math import models import tensorflow as tf import utils from tensorflow import flags import tensorflow.contrib.slim as slim FLAGS = flags.FLAGS class ChainSupportReluMoeModel(models.BaseModel): """A softmax over a mixture of logistic models (with L2 regularization).""" def create_model(self, model_input, vocab_size, num_mixtures=None, l2_penalty=1e-8, sub_scope="", original_input=None, **unused_params): num_supports = FLAGS.num_supports input_size = model_input.shape.as_list()[1] support_relu = slim.fully_connected( model_input, input_size, activation_fn=tf.nn.relu, weights_regularizer=slim.l2_regularizer(l2_penalty), scope="main-relu-"+sub_scope) support_predictions = self.sub_model(support_relu, num_supports, sub_scope=sub_scope+"-support") main_input = tf.concat([model_input, support_predictions], axis=1) main_predictions = self.sub_model(main_input, vocab_size, sub_scope=sub_scope+"-main") return {"predictions": main_predictions, "support_predictions": support_predictions} def sub_model(self, model_input, vocab_size, num_mixtures=None, l2_penalty=1e-8, sub_scope="", **unused_params): num_mixtures = num_mixtures or FLAGS.moe_num_mixtures gate_activations = slim.fully_connected( model_input, vocab_size * (num_mixtures + 1), activation_fn=None, biases_initializer=None, weights_regularizer=slim.l2_regularizer(l2_penalty), scope="gates-"+sub_scope) expert_activations = slim.fully_connected( model_input, vocab_size * num_mixtures, activation_fn=None, weights_regularizer=slim.l2_regularizer(l2_penalty), scope="experts-"+sub_scope) gating_distribution = tf.nn.softmax(tf.reshape( gate_activations, [-1, num_mixtures + 1])) # (Batch * #Labels) x (num_mixtures + 1) expert_distribution = tf.nn.sigmoid(tf.reshape( expert_activations, [-1, num_mixtures])) # (Batch * #Labels) x num_mixtures final_probabilities_by_class_and_batch = tf.reduce_sum( gating_distribution[:, :num_mixtures] * expert_distribution, 1) final_probabilities = tf.reshape(final_probabilities_by_class_and_batch, [-1, vocab_size]) return final_probabilities ================================================ FILE: youtube-8m-wangheda/all_video_models/deep_chain_model.py ================================================ import math import models import tensorflow as tf import utils from tensorflow import flags import tensorflow.contrib.slim as slim FLAGS = flags.FLAGS class DeepChainModel(models.BaseModel): """A softmax over a mixture of logistic models (with L2 regularization).""" def create_model(self, model_input, vocab_size, num_mixtures=None, l2_penalty=1e-8, sub_scope="", original_input=None, num_frames=None, **unused_params): num_supports = FLAGS.num_supports num_layers = FLAGS.deep_chain_layers relu_cells = FLAGS.deep_chain_relu_cells use_length = FLAGS.deep_chain_use_length if use_length: model_input = tf.concat([model_input, self.get_length_code(num_frames)], axis=1) next_input = model_input support_predictions = [] for layer in xrange(num_layers): sub_prediction = self.sub_model(next_input, vocab_size, sub_scope=sub_scope+"prediction-%d"%layer) sub_relu = slim.fully_connected( sub_prediction, relu_cells, activation_fn=tf.nn.relu, weights_regularizer=slim.l2_regularizer(l2_penalty), scope=sub_scope+"relu-%d"%layer) relu_norm = tf.nn.l2_normalize(sub_relu, dim=1) next_input = tf.concat([model_input, relu_norm], axis=1) support_predictions.append(sub_prediction) main_predictions = self.sub_model(next_input, vocab_size, sub_scope=sub_scope+"-main") support_predictions = tf.concat(support_predictions, axis=1) return {"predictions": main_predictions, "support_predictions": support_predictions} def sub_model(self, model_input, vocab_size, num_mixtures=None, l2_penalty=1e-8, sub_scope="", **unused_params): num_mixtures = num_mixtures or FLAGS.moe_num_mixtures gate_activations = slim.fully_connected( model_input, vocab_size * (num_mixtures + 1), activation_fn=None, biases_initializer=None, weights_regularizer=slim.l2_regularizer(l2_penalty), scope="gates-"+sub_scope) expert_activations = slim.fully_connected( model_input, vocab_size * num_mixtures, activation_fn=None, weights_regularizer=slim.l2_regularizer(l2_penalty), scope="experts-"+sub_scope) gating_distribution = tf.nn.softmax(tf.reshape( gate_activations, [-1, num_mixtures + 1])) # (Batch * #Labels) x (num_mixtures + 1) expert_distribution = tf.nn.sigmoid(tf.reshape( expert_activations, [-1, num_mixtures])) # (Batch * #Labels) x num_mixtures final_probabilities_by_class_and_batch = tf.reduce_sum( gating_distribution[:, :num_mixtures] * expert_distribution, 1) final_probabilities = tf.reshape(final_probabilities_by_class_and_batch, [-1, vocab_size]) return final_probabilities def get_length_code(self, num_frames): code_0 = tf.cast(num_frames <= 60, dtype=tf.int32) code_1 = tf.cast(num_frames > 60, dtype=tf.int32) * tf.cast(num_frames <= 120, dtype=tf.int32) code_2 = tf.cast(num_frames > 120, dtype=tf.int32) * tf.cast(num_frames <= 180, dtype=tf.int32) code_3 = tf.cast(num_frames > 180, dtype=tf.int32) * tf.cast(num_frames <= 240, dtype=tf.int32) code_4 = tf.cast(num_frames > 240, dtype=tf.int32) codes = map(lambda x: tf.expand_dims(x, dim=1), [code_0, code_1, code_2, code_3, code_4]) length_code = tf.cast(tf.concat(codes, axis=1), dtype=tf.float32) return length_code ================================================ FILE: youtube-8m-wangheda/all_video_models/deep_combine_chain_model.py ================================================ import math import models import tensorflow as tf import utils from tensorflow import flags import tensorflow.contrib.slim as slim FLAGS = flags.FLAGS class DeepCombineChainModel(models.BaseModel): """A softmax over a mixture of logistic models (with L2 regularization).""" def create_model(self, model_input, vocab_size, num_mixtures=None, l2_penalty=1e-8, sub_scope="", original_input=None, dropout=False, keep_prob=None, noise_level=None, num_frames=None, **unused_params): num_supports = FLAGS.num_supports num_layers = FLAGS.deep_chain_layers relu_cells = FLAGS.deep_chain_relu_cells relu_type = FLAGS.deep_chain_relu_type use_length = FLAGS.deep_chain_use_length next_input = model_input support_predictions = [] for layer in xrange(num_layers): sub_prediction = self.sub_model(next_input, vocab_size, sub_scope=sub_scope+"prediction-%d"%layer, dropout=dropout, keep_prob=keep_prob, noise_level=noise_level) sub_activation = slim.fully_connected( sub_prediction, relu_cells, activation_fn=None, weights_regularizer=slim.l2_regularizer(l2_penalty), scope=sub_scope+"relu-%d"%layer) if relu_type == "elu": sub_relu = tf.nn.elu(sub_activation) else: # default: relu sub_relu = tf.nn.relu(sub_activation) if noise_level is not None: print "adding noise to sub_relu, level = ", noise_level sub_relu = sub_relu + tf.random_normal(tf.shape(sub_relu), mean=0.0, stddev=noise_level) relu_norm = tf.nn.l2_normalize(sub_relu, dim=1) next_input = tf.concat([next_input, relu_norm], axis=1) support_predictions.append(sub_prediction) main_predictions = self.sub_model(next_input, vocab_size, sub_scope=sub_scope+"-main") support_predictions = tf.concat(support_predictions, axis=1) return {"predictions": main_predictions, "support_predictions": support_predictions} def sub_model(self, model_input, vocab_size, num_mixtures=None, l2_penalty=1e-8, sub_scope="", dropout=False, keep_prob=None, noise_level=None, **unused_params): num_mixtures = num_mixtures or FLAGS.moe_num_mixtures if dropout: model_input = tf.nn.dropout(model_input, keep_prob=keep_prob) gate_activations = slim.fully_connected( model_input, vocab_size * (num_mixtures + 1), activation_fn=None, biases_initializer=None, weights_regularizer=slim.l2_regularizer(l2_penalty), scope="gates-"+sub_scope) expert_activations = slim.fully_connected( model_input, vocab_size * num_mixtures, activation_fn=None, weights_regularizer=slim.l2_regularizer(l2_penalty), scope="experts-"+sub_scope) gating_distribution = tf.nn.softmax(tf.reshape( gate_activations, [-1, num_mixtures + 1])) # (Batch * #Labels) x (num_mixtures + 1) expert_distribution = tf.nn.sigmoid(tf.reshape( expert_activations, [-1, num_mixtures])) # (Batch * #Labels) x num_mixtures final_probabilities_by_class_and_batch = tf.reduce_sum( gating_distribution[:, :num_mixtures] * expert_distribution, 1) final_probabilities = tf.reshape(final_probabilities_by_class_and_batch, [-1, vocab_size]) return final_probabilities ================================================ FILE: youtube-8m-wangheda/all_video_models/distillchain_deep_combine_chain_model.py ================================================ import math import models import tensorflow as tf import utils from tensorflow import flags import tensorflow.contrib.slim as slim FLAGS = flags.FLAGS class DistillchainDeepCombineChainModel(models.BaseModel): """A softmax over a mixture of logistic models (with L2 regularization).""" def create_model(self, model_input, vocab_size, num_mixtures=None, l2_penalty=1e-8, sub_scope="", original_input=None, dropout=False, keep_prob=None, noise_level=None, distillation_predictions=None, num_frames=None, **unused_params): assert distillation_predictions is not None, "distillation feature must be used" num_supports = FLAGS.num_supports num_layers = FLAGS.deep_chain_layers relu_cells = FLAGS.deep_chain_relu_cells relu_type = FLAGS.deep_chain_relu_type use_length = FLAGS.deep_chain_use_length distill_relu = slim.fully_connected( distillation_predictions, relu_cells, activation_fn=tf.nn.relu, weights_regularizer=slim.l2_regularizer(l2_penalty), scope=sub_scope+"distillrelu") distill_norm = tf.nn.l2_normalize(distill_relu, dim=1) next_input = tf.concat([model_input, distill_norm], axis=1) support_predictions = [] for layer in xrange(num_layers): sub_prediction = self.sub_model(next_input, vocab_size, sub_scope=sub_scope+"prediction-%d"%layer, dropout=dropout, keep_prob=keep_prob, noise_level=noise_level) sub_activation = slim.fully_connected( sub_prediction, relu_cells, activation_fn=None, weights_regularizer=slim.l2_regularizer(l2_penalty), scope=sub_scope+"relu-%d"%layer) if relu_type == "elu": sub_relu = tf.nn.elu(sub_activation) else: # default: relu sub_relu = tf.nn.relu(sub_activation) if noise_level is not None: print "adding noise to sub_relu, level = ", noise_level sub_relu = sub_relu + tf.random_normal(tf.shape(sub_relu), mean=0.0, stddev=noise_level) relu_norm = tf.nn.l2_normalize(sub_relu, dim=1) next_input = tf.concat([next_input, relu_norm], axis=1) support_predictions.append(sub_prediction) main_predictions = self.sub_model(next_input, vocab_size, sub_scope=sub_scope+"-main") support_predictions = tf.concat(support_predictions, axis=1) return {"predictions": main_predictions, "support_predictions": support_predictions} def sub_model(self, model_input, vocab_size, num_mixtures=None, l2_penalty=1e-8, sub_scope="", dropout=False, keep_prob=None, noise_level=None, **unused_params): num_mixtures = num_mixtures or FLAGS.moe_num_mixtures if dropout: model_input = tf.nn.dropout(model_input, keep_prob=keep_prob) gate_activations = slim.fully_connected( model_input, vocab_size * (num_mixtures + 1), activation_fn=None, biases_initializer=None, weights_regularizer=slim.l2_regularizer(l2_penalty), scope="gates-"+sub_scope) expert_activations = slim.fully_connected( model_input, vocab_size * num_mixtures, activation_fn=None, weights_regularizer=slim.l2_regularizer(l2_penalty), scope="experts-"+sub_scope) gating_distribution = tf.nn.softmax(tf.reshape( gate_activations, [-1, num_mixtures + 1])) # (Batch * #Labels) x (num_mixtures + 1) expert_distribution = tf.nn.sigmoid(tf.reshape( expert_activations, [-1, num_mixtures])) # (Batch * #Labels) x num_mixtures final_probabilities_by_class_and_batch = tf.reduce_sum( gating_distribution[:, :num_mixtures] * expert_distribution, 1) final_probabilities = tf.reshape(final_probabilities_by_class_and_batch, [-1, vocab_size]) return final_probabilities ================================================ FILE: youtube-8m-wangheda/all_video_models/hidden_chain_model.py ================================================ import math import models import tensorflow as tf import utils from tensorflow import flags import tensorflow.contrib.slim as slim FLAGS = flags.FLAGS class HiddenChainModel(models.BaseModel): """A softmax over a mixture of logistic models (with L2 regularization).""" def create_model(self, model_input, vocab_size, num_mixtures=None, l2_penalty=1e-8, sub_scope="", original_input=None, **unused_params): num_supports = FLAGS.num_supports num_layers = FLAGS.hidden_chain_layers relu_cells = FLAGS.hidden_chain_relu_cells next_input = model_input support_predictions = [] for layer in xrange(num_layers): sub_relu = slim.fully_connected( next_input, relu_cells, activation_fn=tf.nn.relu, weights_regularizer=slim.l2_regularizer(l2_penalty), scope=sub_scope+"relu-%d"%layer) sub_prediction = self.sub_model(sub_relu, vocab_size, sub_scope=sub_scope+"prediction-%d"%layer) relu_norm = tf.nn.l2_normalize(sub_relu, dim=1) next_input = tf.concat([model_input, relu_norm], axis=1) support_predictions.append(sub_prediction) main_predictions = self.sub_model(next_input, vocab_size, sub_scope=sub_scope+"-main") support_predictions = tf.concat(support_predictions, axis=1) return {"predictions": main_predictions, "support_predictions": support_predictions} def sub_model(self, model_input, vocab_size, num_mixtures=None, l2_penalty=1e-8, sub_scope="", **unused_params): num_mixtures = num_mixtures or FLAGS.moe_num_mixtures gate_activations = slim.fully_connected( model_input, vocab_size * (num_mixtures + 1), activation_fn=None, biases_initializer=None, weights_regularizer=slim.l2_regularizer(l2_penalty), scope="gates-"+sub_scope) expert_activations = slim.fully_connected( model_input, vocab_size * num_mixtures, activation_fn=None, weights_regularizer=slim.l2_regularizer(l2_penalty), scope="experts-"+sub_scope) gating_distribution = tf.nn.softmax(tf.reshape( gate_activations, [-1, num_mixtures + 1])) # (Batch * #Labels) x (num_mixtures + 1) expert_distribution = tf.nn.sigmoid(tf.reshape( expert_activations, [-1, num_mixtures])) # (Batch * #Labels) x num_mixtures final_probabilities_by_class_and_batch = tf.reduce_sum( gating_distribution[:, :num_mixtures] * expert_distribution, 1) final_probabilities = tf.reshape(final_probabilities_by_class_and_batch, [-1, vocab_size]) return final_probabilities ================================================ FILE: youtube-8m-wangheda/all_video_models/hidden_combine_chain_model.py ================================================ import math import models import tensorflow as tf import utils from tensorflow import flags import tensorflow.contrib.slim as slim FLAGS = flags.FLAGS class HiddenCombineChainModel(models.BaseModel): """A softmax over a mixture of logistic models (with L2 regularization).""" def create_model(self, model_input, vocab_size, num_mixtures=None, l2_penalty=1e-8, sub_scope="", original_input=None, **unused_params): num_supports = FLAGS.num_supports num_layers = FLAGS.hidden_chain_layers relu_cells = FLAGS.hidden_chain_relu_cells next_input = model_input support_predictions = [] for layer in xrange(num_layers): sub_relu = slim.fully_connected( next_input, relu_cells, activation_fn=tf.nn.relu, weights_regularizer=slim.l2_regularizer(l2_penalty), scope=sub_scope+"relu-%d"%layer) sub_prediction = self.sub_model(sub_relu, vocab_size, sub_scope=sub_scope+"prediction-%d"%layer) relu_norm = tf.nn.l2_normalize(sub_relu, dim=1) next_input = tf.concat([next_input, relu_norm], axis=1) support_predictions.append(sub_prediction) main_predictions = self.sub_model(next_input, vocab_size, sub_scope=sub_scope+"-main") support_predictions = tf.concat(support_predictions, axis=1) return {"predictions": main_predictions, "support_predictions": support_predictions} def sub_model(self, model_input, vocab_size, num_mixtures=None, l2_penalty=1e-8, sub_scope="", **unused_params): num_mixtures = num_mixtures or FLAGS.moe_num_mixtures gate_activations = slim.fully_connected( model_input, vocab_size * (num_mixtures + 1), activation_fn=None, biases_initializer=None, weights_regularizer=slim.l2_regularizer(l2_penalty), scope="gates-"+sub_scope) expert_activations = slim.fully_connected( model_input, vocab_size * num_mixtures, activation_fn=None, weights_regularizer=slim.l2_regularizer(l2_penalty), scope="experts-"+sub_scope) gating_distribution = tf.nn.softmax(tf.reshape( gate_activations, [-1, num_mixtures + 1])) # (Batch * #Labels) x (num_mixtures + 1) expert_distribution = tf.nn.sigmoid(tf.reshape( expert_activations, [-1, num_mixtures])) # (Batch * #Labels) x num_mixtures final_probabilities_by_class_and_batch = tf.reduce_sum( gating_distribution[:, :num_mixtures] * expert_distribution, 1) final_probabilities = tf.reshape(final_probabilities_by_class_and_batch, [-1, vocab_size]) return final_probabilities ================================================ FILE: youtube-8m-wangheda/all_video_models/logistic_model.py ================================================ import math import models import tensorflow as tf import utils from tensorflow import flags import tensorflow.contrib.slim as slim FLAGS = flags.FLAGS class LogisticModel(models.BaseModel): """Logistic model with L2 regularization.""" def create_model(self, model_input, vocab_size, l2_penalty=1e-8, original_input=None, **unused_params): """Creates a logistic model. Args: model_input: 'batch' x 'num_features' matrix of input features. vocab_size: The number of classes in the dataset. Returns: A dictionary with a tensor containing the probability predictions of the model in the 'predictions' key. The dimensions of the tensor are batch_size x num_classes.""" output = slim.fully_connected( model_input, vocab_size, activation_fn=tf.nn.sigmoid, weights_regularizer=slim.l2_regularizer(l2_penalty)) return {"predictions": output} ================================================ FILE: youtube-8m-wangheda/all_video_models/mlp_moe_model.py ================================================ import math import models import tensorflow as tf import utils from tensorflow import flags import tensorflow.contrib.slim as slim FLAGS = flags.FLAGS class MlpMoeModel(models.BaseModel): """A softmax over a mixture of logistic models (with L2 regularization).""" def create_model(self, model_input, vocab_size, num_mixtures=None, l2_penalty=1e-8, sub_scope="", original_input=None, **unused_params): """Creates a Mixture of (Logistic) Experts model. The model consists of a per-class softmax distribution over a configurable number of logistic classifiers. One of the classifiers in the mixture is not trained, and always predicts 0. Args: model_input: 'batch_size' x 'num_features' matrix of input features. vocab_size: The number of classes in the dataset. num_mixtures: The number of mixtures (excluding a dummy 'expert' that always predicts the non-existence of an entity). l2_penalty: How much to penalize the squared magnitudes of parameter values. Returns: A dictionary with a tensor containing the probability predictions of the model in the 'predictions' key. The dimensions of the tensor are batch_size x num_classes. """ num_mixtures = num_mixtures or FLAGS.moe_num_mixtures feature_dim = len(model_input.get_shape().as_list()) - 1 num_features = model_input.get_shape().as_list()[feature_dim] layer1_size = 2048 gate_activations = slim.fully_connected( model_input, vocab_size * (num_mixtures + 1), activation_fn=None, biases_initializer=None, weights_regularizer=slim.l2_regularizer(l2_penalty), scope="gates"+sub_scope) expert_activations = slim.fully_connected( slim.fully_connected( model_input, num_features, activation_fn=tf.nn.relu, weights_initializer=tf.truncated_normal_initializer(mean=0.0, stddev=0.1), biases_initializer=tf.zeros_initializer(), weights_regularizer=slim.l2_regularizer(l2_penalty), scope="relu"+sub_scope), vocab_size * num_mixtures, activation_fn=None, weights_regularizer=slim.l2_regularizer(l2_penalty), scope="experts"+sub_scope) gating_distribution = tf.nn.softmax(tf.reshape( gate_activations, [-1, num_mixtures + 1])) # (Batch * #Labels) x (num_mixtures + 1) expert_distribution = tf.nn.sigmoid(tf.reshape( expert_activations, [-1, num_mixtures])) # (Batch * #Labels) x num_mixtures final_probabilities_by_class_and_batch = tf.reduce_sum( gating_distribution[:, :num_mixtures] * expert_distribution, 1) final_probabilities = tf.reshape(final_probabilities_by_class_and_batch, [-1, vocab_size]) return {"predictions": final_probabilities} ================================================ FILE: youtube-8m-wangheda/all_video_models/moe_model.py ================================================ import math import models import tensorflow as tf import utils from tensorflow import flags import tensorflow.contrib.slim as slim FLAGS = flags.FLAGS class MoeModel(models.BaseModel): """A softmax over a mixture of logistic models (with L2 regularization).""" def create_model(self, model_input, vocab_size, num_mixtures=None, l2_penalty=1e-8, sub_scope="", original_input=None, **unused_params): """Creates a Mixture of (Logistic) Experts model. The model consists of a per-class softmax distribution over a configurable number of logistic classifiers. One of the classifiers in the mixture is not trained, and always predicts 0. Args: model_input: 'batch_size' x 'num_features' matrix of input features. vocab_size: The number of classes in the dataset. num_mixtures: The number of mixtures (excluding a dummy 'expert' that always predicts the non-existence of an entity). l2_penalty: How much to penalize the squared magnitudes of parameter values. Returns: A dictionary with a tensor containing the probability predictions of the model in the 'predictions' key. The dimensions of the tensor are batch_size x num_classes. """ num_mixtures = num_mixtures or FLAGS.moe_num_mixtures gate_activations = slim.fully_connected( model_input, vocab_size * (num_mixtures + 1), activation_fn=None, biases_initializer=None, weights_regularizer=slim.l2_regularizer(l2_penalty), scope="gates"+sub_scope) expert_activations = slim.fully_connected( model_input, vocab_size * num_mixtures, activation_fn=None, weights_regularizer=slim.l2_regularizer(l2_penalty), scope="experts"+sub_scope) gating_distribution = tf.nn.softmax(tf.reshape( gate_activations, [-1, num_mixtures + 1])) # (Batch * #Labels) x (num_mixtures + 1) expert_distribution = tf.nn.sigmoid(tf.reshape( expert_activations, [-1, num_mixtures])) # (Batch * #Labels) x num_mixtures final_probabilities_by_class_and_batch = tf.reduce_sum( gating_distribution[:, :num_mixtures] * expert_distribution, 1) final_probabilities = tf.reshape(final_probabilities_by_class_and_batch, [-1, vocab_size]) return {"predictions": final_probabilities} ================================================ FILE: youtube-8m-wangheda/all_video_models/multitask_divergence_deep_combine_chain_model.py ================================================ import math import models import tensorflow as tf import utils from tensorflow import flags import tensorflow.contrib.slim as slim FLAGS = flags.FLAGS class MultiTaskDivergenceDeepCombineChainModel(models.BaseModel): """A softmax over a mixture of logistic models (with L2 regularization).""" def create_model(self, model_input, vocab_size, num_mixtures=None, l2_penalty=1e-8, sub_scope="ddcc", original_input=None, dropout=False, keep_prob=None, noise_level=None, num_frames=None, **unused_params): num_supports = FLAGS.num_supports num_models = FLAGS.divergence_model_count support_predictions = [] for i in xrange(num_models): sub_prediction = self.sub_chain_model(model_input,vocab_size, num_mixtures, l2_penalty, sub_scope+"%d"%i, original_input, dropout, keep_prob, noise_level) support_predictions.append(sub_prediction) support_predictions = tf.stack(support_predictions, axis=1) main_predictions = tf.reduce_mean(support_predictions, axis=1) return {"predictions": main_predictions, "support_predictions": support_predictions} def sub_chain_model(self, model_input, vocab_size, num_mixtures=None, l2_penalty=1e-8, sub_scope="", original_input=None, dropout=False, keep_prob=None, noise_level=None, num_frames=None, **unused_params): num_supports = FLAGS.num_supports num_layers = FLAGS.deep_chain_layers relu_cells = FLAGS.deep_chain_relu_cells relu_type = FLAGS.deep_chain_relu_type use_length = FLAGS.deep_chain_use_length input_list = [model_input] mean_activation = slim.fully_connected( model_input, relu_cells, activation_fn=None, weights_regularizer=slim.l2_regularizer(l2_penalty), scope=sub_scope+"relu-mean") mean_relu = tf.nn.relu(mean_activation) if noise_level is not None: print "adding noise to sub_relu, level = ", noise_level mean_relu = mean_relu + tf.random_normal(tf.shape(mean_relu), mean=0.0, stddev=noise_level) mean_relu_norm = tf.nn.l2_normalize(mean_relu, dim=1) next_input = mean_relu_norm support_predictions = [] for layer in xrange(num_layers): sub_prediction = self.sub_model(next_input, vocab_size, sub_scope=sub_scope+"prediction-%d"%layer, dropout=dropout, keep_prob=keep_prob, noise_level=noise_level) sub_activation = slim.fully_connected( sub_prediction, relu_cells, activation_fn=None, weights_regularizer=slim.l2_regularizer(l2_penalty), scope=sub_scope+"relu-%d"%layer) sub_relu = tf.nn.relu(sub_activation) if noise_level is not None: print "adding noise to sub_relu, level = ", noise_level sub_relu = sub_relu + tf.random_normal(tf.shape(sub_relu), mean=0.0, stddev=noise_level) relu_norm = tf.nn.l2_normalize(sub_relu, dim=1) next_input = tf.concat([next_input, relu_norm], axis=1) support_predictions.append(sub_prediction) main_predictions = self.sub_model(next_input, vocab_size, sub_scope=sub_scope+"-main") support_predictions = tf.reduce_mean(tf.stack(support_predictions, axis=0), axis=0) ratio = min(FLAGS.support_loss_percent, 0.2) final_predictions = main_predictions * (1.0 - ratio) + support_predictions * ratio return final_predictions def sub_model(self, model_input, vocab_size, num_mixtures=None, l2_penalty=1e-8, sub_scope="", dropout=False, keep_prob=None, noise_level=None, **unused_params): num_mixtures = num_mixtures or FLAGS.moe_num_mixtures if dropout: model_input = tf.nn.dropout(model_input, keep_prob=keep_prob) gate_activations = slim.fully_connected( model_input, vocab_size * (num_mixtures + 1), activation_fn=None, biases_initializer=None, weights_regularizer=slim.l2_regularizer(l2_penalty), scope="gates-"+sub_scope) expert_activations = slim.fully_connected( model_input, vocab_size * num_mixtures, activation_fn=None, weights_regularizer=slim.l2_regularizer(l2_penalty), scope="experts-"+sub_scope) gating_distribution = tf.nn.softmax(tf.reshape( gate_activations, [-1, num_mixtures + 1])) # (Batch * #Labels) x (num_mixtures + 1) expert_distribution = tf.nn.sigmoid(tf.reshape( expert_activations, [-1, num_mixtures])) # (Batch * #Labels) x num_mixtures final_probabilities_by_class_and_batch = tf.reduce_sum( gating_distribution[:, :num_mixtures] * expert_distribution, 1) final_probabilities = tf.reshape(final_probabilities_by_class_and_batch, [-1, vocab_size]) return final_probabilities ================================================ FILE: youtube-8m-wangheda/all_video_models/multitask_divergence_moe_model.py ================================================ import math import models import tensorflow as tf import utils from tensorflow import flags import tensorflow.contrib.slim as slim FLAGS = flags.FLAGS class MultiTaskDivergenceMoeModel(models.BaseModel): """A softmax over a mixture of logistic models (with L2 regularization).""" def create_model(self, model_input, vocab_size, num_mixtures=None, l2_penalty=1e-8, sub_scope="ddcc", original_input=None, dropout=False, keep_prob=None, noise_level=None, num_frames=None, **unused_params): num_supports = FLAGS.num_supports num_models = FLAGS.divergence_model_count support_predictions = [] for i in xrange(num_models): sub_prediction = self.sub_model(model_input,vocab_size, num_mixtures, l2_penalty, sub_scope+"%d"%i, dropout, keep_prob, noise_level) support_predictions.append(sub_prediction) support_predictions = tf.stack(support_predictions, axis=1) main_predictions = tf.reduce_mean(support_predictions, axis=1) return {"predictions": main_predictions, "support_predictions": support_predictions} def sub_model(self, model_input, vocab_size, num_mixtures=None, l2_penalty=1e-8, sub_scope="", dropout=False, keep_prob=None, noise_level=None, **unused_params): num_mixtures = num_mixtures or FLAGS.moe_num_mixtures if dropout: model_input = tf.nn.dropout(model_input, keep_prob=keep_prob) gate_activations = slim.fully_connected( model_input, vocab_size * (num_mixtures + 1), activation_fn=None, biases_initializer=None, weights_regularizer=slim.l2_regularizer(l2_penalty), scope="gates-"+sub_scope) expert_activations = slim.fully_connected( model_input, vocab_size * num_mixtures, activation_fn=None, weights_regularizer=slim.l2_regularizer(l2_penalty), scope="experts-"+sub_scope) gating_distribution = tf.nn.softmax(tf.reshape( gate_activations, [-1, num_mixtures + 1])) # (Batch * #Labels) x (num_mixtures + 1) expert_distribution = tf.nn.sigmoid(tf.reshape( expert_activations, [-1, num_mixtures])) # (Batch * #Labels) x num_mixtures final_probabilities_by_class_and_batch = tf.reduce_sum( gating_distribution[:, :num_mixtures] * expert_distribution, 1) final_probabilities = tf.reshape(final_probabilities_by_class_and_batch, [-1, vocab_size]) return final_probabilities ================================================ FILE: youtube-8m-wangheda/all_video_models/multitask_moe_model.py ================================================ import math import models import tensorflow as tf import utils from tensorflow import flags import tensorflow.contrib.slim as slim FLAGS = flags.FLAGS class MultiTaskMoeModel(models.BaseModel): """A softmax over a mixture of logistic models (with L2 regularization).""" def create_model(self, model_input, vocab_size, num_mixtures=None, l2_penalty=1e-8, sub_scope="", original_input=None, **unused_params): num_supports = FLAGS.num_supports support_predictions = self.sub_model(model_input, num_supports, sub_scope=sub_scope+"-support") main_predictions = self.sub_model(model_input, vocab_size, sub_scope=sub_scope+"-main") return {"predictions": main_predictions, "support_predictions": support_predictions} def sub_model(self, model_input, vocab_size, num_mixtures=None, l2_penalty=1e-8, sub_scope="", **unused_params): num_mixtures = num_mixtures or FLAGS.moe_num_mixtures gate_activations = slim.fully_connected( model_input, vocab_size * (num_mixtures + 1), activation_fn=None, biases_initializer=None, weights_regularizer=slim.l2_regularizer(l2_penalty), scope="gates"+sub_scope) expert_activations = slim.fully_connected( model_input, vocab_size * num_mixtures, activation_fn=None, weights_regularizer=slim.l2_regularizer(l2_penalty), scope="experts"+sub_scope) gating_distribution = tf.nn.softmax(tf.reshape( gate_activations, [-1, num_mixtures + 1])) # (Batch * #Labels) x (num_mixtures + 1) expert_distribution = tf.nn.sigmoid(tf.reshape( expert_activations, [-1, num_mixtures])) # (Batch * #Labels) x num_mixtures final_probabilities_by_class_and_batch = tf.reduce_sum( gating_distribution[:, :num_mixtures] * expert_distribution, 1) final_probabilities = tf.reshape(final_probabilities_by_class_and_batch, [-1, vocab_size]) return final_probabilities ================================================ FILE: youtube-8m-wangheda/all_video_models/shortcut_chain_support_relu_moe_model.py ================================================ import math import models import tensorflow as tf import utils from tensorflow import flags import tensorflow.contrib.slim as slim FLAGS = flags.FLAGS class ShortcutChainSupportReluMoeModel(models.BaseModel): """A softmax over a mixture of logistic models (with L2 regularization).""" def create_model(self, model_input, vocab_size, num_mixtures=None, l2_penalty=1e-8, sub_scope="", original_input=None, **unused_params): num_supports = FLAGS.num_supports input_size = model_input.shape.as_list()[1] support_input = tf.reduce_mean(original_input, axis=1) support_relu = slim.fully_connected( support_input, input_size, activation_fn=tf.nn.relu, weights_regularizer=slim.l2_regularizer(l2_penalty), scope="main-relu-"+sub_scope) support_predictions = self.sub_model(support_relu, num_supports, sub_scope=sub_scope+"-support") main_input = tf.concat([model_input, support_predictions], axis=1) main_predictions = self.sub_model(main_input, vocab_size, sub_scope=sub_scope+"-main") return {"predictions": main_predictions, "support_predictions": support_predictions} def sub_model(self, model_input, vocab_size, num_mixtures=None, l2_penalty=1e-8, sub_scope="", **unused_params): num_mixtures = num_mixtures or FLAGS.moe_num_mixtures gate_activations = slim.fully_connected( model_input, vocab_size * (num_mixtures + 1), activation_fn=None, biases_initializer=None, weights_regularizer=slim.l2_regularizer(l2_penalty), scope="gates-"+sub_scope) expert_activations = slim.fully_connected( model_input, vocab_size * num_mixtures, activation_fn=None, weights_regularizer=slim.l2_regularizer(l2_penalty), scope="experts-"+sub_scope) gating_distribution = tf.nn.softmax(tf.reshape( gate_activations, [-1, num_mixtures + 1])) # (Batch * #Labels) x (num_mixtures + 1) expert_distribution = tf.nn.sigmoid(tf.reshape( expert_activations, [-1, num_mixtures])) # (Batch * #Labels) x num_mixtures final_probabilities_by_class_and_batch = tf.reduce_sum( gating_distribution[:, :num_mixtures] * expert_distribution, 1) final_probabilities = tf.reshape(final_probabilities_by_class_and_batch, [-1, vocab_size]) return final_probabilities ================================================ FILE: youtube-8m-wangheda/all_video_models/stage2_logistic_model.py ================================================ import math import models import tensorflow as tf import utils from tensorflow import flags import tensorflow.contrib.slim as slim FLAGS = flags.FLAGS class Stage2LogisticModel(models.BaseModel): def create_model(self, model_input, vocab_size, l2_penalty=1e-8, original_input=None, **unused_params): output = tf.nn.sigmoid(model_input + slim.fully_connected( model_input, vocab_size, activation_fn=None, weights_regularizer=slim.l2_regularizer(l2_penalty))) return {"predictions": output} ================================================ FILE: youtube-8m-wangheda/average_precision_calculator.py ================================================ # Copyright 2016 Google Inc. All Rights Reserved. # # Licensed under the Apache License, Version 2.0 (the "License"); # you may not use this file except in compliance with the License. # You may obtain a copy of the License at # # http://www.apache.org/licenses/LICENSE-2.0 # # Unless required by applicable law or agreed to in writing, software # distributed under the License is distributed on an "AS-IS" BASIS, # WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied. # See the License for the specific language governing permissions and # limitations under the License. """Calculate or keep track of the interpolated average precision. It provides an interface for calculating interpolated average precision for an entire list or the top-n ranked items. For the definition of the (non-)interpolated average precision: http://trec.nist.gov/pubs/trec15/appendices/CE.MEASURES06.pdf Example usages: 1) Use it as a static function call to directly calculate average precision for a short ranked list in the memory. ``` import random p = np.array([random.random() for _ in xrange(10)]) a = np.array([random.choice([0, 1]) for _ in xrange(10)]) ap = average_precision_calculator.AveragePrecisionCalculator.ap(p, a) ``` 2) Use it as an object for long ranked list that cannot be stored in memory or the case where partial predictions can be observed at a time (Tensorflow predictions). In this case, we first call the function accumulate many times to process parts of the ranked list. After processing all the parts, we call peek_interpolated_ap_at_n. ``` p1 = np.array([random.random() for _ in xrange(5)]) a1 = np.array([random.choice([0, 1]) for _ in xrange(5)]) p2 = np.array([random.random() for _ in xrange(5)]) a2 = np.array([random.choice([0, 1]) for _ in xrange(5)]) # interpolated average precision at 10 using 1000 break points calculator = average_precision_calculator.AveragePrecisionCalculator(10) calculator.accumulate(p1, a1) calculator.accumulate(p2, a2) ap3 = calculator.peek_ap_at_n() ``` """ import heapq import random import numbers import numpy class AveragePrecisionCalculator(object): """Calculate the average precision and average precision at n.""" def __init__(self, top_n=None): """Construct an AveragePrecisionCalculator to calculate average precision. This class is used to calculate the average precision for a single label. Args: top_n: A positive Integer specifying the average precision at n, or None to use all provided data points. Raises: ValueError: An error occurred when the top_n is not a positive integer. """ if not ((isinstance(top_n, int) and top_n >= 0) or top_n is None): raise ValueError("top_n must be a positive integer or None.") self._top_n = top_n # average precision at n self._total_positives = 0 # total number of positives have seen self._heap = [] # max heap of (prediction, actual) @property def heap_size(self): """Gets the heap size maintained in the class.""" return len(self._heap) @property def num_accumulated_positives(self): """Gets the number of positive samples that have been accumulated.""" return self._total_positives def accumulate(self, predictions, actuals, num_positives=None): """Accumulate the predictions and their ground truth labels. After the function call, we may call peek_ap_at_n to actually calculate the average precision. Note predictions and actuals must have the same shape. Args: predictions: a list storing the prediction scores. actuals: a list storing the ground truth labels. Any value larger than 0 will be treated as positives, otherwise as negatives. num_positives = If the 'predictions' and 'actuals' inputs aren't complete, then it's possible some true positives were missed in them. In that case, you can provide 'num_positives' in order to accurately track recall. Raises: ValueError: An error occurred when the format of the input is not the numpy 1-D array or the shape of predictions and actuals does not match. """ if len(predictions) != len(actuals): raise ValueError("the shape of predictions and actuals does not match.") if not num_positives is None: if not isinstance(num_positives, numbers.Number) or num_positives < 0: raise ValueError("'num_positives' was provided but it wan't a nonzero number.") if not num_positives is None: self._total_positives += num_positives else: self._total_positives += numpy.size(numpy.where(actuals > 0)) topk = self._top_n heap = self._heap for i in range(numpy.size(predictions)): if topk is None or len(heap) < topk: heapq.heappush(heap, (predictions[i], actuals[i])) else: if predictions[i] > heap[0][0]: # heap[0] is the smallest heapq.heappop(heap) heapq.heappush(heap, (predictions[i], actuals[i])) def clear(self): """Clear the accumulated predictions.""" self._heap = [] self._total_positives = 0 def peek_ap_at_n(self): """Peek the non-interpolated average precision at n. Returns: The non-interpolated average precision at n (default 0). If n is larger than the length of the ranked list, the average precision will be returned. """ if self.heap_size <= 0: return 0 predlists = numpy.array(list(zip(*self._heap))) ap = self.ap_at_n(predlists[0], predlists[1], n=self._top_n, total_num_positives=self._total_positives) return ap @staticmethod def ap(predictions, actuals): """Calculate the non-interpolated average precision. Args: predictions: a numpy 1-D array storing the sparse prediction scores. actuals: a numpy 1-D array storing the ground truth labels. Any value larger than 0 will be treated as positives, otherwise as negatives. Returns: The non-interpolated average precision at n. If n is larger than the length of the ranked list, the average precision will be returned. Raises: ValueError: An error occurred when the format of the input is not the numpy 1-D array or the shape of predictions and actuals does not match. """ return AveragePrecisionCalculator.ap_at_n(predictions, actuals, n=None) @staticmethod def ap_at_n(predictions, actuals, n=20, total_num_positives=None): """Calculate the non-interpolated average precision. Args: predictions: a numpy 1-D array storing the sparse prediction scores. actuals: a numpy 1-D array storing the ground truth labels. Any value larger than 0 will be treated as positives, otherwise as negatives. n: the top n items to be considered in ap@n. total_num_positives : (optionally) you can specify the number of total positive in the list. If specified, it will be used in calculation. Returns: The non-interpolated average precision at n. If n is larger than the length of the ranked list, the average precision will be returned. Raises: ValueError: An error occurred when 1) the format of the input is not the numpy 1-D array; 2) the shape of predictions and actuals does not match; 3) the input n is not a positive integer. """ if len(predictions) != len(actuals): raise ValueError("the shape of predictions and actuals does not match.") if n is not None: if not isinstance(n, int) or n <= 0: raise ValueError("n must be 'None' or a positive integer." " It was '%s'." % n) ap = 0.0 predictions = numpy.array(predictions) actuals = numpy.array(actuals) # add a shuffler to avoid overestimating the ap predictions, actuals = AveragePrecisionCalculator._shuffle(predictions, actuals) sortidx = sorted( range(len(predictions)), key=lambda k: predictions[k], reverse=True) if total_num_positives is None: numpos = numpy.size(numpy.where(actuals > 0)) else: numpos = total_num_positives if numpos == 0: return 0 if n is not None: numpos = min(numpos, n) delta_recall = 1.0 / numpos poscount = 0.0 # calculate the ap r = len(sortidx) if n is not None: r = min(r, n) for i in range(r): if actuals[sortidx[i]] > 0: poscount += 1 ap += poscount / (i + 1) * delta_recall return ap @staticmethod def _shuffle(predictions, actuals): random.seed(0) suffidx = random.sample(range(len(predictions)), len(predictions)) predictions = predictions[suffidx] actuals = actuals[suffidx] return predictions, actuals @staticmethod def _zero_one_normalize(predictions, epsilon=1e-7): """Normalize the predictions to the range between 0.0 and 1.0. For some predictions like SVM predictions, we need to normalize them before calculate the interpolated average precision. The normalization will not change the rank in the original list and thus won't change the average precision. Args: predictions: a numpy 1-D array storing the sparse prediction scores. epsilon: a small constant to avoid denominator being zero. Returns: The normalized prediction. """ denominator = numpy.max(predictions) - numpy.min(predictions) ret = (predictions - numpy.min(predictions)) / numpy.max(denominator, epsilon) return ret ================================================ FILE: youtube-8m-wangheda/bagging_scripts/cnn-deep-combine-chain-bagging.sh ================================================ #!/bin/bash DEFAULT_GPU_ID=0 if [ -z ${CUDA_VISIBLE_DEVICES+x} ]; then GPU_ID=$DEFAULT_GPU_ID echo "set CUDA_VISIBLE_DEVICES to default('$GPU_ID')" else GPU_ID=$CUDA_VISIBLE_DEVICES echo "set CUDA_VISIBLE_DEVICES to external('$GPU_ID')" fi # base_model or sub_model_1 or sub_model_2 or so on model_type="$1" model_name="cnn_deep_combine_chain_bagging" MODEL_DIR="../model/${model_name}" vocab_file="resources/train.video_id.vocab" default_freq_file="resources/train.video_id.freq" if [ ! -f $vocab_file ]; then cd resources wget http://us.data.yt8m.org/1/ground_truth_labels/train_labels.csv echo "OOV" > train.video_id.vocab cat train_labels.csv | cut -d ',' -f 1 >> train.video_id.vocab cd .. fi vocab_checksum=$(md5sum $vocab_file | cut -d ' ' -f 1) if [ "$vocab_checksum" == "b74b8f2592cad5dd21bf614d1438db98" ]; then echo $vocab_file is valid else echo $vocab_file is corrupted exit 1 fi if [ ! -f $default_freq_file ]; then cat $vocab_file | awk '{print 1}' > $default_freq_file fi base_model_dir="${MODEL_DIR}/base_model" if [ $model_type == "base_model" ]; then # base model rm ${MODEL_DIR}/ensemble.conf mkdir -p $base_model_dir for j in 1 2; do CUDA_VISIBLE_DEVICES="$GPU_ID" python train.py \ --train_dir="$base_model_dir" \ --train_data_pattern="/Youtube-8M/data/frame/train/train*" \ --frame_features=True \ --feature_names="rgb,audio" \ --feature_sizes="1024,128" \ --reweight=True \ --sample_vocab_file="$vocab_file" \ --sample_freq_file="$default_freq_file" \ --model=CnnDeepCombineChainModel \ --moe_num_mixtures=4 \ --deep_chain_layers=4 \ --deep_chain_relu_cells=128 \ --label_loss=MultiTaskCrossEntropyLoss \ --multitask=True \ --support_type="label,label,label,label" \ --support_loss_percent=0.05 \ --batch_size=128 \ --base_learning_rate=0.001 \ --num_readers=2 \ --num_epochs=2 \ --keep_checkpoint_every_n_hour=72.0 done elif [[ $model_type =~ ^sub_model ]]; then # sub model sub_model_dir="${MODEL_DIR}/${model_type}" if [ ! -d $sub_model_dir ]; then cp -r $base_model_dir $sub_model_dir # generate freq file python training_utils/sample_freq.py \ --video_id_file="$vocab_file" \ --output_freq_file="${sub_model_dir}/train.video_id.freq" # train N models with re-weighted samples CUDA_VISIBLE_DEVICES="$GPU_ID" python train.py \ --train_dir="$sub_model_dir" \ --train_data_pattern="/Youtube-8M/data/frame/train/train*" \ --frame_features=True \ --feature_names="rgb,audio" \ --feature_sizes="1024,128" \ --reweight=True \ --sample_vocab_file="$vocab_file" \ --sample_freq_file="$default_freq_file" \ --model=CnnDeepCombineChainModel \ --moe_num_mixtures=4 \ --deep_chain_layers=4 \ --deep_chain_relu_cells=128 \ --label_loss=MultiTaskCrossEntropyLoss \ --multitask=True \ --support_type="label,label,label,label" \ --support_loss_percent=0.05 \ --batch_size=128 \ --base_learning_rate=0.001 \ --num_readers=2 \ --num_epochs=2 \ --keep_checkpoint_every_n_hour=72.0 fi # inference-pre-ensemble for part in test ensemble_validate ensemble_train; do output_dir="/Youtube-8M/model_predictions/${part}/${model_name}/${model_type}" if [ ! -d $output_dir ]; then CUDA_VISIBLE_DEVICES="$GPU_ID" python inference-pre-ensemble.py \ --output_dir=$output_dir \ --train_dir="${sub_model_dir}" \ --input_data_pattern="/Youtube-8M/data/frame/${part}/*.tfrecord" \ --frame_features=True \ --feature_names="rgb,audio" \ --feature_sizes="1024,128" \ --model=CnnDeepCombineChainModel \ --moe_num_mixtures=4 \ --deep_chain_layers=4 \ --deep_chain_relu_cells=128 \ --batch_size=32 \ --file_size=4096 fi done echo "${model_name}/${model_type}" >> ${MODEL_DIR}/ensemble.conf fi # on ensemble server #cd ../youtube-8m-ensemble #bash ensemble_scripts/eval-mean_model.sh ${model_name}/ensemble_mean_model ${MODEL_DIR}/ensemble.conf #bash ensemble_scripts/infer-mean_model.sh ${model_name}/ensemble_mean_model ${MODEL_DIR}/ensemble.conf ================================================ FILE: youtube-8m-wangheda/bagging_scripts/distillation-video-dcc-bagging.sh ================================================ #!/bin/bash DEFAULT_GPU_ID=0 if [ -z ${CUDA_VISIBLE_DEVICES+x} ]; then GPU_ID=$DEFAULT_GPU_ID echo "set CUDA_VISIBLE_DEVICES to default('$GPU_ID')" else GPU_ID=$CUDA_VISIBLE_DEVICES echo "set CUDA_VISIBLE_DEVICES to external('$GPU_ID')" fi model_name="distillation_video_dcc_bagging" MODEL_DIR="../model/${model_name}" rm ${MODEL_DIR}/ensemble.conf vocab_file="resources/train.video_id.vocab" if [ ! -f $vocab_file ]; then cd resources wget http://us.data.yt8m.org/1/ground_truth_labels/train_labels.csv echo "OOV" > train.video_id.vocab cat train_labels.csv | cut -d ',' -f 1 >> train.video_id.vocab cd .. fi vocab_checksum=$(md5sum $vocab_file | cut -d ' ' -f 1) if [ "$vocab_checksum" == "b74b8f2592cad5dd21bf614d1438db98" ]; then echo $vocab_file is valid else echo $vocab_file is corrupted exit 1 fi if [ ! -f $default_freq_file ]; then cat $vocab_file | awk '{print 1}' > $default_freq_file fi for i in {1..4}; do sub_model_dir="${MODEL_DIR}/sub_model_${i}" if [ ! -d $sub_model_dir ]; then mkdir -p $sub_model_dir # generate freq file if [ ! -f ${sub_model_dir}/train.video_id.freq ]; then python training_utils/sample_freq.py \ --video_id_file="$vocab_file" \ --output_freq_file="${sub_model_dir}/train.video_id.freq" fi # train N models with re-weighted samples CUDA_VISIBLE_DEVICES="$GPU_ID" python train.py \ --train_dir="$sub_model_dir" \ --train_data_pattern="/Youtube-8M/distillation/video/train/*.tfrecord" \ --frame_features=False \ --feature_names="mean_rgb,mean_audio" \ --feature_sizes="1024,128" \ --distillation_features=True \ --distillation_type=1 \ --distillation_percent=0.5 \ --model=DeepCombineChainModel \ --moe_num_mixtures=4 \ --deep_chain_relu_cells=256 \ --deep_chain_layers=4 \ --multitask=True \ --label_loss=MultiTaskCrossEntropyLoss \ --support_type="label,label,label,label" \ --num_supports=18864 \ --support_loss_percent=0.05 \ --reweight=True \ --sample_vocab_file="resources/train.video_id.vocab" \ --sample_freq_file="${sub_model_dir}/train.video_id.freq" \ --keep_checkpoint_every_n_hour=10 \ --keep_checkpoint_interval=6 \ --base_learning_rate=0.01 \ --data_augmenter=NoiseAugmenter \ --input_noise_level=0.1 \ --num_readers=4 \ --num_epochs=6 \ --batch_size=512 fi # inference-pre-ensemble for part in test ensemble_validate ensemble_train; do output_dir="/Youtube-8M/model_predictions/${part}/${model_name}/sub_model_$i" if [ ! -d $output_dir ]; then CUDA_VISIBLE_DEVICES="$GPU_ID" python inference-pre-ensemble.py \ --output_dir="$output_dir" \ --train_dir="${sub_model_dir}" \ --input_data_pattern="/Youtube-8M/data/video/${part}/*.tfrecord" \ --frame_features=False \ --feature_names="mean_rgb,mean_audio" \ --feature_sizes="1024,128" \ --model=DeepCombineChainModel \ --moe_num_mixtures=4 \ --deep_chain_relu_cells=256 \ --deep_chain_layers=4 \ --batch_size=1024 \ --file_size=4096 fi done echo "${model_name}/sub_model_$i" >> ${MODEL_DIR}/ensemble.conf done #cd ../youtube-8m-ensemble #bash ensemble_scripts/eval-mean_model.sh video_bagging/ensemble_mean_model ${MODEL_DIR}/ensemble.conf #bash ensemble_scripts/infer-mean_model.sh video_bagging/ensemble_mean_model ${MODEL_DIR}/ensemble.conf ================================================ FILE: youtube-8m-wangheda/bagging_scripts/lstmattention8max-bagging.sh ================================================ #!/bin/bash DEFAULT_GPU_ID=1 if [ -z ${CUDA_VISIBLE_DEVICES+x} ]; then GPU_ID=$DEFAULT_GPU_ID echo "set CUDA_VISIBLE_DEVICES to default('$GPU_ID')" else GPU_ID=$CUDA_VISIBLE_DEVICES echo "set CUDA_VISIBLE_DEVICES to external('$GPU_ID')" fi # base_model or sub_model_1 or sub_model_2 or so on model_type="$1" model_name="lstmattention8max_bagging" MODEL_DIR="../model/${model_name}" vocab_file="resources/train.video_id.vocab" default_freq_file="resources/train.video_id.freq" if [ ! -f $vocab_file ]; then cd resources wget http://us.data.yt8m.org/1/ground_truth_labels/train_labels.csv echo "OOV" > train.video_id.vocab cat train_labels.csv | cut -d ',' -f 1 >> train.video_id.vocab cd .. fi vocab_checksum=$(md5sum $vocab_file | cut -d ' ' -f 1) if [ "$vocab_checksum" == "b74b8f2592cad5dd21bf614d1438db98" ]; then echo $vocab_file is valid else echo $vocab_file is corrupted exit 1 fi if [ ! -f $default_freq_file ]; then cat $vocab_file | awk '{print 1}' > $default_freq_file fi base_model_dir="${MODEL_DIR}/base_model" if [ $model_type == "base_model" ]; then # base model rm ${MODEL_DIR}/ensemble.conf mkdir -p $base_model_dir CUDA_VISIBLE_DEVICES="$GPU_ID" python train.py \ --train_dir="$base_model_dir" \ --train_data_pattern="/Youtube-8M/data/frame/train/train*" \ --frame_features=True \ --feature_names="rgb,audio" \ --feature_sizes="1024,128" \ --reweight=True \ --sample_vocab_file="$vocab_file" \ --sample_freq_file="$default_freq_file" \ --model=LstmAttentionMaxPoolingModel \ --moe_num_mixtures=8 \ --lstm_attentions=8 \ --rnn_swap_memory=True \ --base_learning_rate=0.001 \ --num_readers=2 \ --num_epochs=5 \ --batch_size=128 \ --keep_checkpoint_every_n_hour=0.5 elif [[ $model_type =~ ^sub_model ]]; then # sub model sub_model_dir="${MODEL_DIR}/${model_type}" if [ ! -d $sub_model_dir ]; then cp -r $base_model_dir $sub_model_dir # generate freq file python training_utils/sample_freq.py \ --video_id_file="$vocab_file" \ --output_freq_file="${sub_model_dir}/train.video_id.freq" # train N models with re-weighted samples CUDA_VISIBLE_DEVICES="$GPU_ID" python train-with-rebuild.py \ --train_dir="$sub_model_dir" \ --train_data_pattern="/Youtube-8M/data/frame/train/train*" \ --frame_features=True \ --feature_names="rgb,audio" \ --feature_sizes="1024,128" \ --reweight=True \ --sample_vocab_file="resources/train.video_id.vocab" \ --sample_freq_file="${sub_model_dir}/train.video_id.freq" \ --model=LstmAttentionMaxPoolingModel \ --moe_num_mixtures=8 \ --lstm_attentions=8 \ --rnn_swap_memory=True \ --base_learning_rate=0.001 \ --num_readers=4 \ --num_epochs=1 \ --batch_size=128 \ --keep_checkpoint_every_n_hour=72.0 fi # inference-pre-ensemble for part in test ensemble_validate ensemble_train; do output_dir="/Youtube-8M/model_predictions/${part}/${model_name}/${model_type}" if [ ! -d $output_dir ]; then CUDA_VISIBLE_DEVICES="$GPU_ID" python inference-pre-ensemble.py \ --output_dir="$output_dir" \ --train_dir="${sub_model_dir}" \ --input_data_pattern="/Youtube-8M/data/frame/${part}/*.tfrecord" \ --frame_features=True \ --feature_names="rgb,audio" \ --feature_sizes="1024,128" \ --model=LstmAttentionMaxPoolingModel \ --moe_num_mixtures=8 \ --lstm_attentions=8 \ --rnn_swap_memory=True \ --batch_size=32 \ --file_size=4096 fi done echo "${model_name}/${model_type}" >> ${MODEL_DIR}/ensemble.conf fi # on ensemble server #cd ../youtube-8m-ensemble #bash ensemble_scripts/eval-mean_model.sh ${model_name}/ensemble_mean_model ${MODEL_DIR}/ensemble.conf #bash ensemble_scripts/infer-mean_model.sh ${model_name}/ensemble_mean_model ${MODEL_DIR}/ensemble.conf ================================================ FILE: youtube-8m-wangheda/bagging_scripts/lstmparalleloutput-bagging.sh ================================================ #!/bin/bash DEFAULT_GPU_ID=1 if [ -z ${CUDA_VISIBLE_DEVICES+x} ]; then GPU_ID=$DEFAULT_GPU_ID echo "set CUDA_VISIBLE_DEVICES to default('$GPU_ID')" else GPU_ID=$CUDA_VISIBLE_DEVICES echo "set CUDA_VISIBLE_DEVICES to external('$GPU_ID')" fi # base_model or sub_model_1 or sub_model_2 or so on model_type="$1" model_name="lstmparalleloutput_bagging" MODEL_DIR="../model/${model_name}" vocab_file="resources/train.video_id.vocab" default_freq_file="resources/train.video_id.freq" if [ ! -f $vocab_file ]; then cd resources wget http://us.data.yt8m.org/1/ground_truth_labels/train_labels.csv echo "OOV" > train.video_id.vocab cat train_labels.csv | cut -d ',' -f 1 >> train.video_id.vocab cd .. fi vocab_checksum=$(md5sum $vocab_file | cut -d ' ' -f 1) if [ "$vocab_checksum" == "b74b8f2592cad5dd21bf614d1438db98" ]; then echo $vocab_file is valid else echo $vocab_file is corrupted exit 1 fi if [ ! -f $default_freq_file ]; then cat $vocab_file | awk '{print 1}' > $default_freq_file fi base_model_dir="${MODEL_DIR}/base_model" if [ $model_type == "base_model" ]; then # base model rm ${MODEL_DIR}/ensemble.conf mkdir -p $base_model_dir CUDA_VISIBLE_DEVICES="$GPU_ID" python train.py \ --train_dir="$base_model_dir" \ --train_data_pattern="/Youtube-8M/data/frame/train/train*" \ --frame_features=True \ --feature_names="rgb,audio" \ --feature_sizes="1024,128" \ --reweight=True \ --sample_vocab_file="$vocab_file" \ --sample_freq_file="$default_freq_file" \ --model=LstmParallelFinaloutputModel \ --lstm_cells="1024,128" \ --moe_num_mixtures=8 \ --rnn_swap_memory=True \ --base_learning_rate=0.0008 \ --num_readers=2 \ --num_epochs=3 \ --batch_size=128 \ --keep_checkpoint_every_n_hour=2.0 elif [[ $model_type =~ ^sub_model ]]; then # sub model sub_model_dir="${MODEL_DIR}/${model_type}" if [ ! -d $sub_model_dir ]; then cp -r $base_model_dir $sub_model_dir # generate freq file python training_utils/sample_freq.py \ --video_id_file="$vocab_file" \ --output_freq_file="${sub_model_dir}/train.video_id.freq" # train N models with re-weighted samples CUDA_VISIBLE_DEVICES="$GPU_ID" python train-with-rebuild.py \ --train_dir="$sub_model_dir" \ --train_data_pattern="/Youtube-8M/data/frame/train/train*" \ --frame_features=True \ --feature_names="rgb,audio" \ --feature_sizes="1024,128" \ --reweight=True \ --sample_vocab_file="resources/train.video_id.vocab" \ --sample_freq_file="${sub_model_dir}/train.video_id.freq" \ --model=LstmParallelFinaloutputModel \ --lstm_cells="1024,128" \ --moe_num_mixtures=8 \ --rnn_swap_memory=True \ --base_learning_rate=0.0008 \ --num_readers=4 \ --num_epochs=1 \ --batch_size=128 \ --keep_checkpoint_every_n_hour=72.0 fi # inference-pre-ensemble for part in test ensemble_validate ensemble_train; do output_dir="/Youtube-8M/model_predictions/${part}/${model_name}/${model_type}" if [ ! -d $output_dir ]; then CUDA_VISIBLE_DEVICES="$GPU_ID" python inference-pre-ensemble.py \ --output_dir=$output_dir \ --train_dir="${sub_model_dir}" \ --input_data_pattern="/Youtube-8M/data/frame/${part}/*.tfrecord" \ --frame_features=True \ --feature_names="rgb,audio" \ --feature_sizes="1024,128" \ --batch_size=32 \ --model=LstmParallelFinaloutputModel \ --lstm_cells="1024,128" \ --moe_num_mixtures=8 \ --rnn_swap_memory=True \ --file_size=4096 fi done echo "${model_name}/${model_type}" >> ${MODEL_DIR}/ensemble.conf fi # on ensemble server #cd ../youtube-8m-ensemble #bash ensemble_scripts/eval-mean_model.sh ${model_name}/ensemble_mean_model ${MODEL_DIR}/ensemble.conf #bash ensemble_scripts/infer-mean_model.sh ${model_name}/ensemble_mean_model ${MODEL_DIR}/ensemble.conf ================================================ FILE: youtube-8m-wangheda/bagging_scripts/video-deep-combine-chain-bagging.sh ================================================ #!/bin/bash DEFAULT_GPU_ID=0 if [ -z ${CUDA_VISIBLE_DEVICES+x} ]; then GPU_ID=$DEFAULT_GPU_ID echo "set CUDA_VISIBLE_DEVICES to default('$GPU_ID')" else GPU_ID=$CUDA_VISIBLE_DEVICES echo "set CUDA_VISIBLE_DEVICES to external('$GPU_ID')" fi MODEL_DIR="../model/video_bagging" rm ${MODEL_DIR}/ensemble.conf vocab_file="resources/train.video_id.vocab" if [ ! -f $vocab_file ]; then cd resources wget http://us.data.yt8m.org/1/ground_truth_labels/train_labels.csv echo "OOV" > train.video_id.vocab cat train_labels.csv | cut -d ',' -f 1 >> train.video_id.vocab cd .. fi vocab_checksum=$(md5sum $vocab_file | cut -d ' ' -f 1) if [ "$vocab_checksum" == "b74b8f2592cad5dd21bf614d1438db98" ]; then echo $vocab_file is valid else echo $vocab_file is corrupted exit 1 fi if [ ! -f $default_freq_file ]; then cat $vocab_file | awk '{print 1}' > $default_freq_file fi for i in {1..8}; do sub_model_dir="${MODEL_DIR}/sub_model_${i}" mkdir -p $sub_model_dir # generate freq file python training_utils/sample_freq.py \ --video_id_file="$vocab_file" \ --output_freq_file="${sub_model_dir}/train.video_id.freq" # train N models with re-weighted samples CUDA_VISIBLE_DEVICES="$GPU_ID" python train.py \ --train_dir="$sub_model_dir" \ --train_data_pattern="/Youtube-8M/data/video/train/train*" \ --frame_features=False \ --feature_names="mean_rgb,mean_audio" \ --feature_sizes="1024,128" \ --model=DeepCombineChainModel \ --moe_num_mixtures=4 \ --deep_chain_relu_cells=256 \ --deep_chain_layers=4 \ --label_loss=MultiTaskCrossEntropyLoss \ --multitask=True \ --support_type="label,label,label,label" \ --num_supports=18864 \ --support_loss_percent=0.05 \ --reweight=True \ --sample_vocab_file="resources/train.video_id.vocab" \ --sample_freq_file="${sub_model_dir}/train.video_id.freq" \ --keep_checkpoint_every_n_hour=0.25 \ --base_learning_rate=0.01 \ --data_augmenter=NoiseAugmenter \ --input_noise_level=0.2 \ --num_readers=2 \ --num_epochs=5 \ --batch_size=1024 # inference-pre-ensemble for part in test ensemble_validate ensemble_train; do CUDA_VISIBLE_DEVICES="$GPU_ID" python inference-pre-ensemble.py \ --output_dir="/Youtube-8M/model_predictions/${part}/video_deep_combine_chain_bagging/sub_model_$i" \ --train_dir="${sub_model_dir}" \ --input_data_pattern="/Youtube-8M/data/video/${part}/*.tfrecord" \ --frame_features=False \ --feature_names="mean_rgb,mean_audio" \ --feature_sizes="1024,128" \ --batch_size=128 \ --file_size=4096 done echo "video_deep_combine_chain_bagging/sub_model_$i" >> ${MODEL_DIR}/ensemble.conf done cd ../youtube-8m-ensemble CUDA_VISIBLE_DEVICES="$GPU_ID" bash ensemble_scripts/eval-mean_model.sh video_bagging/ensemble_mean_model ${MODEL_DIR}/ensemble.conf CUDA_VISIBLE_DEVICES="$GPU_ID" bash ensemble_scripts/infer-mean_model.sh video_bagging/ensemble_mean_model ${MODEL_DIR}/ensemble.conf ================================================ FILE: youtube-8m-wangheda/boosting_scripts/cnn-deep-combine-chain-boosting.sh ================================================ #!/bin/bash DEFAULT_GPU_ID=0 if [ -z ${CUDA_VISIBLE_DEVICES+x} ]; then GPU_ID=$DEFAULT_GPU_ID echo "set CUDA_VISIBLE_DEVICES to default('$GPU_ID')" else GPU_ID=$CUDA_VISIBLE_DEVICES echo "set CUDA_VISIBLE_DEVICES to external('$GPU_ID')" fi # base_model or sub_model_1 or sub_model_2 or so on model_type="$1" model_name="cnn_deep_combine_chain_boosting" MODEL_DIR="../model/${model_name}" vocab_file="resources/train.video_id.vocab" default_freq_file="resources/train.video_id.freq" if [ ! -f $vocab_file ]; then cd resources wget http://us.data.yt8m.org/1/ground_truth_labels/train_labels.csv echo "OOV" > train.video_id.vocab cat train_labels.csv | cut -d ',' -f 1 >> train.video_id.vocab cd .. fi vocab_checksum=$(md5sum $vocab_file | cut -d ' ' -f 1) if [ "$vocab_checksum" == "b74b8f2592cad5dd21bf614d1438db98" ]; then echo $vocab_file is valid else echo $vocab_file is corrupted exit 1 fi if [ ! -f $default_freq_file ]; then cat $vocab_file | awk '{print 1}' > $default_freq_file fi base_model_dir="${MODEL_DIR}/base_model" if [ $model_type == "base_model" ]; then # base model rm ${MODEL_DIR}/ensemble.conf mkdir -p $base_model_dir CUDA_VISIBLE_DEVICES="$GPU_ID" python train.py \ --train_dir="$base_model_dir" \ --train_data_pattern="/Youtube-8M/data/frame/train/train*" \ --frame_features=True \ --feature_names="rgb,audio" \ --feature_sizes="1024,128" \ --reweight=True \ --sample_vocab_file="$vocab_file" \ --sample_freq_file="$default_freq_file" \ --model=CnnDeepCombineChainModel \ --moe_num_mixtures=4 \ --deep_chain_layers=4 \ --deep_chain_relu_cells=128 \ --label_loss=MultiTaskCrossEntropyLoss \ --multitask=True \ --support_type="label,label,label,label" \ --support_loss_percent=0.05 \ --batch_size=128 \ --base_learning_rate=0.001 \ --num_readers=2 \ --num_epochs=4 \ --keep_checkpoint_every_n_hour=72.0 elif [[ $model_type =~ ^sub_model ]]; then last_freq_file=$default_freq_file # sub model for i in {1..8}; do sub_model_dir="${MODEL_DIR}/sub_model_${i}" if [ ! -d $sub_model_dir ]; then cp -r $base_model_dir $sub_model_dir echo "training model #$i, reweighting with $last_freq_file" # train N models with re-weighted samples CUDA_VISIBLE_DEVICES="$GPU_ID" python train-with-rebuild.py \ --train_dir="$sub_model_dir" \ --train_data_pattern="/Youtube-8M/data/frame/train/train*" \ --frame_features=True \ --feature_names="rgb,audio" \ --feature_sizes="1024,128" \ --reweight=True \ --sample_vocab_file="$vocab_file" \ --sample_freq_file="$last_freq_file" \ --model=CnnDeepCombineChainModel \ --moe_num_mixtures=4 \ --deep_chain_layers=4 \ --deep_chain_relu_cells=128 \ --label_loss=MultiTaskCrossEntropyLoss \ --multitask=True \ --support_type="label,label,label,label" \ --support_loss_percent=0.05 \ --batch_size=128 \ --base_learning_rate=0.001 \ --num_readers=2 \ --num_epochs=2 \ --keep_checkpoint_every_n_hour=72.0 fi # get error mapping sample_error_file="${sub_model_dir}/train.video_id.error" if [ ! -f $sample_error_file ]; then CUDA_VISIBLE_DEVICES="$GPU_ID" python inference-sample-error.py \ --output_file="$sample_error_file" \ --train_dir="${sub_model_dir}" \ --input_data_pattern="/Youtube-8M/data/frame/train/train*.tfrecord" \ --frame_features=True \ --feature_names="rgb,audio" \ --feature_sizes="1024,128" \ --model=CnnDeepCombineChainModel \ --moe_num_mixtures=4 \ --deep_chain_layers=4 \ --deep_chain_relu_cells=128 \ --num_readers=1 \ --batch_size=64 fi # generate resample freq file output_freq_file="${sub_model_dir}/train.video_id.next_freq" if [ ! -f $output_freq_file ]; then python training_utils/reweight_sample_freq.py \ --clip_weight=5.0 \ --video_id_file="$vocab_file" \ --input_freq_file="$last_freq_file" \ --input_error_file="${sub_model_dir}/train.video_id.error" \ --output_freq_file="$output_freq_file" fi last_freq_file="${sub_model_dir}/train.video_id.next_freq" echo "${model_name}/sub_model_$i" >> ${MODEL_DIR}/ensemble.conf done elif [[ $model_type =~ ^ensemble ]]; then for i in {1..8}; do sub_model_dir="${MODEL_DIR}/sub_model_${i}" # inference-pre-ensemble for part in test ensemble_validate ensemble_train; do output_dir="/Youtube-8M/model_predictions/${part}/${model_name}/sub_model_${i}" if [ ! -d $output_dir ]; then CUDA_VISIBLE_DEVICES="$GPU_ID" python inference-pre-ensemble.py \ --output_dir="$output_dir" \ --train_dir="${sub_model_dir}" \ --input_data_pattern="/Youtube-8M/data/frame/${part}/*.tfrecord" \ --frame_features=True \ --feature_names="rgb,audio" \ --feature_sizes="1024,128" \ --model=CnnDeepCombineChainModel \ --moe_num_mixtures=4 \ --deep_chain_layers=4 \ --deep_chain_relu_cells=128 \ --batch_size=64 \ --file_size=4096 fi done done # on ensemble server cd ../youtube-8m-ensemble CUDA_VISIBLE_DEVICES="$GPU_ID" bash ensemble_scripts/train-attention_matrix_model.sh ${model_name}/ensemble_attention_matrix_model ${MODEL_DIR}/ensemble.conf CUDA_VISIBLE_DEVICES="$GPU_ID" bash ensemble_scripts/eval-attention_matrix_model.sh ${model_name}/ensemble_attention_matrix_model ${MODEL_DIR}/ensemble.conf #CUDA_VISIBLE_DEVICES="$GPU_ID" bash ensemble_scripts/infer-matrix_model.sh ${model_name}/ensemble_matrix_model ${MODEL_DIR}/ensemble.conf fi ================================================ FILE: youtube-8m-wangheda/boosting_scripts/distillation-cnn-dcc-boosting.sh ================================================ #!/bin/bash DEFAULT_GPU_ID=0 if [ -z ${CUDA_VISIBLE_DEVICES+x} ]; then GPU_ID=$DEFAULT_GPU_ID echo "set CUDA_VISIBLE_DEVICES to default('$GPU_ID')" else GPU_ID=$CUDA_VISIBLE_DEVICES echo "set CUDA_VISIBLE_DEVICES to external('$GPU_ID')" fi # base_model or sub_model_1 or sub_model_2 or so on model_type="$1" model_name="distillation_cnn_dcc_boosting" MODEL="CnnDeepCombineChainModel" MODEL_DIR="../model/${model_name}" vocab_file="resources/train.video_id.vocab" default_freq_file="resources/train.video_id.freq" if [ ! -f $vocab_file ]; then cd resources wget http://us.data.yt8m.org/1/ground_truth_labels/train_labels.csv echo "OOV" > train.video_id.vocab cat train_labels.csv | cut -d ',' -f 1 >> train.video_id.vocab cd .. fi vocab_checksum=$(md5sum $vocab_file | cut -d ' ' -f 1) if [ "$vocab_checksum" == "b74b8f2592cad5dd21bf614d1438db98" ]; then echo $vocab_file is valid else echo $vocab_file is corrupted exit 1 fi if [ ! -f $default_freq_file ]; then cat $vocab_file | awk '{print 1}' > $default_freq_file fi base_model_dir="${MODEL_DIR}/base_model" if [ $model_type == "base_model" ]; then # base model rm ${MODEL_DIR}/ensemble.conf mkdir -p $base_model_dir CUDA_VISIBLE_DEVICES="$GPU_ID" python train.py \ --train_dir="$base_model_dir" \ --train_data_pattern="/Youtube-8M/distillation/frame/train/*.tfrecord" \ --frame_features=True \ --feature_names="rgb,audio" \ --feature_sizes="1024,128" \ --distillation_features=True \ --distillation_type=1 \ --distillation_percent=0.5 \ --reweight=True \ --sample_vocab_file="$vocab_file" \ --sample_freq_file="$default_freq_file" \ --model=$MODEL \ --moe_num_mixtures=4 \ --deep_chain_layers=4 \ --deep_chain_relu_cells=128 \ --label_loss=MultiTaskCrossEntropyLoss \ --multitask=True \ --support_type="label,label,label,label" \ --support_loss_percent=0.05 \ --base_learning_rate=0.001 \ --num_readers=2 \ --num_epochs=4 \ --batch_size=96 \ --keep_checkpoint_every_n_hour=72.0 elif [[ $model_type =~ ^sub_model ]]; then last_freq_file=$default_freq_file # sub model for i in {1..3}; do sub_model_dir="${MODEL_DIR}/sub_model_${i}" if [ ! -d $sub_model_dir ]; then cp -r $base_model_dir $sub_model_dir echo "training model #$i, reweighting with $last_freq_file" # train N models with re-weighted samples CUDA_VISIBLE_DEVICES="$GPU_ID" python train-with-rebuild.py \ --train_dir="$sub_model_dir" \ --train_data_pattern="/Youtube-8M/distillation/frame/train/*.tfrecord" \ --frame_features=True \ --feature_names="rgb,audio" \ --feature_sizes="1024,128" \ --distillation_features=True \ --distillation_type=1 \ --distillation_percent=0.5 \ --reweight=True \ --sample_vocab_file="resources/train.video_id.vocab" \ --sample_freq_file="$last_freq_file" \ --model=$MODEL \ --moe_num_mixtures=4 \ --deep_chain_layers=4 \ --deep_chain_relu_cells=128 \ --label_loss=MultiTaskCrossEntropyLoss \ --multitask=True \ --support_type="label,label,label,label" \ --support_loss_percent=0.05 \ --base_learning_rate=0.001 \ --num_readers=2 \ --num_epochs=2 \ --batch_size=96 \ --keep_checkpoint_every_n_hour=72.0 fi # get error mapping sample_error_file="${sub_model_dir}/train.video_id.error" if [ ! -f $sample_error_file ]; then CUDA_VISIBLE_DEVICES="$GPU_ID" python inference-sample-error.py \ --output_file="$sample_error_file" \ --train_dir="${sub_model_dir}" \ --input_data_pattern="/Youtube-8M/data/frame/train/*.tfrecord" \ --frame_features=True \ --feature_names="rgb,audio" \ --feature_sizes="1024,128" \ --model=$MODEL \ --moe_num_mixtures=4 \ --deep_chain_layers=4 \ --deep_chain_relu_cells=128 \ --batch_size=128 fi # generate resample freq file output_freq_file="${sub_model_dir}/train.video_id.next_freq" if [ ! -f $output_freq_file ]; then python training_utils/reweight_sample_freq.py \ --clip_weight=5.0 \ --video_id_file="$vocab_file" \ --input_freq_file="$last_freq_file" \ --input_error_file="${sub_model_dir}/train.video_id.error" \ --output_freq_file="$output_freq_file" fi last_freq_file="${sub_model_dir}/train.video_id.next_freq" echo "${model_name}/sub_model_$i" >> ${MODEL_DIR}/ensemble.conf done elif [[ $model_type =~ ^ensemble ]]; then for i in {1..4}; do sub_model_dir="${MODEL_DIR}/sub_model_${i}" # inference-pre-ensemble for part in test ensemble_validate ensemble_train; do output_dir="/Youtube-8M/model_predictions_local/${part}/${model_name}/sub_model_${i}" if [ ! -d $output_dir ]; then CUDA_VISIBLE_DEVICES="$GPU_ID" python inference-pre-ensemble.py \ --output_dir="$output_dir" \ --train_dir="${sub_model_dir}" \ --input_data_pattern="/Youtube-8M/data/frame/${part}/*.tfrecord" \ --frame_features=True \ --feature_names="rgb,audio" \ --feature_sizes="1024,128" \ --model=$MODEL \ --moe_num_mixtures=4 \ --deep_chain_layers=4 \ --deep_chain_relu_cells=128 \ --batch_size=64 \ --file_size=4096 fi done done # on ensemble server cd ../youtube-8m-ensemble #CUDA_VISIBLE_DEVICES="$GPU_ID" bash ensemble_scripts/train-matrix_model.sh ${model_name}/ensemble_matrix_model ${MODEL_DIR}/ensemble.conf #CUDA_VISIBLE_DEVICES="$GPU_ID" bash ensemble_scripts/eval-matrix_model.sh ${model_name}/ensemble_matrix_model ${MODEL_DIR}/ensemble.conf #CUDA_VISIBLE_DEVICES="$GPU_ID" bash ensemble_scripts/preensemble-matrix_model.sh ${model_name}/ensemble_matrix_model ${MODEL_DIR}/ensemble.conf fi ================================================ FILE: youtube-8m-wangheda/boosting_scripts/distillation-lstmcnn-dcc-boosting.sh ================================================ #!/bin/bash DEFAULT_GPU_ID=0 if [ -z ${CUDA_VISIBLE_DEVICES+x} ]; then GPU_ID=$DEFAULT_GPU_ID echo "set CUDA_VISIBLE_DEVICES to default('$GPU_ID')" else GPU_ID=$CUDA_VISIBLE_DEVICES echo "set CUDA_VISIBLE_DEVICES to external('$GPU_ID')" fi # base_model or sub_model_1 or sub_model_2 or so on model_type="$1" model_name="distillation_lstmcnn_dcc_boosting" MODEL="LstmCnnDeepCombineChainModel" MODEL_DIR="../model/${model_name}" vocab_file="resources/train.video_id.vocab" default_freq_file="resources/train.video_id.freq" if [ ! -f $vocab_file ]; then cd resources wget http://us.data.yt8m.org/1/ground_truth_labels/train_labels.csv echo "OOV" > train.video_id.vocab cat train_labels.csv | cut -d ',' -f 1 >> train.video_id.vocab cd .. fi vocab_checksum=$(md5sum $vocab_file | cut -d ' ' -f 1) if [ "$vocab_checksum" == "b74b8f2592cad5dd21bf614d1438db98" ]; then echo $vocab_file is valid else echo $vocab_file is corrupted exit 1 fi if [ ! -f $default_freq_file ]; then cat $vocab_file | awk '{print 1}' > $default_freq_file fi base_model_dir="${MODEL_DIR}/base_model" if [ $model_type == "base_model" ]; then # base model rm ${MODEL_DIR}/ensemble.conf mkdir -p $base_model_dir CUDA_VISIBLE_DEVICES="$GPU_ID" python train.py \ --train_dir="$base_model_dir" \ --train_data_pattern="/Youtube-8M/distillation/frame/train/*.tfrecord" \ --frame_features=True \ --feature_names="rgb,audio" \ --feature_sizes="1024,128" \ --distillation_features=True \ --distillation_type=1 \ --distillation_percent=0.5 \ --reweight=True \ --sample_vocab_file="$vocab_file" \ --sample_freq_file="$default_freq_file" \ --model=$MODEL \ --deep_chain_layers=2 \ --deep_chain_relu_cells=128 \ --moe_num_mixtures=4 \ --lstm_layers=1 \ --lstm_cells="1024,128" \ --rnn_swap_memory=True \ --multitask=True \ --label_loss=MultiTaskCrossEntropyLoss \ --support_type="label,label" \ --support_loss_percent=0.07 \ --base_learning_rate=0.001 \ --num_readers=4 \ --num_epochs=4 \ --batch_size=128 \ --keep_checkpoint_every_n_hour=6.0 elif [[ $model_type =~ ^sub_model ]]; then last_freq_file=$default_freq_file # sub model for i in {1..4}; do sub_model_dir="${MODEL_DIR}/sub_model_${i}" if [ ! -d $sub_model_dir ]; then cp -r $base_model_dir $sub_model_dir echo "training model #$i, reweighting with $last_freq_file" # train N models with re-weighted samples CUDA_VISIBLE_DEVICES="$GPU_ID" python train-with-rebuild.py \ --train_dir="$sub_model_dir" \ --train_data_pattern="/Youtube-8M/distillation/frame/train/*.tfrecord" \ --frame_features=True \ --feature_names="rgb,audio" \ --feature_sizes="1024,128" \ --distillation_features=True \ --distillation_type=1 \ --distillation_percent=0.5 \ --reweight=True \ --sample_vocab_file="resources/train.video_id.vocab" \ --sample_freq_file="$last_freq_file" \ --model=$MODEL \ --deep_chain_layers=2 \ --deep_chain_relu_cells=128 \ --moe_num_mixtures=4 \ --lstm_layers=1 \ --lstm_cells="1024,128" \ --rnn_swap_memory=True \ --multitask=True \ --label_loss=MultiTaskCrossEntropyLoss \ --support_type="label,label" \ --support_loss_percent=0.07 \ --base_learning_rate=0.001 \ --num_readers=4 \ --num_epochs=2 \ --batch_size=128 \ --keep_checkpoint_every_n_hour=6.0 fi # get error mapping sample_error_file="${sub_model_dir}/train.video_id.error" if [ ! -f $sample_error_file ]; then CUDA_VISIBLE_DEVICES="$GPU_ID" python inference-sample-error.py \ --output_file="$sample_error_file" \ --train_dir="${sub_model_dir}" \ --input_data_pattern="/Youtube-8M/data/frame/train/*.tfrecord" \ --frame_features=True \ --feature_names="rgb,audio" \ --feature_sizes="1024,128" \ --model=$MODEL \ --deep_chain_layers=2 \ --deep_chain_relu_cells=128 \ --moe_num_mixtures=4 \ --lstm_layers=1 \ --lstm_cells="1024,128" \ --rnn_swap_memory=True \ --batch_size=128 fi # generate resample freq file output_freq_file="${sub_model_dir}/train.video_id.next_freq" if [ ! -f $output_freq_file ]; then python training_utils/reweight_sample_freq.py \ --clip_weight=5.0 \ --video_id_file="$vocab_file" \ --input_freq_file="$last_freq_file" \ --input_error_file="${sub_model_dir}/train.video_id.error" \ --output_freq_file="$output_freq_file" fi last_freq_file="${sub_model_dir}/train.video_id.next_freq" echo "${model_name}/sub_model_$i" >> ${MODEL_DIR}/ensemble.conf done elif [[ $model_type =~ ^ensemble ]]; then for i in {1..4}; do sub_model_dir="${MODEL_DIR}/sub_model_${i}" # inference-pre-ensemble for part in test ensemble_validate ensemble_train; do output_dir="/Youtube-8M/model_predictions_local/${part}/${model_name}/sub_model_${i}" if [ ! -d $output_dir ]; then CUDA_VISIBLE_DEVICES="$GPU_ID" python inference-pre-ensemble.py \ --output_dir="$output_dir" \ --train_dir="${sub_model_dir}" \ --input_data_pattern="/Youtube-8M/data/frame/${part}/*.tfrecord" \ --frame_features=True \ --feature_names="rgb,audio" \ --feature_sizes="1024,128" \ --model=$MODEL \ --deep_chain_layers=2 \ --deep_chain_relu_cells=128 \ --moe_num_mixtures=4 \ --lstm_layers=1 \ --lstm_cells="1024,128" \ --rnn_swap_memory=True \ --batch_size=64 \ --file_size=4096 fi done done # on ensemble server cd ../youtube-8m-ensemble #CUDA_VISIBLE_DEVICES="$GPU_ID" bash ensemble_scripts/train-matrix_model.sh ${model_name}/ensemble_matrix_model ${MODEL_DIR}/ensemble.conf #CUDA_VISIBLE_DEVICES="$GPU_ID" bash ensemble_scripts/eval-matrix_model.sh ${model_name}/ensemble_matrix_model ${MODEL_DIR}/ensemble.conf #CUDA_VISIBLE_DEVICES="$GPU_ID" bash ensemble_scripts/preensemble-matrix_model.sh ${model_name}/ensemble_matrix_model ${MODEL_DIR}/ensemble.conf fi ================================================ FILE: youtube-8m-wangheda/boosting_scripts/distillation-lstmparalleloutput-boosting.sh ================================================ #!/bin/bash DEFAULT_GPU_ID=1 if [ -z ${CUDA_VISIBLE_DEVICES+x} ]; then GPU_ID=$DEFAULT_GPU_ID echo "set CUDA_VISIBLE_DEVICES to default('$GPU_ID')" else GPU_ID=$CUDA_VISIBLE_DEVICES echo "set CUDA_VISIBLE_DEVICES to external('$GPU_ID')" fi # base_model or sub_model_1 or sub_model_2 or so on model_type="$1" model_name="distillation_lstmparalleloutput_boosting" MODEL="LstmParallelFinaloutputModel" MODEL_DIR="../model/${model_name}" vocab_file="resources/train.video_id.vocab" default_freq_file="resources/train.video_id.freq" if [ ! -f $vocab_file ]; then cd resources wget http://us.data.yt8m.org/1/ground_truth_labels/train_labels.csv echo "OOV" > train.video_id.vocab cat train_labels.csv | cut -d ',' -f 1 >> train.video_id.vocab cd .. fi vocab_checksum=$(md5sum $vocab_file | cut -d ' ' -f 1) if [ "$vocab_checksum" == "b74b8f2592cad5dd21bf614d1438db98" ]; then echo $vocab_file is valid else echo $vocab_file is corrupted exit 1 fi if [ ! -f $default_freq_file ]; then cat $vocab_file | awk '{print 1}' > $default_freq_file fi base_model_dir="${MODEL_DIR}/base_model" if [ $model_type == "base_model" ]; then # base model rm ${MODEL_DIR}/ensemble.conf mkdir -p $base_model_dir CUDA_VISIBLE_DEVICES="$GPU_ID" python train.py \ --train_dir="$base_model_dir" \ --train_data_pattern="/Youtube-8M/distillation/frame/train/*.tfrecord" \ --frame_features=True \ --feature_names="rgb,audio" \ --feature_sizes="1024,128" \ --distillation_features=True \ --distillation_type=1 \ --distillation_percent=0.5 \ --reweight=True \ --sample_vocab_file="$vocab_file" \ --sample_freq_file="$default_freq_file" \ --model=$MODEL \ --lstm_cells="1024,128" \ --moe_num_mixtures=8 \ --rnn_swap_memory=True \ --base_learning_rate=0.0008 \ --num_readers=2 \ --num_epochs=2 \ --batch_size=128 \ --keep_checkpoint_every_n_hour=72.0 elif [[ $model_type =~ ^sub_model ]]; then last_freq_file=$default_freq_file # sub model for i in {1..3}; do sub_model_dir="${MODEL_DIR}/sub_model_${i}" if [ ! -d $sub_model_dir ]; then cp -r $base_model_dir $sub_model_dir echo "training model #$i, reweighting with $last_freq_file" # train N models with re-weighted samples CUDA_VISIBLE_DEVICES="$GPU_ID" python train-with-rebuild.py \ --train_dir="$sub_model_dir" \ --train_data_pattern="/Youtube-8M/distillation/frame/train/*.tfrecord" \ --frame_features=True \ --feature_names="rgb,audio" \ --feature_sizes="1024,128" \ --distillation_features=True \ --distillation_type=1 \ --distillation_percent=0.5 \ --reweight=True \ --sample_vocab_file="resources/train.video_id.vocab" \ --sample_freq_file="$last_freq_file" \ --model=$MODEL \ --lstm_cells="1024,128" \ --moe_num_mixtures=8 \ --rnn_swap_memory=True \ --base_learning_rate=0.0008 \ --num_readers=2 \ --num_epochs=2 \ --batch_size=128 \ --keep_checkpoint_every_n_hour=72.0 fi # get error mapping sample_error_file="${sub_model_dir}/train.video_id.error" if [ ! -f $sample_error_file ]; then CUDA_VISIBLE_DEVICES="$GPU_ID" python inference-sample-error.py \ --output_file="$sample_error_file" \ --train_dir="${sub_model_dir}" \ --input_data_pattern="/Youtube-8M/data/frame/train/*.tfrecord" \ --frame_features=True \ --feature_names="rgb,audio" \ --feature_sizes="1024,128" \ --model=$MODEL \ --lstm_cells="1024,128" \ --moe_num_mixtures=8 \ --rnn_swap_memory=True \ --batch_size=128 fi # generate resample freq file output_freq_file="${sub_model_dir}/train.video_id.next_freq" if [ ! -f $output_freq_file ]; then python training_utils/reweight_sample_freq.py \ --clip_weight=5.0 \ --video_id_file="$vocab_file" \ --input_freq_file="$last_freq_file" \ --input_error_file="${sub_model_dir}/train.video_id.error" \ --output_freq_file="$output_freq_file" fi last_freq_file="${sub_model_dir}/train.video_id.next_freq" echo "${model_name}/sub_model_$i" >> ${MODEL_DIR}/ensemble.conf done elif [[ $model_type =~ ^ensemble ]]; then for i in {1..4}; do sub_model_dir="${MODEL_DIR}/sub_model_${i}" # inference-pre-ensemble for part in test ensemble_validate ensemble_train; do output_dir="/Youtube-8M/model_predictions_local/${part}/${model_name}/sub_model_${i}" if [ ! -d $output_dir ]; then CUDA_VISIBLE_DEVICES="$GPU_ID" python inference-pre-ensemble.py \ --output_dir="$output_dir" \ --train_dir="${sub_model_dir}" \ --input_data_pattern="/Youtube-8M/data/frame/${part}/*.tfrecord" \ --frame_features=True \ --feature_names="rgb,audio" \ --feature_sizes="1024,128" \ --model=$MODEL \ --lstm_cells="1024,128" \ --moe_num_mixtures=8 \ --rnn_swap_memory=True \ --batch_size=64 \ --file_size=4096 fi done done # on ensemble server cd ../youtube-8m-ensemble #CUDA_VISIBLE_DEVICES="$GPU_ID" bash ensemble_scripts/train-matrix_model.sh ${model_name}/ensemble_matrix_model ${MODEL_DIR}/ensemble.conf #CUDA_VISIBLE_DEVICES="$GPU_ID" bash ensemble_scripts/eval-matrix_model.sh ${model_name}/ensemble_matrix_model ${MODEL_DIR}/ensemble.conf #CUDA_VISIBLE_DEVICES="$GPU_ID" bash ensemble_scripts/preensemble-matrix_model.sh ${model_name}/ensemble_matrix_model ${MODEL_DIR}/ensemble.conf fi ================================================ FILE: youtube-8m-wangheda/boosting_scripts/distillation-multilstm-dcc-boosting.sh ================================================ #!/bin/bash DEFAULT_GPU_ID=0 if [ -z ${CUDA_VISIBLE_DEVICES+x} ]; then GPU_ID=$DEFAULT_GPU_ID echo "set CUDA_VISIBLE_DEVICES to default('$GPU_ID')" else GPU_ID=$CUDA_VISIBLE_DEVICES echo "set CUDA_VISIBLE_DEVICES to external('$GPU_ID')" fi # base_model or sub_model_1 or sub_model_2 or so on model_type="$1" model_name="distillation_multilstm_dcc_boosting" MODEL_DIR="../model/${model_name}" vocab_file="resources/train.video_id.vocab" default_freq_file="resources/train.video_id.freq" if [ ! -f $vocab_file ]; then cd resources wget http://us.data.yt8m.org/1/ground_truth_labels/train_labels.csv echo "OOV" > train.video_id.vocab cat train_labels.csv | cut -d ',' -f 1 >> train.video_id.vocab cd .. fi vocab_checksum=$(md5sum $vocab_file | cut -d ' ' -f 1) if [ "$vocab_checksum" == "b74b8f2592cad5dd21bf614d1438db98" ]; then echo $vocab_file is valid else echo $vocab_file is corrupted exit 1 fi if [ ! -f $default_freq_file ]; then cat $vocab_file | awk '{print 1}' > $default_freq_file fi base_model_dir="${MODEL_DIR}/base_model" if [ $model_type == "base_model" ]; then # base model rm ${MODEL_DIR}/ensemble.conf mkdir -p $base_model_dir CUDA_VISIBLE_DEVICES="$GPU_ID" python train.py \ --train_dir="$base_model_dir" \ --train_data_pattern="/Youtube-8M/distillation/frame/train/*.tfrecord" \ --frame_features=True \ --feature_names="rgb,audio" \ --feature_sizes="1024,128" \ --distillation_features=True \ --distillation_type=1 \ --distillation_percent=0.5 \ --reweight=True \ --sample_vocab_file="$vocab_file" \ --sample_freq_file="$default_freq_file" \ --model=LstmMemoryDeepChainModel \ --lstm_layers=2 \ --lstm_cells=1024 \ --moe_num_mixtures=4 \ --deep_chain_relu_cells=256 \ --deep_chain_layers=1 \ --rnn_swap_memory=True \ --multitask=True \ --label_loss=MultiTaskCrossEntropyLoss \ --support_loss_percent=0.2 \ --support_type="label" \ --num_supports=4716 \ --base_learning_rate=0.0005 \ --num_readers=2 \ --num_epochs=2 \ --batch_size=128 \ --keep_checkpoint_every_n_hour=72.0 elif [[ $model_type =~ ^sub_model ]]; then last_freq_file=$default_freq_file # sub model for i in {1..3}; do sub_model_dir="${MODEL_DIR}/sub_model_${i}" if [ ! -d $sub_model_dir ]; then cp -r $base_model_dir $sub_model_dir echo "training model #$i, reweighting with $last_freq_file" # train N models with re-weighted samples CUDA_VISIBLE_DEVICES="$GPU_ID" python train-with-rebuild.py \ --train_dir="$sub_model_dir" \ --train_data_pattern="/Youtube-8M/distillation/frame/train/*.tfrecord" \ --frame_features=True \ --feature_names="rgb,audio" \ --feature_sizes="1024,128" \ --distillation_features=True \ --distillation_type=1 \ --distillation_percent=0.5 \ --reweight=True \ --sample_vocab_file="resources/train.video_id.vocab" \ --sample_freq_file="$last_freq_file" \ --model=LstmMemoryDeepChainModel \ --lstm_layers=2 \ --lstm_cells=1024 \ --moe_num_mixtures=4 \ --deep_chain_relu_cells=256 \ --deep_chain_layers=1 \ --rnn_swap_memory=True \ --multitask=True \ --label_loss=MultiTaskCrossEntropyLoss \ --support_type="label" \ --support_loss_percent=0.2 \ --num_supports=4716 \ --base_learning_rate=0.0005 \ --num_readers=2 \ --num_epochs=2 \ --batch_size=128 \ --keep_checkpoint_every_n_hour=72.0 fi # get error mapping sample_error_file="${sub_model_dir}/train.video_id.error" if [ ! -f $sample_error_file ]; then CUDA_VISIBLE_DEVICES="$GPU_ID" python inference-sample-error.py \ --output_file="$sample_error_file" \ --train_dir="${sub_model_dir}" \ --input_data_pattern="/Youtube-8M/data/frame/train/*.tfrecord" \ --frame_features=True \ --feature_names="rgb,audio" \ --feature_sizes="1024,128" \ --model=LstmMemoryDeepChainModel \ --lstm_layers=2 \ --lstm_cells=1024 \ --moe_num_mixtures=4 \ --deep_chain_relu_cells=256 \ --deep_chain_layers=1 \ --rnn_swap_memory=True \ --batch_size=128 fi # generate resample freq file output_freq_file="${sub_model_dir}/train.video_id.next_freq" if [ ! -f $output_freq_file ]; then python training_utils/reweight_sample_freq.py \ --clip_weight=5.0 \ --video_id_file="$vocab_file" \ --input_freq_file="$last_freq_file" \ --input_error_file="${sub_model_dir}/train.video_id.error" \ --output_freq_file="$output_freq_file" fi last_freq_file="${sub_model_dir}/train.video_id.next_freq" echo "${model_name}/sub_model_$i" >> ${MODEL_DIR}/ensemble.conf done elif [[ $model_type =~ ^ensemble ]]; then for i in {1..3}; do sub_model_dir="${MODEL_DIR}/sub_model_${i}" # inference-pre-ensemble for part in test ensemble_validate ensemble_train; do output_dir="/Youtube-8M/model_predictions/${part}/${model_name}/sub_model_${i}" if [ ! -d $output_dir ]; then CUDA_VISIBLE_DEVICES="$GPU_ID" python inference-pre-ensemble.py \ --output_dir="$output_dir" \ --train_dir="${sub_model_dir}" \ --input_data_pattern="/Youtube-8M/data/frame/${part}/*.tfrecord" \ --frame_features=True \ --feature_names="rgb,audio" \ --feature_sizes="1024,128" \ --model=LstmMemoryDeepChainModel \ --lstm_layers=2 \ --lstm_cells=1024 \ --moe_num_mixtures=4 \ --deep_chain_relu_cells=256 \ --deep_chain_layers=1 \ --rnn_swap_memory=True \ --batch_size=64 \ --file_size=4096 fi done done # on ensemble server cd ../youtube-8m-ensemble #CUDA_VISIBLE_DEVICES="$GPU_ID" bash ensemble_scripts/train-matrix_model.sh ${model_name}/ensemble_matrix_model ${MODEL_DIR}/ensemble.conf #CUDA_VISIBLE_DEVICES="$GPU_ID" bash ensemble_scripts/eval-matrix_model.sh ${model_name}/ensemble_matrix_model ${MODEL_DIR}/ensemble.conf #CUDA_VISIBLE_DEVICES="$GPU_ID" bash ensemble_scripts/preensemble-matrix_model.sh ${model_name}/ensemble_matrix_model ${MODEL_DIR}/ensemble.conf fi ================================================ FILE: youtube-8m-wangheda/boosting_scripts/distillation-multiscale-cnn-lstm-boosting.sh ================================================ #!/bin/bash DEFAULT_GPU_ID=0 if [ -z ${CUDA_VISIBLE_DEVICES+x} ]; then GPU_ID=$DEFAULT_GPU_ID echo "set CUDA_VISIBLE_DEVICES to default('$GPU_ID')" else GPU_ID=$CUDA_VISIBLE_DEVICES echo "set CUDA_VISIBLE_DEVICES to external('$GPU_ID')" fi # base_model or sub_model_1 or sub_model_2 or so on model_type="$1" model_name="distillation_multiscale_cnnlstm_boosting" MODEL="MultiscaleCnnLstmModel" MODEL_DIR="../model/${model_name}" vocab_file="resources/train.video_id.vocab" default_freq_file="resources/train.video_id.freq" if [ ! -f $vocab_file ]; then cd resources wget http://us.data.yt8m.org/1/ground_truth_labels/train_labels.csv echo "OOV" > train.video_id.vocab cat train_labels.csv | cut -d ',' -f 1 >> train.video_id.vocab cd .. fi vocab_checksum=$(md5sum $vocab_file | cut -d ' ' -f 1) if [ "$vocab_checksum" == "b74b8f2592cad5dd21bf614d1438db98" ]; then echo $vocab_file is valid else echo $vocab_file is corrupted exit 1 fi if [ ! -f $default_freq_file ]; then cat $vocab_file | awk '{print 1}' > $default_freq_file fi base_model_dir="${MODEL_DIR}/base_model" if [ $model_type == "base_model" ]; then # base model rm ${MODEL_DIR}/ensemble.conf mkdir -p $base_model_dir CUDA_VISIBLE_DEVICES="$GPU_ID" python train.py \ --train_dir="$base_model_dir" \ --train_data_pattern="/Youtube-8M/distillation/frame/train/*.tfrecord" \ --frame_features=True \ --feature_names="rgb,audio" \ --feature_sizes="1024,128" \ --distillation_features=True \ --distillation_type=1 \ --distillation_percent=0.5 \ --reweight=True \ --sample_vocab_file="$vocab_file" \ --sample_freq_file="$default_freq_file" \ --model=$MODEL \ --multiscale_cnn_lstm_layers=4 \ --moe_num_mixtures=4 \ --is_training=True \ --multitask=True \ --label_loss=MultiTaskCrossEntropyLoss \ --support_loss_percent=1.0 \ --support_type="label,label,label,label" \ --base_learning_rate=0.001 \ --num_readers=4 \ --num_epochs=3 \ --batch_size=128 \ --keep_checkpoint_every_n_hour=72.0 elif [[ $model_type =~ ^sub_model ]]; then last_freq_file=$default_freq_file # sub model for i in {1..4}; do sub_model_dir="${MODEL_DIR}/sub_model_${i}" if [ ! -d $sub_model_dir ]; then cp -r $base_model_dir $sub_model_dir echo "training model #$i, reweighting with $last_freq_file" # train N models with re-weighted samples CUDA_VISIBLE_DEVICES="$GPU_ID" python train-with-rebuild.py \ --train_dir="$sub_model_dir" \ --train_data_pattern="/Youtube-8M/distillation/frame/train/*.tfrecord" \ --frame_features=True \ --feature_names="rgb,audio" \ --feature_sizes="1024,128" \ --distillation_features=True \ --distillation_type=1 \ --distillation_percent=0.5 \ --reweight=True \ --sample_vocab_file="resources/train.video_id.vocab" \ --sample_freq_file="$last_freq_file" \ --model=$MODEL \ --multiscale_cnn_lstm_layers=4 \ --moe_num_mixtures=4 \ --is_training=True \ --multitask=True \ --label_loss=MultiTaskCrossEntropyLoss \ --support_loss_percent=1.0 \ --support_type="label,label,label,label" \ --base_learning_rate=0.001 \ --num_readers=4 \ --num_epochs=1 \ --batch_size=128 \ --keep_checkpoint_every_n_hour=72.0 fi # get error mapping sample_error_file="${sub_model_dir}/train.video_id.error" if [ ! -f $sample_error_file ]; then CUDA_VISIBLE_DEVICES="$GPU_ID" python inference-sample-error.py \ --output_file="$sample_error_file" \ --train_dir="${sub_model_dir}" \ --input_data_pattern="/Youtube-8M/data/frame/train/*.tfrecord" \ --frame_features=True \ --feature_names="rgb,audio" \ --feature_sizes="1024,128" \ --model=$MODEL \ --multiscale_cnn_lstm_layers=4 \ --moe_num_mixtures=4 \ --batch_size=128 fi # generate resample freq file output_freq_file="${sub_model_dir}/train.video_id.next_freq" if [ ! -f $output_freq_file ]; then python training_utils/reweight_sample_freq.py \ --clip_weight=5.0 \ --video_id_file="$vocab_file" \ --input_freq_file="$last_freq_file" \ --input_error_file="${sub_model_dir}/train.video_id.error" \ --output_freq_file="$output_freq_file" fi last_freq_file="${sub_model_dir}/train.video_id.next_freq" echo "${model_name}/sub_model_$i" >> ${MODEL_DIR}/ensemble.conf done elif [[ $model_type =~ ^ensemble ]]; then for i in {1..4}; do sub_model_dir="${MODEL_DIR}/sub_model_${i}" # inference-pre-ensemble for part in test ensemble_validate ensemble_train; do output_dir="/Youtube-8M/model_predictions/${part}/${model_name}/sub_model_${i}" if [ ! -d $output_dir ]; then CUDA_VISIBLE_DEVICES="$GPU_ID" python inference-pre-ensemble.py \ --output_dir="$output_dir" \ --train_dir="${sub_model_dir}" \ --input_data_pattern="/Youtube-8M/data/frame/${part}/*.tfrecord" \ --frame_features=True \ --feature_names="rgb,audio" \ --feature_sizes="1024,128" \ --model=$MODEL \ --multiscale_cnn_lstm_layers=4 \ --moe_num_mixtures=4 \ --batch_size=64 \ --file_size=4096 fi done done # on ensemble server cd ../youtube-8m-ensemble #CUDA_VISIBLE_DEVICES="$GPU_ID" bash ensemble_scripts/train-matrix_model.sh ${model_name}/ensemble_matrix_model ${MODEL_DIR}/ensemble.conf #CUDA_VISIBLE_DEVICES="$GPU_ID" bash ensemble_scripts/eval-matrix_model.sh ${model_name}/ensemble_matrix_model ${MODEL_DIR}/ensemble.conf #CUDA_VISIBLE_DEVICES="$GPU_ID" bash ensemble_scripts/preensemble-matrix_model.sh ${model_name}/ensemble_matrix_model ${MODEL_DIR}/ensemble.conf fi ================================================ FILE: youtube-8m-wangheda/boosting_scripts/distillation-positional-lstmattention8max-boosting.sh ================================================ #!/bin/bash DEFAULT_GPU_ID=1 if [ -z ${CUDA_VISIBLE_DEVICES+x} ]; then GPU_ID=$DEFAULT_GPU_ID echo "set CUDA_VISIBLE_DEVICES to default('$GPU_ID')" else GPU_ID=$CUDA_VISIBLE_DEVICES echo "set CUDA_VISIBLE_DEVICES to external('$GPU_ID')" fi # base_model or sub_model_1 or sub_model_2 or so on model_type="$1" model_name="distillation_positional_lstmattention8max_boosting" MODEL="LstmPositionalAttentionMaxPoolingModel" MODEL_DIR="../model/${model_name}" vocab_file="resources/train.video_id.vocab" default_freq_file="resources/train.video_id.freq" if [ ! -f $vocab_file ]; then cd resources wget http://us.data.yt8m.org/1/ground_truth_labels/train_labels.csv echo "OOV" > train.video_id.vocab cat train_labels.csv | cut -d ',' -f 1 >> train.video_id.vocab cd .. fi vocab_checksum=$(md5sum $vocab_file | cut -d ' ' -f 1) if [ "$vocab_checksum" == "b74b8f2592cad5dd21bf614d1438db98" ]; then echo $vocab_file is valid else echo $vocab_file is corrupted exit 1 fi if [ ! -f $default_freq_file ]; then cat $vocab_file | awk '{print 1}' > $default_freq_file fi base_model_dir="${MODEL_DIR}/base_model" if [ $model_type == "base_model" ]; then # base model rm ${MODEL_DIR}/ensemble.conf mkdir -p $base_model_dir CUDA_VISIBLE_DEVICES="$GPU_ID" python train.py \ --train_dir="$base_model_dir" \ --train_data_pattern="/Youtube-8M/distillation/frame/train/*.tfrecord" \ --frame_features=True \ --feature_names="rgb,audio" \ --feature_sizes="1024,128" \ --distillation_features=True \ --distillation_type=1 \ --distillation_percent=0.5 \ --reweight=True \ --sample_vocab_file="$vocab_file" \ --sample_freq_file="$default_freq_file" \ --model=$MODEL \ --moe_num_mixtures=8 \ --lstm_attentions=8 \ --positional_embedding_size=32 \ --rnn_swap_memory=True \ --base_learning_rate=0.001 \ --num_readers=4 \ --num_epochs=3 \ --batch_size=96 \ --keep_checkpoint_every_n_hour=6.0 elif [[ $model_type =~ ^sub_model ]]; then last_freq_file=$default_freq_file # sub model for i in {1..4}; do sub_model_dir="${MODEL_DIR}/sub_model_${i}" if [ ! -d $sub_model_dir ]; then cp -r $base_model_dir $sub_model_dir echo "training model #$i, reweighting with $last_freq_file" # train N models with re-weighted samples CUDA_VISIBLE_DEVICES="$GPU_ID" python train-with-rebuild.py \ --train_dir="$sub_model_dir" \ --train_data_pattern="/Youtube-8M/distillation/frame/train/*.tfrecord" \ --frame_features=True \ --feature_names="rgb,audio" \ --feature_sizes="1024,128" \ --distillation_features=True \ --distillation_type=1 \ --distillation_percent=0.5 \ --reweight=True \ --sample_vocab_file="resources/train.video_id.vocab" \ --sample_freq_file="$last_freq_file" \ --model=$MODEL \ --moe_num_mixtures=8 \ --lstm_attentions=8 \ --positional_embedding_size=32 \ --rnn_swap_memory=True \ --base_learning_rate=0.001 \ --num_readers=2 \ --num_epochs=2 \ --batch_size=96 \ --keep_checkpoint_every_n_hour=6.0 fi # get error mapping sample_error_file="${sub_model_dir}/train.video_id.error" if [ ! -f $sample_error_file ]; then CUDA_VISIBLE_DEVICES="$GPU_ID" python inference-sample-error.py \ --output_file="$sample_error_file" \ --train_dir="${sub_model_dir}" \ --input_data_pattern="/Youtube-8M/data/frame/train/*.tfrecord" \ --frame_features=True \ --feature_names="rgb,audio" \ --feature_sizes="1024,128" \ --model=$MODEL \ --moe_num_mixtures=8 \ --lstm_attentions=8 \ --positional_embedding_size=32 \ --rnn_swap_memory=True \ --batch_size=128 fi # generate resample freq file output_freq_file="${sub_model_dir}/train.video_id.next_freq" if [ ! -f $output_freq_file ]; then python training_utils/reweight_sample_freq.py \ --clip_weight=5.0 \ --video_id_file="$vocab_file" \ --input_freq_file="$last_freq_file" \ --input_error_file="${sub_model_dir}/train.video_id.error" \ --output_freq_file="$output_freq_file" fi last_freq_file="${sub_model_dir}/train.video_id.next_freq" echo "${model_name}/sub_model_$i" >> ${MODEL_DIR}/ensemble.conf done elif [[ $model_type =~ ^ensemble ]]; then for i in {1..4}; do sub_model_dir="${MODEL_DIR}/sub_model_${i}" # inference-pre-ensemble for part in test ensemble_validate ensemble_train; do output_dir="/Youtube-8M/model_predictions/${part}/${model_name}/sub_model_${i}" if [ ! -d $output_dir ]; then CUDA_VISIBLE_DEVICES="$GPU_ID" python inference-pre-ensemble.py \ --output_dir="$output_dir" \ --train_dir="${sub_model_dir}" \ --input_data_pattern="/Youtube-8M/data/frame/${part}/*.tfrecord" \ --frame_features=True \ --feature_names="rgb,audio" \ --feature_sizes="1024,128" \ --model=$MODEL \ --moe_num_mixtures=8 \ --lstm_attentions=8 \ --positional_embedding_size=32 \ --rnn_swap_memory=True \ --batch_size=64 \ --file_size=4096 fi done done # on ensemble server cd ../youtube-8m-ensemble #CUDA_VISIBLE_DEVICES="$GPU_ID" bash ensemble_scripts/train-matrix_model.sh ${model_name}/ensemble_matrix_model ${MODEL_DIR}/ensemble.conf #CUDA_VISIBLE_DEVICES="$GPU_ID" bash ensemble_scripts/eval-matrix_model.sh ${model_name}/ensemble_matrix_model ${MODEL_DIR}/ensemble.conf #CUDA_VISIBLE_DEVICES="$GPU_ID" bash ensemble_scripts/preensemble-matrix_model.sh ${model_name}/ensemble_matrix_model ${MODEL_DIR}/ensemble.conf fi ================================================ FILE: youtube-8m-wangheda/boosting_scripts/distillation-video-dcc-boosting.sh ================================================ #!/bin/bash DEFAULT_GPU_ID=0 if [ -z ${CUDA_VISIBLE_DEVICES+x} ]; then GPU_ID=$DEFAULT_GPU_ID echo "set CUDA_VISIBLE_DEVICES to default('$GPU_ID')" else GPU_ID=$CUDA_VISIBLE_DEVICES echo "set CUDA_VISIBLE_DEVICES to external('$GPU_ID')" fi model_name="distillation_video_dcc_boosting" MODEL="DeepCombineChainModel" MODEL_DIR="../model/${model_name}" vocab_file="resources/train.video_id.vocab" default_freq_file="resources/train.video_id.freq" if [ ! -f $vocab_file ]; then cd resources wget http://us.data.yt8m.org/1/ground_truth_labels/train_labels.csv echo "OOV" > train.video_id.vocab cat train_labels.csv | cut -d ',' -f 1 >> train.video_id.vocab cd .. fi vocab_checksum=$(md5sum $vocab_file | cut -d ' ' -f 1) if [ "$vocab_checksum" == "b74b8f2592cad5dd21bf614d1438db98" ]; then echo $vocab_file is valid else echo $vocab_file is corrupted exit 1 fi if [ ! -f $default_freq_file ]; then cat $vocab_file | awk '{print 1}' > $default_freq_file fi rm ${MODEL_DIR}/ensemble.conf # base model last_freq_file=$default_freq_file # sub model for i in {1..4}; do sub_model_dir="${MODEL_DIR}/sub_model_${i}" if [ ! -d $sub_model_dir ]; then echo "training model #$i, reweighting with $last_freq_file" # train N models with re-weighted samples CUDA_VISIBLE_DEVICES="$GPU_ID" python train-with-rebuild.py \ --train_dir="$sub_model_dir" \ --train_data_pattern="/Youtube-8M/distillation/video/train/*.tfrecord" \ --frame_features=False \ --feature_names="mean_rgb,mean_audio" \ --feature_sizes="1024,128" \ --distillation_features=True \ --distillation_type=1 \ --distillation_percent=0.5 \ --reweight=True \ --sample_vocab_file="resources/train.video_id.vocab" \ --sample_freq_file="$last_freq_file" \ --model=$MODEL \ --moe_num_mixtures=4 \ --deep_chain_relu_cells=256 \ --deep_chain_layers=4 \ --label_loss=MultiTaskCrossEntropyLoss \ --multitask=True \ --support_type="label,label,label,label" \ --support_loss_percent=0.05 \ --base_learning_rate=0.008 \ --num_readers=4 \ --num_epochs=5 \ --batch_size=512 \ --keep_checkpoint_interval=6 \ --keep_checkpoint_every_n_hour=72.0 fi # get error mapping sample_error_file="${sub_model_dir}/train.video_id.error" if [ ! -f $sample_error_file ]; then CUDA_VISIBLE_DEVICES="$GPU_ID" python inference-sample-error.py \ --output_file="$sample_error_file" \ --train_dir="${sub_model_dir}" \ --input_data_pattern="/Youtube-8M/data/video/train/*.tfrecord" \ --frame_features=False \ --feature_names="mean_rgb,mean_audio" \ --feature_sizes="1024,128" \ --model=$MODEL \ --moe_num_mixtures=4 \ --deep_chain_relu_cells=256 \ --deep_chain_layers=4 \ --batch_size=128 fi # generate resample freq file output_freq_file="${sub_model_dir}/train.video_id.next_freq" if [ ! -f $output_freq_file ]; then python training_utils/reweight_sample_freq.py \ --clip_weight=8.0 \ --video_id_file="$vocab_file" \ --input_freq_file="$last_freq_file" \ --input_error_file="${sub_model_dir}/train.video_id.error" \ --output_freq_file="$output_freq_file" fi # inference-pre-ensemble for part in test ensemble_validate ensemble_train; do output_dir="/Youtube-8M/model_predictions_local/${part}/${model_name}/sub_model_${i}" if [ ! -d $output_dir ]; then CUDA_VISIBLE_DEVICES="$GPU_ID" python inference-pre-ensemble.py \ --output_dir="$output_dir" \ --train_dir="${sub_model_dir}" \ --input_data_pattern="/Youtube-8M/data/video/${part}/*.tfrecord" \ --frame_features=False \ --feature_names="mean_rgb,mean_audio" \ --feature_sizes="1024,128" \ --model=$MODEL \ --moe_num_mixtures=4 \ --deep_chain_relu_cells=256 \ --deep_chain_layers=4 \ --batch_size=64 \ --file_size=4096 fi done last_freq_file="${sub_model_dir}/train.video_id.next_freq" echo "${model_name}/sub_model_$i" >> ${MODEL_DIR}/ensemble.conf done # on ensemble server cd ../youtube-8m-ensemble CUDA_VISIBLE_DEVICES="$GPU_ID" bash ensemble_scripts/train-matrix_model.sh ${model_name}/ensemble_matrix_model ${MODEL_DIR}/ensemble.conf _local CUDA_VISIBLE_DEVICES="$GPU_ID" bash ensemble_scripts/eval-matrix_model.sh ${model_name}/ensemble_matrix_model ${MODEL_DIR}/ensemble.conf _local for part in ensemble_train ensemble_validate test; do CUDA_VISIBLE_DEVICES="$GPU_ID" bash ensemble_scripts/auto-preensemble-matrix_model.sh ${model_name}/ensemble_matrix_model ${MODEL_DIR}/ensemble.conf $part _local done ================================================ FILE: youtube-8m-wangheda/boosting_scripts/lstmattention8max-boosting-weightclip.sh ================================================ #!/bin/bash DEFAULT_GPU_ID=1 if [ -z ${CUDA_VISIBLE_DEVICES+x} ]; then GPU_ID=$DEFAULT_GPU_ID echo "set CUDA_VISIBLE_DEVICES to default('$GPU_ID')" else GPU_ID=$CUDA_VISIBLE_DEVICES echo "set CUDA_VISIBLE_DEVICES to external('$GPU_ID')" fi # base_model or sub_model_1 or sub_model_2 or so on model_type="$1" model_name="lstmattention8max_boosting" MODEL_DIR="../model/${model_name}" vocab_file="resources/train.video_id.vocab" default_freq_file="resources/train.video_id.freq" if [ ! -f $vocab_file ]; then cd resources wget http://us.data.yt8m.org/1/ground_truth_labels/train_labels.csv echo "OOV" > train.video_id.vocab cat train_labels.csv | cut -d ',' -f 1 >> train.video_id.vocab cd .. fi vocab_checksum=$(md5sum $vocab_file | cut -d ' ' -f 1) if [ "$vocab_checksum" == "b74b8f2592cad5dd21bf614d1438db98" ]; then echo $vocab_file is valid else echo $vocab_file is corrupted exit 1 fi if [ ! -f $default_freq_file ]; then cat $vocab_file | awk '{print 1}' > $default_freq_file fi base_model_dir="${MODEL_DIR}/base_model" if [ $model_type == "base_model" ]; then # base model rm ${MODEL_DIR}/ensemble.conf mkdir -p $base_model_dir CUDA_VISIBLE_DEVICES="$GPU_ID" python train.py \ --train_dir="$base_model_dir" \ --train_data_pattern="/Youtube-8M/data/frame/train/train*" \ --frame_features=True \ --feature_names="rgb,audio" \ --feature_sizes="1024,128" \ --reweight=True \ --sample_vocab_file="$vocab_file" \ --sample_freq_file="$default_freq_file" \ --model=LstmAttentionMaxPoolingModel \ --moe_num_mixtures=8 \ --lstm_attentions=8 \ --rnn_swap_memory=True \ --base_learning_rate=0.001 \ --num_readers=2 \ --num_epochs=4 \ --batch_size=128 \ --keep_checkpoint_every_n_hour=0.5 elif [[ $model_type =~ ^sub_model ]]; then last_freq_file=$default_freq_file # sub model for i in {1..8}; do sub_model_dir="${MODEL_DIR}/sub_model_${i}" if [ ! -d $sub_model_dir ]; then cp -r $base_model_dir $sub_model_dir echo "training model #$i, reweighting with $last_freq_file" # train N models with re-weighted samples CUDA_VISIBLE_DEVICES="$GPU_ID" python train-with-rebuild.py \ --train_dir="$sub_model_dir" \ --train_data_pattern="/Youtube-8M/data/frame/train/train*" \ --frame_features=True \ --feature_names="rgb,audio" \ --feature_sizes="1024,128" \ --reweight=True \ --sample_vocab_file="resources/train.video_id.vocab" \ --sample_freq_file="$last_freq_file" \ --model=LstmAttentionMaxPoolingModel \ --moe_num_mixtures=8 \ --lstm_attentions=8 \ --rnn_swap_memory=True \ --base_learning_rate=0.001 \ --num_readers=4 \ --num_epochs=1 \ --batch_size=128 \ --keep_checkpoint_every_n_hour=72.0 fi # get error mapping sample_error_file="${sub_model_dir}/train.video_id.error" if [ ! -f $sample_error_file ]; then CUDA_VISIBLE_DEVICES="$GPU_ID" python inference-sample-error.py \ --output_file="$sample_error_file" \ --train_dir="${sub_model_dir}" \ --input_data_pattern="/Youtube-8M/data/frame/train/*.tfrecord" \ --frame_features=True \ --feature_names="rgb,audio" \ --feature_sizes="1024,128" \ --model=LstmAttentionMaxPoolingModel \ --moe_num_mixtures=8 \ --lstm_attentions=8 \ --rnn_swap_memory=True \ --batch_size=128 fi # generate resample freq file output_freq_file="${sub_model_dir}/train.video_id.next_freq" if [ ! -f $output_freq_file ]; then python training_utils/reweight_sample_freq.py \ --clip_weight=5.0 \ --video_id_file="$vocab_file" \ --input_freq_file="$last_freq_file" \ --input_error_file="${sub_model_dir}/train.video_id.error" \ --output_freq_file="$output_freq_file" fi last_freq_file="${sub_model_dir}/train.video_id.next_freq" echo "${model_name}/sub_model_$i" >> ${MODEL_DIR}/ensemble.conf done elif [[ $model_type =~ ^ensemble ]]; then for i in {1..2}; do sub_model_dir="${MODEL_DIR}/sub_model_${i}" # inference-pre-ensemble for part in test ensemble_validate ensemble_train; do CUDA_VISIBLE_DEVICES="$GPU_ID" python inference-pre-ensemble.py \ --output_dir="/Youtube-8M/model_predictions/${part}/${model_name}/sub_model_${i}" \ --train_dir="${sub_model_dir}" \ --input_data_pattern="/Youtube-8M/data/frame/${part}/*.tfrecord" \ --frame_features=True \ --feature_names="rgb,audio" \ --feature_sizes="1024,128" \ --model=LstmAttentionMaxPoolingModel \ --moe_num_mixtures=8 \ --lstm_attentions=8 \ --rnn_swap_memory=True \ --batch_size=32 \ --file_size=4096 done done # on ensemble server #cd ../youtube-8m-ensemble #CUDA_VISIBLE_DEVICES="$GPU_ID" bash ensemble_scripts/train-matrix_model.sh ${model_name}/ensemble_matrix_model ${MODEL_DIR}/ensemble.conf #CUDA_VISIBLE_DEVICES="$GPU_ID" bash ensemble_scripts/eval-matrix_model.sh ${model_name}/ensemble_matrix_model ${MODEL_DIR}/ensemble.conf #CUDA_VISIBLE_DEVICES="$GPU_ID" bash ensemble_scripts/infer-matrix_model.sh ${model_name}/ensemble_matrix_model ${MODEL_DIR}/ensemble.conf fi ================================================ FILE: youtube-8m-wangheda/boosting_scripts/lstmparalleloutput-boosting-weightclip.sh ================================================ #!/bin/bash DEFAULT_GPU_ID=1 if [ -z ${CUDA_VISIBLE_DEVICES+x} ]; then GPU_ID=$DEFAULT_GPU_ID echo "set CUDA_VISIBLE_DEVICES to default('$GPU_ID')" else GPU_ID=$CUDA_VISIBLE_DEVICES echo "set CUDA_VISIBLE_DEVICES to external('$GPU_ID')" fi # base_model or sub_model_1 or sub_model_2 or so on model_type="$1" model_name="lstmparalleloutput_boosting_weightclip" MODEL_DIR="../model/${model_name}" vocab_file="resources/train.video_id.vocab" default_freq_file="resources/train.video_id.freq" if [ ! -f $vocab_file ]; then cd resources wget http://us.data.yt8m.org/1/ground_truth_labels/train_labels.csv echo "OOV" > train.video_id.vocab cat train_labels.csv | cut -d ',' -f 1 >> train.video_id.vocab cd .. fi vocab_checksum=$(md5sum $vocab_file | cut -d ' ' -f 1) if [ "$vocab_checksum" == "b74b8f2592cad5dd21bf614d1438db98" ]; then echo $vocab_file is valid else echo $vocab_file is corrupted exit 1 fi if [ ! -f $default_freq_file ]; then cat $vocab_file | awk '{print 1}' > $default_freq_file fi base_model_dir="${MODEL_DIR}/base_model" if [ $model_type == "base_model" ]; then # base model rm ${MODEL_DIR}/ensemble.conf mkdir -p $base_model_dir CUDA_VISIBLE_DEVICES="$GPU_ID" python train.py \ --train_dir="$base_model_dir" \ --train_data_pattern="/Youtube-8M/data/frame/train/train*" \ --frame_features=True \ --feature_names="rgb,audio" \ --feature_sizes="1024,128" \ --reweight=True \ --sample_vocab_file="$vocab_file" \ --sample_freq_file="$default_freq_file" \ --model=LstmParallelFinaloutputModel \ --lstm_cells="1024,128" \ --moe_num_mixtures=8 \ --rnn_swap_memory=True \ --base_learning_rate=0.0008 \ --num_readers=2 \ --num_epochs=3 \ --batch_size=128 \ --keep_checkpoint_every_n_hour=2.0 elif [[ $model_type =~ ^sub_model ]]; then last_freq_file=$default_freq_file # sub model for i in {1..8}; do sub_model_dir="${MODEL_DIR}/sub_model_${i}" if [ ! -d $sub_model_dir ]; then cp -r $base_model_dir $sub_model_dir echo "training model #$i, reweighting with $last_freq_file" # train N models with re-weighted samples CUDA_VISIBLE_DEVICES="$GPU_ID" python train-with-rebuild.py \ --train_dir="$sub_model_dir" \ --train_data_pattern="/Youtube-8M/data/frame/train/train*" \ --frame_features=True \ --feature_names="rgb,audio" \ --feature_sizes="1024,128" \ --reweight=True \ --sample_vocab_file="resources/train.video_id.vocab" \ --sample_freq_file="$last_freq_file" \ --model=LstmParallelFinaloutputModel \ --lstm_cells="1024,128" \ --moe_num_mixtures=8 \ --rnn_swap_memory=True \ --base_learning_rate=0.0008 \ --num_readers=2 \ --num_epochs=1 \ --batch_size=128 \ --keep_checkpoint_every_n_hour=72.0 fi # get error mapping sample_error_file="${sub_model_dir}/train.video_id.error" if [ ! -f $sample_error_file ]; then CUDA_VISIBLE_DEVICES="$GPU_ID" python inference-sample-error.py \ --output_file="$sample_error_file" \ --train_dir="${sub_model_dir}" \ --input_data_pattern="/Youtube-8M/data/frame/train/*.tfrecord" \ --frame_features=True \ --feature_names="rgb,audio" \ --feature_sizes="1024,128" \ --model=LstmParallelFinaloutputModel \ --lstm_cells="1024,128" \ --moe_num_mixtures=8 \ --rnn_swap_memory=True \ --batch_size=128 fi # generate resample freq file output_freq_file="${sub_model_dir}/train.video_id.next_freq" if [ ! -f $output_freq_file ]; then python training_utils/reweight_sample_freq.py \ --clip_weight=5.0 \ --video_id_file="$vocab_file" \ --input_freq_file="$last_freq_file" \ --input_error_file="${sub_model_dir}/train.video_id.error" \ --output_freq_file="$output_freq_file" fi last_freq_file="${sub_model_dir}/train.video_id.next_freq" echo "${model_name}/sub_model_$i" >> ${MODEL_DIR}/ensemble.conf done elif [[ $model_type =~ ^ensemble ]]; then for i in {1..8}; do sub_model_dir="${MODEL_DIR}/sub_model_${i}" # inference-pre-ensemble for part in test ensemble_validate ensemble_train; do output_dir="/Youtube-8M/model_predictions/${part}/${model_name}/sub_model_${i}" if [ ! -d $output_dir ]; then CUDA_VISIBLE_DEVICES="$GPU_ID" python inference-pre-ensemble.py \ --output_dir="$output_dir" \ --train_dir="${sub_model_dir}" \ --input_data_pattern="/Youtube-8M/data/frame/${part}/*.tfrecord" \ --frame_features=True \ --feature_names="rgb,audio" \ --feature_sizes="1024,128" \ --model=LstmParallelFinaloutputModel \ --lstm_cells="1024,128" \ --moe_num_mixtures=8 \ --rnn_swap_memory=True \ --batch_size=64 \ --file_size=4096 fi done done # on ensemble server #cd ../youtube-8m-ensemble #CUDA_VISIBLE_DEVICES="$GPU_ID" bash ensemble_scripts/train-matrix_model.sh ${model_name}/ensemble_matrix_model ${MODEL_DIR}/ensemble.conf #CUDA_VISIBLE_DEVICES="$GPU_ID" bash ensemble_scripts/eval-matrix_model.sh ${model_name}/ensemble_matrix_model ${MODEL_DIR}/ensemble.conf #CUDA_VISIBLE_DEVICES="$GPU_ID" bash ensemble_scripts/infer-matrix_model.sh ${model_name}/ensemble_matrix_model ${MODEL_DIR}/ensemble.conf fi ================================================ FILE: youtube-8m-wangheda/boosting_scripts/video-deep-combine-chain-boosting-discardhopeless.sh ================================================ #!/bin/bash # base_model or sub_model_1 or sub_model_2 or so on #model_type="$1" model_name="video_dcc_boosting_discardhopeless" MODEL_DIR="../model/${model_name}" vocab_file="resources/train.video_id.vocab" default_freq_file="resources/train.video_id.freq" rm ${MODEL_DIR}/ensemble.conf if [ ! -f $vocab_file ]; then cd resources wget http://us.data.yt8m.org/1/ground_truth_labels/train_labels.csv echo "OOV" > train.video_id.vocab cat train_labels.csv | cut -d ',' -f 1 >> train.video_id.vocab cd .. fi vocab_checksum=$(md5sum $vocab_file | cut -d ' ' -f 1) if [ "$vocab_checksum" == "b74b8f2592cad5dd21bf614d1438db98" ]; then echo $vocab_file is valid else echo $vocab_file is corrupted exit 1 fi if [ ! -f $default_freq_file ]; then cat $vocab_file | awk '{print 1}' > $default_freq_file fi base_model_dir="${MODEL_DIR}/base_model" # base model (4 epochs) if [ ! -d $base_model_dir ]; then mkdir -p $base_model_dir for j in {1..2}; do CUDA_VISIBLE_DEVICES=0 python train.py \ --train_dir="$base_model_dir" \ --train_data_pattern="/Youtube-8M/data/video/train/train*" \ --frame_features=False \ --feature_names="mean_rgb,mean_audio" \ --feature_sizes="1024,128" \ --model=DeepCombineChainModel \ --moe_num_mixtures=4 \ --deep_chain_relu_cells=256 \ --deep_chain_layers=4 \ --label_loss=MultiTaskCrossEntropyLoss \ --multitask=True \ --support_type="label,label,label,label" \ --num_supports=18864 \ --support_loss_percent=0.05 \ --reweight=True \ --sample_vocab_file="$vocab_file" \ --sample_freq_file="$default_freq_file" \ --keep_checkpoint_every_n_hour=8.0 \ --keep_checkpoint_interval=6 \ --base_learning_rate=0.01 \ --data_augmenter=NoiseAugmenter \ --input_noise_level=0.2 \ --num_readers=2 \ --num_epochs=2 \ --batch_size=1024 done fi last_freq_file=$default_freq_file for i in {1..8}; do sub_model_dir="${MODEL_DIR}/sub_model_${i}" if [ ! -d $sub_model_dir ]; then cp -r $base_model_dir $sub_model_dir echo "training model #$i, reweighting with $last_freq_file" # train N models with re-weighted samples CUDA_VISIBLE_DEVICES=0 python train.py \ --train_dir="$sub_model_dir" \ --train_data_pattern="/Youtube-8M/data/video/train/train*" \ --frame_features=False \ --feature_names="mean_rgb,mean_audio" \ --feature_sizes="1024,128" \ --model=DeepCombineChainModel \ --moe_num_mixtures=4 \ --deep_chain_relu_cells=256 \ --deep_chain_layers=4 \ --label_loss=MultiTaskCrossEntropyLoss \ --multitask=True \ --support_type="label,label,label,label" \ --num_supports=18864 \ --support_loss_percent=0.05 \ --reweight=True \ --sample_vocab_file="$vocab_file" \ --sample_freq_file="$last_freq_file" \ --keep_checkpoint_every_n_hour=8.0 \ --base_learning_rate=0.01 \ --data_augmenter=NoiseAugmenter \ --input_noise_level=0.2 \ --num_readers=2 \ --num_epochs=2 \ --batch_size=1024 fi # inference-pre-ensemble for part in test ensemble_validate ensemble_train; do output_dir="/Youtube-8M/model_predictions/${part}/${model_name}/sub_model_$i" if [ ! -d $output_dir ]; then CUDA_VISIBLE_DEVICES=0 python inference-pre-ensemble.py \ --output_dir="${output_dir}" \ --train_dir="${sub_model_dir}" \ --input_data_pattern="/Youtube-8M/data/video/${part}/*.tfrecord" \ --frame_features=False \ --feature_names="mean_rgb,mean_audio" \ --feature_sizes="1024,128" \ --model=DeepCombineChainModel \ --moe_num_mixtures=4 \ --deep_chain_relu_cells=256 \ --deep_chain_layers=4 \ --batch_size=128 \ --file_size=4096 fi done # get error mapping output_error_file="${sub_model_dir}/train.video_id.error" if [ ! -f $output_error_file ]; then echo "generating error mapping to $output_error_file" CUDA_VISIBLE_DEVICES=0 python inference-sample-error.py \ --output_file="${output_error_file}" \ --train_dir="${sub_model_dir}" \ --input_data_pattern="/Youtube-8M/data/video/train/*.tfrecord" \ --frame_features=False \ --feature_names="mean_rgb,mean_audio" \ --feature_sizes="1024,128" \ --model=DeepCombineChainModel \ --moe_num_mixtures=4 \ --deep_chain_relu_cells=256 \ --deep_chain_layers=4 \ --batch_size=1024 fi # generate resample freq file output_freq_file="${sub_model_dir}/train.video_id.next_freq" if [ ! -f $output_freq_file ]; then echo "generating reweight freq to $output_freq_file" python training_utils/reweight_sample_freq.py \ --discard_weight=20.0 \ --video_id_file="$vocab_file" \ --input_freq_file="$last_freq_file" \ --input_error_file="${sub_model_dir}/train.video_id.error" \ --output_freq_file="${sub_model_dir}/train.video_id.next_freq" fi last_freq_file="${sub_model_dir}/train.video_id.next_freq" echo "${model_name}/sub_model_$i" >> ${MODEL_DIR}/ensemble.conf done cd ../youtube-8m-ensemble # partly ensemble for i in 1 2 4; do part_conf="${MODEL_DIR}/ensemble${i}.conf" cat ${MODEL_DIR}/ensemble.conf | head -n $i > $part_conf bash ensemble_scripts/train-matrix_model.sh ${model_name}/ensemble${i}_matrix_model $part_conf bash ensemble_scripts/eval-matrix_model.sh ${model_name}/ensemble${i}_matrix_model $part_conf done # all ensemble bash ensemble_scripts/train-matrix_model.sh ${model_name}/ensemble_matrix_model ${MODEL_DIR}/ensemble.conf bash ensemble_scripts/eval-matrix_model.sh ${model_name}/ensemble_matrix_model ${MODEL_DIR}/ensemble.conf #bash ensemble_scripts/infer-matrix_model.sh ${model_name}/ensemble_matrix_model ${MODEL_DIR}/ensemble.conf ================================================ FILE: youtube-8m-wangheda/boosting_scripts/video-deep-combine-chain-boosting-weightclip.sh ================================================ #!/bin/bash DEFAULT_GPU_ID=0 if [ -z ${CUDA_VISIBLE_DEVICES+x} ]; then GPU_ID=$DEFAULT_GPU_ID echo "set CUDA_VISIBLE_DEVICES to default('$GPU_ID')" else GPU_ID=$CUDA_VISIBLE_DEVICES echo "set CUDA_VISIBLE_DEVICES to external('$GPU_ID')" fi # base_model or sub_model_1 or sub_model_2 or so on #model_type="$1" model_name="video_dcc_boosting_weightclip" MODEL_DIR="../model/${model_name}" vocab_file="resources/train.video_id.vocab" default_freq_file="resources/train.video_id.freq" rm ${MODEL_DIR}/ensemble.conf if [ ! -f $vocab_file ]; then cd resources wget http://us.data.yt8m.org/1/ground_truth_labels/train_labels.csv echo "OOV" > train.video_id.vocab cat train_labels.csv | cut -d ',' -f 1 >> train.video_id.vocab cd .. fi vocab_checksum=$(md5sum $vocab_file | cut -d ' ' -f 1) if [ "$vocab_checksum" == "b74b8f2592cad5dd21bf614d1438db98" ]; then echo $vocab_file is valid else echo $vocab_file is corrupted exit 1 fi if [ ! -f $default_freq_file ]; then cat $vocab_file | awk '{print 1}' > $default_freq_file fi base_model_dir="${MODEL_DIR}/base_model" # base model (4 epochs) if [ ! -d $base_model_dir ]; then mkdir -p $base_model_dir CUDA_VISIBLE_DEVICES="$GPU_ID" python train.py \ --train_dir="$base_model_dir" \ --train_data_pattern="/Youtube-8M/data/video/train/train*" \ --frame_features=False \ --feature_names="mean_rgb,mean_audio" \ --feature_sizes="1024,128" \ --model=DeepCombineChainModel \ --moe_num_mixtures=4 \ --deep_chain_relu_cells=256 \ --deep_chain_layers=4 \ --label_loss=MultiTaskCrossEntropyLoss \ --multitask=True \ --support_type="label,label,label,label" \ --num_supports=18864 \ --support_loss_percent=0.05 \ --reweight=True \ --sample_vocab_file="$vocab_file" \ --sample_freq_file="$default_freq_file" \ --keep_checkpoint_every_n_hour=8.0 \ --keep_checkpoint_interval=6 \ --base_learning_rate=0.01 \ --data_augmenter=NoiseAugmenter \ --input_noise_level=0.2 \ --num_readers=2 \ --num_epochs=4 \ --batch_size=1024 fi last_freq_file=$default_freq_file for i in {1..8}; do sub_model_dir="${MODEL_DIR}/sub_model_${i}" if [ ! -d $sub_model_dir ]; then cp -r $base_model_dir $sub_model_dir echo "training model #$i, reweighting with $last_freq_file" # train N models with re-weighted samples CUDA_VISIBLE_DEVICES="$GPU_ID" python train-with-rebuild.py \ --train_dir="$sub_model_dir" \ --train_data_pattern="/Youtube-8M/data/video/train/train*" \ --frame_features=False \ --feature_names="mean_rgb,mean_audio" \ --feature_sizes="1024,128" \ --model=DeepCombineChainModel \ --moe_num_mixtures=4 \ --deep_chain_relu_cells=256 \ --deep_chain_layers=4 \ --label_loss=MultiTaskCrossEntropyLoss \ --multitask=True \ --support_type="label,label,label,label" \ --num_supports=18864 \ --support_loss_percent=0.05 \ --reweight=True \ --sample_vocab_file="$vocab_file" \ --sample_freq_file="$last_freq_file" \ --keep_checkpoint_every_n_hour=8.0 \ --base_learning_rate=0.01 \ --data_augmenter=NoiseAugmenter \ --input_noise_level=0.2 \ --num_readers=2 \ --num_epochs=2 \ --batch_size=1024 fi # inference-pre-ensemble for part in test ensemble_validate ensemble_train; do output_dir="/Youtube-8M/model_predictions/${part}/${model_name}/sub_model_$i" if [ ! -d $output_dir ]; then CUDA_VISIBLE_DEVICES="$GPU_ID" python inference-pre-ensemble.py \ --output_dir="${output_dir}" \ --train_dir="${sub_model_dir}" \ --input_data_pattern="/Youtube-8M/data/video/${part}/*.tfrecord" \ --frame_features=False \ --feature_names="mean_rgb,mean_audio" \ --feature_sizes="1024,128" \ --model=DeepCombineChainModel \ --moe_num_mixtures=4 \ --deep_chain_relu_cells=256 \ --deep_chain_layers=4 \ --batch_size=128 \ --file_size=4096 fi done # get error mapping output_error_file="${sub_model_dir}/train.video_id.error" if [ ! -f $output_error_file ]; then CUDA_VISIBLE_DEVICES="$GPU_ID" python inference-sample-error.py \ --output_file="${output_error_file}" \ --train_dir="${sub_model_dir}" \ --input_data_pattern="/Youtube-8M/data/video/train/*.tfrecord" \ --frame_features=False \ --feature_names="mean_rgb,mean_audio" \ --feature_sizes="1024,128" \ --model=DeepCombineChainModel \ --moe_num_mixtures=4 \ --deep_chain_relu_cells=256 \ --deep_chain_layers=4 \ --batch_size=1024 fi # generate resample freq file output_freq_file="${sub_model_dir}/train.video_id.next_freq" if [ ! -f $output_freq_file ]; then python training_utils/reweight_sample_freq.py \ --clip_weight=5.0 \ --video_id_file="$vocab_file" \ --input_freq_file="$last_freq_file" \ --input_error_file="${sub_model_dir}/train.video_id.error" \ --output_freq_file="${sub_model_dir}/train.video_id.next_freq" fi last_freq_file="${sub_model_dir}/train.video_id.next_freq" echo "${model_name}/sub_model_$i" >> ${MODEL_DIR}/ensemble.conf done cd ../youtube-8m-ensemble # all ensemble CUDA_VISIBLE_DEVICES="$GPU_ID" bash ensemble_scripts/train-matrix_model.sh ${model_name}/ensemble_matrix_model ${MODEL_DIR}/ensemble.conf CUDA_VISIBLE_DEVICES="$GPU_ID" bash ensemble_scripts/eval-matrix_model.sh ${model_name}/ensemble_matrix_model ${MODEL_DIR}/ensemble.conf #CUDA_VISIBLE_DEVICES="$GPU_ID" bash ensemble_scripts/infer-matrix_model.sh ${model_name}/ensemble_matrix_model ${MODEL_DIR}/ensemble.conf ================================================ FILE: youtube-8m-wangheda/boosting_scripts/video-deep-combine-chain-boosting.sh ================================================ #!/bin/bash DEFAULT_GPU_ID=0 if [ -z ${CUDA_VISIBLE_DEVICES+x} ]; then GPU_ID=$DEFAULT_GPU_ID echo "set CUDA_VISIBLE_DEVICES to default('$GPU_ID')" else GPU_ID=$CUDA_VISIBLE_DEVICES echo "set CUDA_VISIBLE_DEVICES to external('$GPU_ID')" fi # base_model or sub_model_1 or sub_model_2 or so on #model_type="$1" model_name="video_dcc_boosting" MODEL_DIR="../model/${model_name}" vocab_file="resources/train.video_id.vocab" default_freq_file="resources/train.video_id.freq" rm ${MODEL_DIR}/ensemble.conf if [ ! -f $vocab_file ]; then cd resources wget http://us.data.yt8m.org/1/ground_truth_labels/train_labels.csv echo "OOV" > train.video_id.vocab cat train_labels.csv | cut -d ',' -f 1 >> train.video_id.vocab cd .. fi vocab_checksum=$(md5sum $vocab_file | cut -d ' ' -f 1) if [ "$vocab_checksum" == "b74b8f2592cad5dd21bf614d1438db98" ]; then echo $vocab_file is valid else echo $vocab_file is corrupted exit 1 fi if [ ! -f $default_freq_file ]; then cat $vocab_file | awk '{print 1}' > $default_freq_file fi base_model_dir="${MODEL_DIR}/base_model" # base model (4 epochs) mkdir -p $base_model_dir for j in {1..2}; do CUDA_VISIBLE_DEVICES="$GPU_ID" python train.py \ --train_dir="$base_model_dir" \ --train_data_pattern="/Youtube-8M/data/video/train/train*" \ --frame_features=False \ --feature_names="mean_rgb,mean_audio" \ --feature_sizes="1024,128" \ --model=DeepCombineChainModel \ --moe_num_mixtures=4 \ --deep_chain_relu_cells=256 \ --deep_chain_layers=4 \ --label_loss=MultiTaskCrossEntropyLoss \ --multitask=True \ --support_type="label,label,label,label" \ --num_supports=18864 \ --support_loss_percent=0.05 \ --reweight=True \ --sample_vocab_file="$vocab_file" \ --sample_freq_file="$default_freq_file" \ --keep_checkpoint_every_n_hour=8.0 \ --keep_checkpoint_interval=6 \ --base_learning_rate=0.01 \ --data_augmenter=NoiseAugmenter \ --input_noise_level=0.2 \ --num_readers=2 \ --num_epochs=2 \ --batch_size=1024 done last_freq_file=$default_freq_file for i in {1..8}; do sub_model_dir="${MODEL_DIR}/sub_model_${i}" cp -r $base_model_dir $sub_model_dir echo "training model #$i, reweighting with $last_freq_file" # train N models with re-weighted samples CUDA_VISIBLE_DEVICES="$GPU_ID" python train.py \ --train_dir="$sub_model_dir" \ --train_data_pattern="/Youtube-8M/data/video/train/train*" \ --frame_features=False \ --feature_names="mean_rgb,mean_audio" \ --feature_sizes="1024,128" \ --model=DeepCombineChainModel \ --moe_num_mixtures=4 \ --deep_chain_relu_cells=256 \ --deep_chain_layers=4 \ --label_loss=MultiTaskCrossEntropyLoss \ --multitask=True \ --support_type="label,label,label,label" \ --num_supports=18864 \ --support_loss_percent=0.05 \ --reweight=True \ --sample_vocab_file="$vocab_file" \ --sample_freq_file="$last_freq_file" \ --keep_checkpoint_every_n_hour=8.0 \ --base_learning_rate=0.01 \ --data_augmenter=NoiseAugmenter \ --input_noise_level=0.2 \ --num_readers=2 \ --num_epochs=2 \ --batch_size=1024 # inference-pre-ensemble for part in test ensemble_validate ensemble_train; do CUDA_VISIBLE_DEVICES="$GPU_ID" python inference-pre-ensemble.py \ --output_dir="/Youtube-8M/model_predictions/${part}/${model_name}/sub_model_$i" \ --train_dir="${sub_model_dir}" \ --input_data_pattern="/Youtube-8M/data/video/${part}/*.tfrecord" \ --frame_features=False \ --feature_names="mean_rgb,mean_audio" \ --feature_sizes="1024,128" \ --batch_size=128 \ --file_size=4096 done # get error mapping CUDA_VISIBLE_DEVICES="$GPU_ID" python inference-sample-error.py \ --output_file="${sub_model_dir}/train.video_id.error" \ --train_dir="${sub_model_dir}" \ --input_data_pattern="/Youtube-8M/data/video/train/*.tfrecord" \ --frame_features=False \ --feature_names="mean_rgb,mean_audio" \ --feature_sizes="1024,128" \ --model=DeepCombineChainModel \ --moe_num_mixtures=4 \ --deep_chain_relu_cells=256 \ --deep_chain_layers=4 \ --batch_size=1024 # generate resample freq file python training_utils/reweight_sample_freq.py \ --video_id_file="$vocab_file" \ --input_freq_file="$last_freq_file" \ --input_error_file="${sub_model_dir}/train.video_id.error" \ --output_freq_file="${sub_model_dir}/train.video_id.next_freq" last_freq_file="${sub_model_dir}/train.video_id.next_freq" echo "${model_name}/sub_model_$i" >> ${MODEL_DIR}/ensemble.conf done cd ../youtube-8m-ensemble CUDA_VISIBLE_DEVICES="$GPU_ID" bash ensemble_scripts/train-matrix_model.sh ${model_name}/ensemble_matrix_model ${MODEL_DIR}/ensemble.conf CUDA_VISIBLE_DEVICES="$GPU_ID" bash ensemble_scripts/eval-matrix_model.sh ${model_name}/ensemble_matrix_model ${MODEL_DIR}/ensemble.conf #CUDA_VISIBLE_DEVICES="$GPU_ID" bash ensemble_scripts/infer-matrix_model.sh ${model_name}/ensemble_matrix_model ${MODEL_DIR}/ensemble.conf ================================================ FILE: youtube-8m-wangheda/cascade_scripts/distillchain-v2-hybridchain.sh ================================================ #!/bin/bash task=$1 model_name="hybridchain" model_path="../model/${model_name}" if [ $task == "train" ]; then predictions_data_pattern="/Youtube-8M/model_predictions/train/distillation/ensemble_v2_matrix_model/*.tfrecord" # train model 1: Video-DCC sub_model_name="distillchain_v2_video_dcc" sub_model_path="${model_path}/${sub_model_name}" if [ ! -d $sub_model_path ]; then mkdir -p $sub_model_path echo predictions = "$predictions_data_pattern" CUDA_VISIBLE_DEVICES=0 python train-with-predictions.py \ --train_dir="$sub_model_path" \ --train_data_pattern="/Youtube-8M/data/video/train/*.tfrecord" \ --predictions_data_pattern="$predictions_data_pattern" \ --distillation_features=False \ --distillation_as_input=True \ --frame_features=False \ --feature_names="mean_rgb,mean_audio" \ --feature_sizes="1024,128" \ --model=DistillchainDeepCombineChainModel \ --moe_num_mixtures=4 \ --deep_chain_layers=3 \ --deep_chain_relu_cells=256 \ --data_augmenter=NoiseAugmenter \ --input_noise_level=0.1 \ --multitask=True \ --label_loss=MultiTaskCrossEntropyLoss \ --support_type="label,label,label" \ --support_loss_percent=0.05 \ --base_learning_rate=0.01 \ --keep_checkpoint_every_n_hour=5.0 \ --keep_checkpoint_interval=6 \ --num_epochs=2 \ --batch_size=1024 fi # inference model 1 infer_data_path="/Youtube-8M/model_predictions/train/${model_name}/${sub_model_name}" if [ ! -d $infer_data_path ]; then echo predictions = "$predictions_data_pattern" CUDA_VISIBLE_DEVICES=0 python inference-pre-ensemble-with-predictions.py \ --output_dir="$infer_data_path" \ --train_dir="$sub_model_path" \ --input_data_pattern="/Youtube-8M/data/video/train/*.tfrecord" \ --distill_data_pattern="$predictions_data_pattern" \ --frame_features=False \ --feature_names="mean_rgb,mean_audio" \ --feature_sizes="1024,128" \ --distillation_features=False \ --distillation_as_input=True \ --model=DistillchainDeepCombineChainModel \ --moe_num_mixtures=4 \ --deep_chain_layers=3 \ --deep_chain_relu_cells=256 \ --batch_size=1024 \ --file_size=4096 fi # next predictions pattern predictions_data_pattern="${predictions_data_pattern},${infer_data_path}/*.tfrecord" # train model 2: CNN-DCC sub_model_name="distillchain_v2_cnn_dcc" sub_model_path="${model_path}/${sub_model_name}" if [ ! -d $sub_model_path ]; then mkdir -p $sub_model_path echo predictions = "$predictions_data_pattern" CUDA_VISIBLE_DEVICES=0 python train-with-predictions.py \ --predictions_data_pattern="$predictions_data_pattern" \ --distillation_features=True \ --distillation_as_input=True \ --train_dir="$sub_model_path" \ --train_data_pattern="/Youtube-8M/data/frame/train/*.tfrecord" \ --frame_features=True \ --feature_names="rgb,audio" \ --feature_sizes="1024,128" \ --model=DistillchainCnnDeepCombineChainModel \ --moe_num_mixtures=4 \ --deep_chain_layers=3 \ --deep_chain_relu_cells=256 \ --multitask=True \ --label_loss=MultiTaskCrossEntropyLoss \ --support_type="label,label,label" \ --support_loss_percent=0.05 \ --num_readers=4 \ --batch_size=128 \ --num_epochs=2 \ --keep_checkpoint_every_n_hours=20 \ --base_learning_rate=0.001 fi # inference model 2 infer_data_path="/Youtube-8M/model_predictions/train/${model_name}/${sub_model_name}" if [ ! -d $infer_data_path ]; then CUDA_VISIBLE_DEVICES=0 python inference-pre-ensemble-with-predictions.py \ --output_dir="$infer_data_path" \ --train_dir="$sub_model_path" \ --input_data_pattern="/Youtube-8M/data/frame/train/*.tfrecord" \ --frame_features=True \ --feature_names="rgb,audio" \ --feature_sizes="1024,128" \ --distill_data_pattern="$predictions_data_pattern" \ --distillation_features=True \ --distillation_as_input=True \ --model=DistillchainCnnDeepCombineChainModel \ --moe_num_mixtures=4 \ --deep_chain_layers=3 \ --deep_chain_relu_cells=256 \ --batch_size=128 \ --file_size=4096 fi # next predictions pattern predictions_data_pattern="${predictions_data_pattern},${infer_data_path}/*.tfrecord" # train model 2: LSTM sub_model_name="distillchain_v2_lstmparalleloutput" sub_model_path="${model_path}/${sub_model_name}" if [ ! -d $sub_model_path ]; then mkdir -p $sub_model_path echo predictions = "$predictions_data_pattern" CUDA_VISIBLE_DEVICES=0 python train-with-predictions.py \ --predictions_data_pattern="$predictions_data_pattern" \ --distillation_features=True \ --distillation_as_input=True \ --train_dir="$sub_model_path" \ --train_data_pattern="/Youtube-8M/data/frame/train/*.tfrecord" \ --frame_features=True \ --feature_names="rgb,audio" \ --feature_sizes="1024,128" \ --model=DistillchainLstmParallelFinaloutputModel \ --rnn_swap_memory=True \ --lstm_layers=1 \ --moe_num_mixtures=8 \ --lstm_cells="1024,128" \ --batch_size=128 \ --num_epochs=1 \ --keep_checkpoint_every_n_hours=20 \ --base_learning_rate=0.0008 fi elif [ $task == "test-lstm" ]; then sub_model_name="distillchain_v2_lstmparalleloutput" sub_model_path="${model_path}/${sub_model_name}" if [ -d $sub_model_path ]; then for part in ensemble_train ensemble_validate test; do infer_data_path="/Youtube-8M/model_predictions/${part}/${model_name}/${sub_model_name}" if [ ! -d $infer_data_path ]; then predictions_data_pattern="/Youtube-8M/model_predictions/${part}/distillation/ensemble_v2_matrix_model/*.tfrecord" for prev_model in distillchain_v2_video_dcc distillchain_v2_cnn_dcc; do predictions_data_pattern="${predictions_data_pattern},/Youtube-8M/model_predictions/${part}/${model_name}/${prev_model}/*.tfrecord" done echo predictions = "$predictions_data_pattern" CUDA_VISIBLE_DEVICES=1 python inference-pre-ensemble-with-predictions.py \ --output_dir="$infer_data_path" \ --train_dir="$sub_model_path" \ --input_data_pattern="/Youtube-8M/data/frame/${part}/*.tfrecord" \ --frame_features=True \ --feature_names="rgb,audio" \ --feature_sizes="1024,128" \ --distill_data_pattern="$predictions_data_pattern" \ --distillation_features=True \ --distillation_as_input=True \ --model=DistillchainLstmParallelFinaloutputModel \ --rnn_swap_memory=True \ --lstm_layers=1 \ --moe_num_mixtures=8 \ --lstm_cells="1024,128" \ --batch_size=128 \ --file_size=4096 fi done fi elif [ $task == "test-cnn" ]; then sub_model_name="distillchain_v2_cnn_dcc" sub_model_path="${model_path}/${sub_model_name}" if [ -d $sub_model_path ]; then for part in ensemble_train ensemble_validate test; do infer_data_path="/Youtube-8M/model_predictions/${part}/${model_name}/${sub_model_name}" if [ ! -d $infer_data_path ]; then predictions_data_pattern="/Youtube-8M/model_predictions/${part}/distillation/ensemble_v2_matrix_model/*.tfrecord" for prev_model in distillchain_v2_video_dcc; do predictions_data_pattern="${predictions_data_pattern},/Youtube-8M/model_predictions/${part}/${model_name}/${prev_model}/*.tfrecord" done echo predictions = "$predictions_data_pattern" CUDA_VISIBLE_DEVICES=1 python inference-pre-ensemble-with-predictions.py \ --output_dir="$infer_data_path" \ --train_dir="$sub_model_path" \ --input_data_pattern="/Youtube-8M/data/frame/${part}/*.tfrecord" \ --frame_features=True \ --feature_names="rgb,audio" \ --feature_sizes="1024,128" \ --distill_data_pattern="$predictions_data_pattern" \ --distillation_features=True \ --distillation_as_input=True \ --model=DistillchainCnnDeepCombineChainModel \ --moe_num_mixtures=4 \ --deep_chain_layers=3 \ --deep_chain_relu_cells=256 \ --batch_size=128 \ --file_size=4096 fi done fi elif [ $task == "test-video" ]; then sub_model_name="distillchain_v2_video_dcc" sub_model_path="${model_path}/${sub_model_name}" if [ -d $sub_model_path ]; then for part in ensemble_train ensemble_validate test; do infer_data_path="/Youtube-8M/model_predictions/${part}/${model_name}/${sub_model_name}" if [ ! -d $infer_data_path ]; then predictions_data_pattern="/Youtube-8M/model_predictions/${part}/distillation/ensemble_v2_matrix_model/*.tfrecord" echo predictions = "$predictions_data_pattern" CUDA_VISIBLE_DEVICES=1 python inference-pre-ensemble-with-predictions.py \ --output_dir="$infer_data_path" \ --train_dir="$sub_model_path" \ --input_data_pattern="/Youtube-8M/data/video/${part}/*.tfrecord" \ --distill_data_pattern="$predictions_data_pattern" \ --frame_features=False \ --feature_names="mean_rgb,mean_audio" \ --feature_sizes="1024,128" \ --distillation_features=False \ --distillation_as_input=True \ --model=DistillchainDeepCombineChainModel \ --moe_num_mixtures=4 \ --deep_chain_layers=3 \ --deep_chain_relu_cells=256 \ --batch_size=1024 \ --file_size=4096 fi done fi fi ================================================ FILE: youtube-8m-wangheda/cascade_scripts/distillchain-v2-hybridchain2.sh ================================================ #!/bin/bash task=$1 model_name="hybridchain2" model_path="../model/${model_name}" if [ $task == "train" ]; then predictions_data_pattern="/Youtube-8M/model_predictions_x32/train/distillation/ensemble_v2_matrix_model/*.tfrecord" # train model 1: LSTM sub_model_name="distillchain_v2_lstmparalleloutput" sub_model_path="${model_path}/${sub_model_name}" if [ ! -d $sub_model_path ]; then mkdir -p $sub_model_path echo predictions = "$predictions_data_pattern" CUDA_VISIBLE_DEVICES=0 python train-with-predictions.py \ --predictions_data_pattern="$predictions_data_pattern" \ --distillation_features=False \ --distillation_as_input=True \ --train_dir="$sub_model_path" \ --train_data_pattern="/Youtube-8M/data/frame/train/*.tfrecord" \ --frame_features=True \ --feature_names="rgb,audio" \ --feature_sizes="1024,128" \ --model=DistillchainLstmParallelFinaloutputModel \ --rnn_swap_memory=True \ --lstm_layers=1 \ --moe_num_mixtures=8 \ --lstm_cells="1024,128" \ --batch_size=128 \ --num_epochs=1 \ --keep_checkpoint_every_n_hours=20 \ --base_learning_rate=0.0008 fi # inference model 2 infer_data_path="/Youtube-8M/model_predictions_local/train/${model_name}/${sub_model_name}" if [ ! -d $infer_data_path ]; then CUDA_VISIBLE_DEVICES=0 python inference-pre-ensemble-with-predictions.py \ --output_dir="$infer_data_path" \ --train_dir="$sub_model_path" \ --input_data_pattern="/Youtube-8M/data/frame/train/*.tfrecord" \ --frame_features=True \ --feature_names="rgb,audio" \ --feature_sizes="1024,128" \ --distill_data_pattern="$predictions_data_pattern" \ --distillation_features=False \ --distillation_as_input=True \ --model=DistillchainLstmParallelFinaloutputModel \ --rnn_swap_memory=True \ --lstm_layers=1 \ --moe_num_mixtures=8 \ --lstm_cells="1024,128" \ --batch_size=128 \ --file_size=4096 fi # next predictions pattern predictions_data_pattern="${predictions_data_pattern},${infer_data_path}/*.tfrecord" # train model 2: CNN-DCC sub_model_name="distillchain_v2_cnn_dcc" sub_model_path="${model_path}/${sub_model_name}" if [ ! -d $sub_model_path ]; then mkdir -p $sub_model_path echo predictions = "$predictions_data_pattern" CUDA_VISIBLE_DEVICES=0 python train-with-predictions.py \ --predictions_data_pattern="$predictions_data_pattern" \ --distillation_features=False \ --distillation_as_input=True \ --train_dir="$sub_model_path" \ --train_data_pattern="/Youtube-8M/data/frame/train/*.tfrecord" \ --frame_features=True \ --feature_names="rgb,audio" \ --feature_sizes="1024,128" \ --model=DistillchainCnnDeepCombineChainModel \ --moe_num_mixtures=4 \ --deep_chain_layers=3 \ --deep_chain_relu_cells=256 \ --multitask=True \ --label_loss=MultiTaskCrossEntropyLoss \ --support_type="label,label,label" \ --support_loss_percent=0.05 \ --num_readers=4 \ --batch_size=128 \ --num_epochs=1 \ --keep_checkpoint_every_n_hours=20 \ --base_learning_rate=0.001 fi elif [ $task == "test-lstm" ]; then sub_model_name="distillchain_v2_lstmparalleloutput" sub_model_path="${model_path}/${sub_model_name}" if [ -d $sub_model_path ]; then for part in ensemble_train ensemble_validate test; do infer_data_path="/Youtube-8M/model_predictions/${part}/${model_name}/${sub_model_name}" if [ ! -d $infer_data_path ]; then predictions_data_pattern="/Youtube-8M/model_predictions/${part}/distillation/ensemble_v2_matrix_model/*.tfrecord" echo predictions = "$predictions_data_pattern" CUDA_VISIBLE_DEVICES=1 python inference-pre-ensemble-with-predictions.py \ --output_dir="$infer_data_path" \ --train_dir="$sub_model_path" \ --input_data_pattern="/Youtube-8M/data/frame/${part}/*.tfrecord" \ --frame_features=True \ --feature_names="rgb,audio" \ --feature_sizes="1024,128" \ --distill_data_pattern="$predictions_data_pattern" \ --distillation_features=False \ --distillation_as_input=True \ --model=DistillchainLstmParallelFinaloutputModel \ --rnn_swap_memory=True \ --lstm_layers=1 \ --moe_num_mixtures=8 \ --lstm_cells="1024,128" \ --batch_size=32 \ --file_size=4096 fi done fi elif [ $task == "test-cnn" ]; then sub_model_name="distillchain_v2_cnn_dcc" sub_model_path="${model_path}/${sub_model_name}" if [ -d $sub_model_path ]; then for part in ensemble_train ensemble_validate test; do infer_data_path="/Youtube-8M/model_predictions/${part}/${model_name}/${sub_model_name}" if [ ! -d $infer_data_path ]; then predictions_data_pattern="/Youtube-8M/model_predictions/${part}/distillation/ensemble_v2_matrix_model/*.tfrecord" for prev_model in distillchain_v2_lstmparalleloutput; do predictions_data_pattern="${predictions_data_pattern},/Youtube-8M/model_predictions_local/${part}/${model_name}/${prev_model}/*.tfrecord" done echo predictions = "$predictions_data_pattern" CUDA_VISIBLE_DEVICES=0 python inference-pre-ensemble-with-predictions.py \ --output_dir="$infer_data_path" \ --train_dir="$sub_model_path" \ --input_data_pattern="/Youtube-8M/data/frame/${part}/*.tfrecord" \ --frame_features=True \ --feature_names="rgb,audio" \ --feature_sizes="1024,128" \ --distill_data_pattern="$predictions_data_pattern" \ --distillation_features=False \ --distillation_as_input=True \ --model=DistillchainCnnDeepCombineChainModel \ --moe_num_mixtures=4 \ --deep_chain_layers=3 \ --deep_chain_relu_cells=256 \ --batch_size=128 \ --file_size=4096 fi done fi fi ================================================ FILE: youtube-8m-wangheda/cascade_scripts/distillchain-v2-videochain.sh ================================================ #!/bin/bash task=$1 model_name="videochain" model_path="../model/${model_name}" if [ $task == "train" ]; then predictions_data_pattern='/Youtube-8M/model_predictions_local/train/distillation/ensemble_v2_matrix_model/*.tfrecord' for i in {1..5}; do sub_model_name="distillchain_v2_video_dcc_$i" sub_model_path="${model_path}/${sub_model_name}" if [ ! -d $sub_model_path ]; then mkdir -p $sub_model_path echo predictions = "$predictions_data_pattern" CUDA_VISIBLE_DEVICES=0 python train-with-predictions.py \ --train_dir="$sub_model_path" \ --train_data_pattern="/Youtube-8M/data/video/train/*.tfrecord" \ --predictions_data_pattern="$predictions_data_pattern" \ --distillation_features=False \ --distillation_as_input=True \ --frame_features=False \ --feature_names="mean_rgb,mean_audio" \ --feature_sizes="1024,128" \ --model=DistillchainDeepCombineChainModel \ --moe_num_mixtures=4 \ --deep_chain_layers=3 \ --deep_chain_relu_cells=256 \ --data_augmenter=NoiseAugmenter \ --input_noise_level=0.1 \ --multitask=True \ --label_loss=MultiTaskCrossEntropyLoss \ --support_type="label,label,label" \ --support_loss_percent=0.05 \ --base_learning_rate=0.01 \ --keep_checkpoint_every_n_hour=5.0 \ --keep_checkpoint_interval=6 \ --num_epochs=2 \ --batch_size=1024 fi if [ "$i" -ne "5" ]; then # inference model 1 infer_data_path="/Youtube-8M/model_predictions_local/train/${model_name}/${sub_model_name}" if [ ! -d $infer_data_path ]; then echo predictions = "$predictions_data_pattern" CUDA_VISIBLE_DEVICES=0 python inference-pre-ensemble-with-predictions.py \ --output_dir="$infer_data_path" \ --train_dir="$sub_model_path" \ --input_data_pattern="/Youtube-8M/data/video/train/*.tfrecord" \ --distill_data_pattern="$predictions_data_pattern" \ --frame_features=False \ --feature_names="mean_rgb,mean_audio" \ --feature_sizes="1024,128" \ --distillation_features=False \ --distillation_as_input=True \ --model=DistillchainDeepCombineChainModel \ --moe_num_mixtures=4 \ --deep_chain_layers=3 \ --deep_chain_relu_cells=256 \ --batch_size=1024 \ --file_size=4096 fi fi # next predictions pattern predictions_data_pattern="${predictions_data_pattern},${infer_data_path}/*.tfrecord" if [ -d $sub_model_path ]; then for part in ensemble_train ensemble_validate test; do infer_data_path="/Youtube-8M/model_predictions_local/${part}/${model_name}/${sub_model_name}" if [ ! -d $infer_data_path ]; then preensemble_data_pattern="/Youtube-8M/model_predictions_local/${part}/distillation/ensemble_v2_matrix_model/*.tfrecord" for j in {1..5}; do if [ "$j" -lt "$i" ]; then prev_model="distillchain_v2_video_dcc_${j}" preensemble_data_pattern="${preensemble_data_pattern},/Youtube-8M/model_predictions_local/${part}/${model_name}/${prev_model}/*.tfrecord" fi done echo preensemble_data_pattern = "${preensemble_data_pattern}" CUDA_VISIBLE_DEVICES=0 python inference-pre-ensemble-with-predictions.py \ --output_dir="$infer_data_path" \ --train_dir="$sub_model_path" \ --input_data_pattern="/Youtube-8M/data/video/${part}/*.tfrecord" \ --distill_data_pattern="$preensemble_data_pattern" \ --frame_features=False \ --feature_names="mean_rgb,mean_audio" \ --feature_sizes="1024,128" \ --distillation_features=False \ --distillation_as_input=True \ --model=DistillchainDeepCombineChainModel \ --moe_num_mixtures=4 \ --deep_chain_layers=3 \ --deep_chain_relu_cells=256 \ --batch_size=1024 \ --file_size=4096 fi done fi done fi ================================================ FILE: youtube-8m-wangheda/cloudml-gpu-distributed.yaml ================================================ trainingInput: runtimeVersion: "1.0" scaleTier: CUSTOM masterType: standard_gpu workerCount: 2 workerType: standard_gpu parameterServerCount: 2 parameterServerType: standard ================================================ FILE: youtube-8m-wangheda/cloudml-gpu.yaml ================================================ trainingInput: scaleTier: CUSTOM # standard_gpu provides 1 GPU. Change to complex_model_m_gpu for 4 GPUs masterType: standard_gpu runtimeVersion: "1.0" ================================================ FILE: youtube-8m-wangheda/data_augmentation.py ================================================ from tensorflow import flags flags.DEFINE_string("data_augmenter", "DefaultAugmenter", "how to preprocess feature, defaults to identical, which means no transform") flags.DEFINE_float("input_noise_level", 0.2, "the standard deviation of normal noise added to input") from all_data_augmentation import * ================================================ FILE: youtube-8m-wangheda/data_augmentation_scripts/eval-chaining-video.sh ================================================ GPU_ID=1 EVERY=100 MODEL=DeepCombineChainModel MODEL_DIR="../model/dataaugmentation_chaining_video" start=$1 DIR="$(pwd)" for checkpoint in $(cd $MODEL_DIR && python ${DIR}/training_utils/select.py $EVERY); do echo $checkpoint; if [ $checkpoint -gt $start ]; then echo $checkpoint; CUDA_VISIBLE_DEVICES=$GPU_ID python eval.py \ --train_dir="$MODEL_DIR" \ --model_checkpoint_path="${MODEL_DIR}/model.ckpt-${checkpoint}" \ --eval_data_pattern="/Youtube-8M/data/video/validate/validatea*" \ --frame_features=False \ --feature_names="mean_rgb,mean_audio" \ --feature_sizes="1024,128" \ --model=$MODEL \ --moe_num_mixtures=2 \ --num_readers=4 \ --deep_chain_layers=8 \ --deep_chain_relu_cells=40 \ --batch_size=1024 \ --run_once=True fi done ================================================ FILE: youtube-8m-wangheda/data_augmentation_scripts/run-chaining-cnn.sh ================================================ CUDA_VISIBLE_DEVICES=0 python train.py \ --train_dir="../model/dataaugmentation_chaining_cnn/" \ --train_data_pattern="/Youtube-8M/data/frame/largetrain/*.tfrecord" \ --frame_features=True \ --feature_names="rgb,audio" \ --feature_sizes="1024,128" \ --model=CnnDeepCombineChainModel \ --deep_chain_layers=4 \ --deep_chain_relu_cells=128 \ --moe_num_mixtures=4 \ --multitask=True \ --label_loss=MultiTaskCrossEntropyLoss \ --support_type="label,label,label,label" \ --support_loss_percent=0.05 \ --num_readers=4 \ --batch_size=40 \ --data_augmenter=HalfAugmenter \ --keep_checkpoint_every_n_hours=1.0 \ --base_learning_rate=0.001 ================================================ FILE: youtube-8m-wangheda/data_augmentation_scripts/run-chaining-lstm.sh ================================================ CUDA_VISIBLE_DEVICES=0 python train.py \ --train_dir="../model/dataaugmentation_chaining_lstm/" \ --train_data_pattern="/Youtube-8M/data/frame/largetrain/*.tfrecord" \ --model=LstmMemoryDeepChainModel \ --moe_num_mixtures=4 \ --deep_chain_relu_cells=200 \ --deep_chain_layers=1 \ --frame_features=True \ --feature_names="rgb,audio" \ --feature_sizes="1024,128" \ --lstm_layers=2 \ --lstm_cells=1024 \ --multitask=True \ --support_type="label" \ --num_supports=4716 \ --label_loss=MultiTaskCrossEntropyLoss \ --base_learning_rate=0.0008 \ --support_loss_percent=0.2 \ --keep_checkpoint_every_n_hours=3.0 \ --rnn_swap_memory=True \ --num_epochs=4 \ --data_augmenter=HalfAugmenter \ --batch_size=40 ================================================ FILE: youtube-8m-wangheda/data_augmentation_scripts/run-chaining-video.sh ================================================ CUDA_VISIBLE_DEVICES="0" python train.py \ --train_dir="../model/dataaugmentation_chaining_video/" \ --train_data_pattern="/Youtube-8M/data/frame/largetrain/*.tfrecord" \ --frame_features=True \ --feature_names="rgb,audio" \ --feature_sizes="1024,128" \ --model=DeepCombineChainModel \ --label_loss=MultiTaskCrossEntropyLoss \ --multitask=True \ --support_type="label,label,label,label,label,label,label,label" \ --support_loss_percent=0.025 \ --moe_num_mixtures=2 \ --keep_checkpoint_every_n_hours=0.25 \ --num_readers=4 \ --deep_chain_layers=8 \ --deep_chain_relu_cells=40 \ --batch_size=200 \ --data_augmenter=HalfVideoAugmenter \ --num_epochs=5 \ --base_learning_rate=0.005 ================================================ FILE: youtube-8m-wangheda/data_augmentation_scripts/run-multiple-attention-pooling-positional-embedding.sh ================================================ CUDA_VISIBLE_DEVICES=0 python train.py \ --train_dir="../model/dataaugmentation_attention_pooling" \ --train_data_pattern="/Youtube-8M/data/frame/largetrain/*.tfrecord" \ --frame_features=True \ --feature_names="rgb,audio" \ --feature_sizes="1024,128" \ --model=LstmPositionalAttentionMaxPoolingModel \ --moe_num_mixtures=8 \ --lstm_attentions=1 \ --positional_embedding_size=32 \ --rnn_swap_memory=True \ --base_learning_rate=0.001 \ --num_readers=4 \ --num_epochs=6 \ --batch_size=40 \ --data_augmenter=HalfAugmenter \ --keep_checkpoint_every_n_hour=2.0 ================================================ FILE: youtube-8m-wangheda/data_augmentation_scripts/run-multiscale-cnn-lstm-model.sh ================================================ CUDA_VISIBLE_DEVICES=0 python train.py \ --train_dir="../model/dataaugmentation_multiscale_cnn_lstm" \ --train_data_pattern='/Youtube-8M/data/frame/largetrain/*.tfrecord' \ --frame_features=True \ --feature_names="rgb,audio" \ --feature_sizes="1024,128" \ --model='MultiscaleCnnLstmModel' \ --multiscale_cnn_lstm_layers=4 \ --moe_num_mixtures=4 \ --multitask=True \ --label_loss=MultiTaskCrossEntropyLoss \ --support_loss_percent=1.0 \ --support_type="label,label,label,label" \ --is_training=True \ --num_readers=4 \ --num_epochs=5 \ --batch_size=40 \ --data_augmenter=HalfAugmenter \ --base_learning_rate=0.001 \ --keep_checkpoint_every_n_hour=2.0 ================================================ FILE: youtube-8m-wangheda/data_augmentation_scripts/run-parallel-lstm-memory.sh ================================================ CUDA_VISIBLE_DEVICES=1 python train.py \ --train_dir="../model/dataaugmentation_parallel_lstm_memory" \ --frame_features=True \ --feature_names="rgb,audio" \ --feature_sizes="1024,128" \ --train_data_pattern="/Youtube-8M/data/frame/largetrain/*.tfrecord" \ --lstm_cells="1024,128" \ --moe_num_mixtures=8 \ --model=LstmParallelMemoryModel \ --rnn_swap_memory=True \ --num_readers=4 \ --batch_size=40 \ --data_augmenter=HalfAugmenter \ --num_epochs=5 \ --base_learning_rate=0.0008 \ --keep_checkpoint_every_n_hour=2.0 ================================================ FILE: youtube-8m-wangheda/eval.py ================================================ # Copyright 2016 Google Inc. All Rights Reserved. # # Licensed under the Apache License, Version 2.0 (the "License"); # you may not use this file except in compliance with the License. # You may obtain a copy of the License at # # http://www.apache.org/licenses/LICENSE-2.0 # # Unless required by applicable law or agreed to in writing, software # distributed under the License is distributed on an "AS-IS" BASIS, # WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied. # See the License for the specific language governing permissions and # limitations under the License. """Binary for evaluating Tensorflow models on the YouTube-8M dataset.""" import time import eval_util import losses import frame_level_models import video_level_models import feature_transform import readers import tensorflow as tf from tensorflow import app from tensorflow import flags from tensorflow import gfile from tensorflow import logging import utils FLAGS = flags.FLAGS if __name__ == "__main__": # Dataset flags. flags.DEFINE_string("train_dir", "/tmp/yt8m_model/", "The directory to load the model files from. " "The tensorboard metrics files are also saved to this " "directory.") flags.DEFINE_string("model_checkpoint_path", "", "The file to load the model files from. ") flags.DEFINE_string( "eval_data_pattern", "", "File glob defining the evaluation dataset in tensorflow.SequenceExample " "format. The SequenceExamples are expected to have an 'rgb' byte array " "sequence feature as well as a 'labels' int64 context feature.") flags.DEFINE_string( "distill_data_pattern", None, "File glob defining the distillation data pattern") flags.DEFINE_string("feature_names", "mean_rgb", "Name of the feature " "to use for training.") flags.DEFINE_string("feature_sizes", "1024", "Length of the feature vectors.") # Model flags. flags.DEFINE_bool( "frame_features", False, "If set, then --eval_data_pattern must be frame-level features. " "Otherwise, --eval_data_pattern must be aggregated video-level " "features. The model must also be set appropriately (i.e. to read 3D " "batches VS 4D batches.") flags.DEFINE_string( "model", "LogisticModel", "Which architecture to use for the model. Options include 'Logistic', " "'SingleMixtureMoe', and 'TwoLayerSigmoid'. See aggregated_models.py and " "frame_level_models.py for the model definitions.") flags.DEFINE_integer("batch_size", 1024, "How many examples to process per batch.") flags.DEFINE_string("label_loss", "CrossEntropyLoss", "Loss computed on validation data") # Other flags. flags.DEFINE_integer("num_readers", 8, "How many threads to use for reading input files.") flags.DEFINE_boolean("run_once", False, "Whether to run eval only once.") flags.DEFINE_integer("top_k", 20, "How many predictions to output per video.") flags.DEFINE_bool( "multitask", False, "Whether to consider support_predictions") flags.DEFINE_bool( "dropout", False, "Whether to consider dropout") flags.DEFINE_float("keep_prob", 1.0, "probability to keep output (used in dropout, keep it unchanged in validationg and test)") flags.DEFINE_float("noise_level", 0.0, "standard deviation of noise (added to hidden nodes)") def find_class_by_name(name, modules): """Searches the provided modules for the named class and returns it.""" modules = [getattr(module, name, None) for module in modules] return next(a for a in modules if a) def get_input_evaluation_tensors(reader, data_pattern, batch_size=1024): logging.info("Using batch size of " + str(batch_size) + " for evaluation.") with tf.name_scope("eval_input"): files = gfile.Glob(data_pattern) if not files: raise IOError("Unable to find the evaluation files.") logging.info("number of evaluation files: " + str(len(files))) files.sort() filename_queue = tf.train.string_input_producer( files, shuffle=False, num_epochs=1) eval_data = reader.prepare_reader(filename_queue) return tf.train.batch( eval_data, batch_size=batch_size, capacity=3 * batch_size, allow_smaller_final_batch=True, enqueue_many=True) def build_graph(reader, model, eval_data_pattern, label_loss_fn, batch_size=1024, transformer_class=feature_transform.DefaultTransformer, distill_reader=None, num_readers=1): """Creates the Tensorflow graph for evaluation. Args: reader: The data file reader. It should inherit from BaseReader. model: The core model (e.g. logistic or neural net). It should inherit from BaseModel. eval_data_pattern: glob path to the evaluation data files. label_loss_fn: What kind of loss to apply to the model. It should inherit from BaseLoss. batch_size: How many examples to process at a time. num_readers: How many threads to use for I/O operations. """ global_step = tf.Variable(0, trainable=False, name="global_step") video_id_batch, model_input_raw, labels_batch, num_frames = get_input_evaluation_tensors( # pylint: disable=g-line-too-long reader, eval_data_pattern, batch_size=batch_size) tf.summary.histogram("model_input_raw", model_input_raw) if distill_reader is not None: unused_video_id_batch, distill_input_raw, unused_labels_batch, unused_num_frames = get_input_evaluation_tensors( # pylint: disable=g-line-too-long distill_reader, FLAGS.distill_data_pattern, batch_size=batch_size) feature_dim = len(model_input_raw.get_shape()) - 1 # Normalize input features. feature_transformer = transformer_class() model_input, num_frames = feature_transformer.transform(model_input_raw, num_frames=num_frames) with tf.name_scope("model"): if FLAGS.noise_level > 0: noise_level_tensor = tf.placeholder_with_default(0.0, shape=[], name="noise_level") else: noise_level_tensor = None if distill_reader is not None: distillation_predictions = distill_input_raw else: distillation_predictions = None if FLAGS.dropout: keep_prob_tensor = tf.placeholder_with_default(1.0, shape=[], name="keep_prob") result = model.create_model(model_input, num_frames=num_frames, vocab_size=reader.num_classes, labels=labels_batch, dropout=FLAGS.dropout, keep_prob=keep_prob_tensor, distillation_predictions=distillation_predictions, is_training=False) else: result = model.create_model(model_input, num_frames=num_frames, vocab_size=reader.num_classes, labels=labels_batch, distillation_predictions=distillation_predictions, is_training=False) predictions = result["predictions"] tf.summary.histogram("model_activations", predictions) if "loss" in result.keys(): label_loss = result["loss"] else: if FLAGS.multitask: support_predictions = result["support_predictions"] label_loss = label_loss_fn.calculate_loss(predictions, support_predictions, labels_batch) else: label_loss = label_loss_fn.calculate_loss(predictions, labels_batch) tf.add_to_collection("global_step", global_step) tf.add_to_collection("loss", label_loss) tf.add_to_collection("predictions", predictions) tf.add_to_collection("input_batch", model_input) tf.add_to_collection("video_id_batch", video_id_batch) tf.add_to_collection("num_frames", num_frames) tf.add_to_collection("labels", tf.cast(labels_batch, tf.float32)) tf.add_to_collection("summary_op", tf.summary.merge_all()) if FLAGS.dropout: tf.add_to_collection("keep_prob", keep_prob_tensor) if FLAGS.noise_level > 0: tf.add_to_collection("noise_level", noise_level_tensor) def evaluation_loop(video_id_batch, prediction_batch, label_batch, loss, summary_op, saver, summary_writer, evl_metrics, last_global_step_val): """Run the evaluation loop once. Args: video_id_batch: a tensor of video ids mini-batch. prediction_batch: a tensor of predictions mini-batch. label_batch: a tensor of label_batch mini-batch. loss: a tensor of loss for the examples in the mini-batch. summary_op: a tensor which runs the tensorboard summary operations. saver: a tensorflow saver to restore the model. summary_writer: a tensorflow summary_writer evl_metrics: an EvaluationMetrics object. last_global_step_val: the global step used in the previous evaluation. Returns: The global_step used in the latest model. """ global_step_val = -1 with tf.Session() as sess: if FLAGS.model_checkpoint_path: checkpoint = FLAGS.model_checkpoint_path else: checkpoint = tf.train.latest_checkpoint(FLAGS.train_dir) if checkpoint: logging.info("Loading checkpoint for eval: " + checkpoint) # Restores from checkpoint saver.restore(sess, checkpoint) # Assuming model_checkpoint_path looks something like: # /my-favorite-path/yt8m_train/model.ckpt-0, extract global_step from it. global_step_val = checkpoint.split("/")[-1].split("-")[-1] else: logging.info("No checkpoint file found.") return global_step_val if global_step_val == last_global_step_val: logging.info("skip this checkpoint global_step_val=%s " "(same as the previous one).", global_step_val) return global_step_val sess.run([tf.local_variables_initializer()]) # Start the queue runners. if FLAGS.dropout: keep_prob_tensor = tf.get_collection("keep_prob")[0] if FLAGS.noise_level > 0: noise_level_tensor = tf.get_collection("noise_level")[0] fetches = [video_id_batch, prediction_batch, label_batch, loss, summary_op] coord = tf.train.Coordinator() try: threads = [] for qr in tf.get_collection(tf.GraphKeys.QUEUE_RUNNERS): threads.extend(qr.create_threads( sess, coord=coord, daemon=True, start=True)) logging.info("enter eval_once loop global_step_val = %s. ", global_step_val) evl_metrics.clear() examples_processed = 0 while not coord.should_stop(): batch_start_time = time.time() custom_feed = {} if FLAGS.dropout: custom_feed[keep_prob_tensor] = FLAGS.keep_prob if FLAGS.noise_level > 0: custom_feed[noise_level_tensor] = FLAGS.noise_level _, predictions_val, labels_val, loss_val, summary_val = sess.run( fetches, feed_dict=custom_feed) seconds_per_batch = time.time() - batch_start_time example_per_second = labels_val.shape[0] / seconds_per_batch examples_processed += labels_val.shape[0] iteration_info_dict = evl_metrics.accumulate(predictions_val, labels_val, loss_val) iteration_info_dict["examples_per_second"] = example_per_second iterinfo = utils.AddGlobalStepSummary( summary_writer, global_step_val, iteration_info_dict, summary_scope="Eval") logging.info("examples_processed: %d | %s", examples_processed, iterinfo) except tf.errors.OutOfRangeError as e: logging.info( "Done with batched inference. Now calculating global performance " "metrics.") # calculate the metrics for the entire epoch epoch_info_dict = evl_metrics.get() epoch_info_dict["epoch_id"] = global_step_val summary_writer.add_summary(summary_val, global_step_val) epochinfo = utils.AddEpochSummary( summary_writer, global_step_val, epoch_info_dict, summary_scope="Eval") logging.info(epochinfo) evl_metrics.clear() except Exception as e: # pylint: disable=broad-except logging.info("Unexpected exception: " + str(e)) coord.request_stop(e) coord.request_stop() coord.join(threads, stop_grace_period_secs=10) return global_step_val def evaluate(): tf.set_random_seed(0) # for reproducibility with tf.Graph().as_default(): # convert feature_names and feature_sizes to lists of values feature_names, feature_sizes = utils.GetListOfFeatureNamesAndSizes( FLAGS.feature_names, FLAGS.feature_sizes) if FLAGS.frame_features: reader = readers.YT8MFrameFeatureReader(feature_names=feature_names, feature_sizes=feature_sizes) else: reader = readers.YT8MAggregatedFeatureReader(feature_names=feature_names, feature_sizes=feature_sizes) if FLAGS.distill_data_pattern is not None: distill_reader = readers.YT8MAggregatedFeatureReader(feature_names=["predictions"], feature_sizes=[4716]) else: distill_reader = None model = find_class_by_name(FLAGS.model, [frame_level_models, video_level_models])() label_loss_fn = find_class_by_name(FLAGS.label_loss, [losses])() transformer_class = find_class_by_name(FLAGS.feature_transformer, [feature_transform]) if FLAGS.eval_data_pattern is "": raise IOError("'eval_data_pattern' was not specified. " + "Nothing to evaluate.") build_graph( reader=reader, model=model, eval_data_pattern=FLAGS.eval_data_pattern, label_loss_fn=label_loss_fn, num_readers=FLAGS.num_readers, transformer_class=transformer_class, distill_reader=distill_reader, batch_size=FLAGS.batch_size) logging.info("built evaluation graph") video_id_batch = tf.get_collection("video_id_batch")[0] prediction_batch = tf.get_collection("predictions")[0] label_batch = tf.get_collection("labels")[0] loss = tf.get_collection("loss")[0] summary_op = tf.get_collection("summary_op")[0] saver = tf.train.Saver(tf.global_variables()) summary_writer = tf.summary.FileWriter( FLAGS.train_dir, graph=tf.get_default_graph()) evl_metrics = eval_util.EvaluationMetrics(reader.num_classes, FLAGS.top_k) last_global_step_val = -1 while True: last_global_step_val = evaluation_loop(video_id_batch, prediction_batch, label_batch, loss, summary_op, saver, summary_writer, evl_metrics, last_global_step_val) if FLAGS.run_once: break def main(unused_argv): logging.set_verbosity(tf.logging.INFO) print("tensorflow version: %s" % tf.__version__) evaluate() if __name__ == "__main__": app.run() ================================================ FILE: youtube-8m-wangheda/eval.sh ================================================ GPU_ID=0 EVERY=500 MODEL=LstmModel MODEL_DIR="../model/attentionlstm_moe4" start=$1 DIR="$(pwd)" for checkpoint in $(cd $MODEL_DIR && python ${DIR}/training_utils/select.py $EVERY); do echo $checkpoint; if [ $checkpoint -gt $start ]; then echo $checkpoint; CUDA_VISIBLE_DEVICES=$GPU_ID python eval.py \ --train_dir="$MODEL_DIR" \ --model_checkpoint_path="${MODEL_DIR}/model.ckpt-${checkpoint}" \ --eval_data_pattern="/Youtube-8M-validate/validatea*" \ --frame_features=True \ --feature_names="rgb,audio" \ --feature_sizes="1024,128" \ --batch_size=128 \ --model=$MODEL \ --lstm_cells=1024 \ --moe_num_mixtures=4 \ --num_readers=1 \ --run_once=True fi done ================================================ FILE: youtube-8m-wangheda/eval_scripts/eval-att-lstm.sh ================================================ GPU_ID=1 EVERY=500 MODEL=LstmAttentionLstmModel MODEL_DIR="../model/lstmattlstm1024_moe8" start=$1 DIR="$(pwd)" for checkpoint in $(cd $MODEL_DIR && python ${DIR}/training_utils/select.py $EVERY); do echo $checkpoint; if [ $checkpoint -gt $start ]; then echo $checkpoint; CUDA_VISIBLE_DEVICES=$GPU_ID python eval.py \ --train_dir="$MODEL_DIR" \ --model_checkpoint_path="${MODEL_DIR}/model.ckpt-${checkpoint}" \ --eval_data_pattern="/Youtube-8M-validate/validatea*" \ --frame_features=True \ --feature_names="rgb,audio" \ --feature_sizes="1024,128" \ --batch_size=128 \ --model=$MODEL \ --lstm_cells=1024 \ --lstm_layers=1 \ --moe_num_mixtures=8 \ --num_readers=1 \ --run_once=True fi done ================================================ FILE: youtube-8m-wangheda/eval_scripts/eval-att.sh ================================================ GPU_ID=1 EVERY=500 MODEL=LstmAttentionModel MODEL_DIR="../model/attentionlstm_moe4" start=$1 DIR="$(pwd)" for checkpoint in $(cd $MODEL_DIR && python ${DIR}/training_utils/select.py $EVERY); do echo $checkpoint; if [ $checkpoint -gt $start ]; then echo $checkpoint; CUDA_VISIBLE_DEVICES=$GPU_ID python eval.py \ --train_dir="$MODEL_DIR" \ --model_checkpoint_path="${MODEL_DIR}/model.ckpt-${checkpoint}" \ --eval_data_pattern="/Youtube-8M-validate/validatea*" \ --frame_features=True \ --feature_names="rgb,audio" \ --feature_sizes="1024,128" \ --batch_size=128 \ --model=$MODEL \ --lstm_cells=1024 \ --moe_num_mixtures=4 \ --num_readers=1 \ --run_once=True fi done ================================================ FILE: youtube-8m-wangheda/eval_scripts/eval-bi-uni-lstm.sh ================================================ GPU_ID=1 EVERY=500 MODEL=BiUniLstmModel MODEL_DIR="../model/biunilstm1024_moe4" start=$1 DIR="$(pwd)" for checkpoint in $(cd $MODEL_DIR && python ${DIR}/training_utils/select.py $EVERY); do echo $checkpoint; if [ $checkpoint -gt $start ]; then echo $checkpoint; CUDA_VISIBLE_DEVICES=$GPU_ID python eval.py \ --train_dir="$MODEL_DIR" \ --model_checkpoint_path="${MODEL_DIR}/model.ckpt-${checkpoint}" \ --eval_data_pattern="/Youtube-8M/data/frame/validate/validatea*" \ --frame_features=True \ --feature_names="rgb,audio" \ --feature_sizes="1024,128" \ --batch_size=32 \ --model=$MODEL \ --lstm_cells=1024 \ --moe_num_mixtures=4 \ --run_once=True fi done ================================================ FILE: youtube-8m-wangheda/eval_scripts/eval-chain-model-relu.sh ================================================ start=$1 GPU_ID=1 MODEL=ChainMainReluMoeModel MODEL_DIR="../model/video_chain_main_relu_moe16_ce" for checkpoint in $(for filename in $MODEL_DIR/model.ckpt-*.meta; do echo $filename | grep -o "ckpt-[0123456789]*.meta" | cut -d '-' -f 2 | cut -d '.' -f 1; done | sort -n); do if [ $checkpoint -gt $start ]; then echo $checkpoint; CUDA_VISIBLE_DEVICES=$GPU_ID python eval.py \ --train_dir="$MODEL_DIR" \ --model_checkpoint_path="${MODEL_DIR}/model.ckpt-${checkpoint}" \ --eval_data_pattern="/Youtube-8M/data/video/validate/validatea*" \ --frame_features=False \ --feature_names="mean_rgb,mean_audio" \ --feature_sizes="1024,128" \ --batch_size=256 \ --model=$MODEL \ --moe_num_mixtures=16 \ --run_once=True fi done ================================================ FILE: youtube-8m-wangheda/eval_scripts/eval-chain-model-suprelu.sh ================================================ start=$1 GPU_ID=1 MODEL=ChainSupportReluMoeModel MODEL_DIR="../model/video_chain_support_relu_moe16_ce" for checkpoint in $(for filename in $MODEL_DIR/model.ckpt-*.meta; do echo $filename | grep -o "ckpt-[0123456789]*.meta" | cut -d '-' -f 2 | cut -d '.' -f 1; done | sort -n); do if [ $checkpoint -gt $start ]; then echo $checkpoint; CUDA_VISIBLE_DEVICES=$GPU_ID python eval.py \ --train_dir="$MODEL_DIR" \ --model_checkpoint_path="${MODEL_DIR}/model.ckpt-${checkpoint}" \ --eval_data_pattern="/Youtube-8M/data/video/validate/validatea*" \ --frame_features=False \ --feature_names="mean_rgb,mean_audio" \ --feature_sizes="1024,128" \ --batch_size=256 \ --model=$MODEL \ --moe_num_mixtures=16 \ --run_once=True fi done ================================================ FILE: youtube-8m-wangheda/eval_scripts/eval-chain-moe-0.4.sh ================================================ GPU_ID="1" EVERY=500 MODEL=ChainMoeModel MODEL_DIR="../model/video_chain_moe16_ce_0.4" start=$1 DIR="$(pwd)" for checkpoint in $(cd $MODEL_DIR && python ${DIR}/training_utils/select.py $EVERY); do echo $checkpoint; if [ $checkpoint -gt $start ]; then echo $checkpoint; CUDA_VISIBLE_DEVICES=$GPU_ID python eval.py \ --train_dir="$MODEL_DIR" \ --model_checkpoint_path="${MODEL_DIR}/model.ckpt-${checkpoint}" \ --eval_data_pattern="/Youtube-8M/data/video/validate/validatea*" \ --frame_features=False \ --feature_names="mean_rgb,mean_audio" \ --feature_sizes="1024,128" \ --batch_size=256 \ --model=$MODEL \ --moe_num_mixtures=16 \ --run_once=True fi done ================================================ FILE: youtube-8m-wangheda/eval_scripts/eval-chain-moe-freq.sh ================================================ GPU_ID="1" EVERY=100 MODEL=ChainSupportReluMoeModel MODEL_DIR="../model/video_chain_moe16_ce_freq" start=$1 DIR="$(pwd)" for checkpoint in $(cd $MODEL_DIR && python ${DIR}/training_utils/select.py $EVERY); do echo $checkpoint; if [ $checkpoint -gt $start ]; then echo $checkpoint; CUDA_VISIBLE_DEVICES=$GPU_ID python eval.py \ --train_dir="$MODEL_DIR" \ --model_checkpoint_path="${MODEL_DIR}/model.ckpt-${checkpoint}" \ --eval_data_pattern="/Youtube-8M/data/video/validate/validatea*" \ --frame_features=False \ --feature_names="mean_rgb,mean_audio" \ --feature_sizes="1024,128" \ --batch_size=256 \ --num_supports=200 \ --support_type="frequent" \ --model=$MODEL \ --moe_num_mixtures=16 \ --run_once=True fi done ================================================ FILE: youtube-8m-wangheda/eval_scripts/eval-chain-moe-suprelu-vert+freq.sh ================================================ start=$1 GPU_ID=1 MODEL=ChainSupportReluMoeModel MODEL_DIR="../model/video_chain_support_relu_moe16_ce_vert+freq" for checkpoint in $(for filename in $MODEL_DIR/model.ckpt-*.meta; do echo $filename | grep -o "ckpt-[0123456789]*.meta" | cut -d '-' -f 2 | cut -d '.' -f 1; done | sort -n); do if [ $checkpoint -gt $start ]; then echo $checkpoint; CUDA_VISIBLE_DEVICES=$GPU_ID python eval.py \ --train_dir="$MODEL_DIR" \ --model_checkpoint_path="${MODEL_DIR}/model.ckpt-${checkpoint}" \ --eval_data_pattern="/Youtube-8M/data/video/validate/validatea*" \ --frame_features=False \ --feature_names="mean_rgb,mean_audio" \ --feature_sizes="1024,128" \ --batch_size=256 \ --support_type="vertical,frequent" \ --num_supports=225 \ --num_verticals=25 \ --num_frequents=200 \ --model=$MODEL \ --moe_num_mixtures=16 \ --run_once=True fi done ================================================ FILE: youtube-8m-wangheda/eval_scripts/eval-chain-moe.sh ================================================ GPU_ID="1" EVERY=500 MODEL=ChainMoeModel MODEL_DIR="../model/video_chain_moe16_ce" start=$1 DIR="$(pwd)" for checkpoint in $(cd $MODEL_DIR && python ${DIR}/training_utils/select.py $EVERY); do echo $checkpoint; if [ $checkpoint -gt $start ]; then echo $checkpoint; CUDA_VISIBLE_DEVICES=$GPU_ID python eval.py \ --train_dir="$MODEL_DIR" \ --model_checkpoint_path="${MODEL_DIR}/model.ckpt-${checkpoint}" \ --eval_data_pattern="/Youtube-8M/data/video/validate/validatea*" \ --frame_features=False \ --feature_names="mean_rgb,mean_audio" \ --feature_sizes="1024,128" \ --batch_size=256 \ --model=$MODEL \ --moe_num_mixtures=16 \ --run_once=True fi done ================================================ FILE: youtube-8m-wangheda/eval_scripts/eval-cnn-deep-combine-chain.sh ================================================ GPU_ID="1" EVERY=1000 MODEL=CnnDeepCombineChainModel MODEL_DIR="../model/cnn_deep_combine_chain" start=$1 DIR="$(pwd)" for checkpoint in $(cd $MODEL_DIR && python ${DIR}/training_utils/select.py $EVERY); do echo $checkpoint; if [ $checkpoint -gt $start ]; then echo $checkpoint; CUDA_VISIBLE_DEVICES=$GPU_ID python eval.py \ --train_dir="$MODEL_DIR" \ --model_checkpoint_path="${MODEL_DIR}/model.ckpt-${checkpoint}" \ --eval_data_pattern="/Youtube-8M/data/frame/validate/validatea*" \ --frame_features=True \ --feature_names="rgb,audio" \ --feature_sizes="1024,128" \ --batch_size=64 \ --model=$MODEL \ --label_loss=MultiTaskCrossEntropyLoss \ --multitask=True \ --support_type="label,label,label,label" \ --support_loss_percent=0.05 \ --moe_num_mixtures=4 \ --deep_chain_layers=4 \ --deep_chain_relu_cells=128 \ --rnn_swap_memory=True \ --run_once=True fi done ================================================ FILE: youtube-8m-wangheda/eval_scripts/eval-cnn-model.sh ================================================ GPU_ID="0" EVERY=4000 MODEL=CnnModel MODEL_DIR="../model/cnn_model" start=$1 DIR="$(pwd)" for checkpoint in $(cd $MODEL_DIR && python ${DIR}/training_utils/select.py $EVERY); do echo $checkpoint; if [ $checkpoint -gt $start ]; then echo $checkpoint; CUDA_VISIBLE_DEVICES=$GPU_ID python eval.py \ --train_dir="$MODEL_DIR" \ --model_checkpoint_path="${MODEL_DIR}/model.ckpt-${checkpoint}" \ --eval_data_pattern="/Youtube-8M/data/frame/validate/validatea*" \ --frame_features=True \ --feature_names="rgb,audio" \ --feature_sizes="1024,128" \ --model=$MODEL \ --cnn_num_filters=512 \ --moe_num_mixtures=4 \ --num_readers=4 \ --batch_size=128 \ --run_once=True fi done ================================================ FILE: youtube-8m-wangheda/eval_scripts/eval-dbof.sh ================================================ GPU_ID=1 EVERY=1000 MODEL=DbofModel MODEL_DIR="../model/dbof_model" start=$1 DIR="$(pwd)" for checkpoint in $(cd $MODEL_DIR && python ${DIR}/training_utils/select.py $EVERY); do echo $checkpoint; if [ $checkpoint -gt $start ]; then echo $checkpoint; CUDA_VISIBLE_DEVICES=$GPU_ID python eval.py \ --train_dir="$MODEL_DIR" \ --model_checkpoint_path="${MODEL_DIR}/model.ckpt-${checkpoint}" \ --eval_data_pattern="/Youtube-8M/data/frame/validate/validatea*" \ --frame_features=True \ --feature_names="rgb,audio" \ --feature_sizes="1024,128" \ --model=$MODEL \ --batch_size=128 \ --num_readers=1 \ --run_once=True fi done ================================================ FILE: youtube-8m-wangheda/eval_scripts/eval-deep-cnn-deep-combine-chain.sh ================================================ GPU_ID=0 EVERY=1000 MODEL=DeepCnnDeepCombineChainModel MODEL_DIR="../model/deep_cnn_deep_combine_chain" start=$1 DIR="$(pwd)" for checkpoint in $(cd $MODEL_DIR && python ${DIR}/training_utils/select.py $EVERY); do echo $checkpoint; if [ $checkpoint -gt $start ]; then echo $checkpoint; CUDA_VISIBLE_DEVICES=$GPU_ID python eval.py \ --train_dir="$MODEL_DIR" \ --model_checkpoint_path="${MODEL_DIR}/model.ckpt-${checkpoint}" \ --eval_data_pattern="/Youtube-8M/data/frame/validate/validatea*" \ --frame_features=True \ --feature_names="rgb,audio" \ --feature_sizes="1024,128" \ --moe_num_mixtures=4 \ --deep_chain_layers=2 \ --deep_chain_relu_cells=256 \ --deep_cnn_base_size=128 \ --label_loss=MultiTaskCrossEntropyLoss \ --multitask=True \ --support_type="label,label" \ --support_loss_percent=0.1 \ --batch_size=256 \ --model=$MODEL \ --run_once=True fi done ================================================ FILE: youtube-8m-wangheda/eval_scripts/eval-distill-video-dcc-noise-scene1.sh ================================================ percent="$1" start=0 GPU_ID=0 EVERY=200 MODEL=DeepCombineChainModel MODEL_DIR="../model/distillation_video_dcc_noise/scene1_percent_${percent}" DIR="$(pwd)" for checkpoint in $(cd $MODEL_DIR && python ${DIR}/training_utils/select.py $EVERY); do echo $checkpoint; if [ $checkpoint -gt $start ]; then echo $checkpoint; CUDA_VISIBLE_DEVICES=$GPU_ID python eval.py \ --train_dir="$MODEL_DIR" \ --model_checkpoint_path="${MODEL_DIR}/model.ckpt-${checkpoint}" \ --eval_data_pattern="/Youtube-8M/data/video/validate/validatea*" \ --frame_features=False \ --feature_names="mean_rgb,mean_audio" \ --feature_sizes="1024,128" \ --model=$MODEL \ --deep_chain_relu_cells=256 \ --deep_chain_layers=4 \ --moe_num_mixtures=4 \ --batch_size=1024 \ --num_readers=1 \ --run_once=True fi done ================================================ FILE: youtube-8m-wangheda/eval_scripts/eval-distill-video-dcc-noise-scene2.sh ================================================ percent="$1" start=0 GPU_ID=0 EVERY=200 MODEL=DeepCombineChainModel MODEL_DIR="../model/distillation_video_dcc_noise/scene2_percent_${percent}" DIR="$(pwd)" for checkpoint in $(cd $MODEL_DIR && python ${DIR}/training_utils/select.py $EVERY); do echo $checkpoint; if [ $checkpoint -gt $start ]; then echo $checkpoint; CUDA_VISIBLE_DEVICES=$GPU_ID python eval.py \ --train_dir="$MODEL_DIR" \ --model_checkpoint_path="${MODEL_DIR}/model.ckpt-${checkpoint}" \ --eval_data_pattern="/Youtube-8M/data/video/validate/validatea*" \ --frame_features=False \ --feature_names="mean_rgb,mean_audio" \ --feature_sizes="1024,128" \ --model=$MODEL \ --deep_chain_relu_cells=256 \ --deep_chain_layers=4 \ --moe_num_mixtures=4 \ --batch_size=1024 \ --num_readers=1 \ --run_once=True fi done ================================================ FILE: youtube-8m-wangheda/eval_scripts/eval-distillchain-cnn-dcc.sh ================================================ start=$1 GPU_ID=0 EVERY=1000 MODEL=DistillchainCnnDeepCombineChainModel MODEL_DIR="../model/distillchain_cnn_dcc" DIR="$(pwd)" for checkpoint in $(cd $MODEL_DIR && python ${DIR}/training_utils/select.py $EVERY); do echo $checkpoint; if [ $checkpoint -gt $start ]; then echo $checkpoint; CUDA_VISIBLE_DEVICES=$GPU_ID python eval.py \ --train_dir="$MODEL_DIR" \ --model_checkpoint_path="${MODEL_DIR}/model.ckpt-${checkpoint}" \ --eval_data_pattern="/Youtube-8M/data/frame/validate/validatea*" \ --distill_data_pattern="/Youtube-8M/model_predictions/validatea/distillation/ensemble_mean_model/*.tfrecord" \ --frame_features=True \ --feature_names="rgb,audio" \ --feature_sizes="1024,128" \ --distillation_features=True \ --distillation_as_input=True \ --model=$MODEL \ --deep_chain_layers=3 \ --deep_chain_relu_cells=256 \ --moe_num_mixtures=4 \ --multitask=True \ --label_loss=MultiTaskCrossEntropyLoss \ --support_type="label,label,label" \ --support_loss_percent=0.05 \ --batch_size=128 \ --num_readers=1 \ --run_once=True fi done ================================================ FILE: youtube-8m-wangheda/eval_scripts/eval-distillchain-lstmcnn.sh ================================================ start=$1 GPU_ID=1 EVERY=2000 MODEL=DistillchainLstmCnnDeepCombineChainModel MODEL_DIR="../model/distillchain_lstmcnn_dcc" DIR="$(pwd)" for checkpoint in $(cd $MODEL_DIR && python ${DIR}/training_utils/select.py $EVERY); do echo $checkpoint; if [ $checkpoint -gt $start ]; then echo $checkpoint; CUDA_VISIBLE_DEVICES=$GPU_ID python eval.py \ --train_dir="$MODEL_DIR" \ --model_checkpoint_path="${MODEL_DIR}/model.ckpt-${checkpoint}" \ --eval_data_pattern="/Youtube-8M/data/frame/validate/validatea*" \ --distill_data_pattern="/Youtube-8M/model_predictions/validatea/distillation/ensemble_mean_model/*.tfrecord" \ --frame_features=True \ --feature_names="rgb,audio" \ --feature_sizes="1024,128" \ --distillation_features=True \ --distillation_as_input=True \ --model=$MODEL \ --deep_chain_layers=2 \ --deep_chain_relu_cells=128 \ --distillchain_relu_cells=256 \ --moe_num_mixtures=4 \ --lstm_layers=1 \ --lstm_cells="1024,128" \ --rnn_swap_memory=True \ --batch_size=32 \ --num_readers=1 \ --run_once=True fi done ================================================ FILE: youtube-8m-wangheda/eval_scripts/eval-distillchain-lstmparalleloutput.sh ================================================ start=$1 GPU_ID=0 EVERY=2000 MODEL=DistillchainLstmParallelFinaloutputModel MODEL_DIR="../model/distillchain_lstmparalleloutput" DIR="$(pwd)" for checkpoint in $(cd $MODEL_DIR && python ${DIR}/training_utils/select.py $EVERY); do echo $checkpoint; if [ $checkpoint -gt $start ]; then echo $checkpoint; CUDA_VISIBLE_DEVICES=$GPU_ID python eval.py \ --train_dir="$MODEL_DIR" \ --model_checkpoint_path="${MODEL_DIR}/model.ckpt-${checkpoint}" \ --eval_data_pattern="/Youtube-8M/data/frame/validate/validatea*" \ --distill_data_pattern="/Youtube-8M/model_predictions/validatea/distillation/ensemble_mean_model/*.tfrecord" \ --frame_features=True \ --feature_names="rgb,audio" \ --feature_sizes="1024,128" \ --distillation_features=True \ --distillation_as_input=True \ --model=$MODEL \ --rnn_swap_memory=True \ --lstm_cells="1024,128" \ --moe_num_mixtures=8 \ --batch_size=128 \ --num_readers=1 \ --run_once=True fi done ================================================ FILE: youtube-8m-wangheda/eval_scripts/eval-distillchain-multilstm.sh ================================================ start=$1 GPU_ID=1 EVERY=100 MODEL=DistillchainLstmMemoryDeepCombineChainModel MODEL_DIR="../model/distillchain_multilstm_dcc" DIR="$(pwd)" for checkpoint in $(cd $MODEL_DIR && python ${DIR}/training_utils/select.py $EVERY); do echo $checkpoint; if [ $checkpoint -gt $start ]; then echo $checkpoint; CUDA_VISIBLE_DEVICES=$GPU_ID python eval.py \ --train_dir="$MODEL_DIR" \ --model_checkpoint_path="${MODEL_DIR}/model.ckpt-${checkpoint}" \ --eval_data_pattern="/Youtube-8M/data/frame/validate/validatea*" \ --distill_data_pattern="/Youtube-8M/model_predictions/validatea/distillation/ensemble_mean_model/*.tfrecord" \ --frame_features=True \ --feature_names="rgb,audio" \ --feature_sizes="1024,128" \ --distillation_features=True \ --distillation_as_input=True \ --model=$MODEL \ --lstm_layers=1 \ --lstm_cells=1024 \ --moe_num_mixtures=4 \ --distillation_relu_cells=256 \ --deep_chain_relu_cells=256 \ --deep_chain_layers=2 \ --rnn_swap_memory=True \ --batch_size=32 \ --num_readers=1 \ --run_once=True fi done ================================================ FILE: youtube-8m-wangheda/eval_scripts/eval-distillchain-v2-boostinglstmparalleloutput.sh ================================================ start=$1 GPU_ID=1 EVERY=2000 MODEL=DistillchainLstmParallelFinaloutputModel MODEL_DIR="../model/distillchain_v2_boost_lstmparalleloutput" DIR="$(pwd)" for checkpoint in $(cd $MODEL_DIR && python ${DIR}/training_utils/select.py $EVERY); do echo $checkpoint; if [ $checkpoint -gt $start ]; then echo $checkpoint; CUDA_VISIBLE_DEVICES=$GPU_ID python eval.py \ --train_dir="$MODEL_DIR" \ --model_checkpoint_path="${MODEL_DIR}/model.ckpt-${checkpoint}" \ --eval_data_pattern="/Youtube-8M/data/frame/validate/validatea*" \ --distill_data_pattern="/Youtube-8M/model_predictions/validatea/distillation/ensemble_v2_matrix_model/*.tfrecord" \ --frame_features=True \ --feature_names="rgb,audio" \ --feature_sizes="1024,128" \ --distillation_features=False \ --distillation_as_input=True \ --model=$MODEL \ --lstm_layers=1 \ --rnn_swap_memory=True \ --lstm_cells="1024,128" \ --moe_num_mixtures=8 \ --batch_size=32 \ --num_readers=1 \ --run_once=True fi done ================================================ FILE: youtube-8m-wangheda/eval_scripts/eval-distillchain-v2-lstmattention8max.sh ================================================ start=$1 GPU_ID=0 EVERY=2000 MODEL=DistillchainLstmAttentionMaxPoolingModel \ MODEL_DIR="../model/distillchain_v2_lstmattention8max" DIR="$(pwd)" for checkpoint in $(cd $MODEL_DIR && python ${DIR}/training_utils/select.py $EVERY); do echo $checkpoint; if [ $checkpoint -gt $start ]; then echo $checkpoint; CUDA_VISIBLE_DEVICES=$GPU_ID python eval.py \ --train_dir="$MODEL_DIR" \ --model_checkpoint_path="${MODEL_DIR}/model.ckpt-${checkpoint}" \ --eval_data_pattern="/Youtube-8M/data/frame/validate/validatea*" \ --distill_data_pattern="/Youtube-8M/model_predictions/validatea/distillation/ensemble_v2_matrix_model/*.tfrecord" \ --frame_features=True \ --feature_names="rgb,audio" \ --feature_sizes="1024,128" \ --distillation_features=False \ --distillation_as_input=True \ --model=$MODEL \ --moe_num_mixtures=8 \ --lstm_attentions=8 \ --lstm_cells=1024 \ --rnn_swap_memory=True \ --moe_num_mixtures=8 \ --batch_size=128 \ --num_readers=1 \ --run_once=True fi done ================================================ FILE: youtube-8m-wangheda/eval_scripts/eval-distillchain-v2-lstmcnn.sh ================================================ start=$1 GPU_ID=0 EVERY=2000 MODEL=DistillchainLstmCnnDeepCombineChainModel MODEL_DIR="../model/distillchain_v2_lstmcnn_dcc" DIR="$(pwd)" for checkpoint in $(cd $MODEL_DIR && python ${DIR}/training_utils/select.py $EVERY); do echo $checkpoint; if [ $checkpoint -gt $start ]; then echo $checkpoint; CUDA_VISIBLE_DEVICES=$GPU_ID python eval.py \ --train_dir="$MODEL_DIR" \ --model_checkpoint_path="${MODEL_DIR}/model.ckpt-${checkpoint}" \ --eval_data_pattern="/Youtube-8M/data/frame/validate/validatea*" \ --distill_data_pattern="/Youtube-8M/model_predictions/validatea/distillation/ensemble_v2_matrix_model/*.tfrecord" \ --frame_features=True \ --feature_names="rgb,audio" \ --feature_sizes="1024,128" \ --distillation_features=False \ --distillation_as_input=True \ --model=$MODEL \ --deep_chain_layers=2 \ --deep_chain_relu_cells=128 \ --distillchain_relu_cells=256 \ --moe_num_mixtures=4 \ --lstm_layers=1 \ --lstm_cells="1024,128" \ --rnn_swap_memory=True \ --batch_size=128 \ --num_readers=1 \ --run_once=True fi done ================================================ FILE: youtube-8m-wangheda/eval_scripts/eval-distillchain-v2-lstmparalleloutput.sh ================================================ start=$1 GPU_ID=1 EVERY=2000 MODEL=DistillchainLstmParallelFinaloutputModel MODEL_DIR="../model/distillchain_v2_lstmparalleloutput" DIR="$(pwd)" for checkpoint in $(cd $MODEL_DIR && python ${DIR}/training_utils/select.py $EVERY); do echo $checkpoint; if [ $checkpoint -gt $start ]; then echo $checkpoint; CUDA_VISIBLE_DEVICES=$GPU_ID python eval.py \ --train_dir="$MODEL_DIR" \ --model_checkpoint_path="${MODEL_DIR}/model.ckpt-${checkpoint}" \ --eval_data_pattern="/Youtube-8M/data/frame/validate/validatea*" \ --distill_data_pattern="/Youtube-8M/model_predictions/validatea/distillation/ensemble_v2_matrix_model/*.tfrecord" \ --frame_features=True \ --feature_names="rgb,audio" \ --feature_sizes="1024,128" \ --distillation_features=False \ --distillation_as_input=True \ --model=$MODEL \ --rnn_swap_memory=True \ --lstm_cells="1024,128" \ --moe_num_mixtures=4 \ --batch_size=32 \ --num_readers=1 \ --run_once=True fi done ================================================ FILE: youtube-8m-wangheda/eval_scripts/eval-distillchain-v2-multilstm.sh ================================================ start=$1 GPU_ID=0 EVERY=500 MODEL=DistillchainLstmMemoryDeepCombineChainModel MODEL_DIR="../model/distillchain_v2_multilstm_dcc" DIR="$(pwd)" for checkpoint in $(cd $MODEL_DIR && python ${DIR}/training_utils/select.py $EVERY); do echo $checkpoint; if [ $checkpoint -gt $start ]; then echo $checkpoint; CUDA_VISIBLE_DEVICES=$GPU_ID python eval.py \ --train_dir="$MODEL_DIR" \ --model_checkpoint_path="${MODEL_DIR}/model.ckpt-${checkpoint}" \ --eval_data_pattern="/Youtube-8M/data/frame/validate/validatea*" \ --distill_data_pattern="/Youtube-8M/model_predictions/validatea/distillation/ensemble_v2_matrix_model/*.tfrecord" \ --frame_features=True \ --feature_names="rgb,audio" \ --feature_sizes="1024,128" \ --distillation_features=False \ --distillation_as_input=True \ --model=$MODEL \ --lstm_layers=1 \ --lstm_cells=1024 \ --moe_num_mixtures=4 \ --distillation_relu_cells=256 \ --deep_chain_relu_cells=256 \ --deep_chain_layers=2 \ --rnn_swap_memory=True \ --batch_size=128 \ --num_readers=1 \ --run_once=True fi done ================================================ FILE: youtube-8m-wangheda/eval_scripts/eval-distillchain-v2-multiscale-cnnlstm.sh ================================================ start=$1 GPU_ID=0 EVERY=2000 MODEL=DistillchainMultiscaleCnnLstmModel MODEL_DIR="../model/distillchain_v2_multiscale_cnnlstm" DIR="$(pwd)" for checkpoint in $(cd $MODEL_DIR && python ${DIR}/training_utils/select.py $EVERY); do echo $checkpoint; if [ $checkpoint -gt $start ]; then echo $checkpoint; CUDA_VISIBLE_DEVICES=$GPU_ID python eval.py \ --train_dir="$MODEL_DIR" \ --model_checkpoint_path="${MODEL_DIR}/model.ckpt-${checkpoint}" \ --eval_data_pattern="/Youtube-8M/data/frame/validate/validatea*" \ --distill_data_pattern="/Youtube-8M/model_predictions/validatea/distillation/ensemble_v2_matrix_model/*.tfrecord" \ --frame_features=True \ --feature_names="rgb,audio" \ --feature_sizes="1024,128" \ --distillation_features=False \ --distillation_as_input=True \ --model=$MODEL \ --multiscale_cnn_lstm_layers=3 \ --moe_num_mixtures=4 \ --rnn_swap_memory=True \ --is_training=False \ --batch_size=128 \ --num_readers=1 \ --run_once=True fi done ================================================ FILE: youtube-8m-wangheda/eval_scripts/eval-distillchain-v2-video-dcc.sh ================================================ noise_level=0.15 start=$1 GPU_ID=0 EVERY=500 MODEL=DistillchainDeepCombineChainModel MODEL_DIR="../model/distillchain_v2_video_dcc" DIR="$(pwd)" for checkpoint in $(cd $MODEL_DIR && python ${DIR}/training_utils/select.py $EVERY); do echo $checkpoint; if [ $checkpoint -gt $start ]; then echo $checkpoint; CUDA_VISIBLE_DEVICES=$GPU_ID python eval.py \ --train_dir="$MODEL_DIR" \ --model_checkpoint_path="${MODEL_DIR}/model.ckpt-${checkpoint}" \ --eval_data_pattern="/Youtube-8M/data/video/validate/validatea*" \ --distill_data_pattern="/Youtube-8M/model_predictions/validatea/distillation/ensemble_v2_matrix_model/*.tfrecord" \ --frame_features=False \ --feature_names="mean_rgb,mean_audio" \ --feature_sizes="1024,128" \ --model=$MODEL \ --deep_chain_relu_cells=256 \ --deep_chain_layers=4 \ --moe_num_mixtures=4 \ --batch_size=1024 \ --num_readers=1 \ --run_once=True fi done ================================================ FILE: youtube-8m-wangheda/eval_scripts/eval-distillchain-video-dcc.sh ================================================ noise_level=0.15 start=$1 GPU_ID=0 EVERY=100 MODEL=DistillchainDeepCombineChainModel MODEL_DIR="../model/distillchain_video_dcc" DIR="$(pwd)" for checkpoint in $(cd $MODEL_DIR && python ${DIR}/training_utils/select.py $EVERY); do echo $checkpoint; if [ $checkpoint -gt $start ]; then echo $checkpoint; CUDA_VISIBLE_DEVICES=$GPU_ID python eval.py \ --train_dir="$MODEL_DIR" \ --model_checkpoint_path="${MODEL_DIR}/model.ckpt-${checkpoint}" \ --eval_data_pattern="/Youtube-8M/data/video/validate/validatea*" \ --distill_data_pattern="/Youtube-8M/model_predictions/validatea/distillation/ensemble_mean_model/*.tfrecord" \ --frame_features=False \ --feature_names="mean_rgb,mean_audio" \ --feature_sizes="1024,128" \ --label_loss=MultiTaskCrossEntropyLoss \ --multitask=True \ --support_type="label,label,label,label" \ --num_supports=18864 \ --support_loss_percent=0.05 \ --model=$MODEL \ --deep_chain_relu_cells=200 \ --deep_chain_layers=4 \ --moe_num_mixtures=4 \ --batch_size=1024 \ --num_readers=1 \ --run_once=True fi done ================================================ FILE: youtube-8m-wangheda/eval_scripts/eval-frame-seg.sh ================================================ GPU_ID="1" EVERY=1000 MODEL=FrameSegModel MODEL_DIR="../model/frame_seg_model" start=$1 DIR="$(pwd)" for checkpoint in $(cd $MODEL_DIR && python ${DIR}/training_utils/select.py $EVERY); do echo $checkpoint; if [ $checkpoint -gt $start ]; then echo $checkpoint; CUDA_VISIBLE_DEVICES=$GPU_ID python eval.py \ --train_dir="$MODEL_DIR" \ --model_checkpoint_path="${MODEL_DIR}/model.ckpt-${checkpoint}" \ --eval_data_pattern="/Youtube-8M/data/frame/validate/validatea*" \ --frame_features=True \ --feature_names="rgb,audio" \ --feature_sizes="1024,128" \ --batch_size=32 \ --feature_transformer=IdenticalTransformer \ --model=$MODEL \ --moe_num_mixtures=16 \ --run_once=True fi done ================================================ FILE: youtube-8m-wangheda/eval_scripts/eval-framehop-lstmmem.sh ================================================ GPU_ID=1 EVERY=500 MODEL=FramehopLstmMemoryModel MODEL_DIR="../model/framehop_lstm" start=$1 DIR="$(pwd)" for checkpoint in $(cd $MODEL_DIR && python ${DIR}/training_utils/select.py $EVERY); do echo $checkpoint; if [ $checkpoint -gt $start ]; then echo $checkpoint; CUDA_VISIBLE_DEVICES=$GPU_ID python eval.py \ --train_dir="$MODEL_DIR" \ --model_checkpoint_path="${MODEL_DIR}/model.ckpt-${checkpoint}" \ --eval_data_pattern="/Youtube-8M-validate/validatea*" \ --frame_features=True \ --feature_names="rgb,audio" \ --feature_sizes="1024,128" \ --batch_size=64 \ --model=$MODEL \ --deep_chain_layers=4 \ --deep_chain_relu_cells=256 \ --feature_transformer=IdenticalTransformer \ --lstm_layers=2 \ --lstm_cells="512,64" \ --moe_num_mixtures=4 \ --rnn_swap_memory=True \ --num_readers=1 \ --run_once=True fi done ================================================ FILE: youtube-8m-wangheda/eval_scripts/eval-layer-chain-moe8-freq.sh ================================================ GPU_ID="1" EVERY=500 MODEL=ChainSupportReluMoeModel MODEL_DIR="../model/layer_chain_moe8_freq" start=$1 DIR="$(pwd)" for checkpoint in $(cd $MODEL_DIR && python ${DIR}/training_utils/select.py $EVERY); do echo $checkpoint; if [ $checkpoint -gt $start ]; then echo $checkpoint; CUDA_VISIBLE_DEVICES=$GPU_ID python eval.py \ --train_dir="$MODEL_DIR" \ --model_checkpoint_path="${MODEL_DIR}/model.ckpt-${checkpoint}" \ --eval_data_pattern="/Youtube-8M/data/layer/validate/validate*0.tfrecord" \ --frame_features=False \ --feature_names="layer" \ --feature_sizes="2048" \ --batch_size=128 \ --num_supports=200 \ --support_type="frequent" \ --model=$MODEL \ --moe_num_mixtures=8 \ --run_once=True fi done ================================================ FILE: youtube-8m-wangheda/eval_scripts/eval-layer-moe-vert.sh ================================================ GPU_ID="1" EVERY=500 MODEL=ChainSupportReluMoeModel MODEL_DIR="../model/layer_chain_moe8_vert" start=$1 DIR="$(pwd)" for checkpoint in $(cd $MODEL_DIR && python ${DIR}/training_utils/select.py $EVERY); do echo $checkpoint; if [ $checkpoint -gt $start ]; then echo $checkpoint; CUDA_VISIBLE_DEVICES=$GPU_ID python eval.py \ --train_dir="$MODEL_DIR" \ --model_checkpoint_path="${MODEL_DIR}/model.ckpt-${checkpoint}" \ --eval_data_pattern="/Youtube-8M/data/layer/validate/validate*0.tfrecord" \ --frame_features=False \ --feature_names="layer" \ --feature_sizes="2048" \ --batch_size=128 \ --num_supports=25 \ --support_type="vertical" \ --model=$MODEL \ --moe_num_mixtures=8 \ --run_once=True fi done ================================================ FILE: youtube-8m-wangheda/eval_scripts/eval-lstm-attention-8max.sh ================================================ GPU_ID="1" EVERY=200 MODEL=LstmAttentionMaxPoolingModel MODEL_DIR="../model/lstmattention8max_bagging/base_model" start=$1 DIR="$(pwd)" for checkpoint in $(cd $MODEL_DIR && python ${DIR}/training_utils/select.py $EVERY); do echo $checkpoint; if [ $checkpoint -gt $start ]; then echo $checkpoint; CUDA_VISIBLE_DEVICES=$GPU_ID python eval.py \ --train_dir="$MODEL_DIR" \ --model_checkpoint_path="${MODEL_DIR}/model.ckpt-${checkpoint}" \ --eval_data_pattern="/Youtube-8M/data/frame/validate/validatea*" \ --frame_features=True \ --feature_names="rgb,audio" \ --feature_sizes="1024,128" \ --model=$MODEL \ --moe_num_mixtures=8 \ --lstm_attentions=8 \ --rnn_swap_memory=True \ --batch_size=32 \ --run_once=True fi done ================================================ FILE: youtube-8m-wangheda/eval_scripts/eval-lstm-cnn-deep-combine-chain.sh ================================================ GPU_ID=0 EVERY=500 MODEL=LstmCnnDeepCombineChainModel MODEL_DIR="../model/lstm_cnn_deep_combine_chain" start=$1 DIR="$(pwd)" for checkpoint in $(cd $MODEL_DIR && python ${DIR}/training_utils/select.py $EVERY); do echo $checkpoint; if [ $checkpoint -gt $start ]; then echo $checkpoint; CUDA_VISIBLE_DEVICES=$GPU_ID python eval.py \ --train_dir="$MODEL_DIR" \ --model_checkpoint_path="${MODEL_DIR}/model.ckpt-${checkpoint}" \ --eval_data_pattern="/Youtube-8M/data/frame/validate/validatea*" \ --frame_features=True \ --feature_names="rgb,audio" \ --feature_sizes="1024,128" \ --model=$MODEL \ --lstm_layers=1 \ --lstm_cells="1024,128" \ --deep_chain_layers=3 \ --deep_chain_relu_cells=128 \ --moe_num_mixtures=4 \ --label_loss=MultiTaskCrossEntropyLoss \ --multitask=True \ --support_type="label,label,label" \ --support_loss_percent=0.05 \ --batch_size=32 \ --rnn_swap_memory=True \ --num_readers=4 \ --run_once=True fi done ================================================ FILE: youtube-8m-wangheda/eval_scripts/eval-lstm-look-back.sh ================================================ GPU_ID="1" EVERY=1000 MODEL=LstmLookBackModel MODEL_DIR="../model/lstmlookback1024_moe8" start=$1 DIR="$(pwd)" for checkpoint in $(cd $MODEL_DIR && python ${DIR}/training_utils/select.py $EVERY); do echo $checkpoint; if [ $checkpoint -gt $start ]; then echo $checkpoint; CUDA_VISIBLE_DEVICES=$GPU_ID python eval.py \ --train_dir="$MODEL_DIR" \ --model_checkpoint_path="${MODEL_DIR}/model.ckpt-${checkpoint}" \ --eval_data_pattern="/Youtube-8M/data/frame/validate/validatea*" \ --frame_features=True \ --feature_names="rgb,audio" \ --feature_sizes="1024,128" \ --batch_size=32 \ --model=$MODEL \ --moe_num_mixtures=8 \ --lstm_layers=2 \ --lstm_cells="1024,128" \ --lstm_look_back=3 \ --rnn_swap_memory=True \ --run_once=True fi done ================================================ FILE: youtube-8m-wangheda/eval_scripts/eval-lstm-positional-attention-8max.sh ================================================ GPU_ID="1" EVERY=1000 MODEL=LstmPositionalAttentionMaxPoolingModel MODEL_DIR="../x27_model/lstm_positional_attention8max" start=70000 DIR="$(pwd)" for checkpoint in $(cd $MODEL_DIR && python ${DIR}/training_utils/select.py $EVERY); do echo $checkpoint; if [ $checkpoint -gt $start ]; then echo $checkpoint; CUDA_VISIBLE_DEVICES=$GPU_ID python eval.py \ --train_dir="$MODEL_DIR" \ --model_checkpoint_path="${MODEL_DIR}/model.ckpt-${checkpoint}" \ --eval_data_pattern="/Youtube-8M/data/frame/validate/validatea*" \ --frame_features=True \ --feature_names="rgb,audio" \ --feature_sizes="1024,128" \ --model=$MODEL \ --moe_num_mixtures=8 \ --lstm_attentions=8 \ --positional_embedding_size=32 \ --rnn_swap_memory=True \ --batch_size=32 \ --run_once=True fi done ================================================ FILE: youtube-8m-wangheda/eval_scripts/eval-lstmmem-augmenter.sh ================================================ GPU_ID=0 EVERY=1000 MODEL=LstmMemoryModel MODEL_DIR="../model/lstmmemory1024_moe8_augment" start=$1 DIR="$(pwd)" for checkpoint in $(cd $MODEL_DIR && python ${DIR}/training_utils/select.py $EVERY); do echo $checkpoint; if [ $checkpoint -gt $start ]; then echo $checkpoint; CUDA_VISIBLE_DEVICES=$GPU_ID python eval.py \ --train_dir="$MODEL_DIR" \ --model_checkpoint_path="${MODEL_DIR}/model.ckpt-${checkpoint}" \ --eval_data_pattern="/Youtube-8M/data/frame/validate/validatea*" \ --frame_features=True \ --feature_names="rgb,audio" \ --feature_sizes="1024,128" \ --model=$MODEL \ --moe_num_mixtures=8 \ --lstm_cells=1024 \ --lstm_layers=2 \ --batch_size=64 \ --rnn_swap_memory=True \ --run_once=True fi done ================================================ FILE: youtube-8m-wangheda/eval_scripts/eval-lstmmem-chain-freq.sh ================================================ GPU_ID=1 EVERY=500 MODEL=LstmMemoryModel MODEL_DIR="../model/lstmmemory1024_moe8_chain_freq" start=$1 DIR="$(pwd)" for checkpoint in $(cd $MODEL_DIR && python ${DIR}/training_utils/select.py $EVERY); do echo $checkpoint; if [ $checkpoint -gt $start ]; then echo $checkpoint; CUDA_VISIBLE_DEVICES=$GPU_ID python eval.py \ --train_dir="$MODEL_DIR" \ --model_checkpoint_path="${MODEL_DIR}/model.ckpt-${checkpoint}" \ --eval_data_pattern="/Youtube-8M/data/frame/validate/validatea*" \ --frame_features=True \ --feature_names="rgb,audio" \ --feature_sizes="1024,128" \ --batch_size=64 \ --model=LstmMemoryModel \ --video_level_classifier_model=ChainSupportReluMoeModel \ --lstm_layers=2 \ --lstm_cells=1024 \ --multitask=True \ --support_type="frequent" \ --num_supports=200 \ --num_frequents=200 \ --label_loss=MultiTaskCrossEntropyLoss \ --base_learning_rate=0.0008 \ --vertical_loss_percent=0.4 \ --moe_num_mixtures=8 \ --num_readers=1 \ --run_once=True fi done ================================================ FILE: youtube-8m-wangheda/eval_scripts/eval-lstmmem-chain.sh ================================================ GPU_ID=1 EVERY=500 MODEL=LstmMemoryChainModel MODEL_DIR="../model/lstmmemory1024_moe8_chain" start=$1 DIR="$(pwd)" for checkpoint in $(cd $MODEL_DIR && python ${DIR}/training_utils/select.py $EVERY); do echo $checkpoint; if [ $checkpoint -gt $start ]; then echo $checkpoint; CUDA_VISIBLE_DEVICES=$GPU_ID python eval.py \ --train_dir="$MODEL_DIR" \ --model_checkpoint_path="${MODEL_DIR}/model.ckpt-${checkpoint}" \ --eval_data_pattern="/Youtube-8M-validate/validatea*" \ --frame_features=True \ --feature_names="rgb,audio" \ --feature_sizes="1024,128" \ --batch_size=128 \ --model=$MODEL \ --lstm_layers=2 \ --lstm_cells=1024 \ --multitask=True \ --num_supports=25 \ --num_verticals=25 \ --vertical_file="resources/vertical.tsv" \ --label_loss=MultiTaskCrossEntropyLoss \ --base_learning_rate=0.0008 \ --vertical_loss_percent=0.4 \ --moe_num_mixtures=8 \ --num_readers=1 \ --run_once=True fi done ================================================ FILE: youtube-8m-wangheda/eval_scripts/eval-lstmmem-cnnlstm.sh ================================================ GPU_ID="1" EVERY=1000 MODEL=CnnLstmMemoryModel MODEL_DIR="../model/cnnlstmmemory1024_moe8" start=$1 DIR="$(pwd)" for checkpoint in $(cd $MODEL_DIR && python ${DIR}/training_utils/select.py $EVERY); do echo $checkpoint; if [ $checkpoint -gt $start ]; then echo $checkpoint; CUDA_VISIBLE_DEVICES=$GPU_ID python eval.py \ --train_dir="$MODEL_DIR" \ --model_checkpoint_path="${MODEL_DIR}/model.ckpt-${checkpoint}" \ --eval_data_pattern="/Youtube-8M/data/frame/validate/validatea*" \ --frame_features=True \ --feature_names="rgb,audio" \ --feature_sizes="1024,128" \ --batch_size=32 \ --model=$MODEL \ --moe_num_mixtures=8 \ --lstm_layers=2 \ --rnn_swap_memory=True \ --run_once=True fi done ================================================ FILE: youtube-8m-wangheda/eval_scripts/eval-lstmmem-deep-chain.sh ================================================ GPU_ID=1 EVERY=500 MODEL=LstmMemoryModel MODEL_DIR="../model/lstmmemory1024_moe4_deep_chain" start=$1 DIR="$(pwd)" for checkpoint in $(cd $MODEL_DIR && python ${DIR}/training_utils/select.py $EVERY); do echo $checkpoint; if [ $checkpoint -gt $start ]; then echo $checkpoint; CUDA_VISIBLE_DEVICES=$GPU_ID python eval.py \ --train_dir="$MODEL_DIR" \ --model_checkpoint_path="${MODEL_DIR}/model.ckpt-${checkpoint}" \ --eval_data_pattern="/run/shm/validate/validatea*" \ --frame_features=True \ --feature_names="rgb,audio" \ --feature_sizes="1024,128" \ --model=LstmMemoryModel \ --lstm_layers=2 \ --lstm_cells=1024 \ --video_level_classifier_model=DeepChainModel \ --deep_chain_relu_cells=200 \ --deep_chain_layers=1 \ --moe_num_mixtures=4 \ --label_loss=MultiTaskCrossEntropyLoss \ --multitask=True \ --support_type="label" \ --num_supports=4716 \ --support_loss_percent=0.1 \ --batch_size=128 \ --num_readers=4 \ --run_once=True fi done ================================================ FILE: youtube-8m-wangheda/eval_scripts/eval-lstmmem-deep-combine-chain-length.sh ================================================ GPU_ID="0" EVERY=1000 MODEL=LstmMemoryModel MODEL_DIR="../model/lstmmem1024_deep_combine_chain_length" \ start=$1 DIR="$(pwd)" for checkpoint in $(cd $MODEL_DIR && python ${DIR}/training_utils/select.py $EVERY); do echo $checkpoint; if [ $checkpoint -gt $start ]; then echo $checkpoint; CUDA_VISIBLE_DEVICES=$GPU_ID python eval.py \ --train_dir="$MODEL_DIR" \ --model_checkpoint_path="${MODEL_DIR}/model.ckpt-${checkpoint}" \ --eval_data_pattern="/Youtube-8M/data/frame/validate/validatea*" \ --frame_features=True \ --feature_names="rgb,audio" \ --feature_sizes="1024,128" \ --model=LstmMemoryModel \ --video_level_classifier_model=DeepCombineChainModel \ --moe_num_mixtures=4 \ --deep_chain_relu_cells=256 \ --deep_chain_layers=1 \ --deep_chain_use_length=True \ --label_loss=MultiTaskCrossEntropyLoss \ --multitask=True \ --support_type="label" \ --num_supports=4716 \ --support_loss_percent=0.05 \ --batch_size=128 \ --model=$MODEL \ --lstm_layers=2 \ --rnn_swap_memory=True \ --run_once=True fi done ================================================ FILE: youtube-8m-wangheda/eval_scripts/eval-lstmmem-dropout.sh ================================================ GPU_ID=0 EVERY=1000 MODEL=LstmMemoryModel MODEL_DIR="../model/lstmmemory1024_moe8_dropout" start=$1 DIR="$(pwd)" for checkpoint in $(cd $MODEL_DIR && python ${DIR}/training_utils/select.py $EVERY); do echo $checkpoint; if [ $checkpoint -gt $start ]; then echo $checkpoint; CUDA_VISIBLE_DEVICES=$GPU_ID python eval.py \ --train_dir="$MODEL_DIR" \ --model_checkpoint_path="${MODEL_DIR}/model.ckpt-${checkpoint}" \ --eval_data_pattern="/Youtube-8M/data/frame/validate/validatea*" \ --frame_features=True \ --feature_names="rgb,audio" \ --feature_sizes="1024,128" \ --model=$MODEL \ --moe_num_mixtures=8 \ --lstm_cells=1024 \ --lstm_layers=2 \ --dropout=True \ --batch_size=64 \ --run_once=True fi done ================================================ FILE: youtube-8m-wangheda/eval_scripts/eval-lstmmem-feature.sh ================================================ GPU_ID="1" EVERY=100 MODEL=LstmMemoryModel MODEL_DIR="../model/lstmmemory_moe8_feature" start=$1 DIR="$(pwd)" for checkpoint in $(cd $MODEL_DIR && python ${DIR}/training_utils/select.py $EVERY); do echo $checkpoint; if [ $checkpoint -gt $start ]; then echo $checkpoint; CUDA_VISIBLE_DEVICES=$GPU_ID python eval.py \ --train_dir="$MODEL_DIR" \ --model_checkpoint_path="${MODEL_DIR}/model.ckpt-${checkpoint}" \ --eval_data_pattern="/Youtube-8M/data/frame/validate/validatea*" \ --frame_features=True \ --feature_names="rgb,audio" \ --feature_sizes="1024,128" \ --feature_transformer=EngineerTransformer \ --engineer_types="identical,avg,std,diff" \ --batch_size=32 \ --model=$MODEL \ --moe_num_mixtures=8 \ --lstm_layers=2 \ --rnn_swap_memory=True \ --run_once=True fi done ================================================ FILE: youtube-8m-wangheda/eval_scripts/eval-lstmmem-input-chain.sh ================================================ GPU_ID=1 EVERY=500 MODEL=LstmMemoryInputChainModel MODEL_DIR="../model/lstmmemory1024_moe8_input_chain" start=$1 DIR="$(pwd)" for checkpoint in $(cd $MODEL_DIR && python ${DIR}/training_utils/select.py $EVERY); do echo $checkpoint; if [ $checkpoint -gt $start ]; then echo $checkpoint; CUDA_VISIBLE_DEVICES=$GPU_ID python eval.py \ --train_dir="$MODEL_DIR" \ --model_checkpoint_path="${MODEL_DIR}/model.ckpt-${checkpoint}" \ --eval_data_pattern="/Youtube-8M/data/frame/validate/validatea*" \ --frame_features=True \ --feature_names="rgb,audio" \ --feature_sizes="1024,128" \ --batch_size=64 \ --model=$MODEL \ --lstm_layers=2 \ --lstm_cells=1024 \ --multitask=True \ --moe_num_mixtures=8 \ --num_supports=25 \ --num_verticals=25 \ --vertical_file="resources/vertical.tsv" \ --label_loss=MultiTaskCrossEntropyLoss \ --run_once=True fi done ================================================ FILE: youtube-8m-wangheda/eval_scripts/eval-lstmmem-input-noise.sh ================================================ noise_level=$1 start=$2 EVERY=1000 MODEL=LstmMemoryModel MODEL_DIR="../model/lstmmemory1024_moe8_input_noise/noise_level_$noise_level" DIR="$(pwd)" for checkpoint in $(cd $MODEL_DIR && python ${DIR}/training_utils/select.py $EVERY); do echo $checkpoint; if [ $checkpoint -gt $start ]; then echo $checkpoint; python eval.py \ --train_dir="$MODEL_DIR" \ --model_checkpoint_path="${MODEL_DIR}/model.ckpt-${checkpoint}" \ --eval_data_pattern="/Youtube-8M/data/frame/validate/validatea*" \ --frame_features=True \ --feature_names="rgb,audio" \ --feature_sizes="1024,128" \ --model=LstmMemoryModel \ --lstm_layers=2 \ --lstm_cells=1024 \ --moe_num_mixtures=8 \ --batch_size=128 \ --num_readers=4 \ --run_once=True fi done ================================================ FILE: youtube-8m-wangheda/eval_scripts/eval-lstmmem-l2norm.sh ================================================ GPU_ID=1 EVERY=500 MODEL=LstmMemoryNormalizationModel MODEL_DIR="../model/lstmmemory1024_moe8_l2norm" start=$1 DIR="$(pwd)" for checkpoint in $(cd $MODEL_DIR && python ${DIR}/training_utils/select.py $EVERY); do echo $checkpoint; if [ $checkpoint -gt $start ]; then echo $checkpoint; CUDA_VISIBLE_DEVICES=$GPU_ID python eval.py \ --train_dir="$MODEL_DIR" \ --model_checkpoint_path="${MODEL_DIR}/model.ckpt-${checkpoint}" \ --eval_data_pattern="/Youtube-8M-validate/validatea*" \ --frame_features=True \ --feature_names="rgb,audio" \ --feature_sizes="1024,128" \ --batch_size=128 \ --model=$MODEL \ --lstm_normalization="l2_normalize" \ --feature_transformer="IdenticalTransformer" \ --lstm_cells=1024 \ --lstm_layers=2 \ --moe_num_mixtures=8 \ --num_readers=1 \ --rnn_swap_memory=True \ --run_once=True fi done ================================================ FILE: youtube-8m-wangheda/eval_scripts/eval-lstmmem-layernorm.sh ================================================ GPU_ID=0 EVERY=500 MODEL=LayerNormLstmMemoryModel MODEL_DIR="../model/layernormlstmmemory1024_moe8" start=$1 DIR="$(pwd)" for checkpoint in $(cd $MODEL_DIR && python ${DIR}/training_utils/select.py $EVERY); do echo $checkpoint; if [ $checkpoint -gt $start ]; then echo $checkpoint; CUDA_VISIBLE_DEVICES=$GPU_ID python eval.py \ --train_dir="$MODEL_DIR" \ --model_checkpoint_path="${MODEL_DIR}/model.ckpt-${checkpoint}" \ --eval_data_pattern="/Youtube-8M-validate/validatea*" \ --frame_features=True \ --feature_names="rgb,audio" \ --feature_sizes="1024,128" \ --batch_size=128 \ --model=$MODEL \ --lstm_cells=1024 \ --lstm_layers=2 \ --moe_num_mixtures=8 \ --num_readers=1 \ --rnn_swap_memory=True \ --run_once=True fi done ================================================ FILE: youtube-8m-wangheda/eval_scripts/eval-lstmmem-lowres.sh ================================================ GPU_ID="0" EVERY=1000 MODEL=LstmMemoryModel MODEL_DIR="../model/lstmmemory_moe8_lowres" start=$1 DIR="$(pwd)" for checkpoint in $(cd $MODEL_DIR && python ${DIR}/training_utils/select.py $EVERY); do echo $checkpoint; if [ $checkpoint -gt $start ]; then echo $checkpoint; CUDA_VISIBLE_DEVICES=$GPU_ID python eval.py \ --train_dir="$MODEL_DIR" \ --model_checkpoint_path="${MODEL_DIR}/model.ckpt-${checkpoint}" \ --eval_data_pattern="/Youtube-8M/data/frame/validate/validatea*" \ --frame_features=True \ --feature_names="rgb,audio" \ --feature_sizes="1024,128" \ --batch_size=32 \ --model=$MODEL \ --moe_num_mixtures=8 \ --lstm_layers=2 \ --feature_transformer=ResolutionTransformer \ --rnn_swap_memory=True \ --run_once=True fi done ================================================ FILE: youtube-8m-wangheda/eval_scripts/eval-lstmmem-no-transform.sh ================================================ GPU_ID=1 EVERY=500 MODEL=LstmMemoryModel MODEL_DIR="../model/lstmmemory1024_moe8_no_transform" start=$1 DIR="$(pwd)" for checkpoint in $(cd $MODEL_DIR && python ${DIR}/training_utils/select.py $EVERY); do echo $checkpoint; if [ $checkpoint -gt $start ]; then echo $checkpoint; CUDA_VISIBLE_DEVICES=$GPU_ID python eval.py \ --train_dir="$MODEL_DIR" \ --model_checkpoint_path="${MODEL_DIR}/model.ckpt-${checkpoint}" \ --eval_data_pattern="/Youtube-8M/data/frame/validate/validatea*" \ --frame_features=True \ --feature_names="rgb,audio" \ --feature_sizes="1024,128" \ --feature_transformer=IdenticalTransformer \ --batch_size=64 \ --model=$MODEL \ --lstm_cells=1024 \ --moe_num_mixtures=8 \ --rnn_swap_memory=True \ --run_once=True fi done ================================================ FILE: youtube-8m-wangheda/eval_scripts/eval-lstmmem-noise.sh ================================================ GPU_ID=1 EVERY=1000 MODEL=LstmMemoryModel MODEL_DIR="../model/lstmmemory1024_moe8_noise" start=$1 DIR="$(pwd)" for checkpoint in $(cd $MODEL_DIR && python ${DIR}/training_utils/select.py $EVERY); do echo $checkpoint; if [ $checkpoint -gt $start ]; then echo $checkpoint; CUDA_VISIBLE_DEVICES=$GPU_ID python eval.py \ --train_dir="$MODEL_DIR" \ --model_checkpoint_path="${MODEL_DIR}/model.ckpt-${checkpoint}" \ --eval_data_pattern="/Youtube-8M/data/frame/validate/validatea*" \ --frame_features=True \ --feature_names="rgb,audio" \ --feature_sizes="1024,128" \ --model=$MODEL \ --moe_num_mixtures=8 \ --lstm_cells=1024 \ --lstm_layers=2 \ --batch_size=64 \ --rnn_swap_memory=True \ --run_once=True fi done ================================================ FILE: youtube-8m-wangheda/eval_scripts/eval-lstmmem-parallel.sh ================================================ start=$1 GPU_ID=0 MODEL=LstmParallelMemoryModel MODEL_DIR="../model/lstmparallelmemory1024_moe8" for checkpoint in $(for filename in $MODEL_DIR/model.ckpt-*.meta; do echo $filename | grep -o "ckpt-[0123456789]*.meta" | cut -d '-' -f 2 | cut -d '.' -f 1; done | sort -n); do if [ $checkpoint -gt $start ]; then echo $checkpoint; CUDA_VISIBLE_DEVICES=$GPU_ID python eval.py \ --train_dir="$MODEL_DIR" \ --model_checkpoint_path="${MODEL_DIR}/model.ckpt-${checkpoint}" \ --eval_data_pattern="/Youtube-8M/data/frame/validate/validatea*" \ --frame_features=True \ --feature_names="rgb,audio" \ --feature_sizes="1024,128" \ --lstm_cells="1024,128" \ --batch_size=128 \ --model=$MODEL \ --moe_num_mixtures=8 \ --rnn_swap_memory=True \ --run_once=True fi done ================================================ FILE: youtube-8m-wangheda/eval_scripts/eval-lstmmem-shortcut-chain-freq.sh ================================================ GPU_ID=1 EVERY=500 MODEL=LstmMemoryModel MODEL_DIR="../model/lstmmemory1024_moe8_shortcut_chain_freq" start=$1 DIR="$(pwd)" for checkpoint in $(cd $MODEL_DIR && python ${DIR}/training_utils/select.py $EVERY); do echo $checkpoint; if [ $checkpoint -gt $start ]; then echo $checkpoint; CUDA_VISIBLE_DEVICES=$GPU_ID python eval.py \ --train_dir="$MODEL_DIR" \ --model_checkpoint_path="${MODEL_DIR}/model.ckpt-${checkpoint}" \ --eval_data_pattern="/Youtube-8M/data/frame/validate/validatea*" \ --frame_features=True \ --feature_names="rgb,audio" \ --feature_sizes="1024,128" \ --batch_size=64 \ --model=LstmMemoryModel \ --video_level_classifier_model=ShortcutChainSupportReluMoeModel \ --lstm_layers=2 \ --lstm_cells=1024 \ --multitask=True \ --support_type="frequent" \ --num_supports=200 \ --num_frequents=200 \ --label_loss=MultiTaskCrossEntropyLoss \ --base_learning_rate=0.0008 \ --vertical_loss_percent=0.4 \ --moe_num_mixtures=8 \ --num_readers=1 \ --run_once=True fi done ================================================ FILE: youtube-8m-wangheda/eval_scripts/eval-lstmmem2048.sh ================================================ GPU_ID=0 EVERY=1000 MODEL=LstmMemoryModel MODEL_DIR="../model/lstmmemory2048_moe4_batch128" start=$1 DIR="$(pwd)" for checkpoint in $(cd $MODEL_DIR && python ${DIR}/training_utils/select.py $EVERY); do echo $checkpoint; if [ $checkpoint -gt $start ]; then echo $checkpoint; CUDA_VISIBLE_DEVICES=$GPU_ID python eval.py \ --train_dir="$MODEL_DIR" \ --model_checkpoint_path="${MODEL_DIR}/model.ckpt-${checkpoint}" \ --eval_data_pattern="/Youtube-8M/data/frame/validate/validatea*" \ --frame_features=True \ --feature_names="rgb,audio" \ --feature_sizes="1024,128" \ --model=$MODEL \ --moe_num_mixtures=4 \ --lstm_cells=2048 \ --lstm_layers=2 \ --batch_size=64 \ --rnn_swap_memory=True \ --run_once=True fi done ================================================ FILE: youtube-8m-wangheda/eval_scripts/eval-lstmmemory-audio.sh ================================================ GPU_ID="0" EVERY=2000 MODEL=LstmMemoryModel MODEL_DIR="../model/audio_lstmmemory1024_layer1_moe8" start=$1 DIR="$(pwd)" for checkpoint in $(cd $MODEL_DIR && python ${DIR}/training_utils/select.py $EVERY); do echo $checkpoint; if [ $checkpoint -gt $start ]; then echo $checkpoint; CUDA_VISIBLE_DEVICES=$GPU_ID python eval.py \ --train_dir="$MODEL_DIR" \ --model_checkpoint_path="${MODEL_DIR}/model.ckpt-${checkpoint}" \ --eval_data_pattern="/Youtube-8M/data/frame/validate/validatea*" \ --frame_features=True \ --feature_names="audio" \ --feature_sizes="128" \ --model=$MODEL \ --lstm_cells=512 \ --lstm_layers=1 \ --moe_num_mixtures=8 \ --rnn_swap_memory=False \ --batch_size=128 \ --run_once=True fi done ================================================ FILE: youtube-8m-wangheda/eval_scripts/eval-lstmmemory-layer1.sh ================================================ GPU_ID="1" EVERY=2000 MODEL=LstmMemoryModel MODEL_DIR="../model/lstmmemory1024_layer1_moe8" start=$1 DIR="$(pwd)" for checkpoint in $(cd $MODEL_DIR && python ${DIR}/training_utils/select.py $EVERY); do echo $checkpoint; if [ $checkpoint -gt $start ]; then echo $checkpoint; CUDA_VISIBLE_DEVICES=$GPU_ID python eval.py \ --train_dir="$MODEL_DIR" \ --model_checkpoint_path="${MODEL_DIR}/model.ckpt-${checkpoint}" \ --eval_data_pattern="/Youtube-8M/data/frame/validate/validatea*" \ --frame_features=True \ --feature_names="rgb,audio" \ --feature_sizes="1024,128" \ --model=$MODEL \ --lstm_cells=1024 \ --lstm_layers=1 \ --moe_num_mixtures=8 \ --rnn_swap_memory=False \ --batch_size=128 \ --run_once=True fi done ================================================ FILE: youtube-8m-wangheda/eval_scripts/eval-lstmmemory.sh ================================================ GPU_ID="1" EVERY=100 MODEL=LstmMemoryModel MODEL_DIR="../model/lstmmemory_moe8_batch256" start=$1 DIR="$(pwd)" for checkpoint in $(cd $MODEL_DIR && python ${DIR}/training_utils/select.py $EVERY); do echo $checkpoint; if [ $checkpoint -gt $start ]; then echo $checkpoint; CUDA_VISIBLE_DEVICES=$GPU_ID python eval.py \ --train_dir="$MODEL_DIR" \ --model_checkpoint_path="${MODEL_DIR}/model.ckpt-${checkpoint}" \ --eval_data_pattern="/Youtube-8M/data/frame/validate/validatea*" \ --frame_features=True \ --feature_names="rgb,audio" \ --feature_sizes="1024,128" \ --batch_size=32 \ --model=$MODEL \ --moe_num_mixtures=8 \ --lstm_layers=2 \ --rnn_swap_memory=True \ --run_once=True fi done ================================================ FILE: youtube-8m-wangheda/eval_scripts/eval-lstmoutput-parallel.sh ================================================ GPU_ID=1 EVERY=500 MODEL=LstmParallelFinaloutputModel MODEL_DIR="../model/lstmparallelfinaloutput1024_moe8" start=$1 DIR="$(pwd)" for checkpoint in $(cd $MODEL_DIR && python ${DIR}/training_utils/select.py $EVERY); do echo $checkpoint; if [ $checkpoint -gt $start ]; then echo $checkpoint; CUDA_VISIBLE_DEVICES=$GPU_ID python eval.py \ --train_dir="$MODEL_DIR" \ --model_checkpoint_path="${MODEL_DIR}/model.ckpt-${checkpoint}" \ --eval_data_pattern="/Youtube-8M/data/frame/validate/validatea*" \ --frame_features=True \ --feature_names="rgb,audio" \ --feature_sizes="1024,128" \ --batch_size=64 \ --model=$MODEL\ --lstm_layers=2 \ --lstm_cells="1024,128" \ --moe_num_mixtures=8 \ --num_readers=1 \ --run_once=True fi done ================================================ FILE: youtube-8m-wangheda/eval_scripts/eval-mem.sh ================================================ GPU_ID=1 EVERY=500 MODEL=LstmMemoryModel MODEL_DIR="../model/lstmmemory1024_moe8" start=$1 DIR="$(pwd)" for checkpoint in $(cd $MODEL_DIR && python ${DIR}/training_utils/select.py $EVERY); do echo $checkpoint; if [ $checkpoint -gt $start ]; then echo $checkpoint; CUDA_VISIBLE_DEVICES=$GPU_ID python eval.py \ --train_dir="$MODEL_DIR" \ --model_checkpoint_path="${MODEL_DIR}/model.ckpt-${checkpoint}" \ --eval_data_pattern="/Youtube-8M-validate/validatea*" \ --frame_features=True \ --feature_names="rgb,audio" \ --feature_sizes="1024,128" \ --batch_size=128 \ --model=$MODEL \ --lstm_cells=1024 \ --moe_num_mixtures=8 \ --num_readers=1 \ --run_once=True fi done ================================================ FILE: youtube-8m-wangheda/eval_scripts/eval-mm-lstm.sh ================================================ start=$1 GPU_ID=0 MODEL=MatchingMatrixLstmMemoryModel MODEL_DIR="../model/mmlstmmemory1024_moe8" for checkpoint in $(for filename in $MODEL_DIR/model.ckpt-*.meta; do echo $filename | grep -o "ckpt-[0123456789]*.meta" | cut -d '-' -f 2 | cut -d '.' -f 1; done | sort -n); do if [ $checkpoint -gt $start ]; then echo $checkpoint; CUDA_VISIBLE_DEVICES=$GPU_ID python eval.py \ --train_dir="$MODEL_DIR" \ --model_checkpoint_path="${MODEL_DIR}/model.ckpt-${checkpoint}" \ --eval_data_pattern="/Youtube-8M/data/frame/validate/validatea*" \ --frame_features=True \ --feature_names="rgb,audio" \ --feature_sizes="1024,128" \ --batch_size=128 \ --model=$MODEL \ --mm_label_embedding=256 \ --moe_num_mixtures=8 \ --lstm_cells=1024 \ --lstm_layers=2 \ --rnn_swap_memory=True \ --run_once=True fi done ================================================ FILE: youtube-8m-wangheda/eval_scripts/eval-moe-baseline.sh ================================================ GPU_ID=1 EVERY=100 MODEL=MoeModel MODEL_DIR="../model/video_moe16_baseline" start=$1 DIR="$(pwd)" for checkpoint in $(cd $MODEL_DIR && python ${DIR}/training_utils/select.py $EVERY); do echo $checkpoint; if [ $checkpoint -gt $start ]; then echo $checkpoint; CUDA_VISIBLE_DEVICES=$GPU_ID python eval.py \ --train_dir="$MODEL_DIR" \ --model_checkpoint_path="${MODEL_DIR}/model.ckpt-${checkpoint}" \ --eval_data_pattern="/Youtube-8M/data/video/validate/validatea*" \ --frame_features=False \ --feature_names="mean_rgb,mean_audio" \ --feature_sizes="1024,128" \ --label_loss=CrossEntropyLoss \ --batch_size=128 \ --model=$MODEL \ --moe_num_mixtures=16 \ --run_once=True fi done ================================================ FILE: youtube-8m-wangheda/eval_scripts/eval-moe-batchagreement1.sh ================================================ GPU_ID=1 EVERY=100 MODEL=MoeModel MODEL_DIR="../model/video_moe16_batchagreement0.001" start=$1 DIR="$(pwd)" for checkpoint in $(cd $MODEL_DIR && python ${DIR}/training_utils/select.py $EVERY); do echo $checkpoint; if [ $checkpoint -gt $start ]; then echo $checkpoint; CUDA_VISIBLE_DEVICES=$GPU_ID python eval.py \ --train_dir="$MODEL_DIR" \ --model_checkpoint_path="${MODEL_DIR}/model.ckpt-${checkpoint}" \ --eval_data_pattern="/Youtube-8M/data/video/validate/validatea*" \ --frame_features=False \ --feature_names="mean_rgb,mean_audio" \ --feature_sizes="1024,128" \ --label_loss=BatchAgreementCrossEntropyLoss \ --batch_agreement=0.001 \ --batch_size=128 \ --model=$MODEL \ --moe_num_mixtures=16 \ --run_once=True fi done ================================================ FILE: youtube-8m-wangheda/eval_scripts/eval-moe-batchagreement2.sh ================================================ GPU_ID=1 EVERY=100 MODEL=MoeModel MODEL_DIR="../model/video_moe16_batchagreement0.003" start=$1 DIR="$(pwd)" for checkpoint in $(cd $MODEL_DIR && python ${DIR}/training_utils/select.py $EVERY); do echo $checkpoint; if [ $checkpoint -gt $start ]; then echo $checkpoint; CUDA_VISIBLE_DEVICES=$GPU_ID python eval.py \ --train_dir="$MODEL_DIR" \ --model_checkpoint_path="${MODEL_DIR}/model.ckpt-${checkpoint}" \ --eval_data_pattern="/Youtube-8M/data/video/validate/validatea*" \ --frame_features=False \ --feature_names="mean_rgb,mean_audio" \ --feature_sizes="1024,128" \ --label_loss=BatchAgreementCrossEntropyLoss \ --batch_agreement=0.003 \ --batch_size=128 \ --model=$MODEL \ --moe_num_mixtures=16 \ --run_once=True fi done ================================================ FILE: youtube-8m-wangheda/eval_scripts/eval-moe-batchagreement3.sh ================================================ GPU_ID=1 EVERY=100 MODEL=MoeModel MODEL_DIR="../model/video_moe16_batchagreement0.01" start=$1 DIR="$(pwd)" for checkpoint in $(cd $MODEL_DIR && python ${DIR}/training_utils/select.py $EVERY); do echo $checkpoint; if [ $checkpoint -gt $start ]; then echo $checkpoint; CUDA_VISIBLE_DEVICES=$GPU_ID python eval.py \ --train_dir="$MODEL_DIR" \ --model_checkpoint_path="${MODEL_DIR}/model.ckpt-${checkpoint}" \ --eval_data_pattern="/Youtube-8M/data/video/validate/validatea*" \ --frame_features=False \ --feature_names="mean_rgb,mean_audio" \ --feature_sizes="1024,128" \ --label_loss=BatchAgreementCrossEntropyLoss \ --batch_agreement=0.01 \ --batch_size=128 \ --model=$MODEL \ --moe_num_mixtures=16 \ --run_once=True fi done ================================================ FILE: youtube-8m-wangheda/eval_scripts/eval-moe-model.sh ================================================ GPU_ID="1" EVERY=100 MODEL=MoeModel MODEL_DIR="../model/video_moe16_model" start=$1 DIR="$(pwd)" for checkpoint in $(cd $MODEL_DIR && python ${DIR}/training_utils/select.py $EVERY); do echo $checkpoint; if [ $checkpoint -gt $start ]; then echo $checkpoint; CUDA_VISIBLE_DEVICES=$GPU_ID python eval.py \ --train_dir="$MODEL_DIR" \ --model_checkpoint_path="${MODEL_DIR}/model.ckpt-${checkpoint}" \ --eval_data_pattern="/Youtube-8M/data/video/validate/validatea*" \ --frame_features=False \ --feature_names="mean_rgb,mean_audio" \ --feature_sizes="1024,128" \ --model=$MODEL \ --moe_num_mixtures=16 \ --batch_size=1024 \ --run_once=True fi done ================================================ FILE: youtube-8m-wangheda/eval_scripts/eval-moe-topk-batchagreement1.sh ================================================ GPU_ID=1 EVERY=100 MODEL=MoeModel MODEL_DIR="../model/video_moe16_topk_batchagreement0.3" start=$1 DIR="$(pwd)" for checkpoint in $(cd $MODEL_DIR && python ${DIR}/training_utils/select.py $EVERY); do echo $checkpoint; if [ $checkpoint -gt $start ]; then echo $checkpoint; CUDA_VISIBLE_DEVICES=$GPU_ID python eval.py \ --train_dir="$MODEL_DIR" \ --model_checkpoint_path="${MODEL_DIR}/model.ckpt-${checkpoint}" \ --eval_data_pattern="/Youtube-8M/data/video/validate/validatea*" \ --frame_features=False \ --feature_names="mean_rgb,mean_audio" \ --feature_sizes="1024,128" \ --label_loss=TopKBatchAgreementCrossEntropyLoss \ --batch_agreement=0.3 \ --batch_size=128 \ --model=$MODEL \ --moe_num_mixtures=16 \ --run_once=True fi done ================================================ FILE: youtube-8m-wangheda/eval_scripts/eval-moe-topk-batchagreement2.sh ================================================ GPU_ID=1 EVERY=100 MODEL=MoeModel MODEL_DIR="../model/video_moe16_topk_batchagreement1" start=$1 DIR="$(pwd)" for checkpoint in $(cd $MODEL_DIR && python ${DIR}/training_utils/select.py $EVERY); do echo $checkpoint; if [ $checkpoint -gt $start ]; then echo $checkpoint; CUDA_VISIBLE_DEVICES=$GPU_ID python eval.py \ --train_dir="$MODEL_DIR" \ --model_checkpoint_path="${MODEL_DIR}/model.ckpt-${checkpoint}" \ --eval_data_pattern="/Youtube-8M/data/video/validate/validatea*" \ --frame_features=False \ --feature_names="mean_rgb,mean_audio" \ --feature_sizes="1024,128" \ --label_loss=TopKBatchAgreementCrossEntropyLoss \ --batch_agreement=1 \ --batch_size=128 \ --model=$MODEL \ --moe_num_mixtures=16 \ --run_once=True fi done ================================================ FILE: youtube-8m-wangheda/eval_scripts/eval-moe-topk-batchagreement3.sh ================================================ GPU_ID=1 EVERY=100 MODEL=MoeModel MODEL_DIR="../model/video_moe16_topk_batchagreement3" start=$1 DIR="$(pwd)" for checkpoint in $(cd $MODEL_DIR && python ${DIR}/training_utils/select.py $EVERY); do echo $checkpoint; if [ $checkpoint -gt $start ]; then echo $checkpoint; CUDA_VISIBLE_DEVICES=$GPU_ID python eval.py \ --train_dir="$MODEL_DIR" \ --model_checkpoint_path="${MODEL_DIR}/model.ckpt-${checkpoint}" \ --eval_data_pattern="/Youtube-8M/data/video/validate/validatea*" \ --frame_features=False \ --feature_names="mean_rgb,mean_audio" \ --feature_sizes="1024,128" \ --label_loss=TopKBatchAgreementCrossEntropyLoss \ --batch_agreement=3 \ --batch_size=128 \ --model=$MODEL \ --moe_num_mixtures=16 \ --run_once=True fi done ================================================ FILE: youtube-8m-wangheda/eval_scripts/eval-multi-lstmmem-deep-chain.sh ================================================ GPU_ID=1 EVERY=500 MODEL=LstmMemoryDeepChainModel MODEL_DIR="../model/multilstmmemory1024_moe4_deep_chain" start=$1 DIR="$(pwd)" for checkpoint in $(cd $MODEL_DIR && python ${DIR}/training_utils/select.py $EVERY); do echo $checkpoint; if [ $checkpoint -gt $start ]; then echo $checkpoint; CUDA_VISIBLE_DEVICES=$GPU_ID python eval.py \ --train_dir="$MODEL_DIR" \ --model_checkpoint_path="${MODEL_DIR}/model.ckpt-${checkpoint}" \ --eval_data_pattern="/Youtube-8M/data/frame/validate/validatea*" \ --frame_features=True \ --feature_names="rgb,audio" \ --feature_sizes="1024,128" \ --batch_size=64 \ --model=$MODEL \ --moe_num_mixtures=4 \ --deep_chain_relu_cells=200 \ --deep_chain_layers=1 \ --lstm_layers=2 \ --lstm_cells=1024 \ --multitask=True \ --support_type="label" \ --num_supports=4716 \ --label_loss=MultiTaskCrossEntropyLoss \ --support_loss_percent=0.2 \ --rnn_swap_memory=True \ --run_once=True fi done ================================================ FILE: youtube-8m-wangheda/eval_scripts/eval-multi-view-cnn-deep-combine-chain.sh ================================================ GPU_ID=0 EVERY=1000 MODEL=MultiViewCnnDeepCombineChainModel MODEL_DIR="../model/multi_view_cnn_deep_combine_chain" start=$1 DIR="$(pwd)" for checkpoint in $(cd $MODEL_DIR && python ${DIR}/training_utils/select.py $EVERY); do echo $checkpoint; if [ $checkpoint -gt $start ]; then echo $checkpoint; CUDA_VISIBLE_DEVICES=$GPU_ID python eval.py \ --train_dir="$MODEL_DIR" \ --model_checkpoint_path="${MODEL_DIR}/model.ckpt-${checkpoint}" \ --eval_data_pattern="/Youtube-8M/data/frame/validate/validatea*" \ --frame_features=True \ --feature_names="rgb,audio" \ --feature_sizes="1024,128" \ --model=$MODEL \ --moe_num_mixtures=4 \ --deep_chain_layers=3 \ --deep_chain_relu_cells=64 \ --label_loss=MultiTaskCrossEntropyLoss \ --multitask=True \ --support_type="label,label,label" \ --support_loss_percent=0.05 \ --batch_size=256 \ --run_once=True fi done ================================================ FILE: youtube-8m-wangheda/eval_scripts/eval-multires-lstm-deep-combine-chain.sh ================================================ GPU_ID=0 EVERY=1000 MODEL=MultiresLstmMemoryDeepCombineChainModel MODEL_DIR="../model/multires_lstm_deep_combine_chain" start=$1 DIR="$(pwd)" for checkpoint in $(cd $MODEL_DIR && python ${DIR}/training_utils/select.py $EVERY); do echo $checkpoint; if [ $checkpoint -gt $start ]; then echo $checkpoint; CUDA_VISIBLE_DEVICES=$GPU_ID python eval.py \ --train_dir="$MODEL_DIR" \ --model_checkpoint_path="${MODEL_DIR}/model.ckpt-${checkpoint}" \ --eval_data_pattern="/Youtube-8M/data/frame/validate/validatea*" \ --frame_features=True \ --feature_names="rgb,audio" \ --feature_sizes="1024,128" \ --model=$MODEL \ --lstm_cells="512,64" \ --deep_chain_layers=4 \ --deep_chain_relu_cells=256 \ --moe_num_mixtures=4 \ --label_loss=CrossEntropyLoss \ --dropout=True \ --batch_size=128 \ --feature_transformer=IdenticalTransformer \ --rnn_swap_memory=True \ --run_once=True fi done ================================================ FILE: youtube-8m-wangheda/eval_scripts/eval-multitask-ce.sh ================================================ GPU_ID=1 EVERY=2000 MODEL=LstmMemoryMultitaskModel MODEL_DIR="../model/lstmmemory1024_moe8_multitask_ce" start=$1 DIR="$(pwd)" for checkpoint in $(cd $MODEL_DIR && python ${DIR}/training_utils/select.py $EVERY); do echo $checkpoint; if [ $checkpoint -gt $start ]; then echo $checkpoint; CUDA_VISIBLE_DEVICES=$GPU_ID python eval.py \ --train_dir="$MODEL_DIR" \ --model_checkpoint_path="${MODEL_DIR}/model.ckpt-${checkpoint}" \ --eval_data_pattern="/Youtube-8M/data/frame/validate/validatea*" \ --frame_features=True \ --feature_names="rgb,audio" \ --feature_sizes="1024,128" \ --batch_size=32 \ --model=$MODEL \ --lstm_layers=2 \ --lstm_cells=1024 \ --multitask=True \ --moe_num_mixtures=8 \ --num_supports=25 \ --num_verticals=25 \ --vertical_file="resources/vertical.tsv" \ --run_once=True fi done ================================================ FILE: youtube-8m-wangheda/eval_scripts/eval-multitask.sh ================================================ GPU_ID=0 EVERY=2000 MODEL=LstmMemoryMultitaskModel MODEL_DIR="../model/lstmmemory1024_moe8_multitask" start=$1 DIR="$(pwd)" for checkpoint in $(cd $MODEL_DIR && python ${DIR}/training_utils/select.py $EVERY); do echo $checkpoint; if [ $checkpoint -gt $start ]; then echo $checkpoint; CUDA_VISIBLE_DEVICES=$GPU_ID python eval.py \ --train_dir="$MODEL_DIR" \ --model_checkpoint_path="${MODEL_DIR}/model.ckpt-${checkpoint}" \ --eval_data_pattern="/Youtube-8M/data/frame/validate/validatea*" \ --frame_features=True \ --feature_names="rgb,audio" \ --feature_sizes="1024,128" \ --batch_size=64 \ --model=$MODEL \ --lstm_layers=2 \ --lstm_cells=1024 \ --multitask=True \ --moe_num_mixtures=8 \ --num_supports=25 \ --num_verticals=25 \ --vertical_file="resources/vertical.tsv" \ --run_once=True fi done ================================================ FILE: youtube-8m-wangheda/eval_scripts/eval-positional-cnn-dcc.sh ================================================ GPU_ID="0" EVERY=1000 MODEL=PositionalCnnDeepCombineChainModel MODEL_DIR="../model/positional_cnn_deep_combine_chain" start=$1 DIR="$(pwd)" for checkpoint in $(cd $MODEL_DIR && python ${DIR}/training_utils/select.py $EVERY); do echo $checkpoint; if [ $checkpoint -gt $start ]; then echo $checkpoint; CUDA_VISIBLE_DEVICES=$GPU_ID python eval.py \ --train_dir="$MODEL_DIR" \ --model_checkpoint_path="${MODEL_DIR}/model.ckpt-${checkpoint}" \ --eval_data_pattern="/Youtube-8M/data/frame/validate/validatea*" \ --frame_features=True \ --feature_names="rgb,audio" \ --feature_sizes="1024,128" \ --batch_size=64 \ --model=$MODEL \ --moe_num_mixtures=4 \ --deep_chain_layers=2 \ --deep_chain_relu_cells=128 \ --label_loss=MultiTaskCrossEntropyLoss \ --multitask=True \ --support_type="label,label" \ --support_loss_percent=0.1 \ --run_once=True fi done ================================================ FILE: youtube-8m-wangheda/eval_scripts/eval-stage2-logistic.sh ================================================ start=$1 GPU_ID=1 MODEL=Stage2LogisticModel MODEL_DIR="../model/video_moe16_stage2logistic" for checkpoint in $(for filename in $MODEL_DIR/model.ckpt-*.meta; do echo $filename | grep -o "ckpt-[0123456789]*.meta" | cut -d '-' -f 2 | cut -d '.' -f 1; done | sort -n); do if [ $checkpoint -gt $start ]; then echo $checkpoint; CUDA_VISIBLE_DEVICES=$GPU_ID python eval.py \ --train_dir="$MODEL_DIR" \ --model_checkpoint_path="${MODEL_DIR}/model.ckpt-${checkpoint}" \ --eval_data_pattern="/Youtube-8M/data/video/validate-validate-part1/validatea*" \ --frame_features=False \ --feature_names="predictions" \ --feature_sizes=4716 \ --batch_size=1024 \ --model=$MODEL \ --run_once=True fi done ================================================ FILE: youtube-8m-wangheda/eval_scripts/eval-stage2-moe.sh ================================================ start=$1 GPU_ID=0 MODEL=MoeModel MODEL_DIR="../model/video_moe16_stage2_moe8" for checkpoint in $(for filename in $MODEL_DIR/model.ckpt-*.meta; do echo $filename | grep -o "ckpt-[0123456789]*.meta" | cut -d '-' -f 2 | cut -d '.' -f 1; done | sort -n); do if [ $checkpoint -gt $start ]; then echo $checkpoint; CUDA_VISIBLE_DEVICES=$GPU_ID python eval.py \ --train_dir="$MODEL_DIR" \ --model_checkpoint_path="${MODEL_DIR}/model.ckpt-${checkpoint}" \ --eval_data_pattern="/Youtube-8M/data/video/validate-validate-part1/validatea*" \ --frame_features=False \ --feature_names="predictions" \ --feature_sizes=4716 \ --batch_size=1024 \ --moe_num_mixtures=8 \ --model=$MODEL \ --run_once=True fi done ================================================ FILE: youtube-8m-wangheda/eval_scripts/eval-video-deep-chain.sh ================================================ GPU_ID=1 EVERY=500 MODEL=DeepChainModel MODEL_DIR="../model/video_deep_chain_moe4" start=$1 DIR="$(pwd)" for checkpoint in $(cd $MODEL_DIR && python ${DIR}/training_utils/select.py $EVERY); do echo $checkpoint; if [ $checkpoint -gt $start ]; then echo $checkpoint; CUDA_VISIBLE_DEVICES=$GPU_ID python eval.py \ --train_dir="$MODEL_DIR" \ --model_checkpoint_path="${MODEL_DIR}/model.ckpt-${checkpoint}" \ --eval_data_pattern="/Youtube-8M/data/video/validate/validatea*" \ --frame_features=False \ --feature_names="mean_rgb,mean_audio" \ --feature_sizes="1024,128" \ --batch_size=128 \ --model=$MODEL \ --deep_chain_relu_cells=200 \ --deep_chain_layers=3 \ --moe_num_mixtures=4 \ --num_readers=1 \ --run_once=True fi done ================================================ FILE: youtube-8m-wangheda/eval_scripts/eval-video-deep-combine-addnoise.sh ================================================ noise_level=$1 start=$2 GPU_ID=0 EVERY=100 MODEL=DeepCombineChainModel MODEL_DIR="../model/video_chain_moe4_deep_combine_addnoise/after_relu_noise_level_$noise_level" DIR="$(pwd)" for checkpoint in $(cd $MODEL_DIR && python ${DIR}/training_utils/select.py $EVERY); do echo $checkpoint; if [ $checkpoint -gt $start ]; then echo $checkpoint; CUDA_VISIBLE_DEVICES=$GPU_ID python eval.py \ --train_dir="$MODEL_DIR" \ --model_checkpoint_path="${MODEL_DIR}/model.ckpt-${checkpoint}" \ --eval_data_pattern="/Youtube-8M/data/video/validate/validatea*" \ --frame_features=False \ --feature_names="mean_rgb,mean_audio" \ --feature_sizes="1024,128" \ --label_loss=MultiTaskCrossEntropyLoss \ --multitask=True \ --support_type="label,label,label,label" \ --num_supports=18864 \ --support_loss_percent=0.05 \ --model=$MODEL \ --deep_chain_relu_cells=256 \ --deep_chain_layers=4 \ --moe_num_mixtures=4 \ --batch_size=128 \ --num_readers=1 \ --run_once=True fi done ================================================ FILE: youtube-8m-wangheda/eval_scripts/eval-video-deep-combine-chain-dropout.sh ================================================ GPU_ID=0 EVERY=100 MODEL=DeepCombineChainModel MODEL_DIR="../model/video_chain_moe4_deep_combine_dropout/input_dropout_0.8" start=$1 DIR="$(pwd)" for checkpoint in $(cd $MODEL_DIR && python ${DIR}/training_utils/select.py $EVERY); do echo $checkpoint; if [ $checkpoint -gt $start ]; then echo $checkpoint; CUDA_VISIBLE_DEVICES=$GPU_ID python eval.py \ --train_dir="$MODEL_DIR" \ --model_checkpoint_path="${MODEL_DIR}/model.ckpt-${checkpoint}" \ --eval_data_pattern="/Youtube-8M/data/video/validate/validatea*" \ --frame_features=False \ --feature_names="mean_rgb,mean_audio" \ --feature_sizes="1024,128" \ --moe_num_mixtures=4 \ --deep_chain_layers=4 \ --deep_chain_relu_cells=256 \ --label_loss=MultiTaskCrossEntropyLoss \ --multitask=True \ --support_type="label,label,label,label" \ --support_loss_percent=0.05 \ --batch_size=128 \ --model=$MODEL \ --run_once=True fi done ================================================ FILE: youtube-8m-wangheda/eval_scripts/eval-video-deep-combine-chain-labelsmoothing.sh ================================================ GPU_ID=0 EVERY=100 MODEL=DeepCombineChainModel MODEL_DIR="../model/video_chain_moe4_deep_combine_labelsmoothing" start=$1 DIR="$(pwd)" for checkpoint in $(cd $MODEL_DIR && python ${DIR}/training_utils/select.py $EVERY); do echo $checkpoint; if [ $checkpoint -gt $start ]; then echo $checkpoint; CUDA_VISIBLE_DEVICES=$GPU_ID python eval.py \ --train_dir="$MODEL_DIR" \ --model_checkpoint_path="${MODEL_DIR}/model.ckpt-${checkpoint}" \ --eval_data_pattern="/Youtube-8M/data/video/validate/validatea*" \ --frame_features=False \ --feature_names="mean_rgb,mean_audio" \ --feature_sizes="1024,128" \ --moe_num_mixtures=4 \ --deep_chain_layers=4 \ --deep_chain_relu_cells=256 \ --label_loss=MultiTaskCrossEntropyLoss \ --multitask=True \ --support_type="label,label,label,label" \ --support_loss_percent=0.05 \ --batch_size=128 \ --model=$MODEL \ --run_once=True fi done ================================================ FILE: youtube-8m-wangheda/eval_scripts/eval-video-deep-combine-chain-noise.sh ================================================ noise_level=$1 start=$2 GPU_ID=0 EVERY=100 MODEL=DeepCombineChainModel MODEL_DIR="../model/video_deep_combine_chain_moe4_noise/input_noise_level_$noise_level" DIR="$(pwd)" for checkpoint in $(cd $MODEL_DIR && python ${DIR}/training_utils/select.py $EVERY); do echo $checkpoint; if [ $checkpoint -gt $start ]; then echo $checkpoint; CUDA_VISIBLE_DEVICES=$GPU_ID python eval.py \ --train_dir="$MODEL_DIR" \ --model_checkpoint_path="${MODEL_DIR}/model.ckpt-${checkpoint}" \ --eval_data_pattern="/Youtube-8M/data/video/validate/validatea*" \ --frame_features=False \ --feature_names="mean_rgb,mean_audio" \ --feature_sizes="1024,128" \ --label_loss=MultiTaskCrossEntropyLoss \ --multitask=True \ --support_type="label,label,label,label" \ --num_supports=18864 \ --support_loss_percent=0.05 \ --model=$MODEL \ --deep_chain_relu_cells=256 \ --deep_chain_layers=4 \ --moe_num_mixtures=4 \ --batch_size=1024 \ --num_readers=1 \ --run_once=True fi done ================================================ FILE: youtube-8m-wangheda/eval_scripts/eval-video-deep-combine-chain.sh ================================================ GPU_ID=0 EVERY=100 MODEL=DeepCombineChainModel MODEL_DIR="../model/video_chain_moe16_deep_combine" start=$1 DIR="$(pwd)" for checkpoint in $(cd $MODEL_DIR && python ${DIR}/training_utils/select.py $EVERY); do echo $checkpoint; if [ $checkpoint -gt $start ]; then echo $checkpoint; CUDA_VISIBLE_DEVICES=$GPU_ID python eval.py \ --train_dir="$MODEL_DIR" \ --model_checkpoint_path="${MODEL_DIR}/model.ckpt-${checkpoint}" \ --eval_data_pattern="/Youtube-8M/data/video/validate/validatea*" \ --frame_features=False \ --feature_names="mean_rgb,mean_audio" \ --feature_sizes="1024,128" \ --moe_num_mixtures=2 \ --deep_chain_layers=16 \ --deep_chain_relu_cells=256 \ --label_loss=MultiTaskCrossEntropyLoss \ --multitask=True \ --support_type="label,label,label,label,label,label,label,label,label,label,label,label,label,label,label,label" \ --support_loss_percent=0.0125 \ --batch_size=512 \ --model=$MODEL \ --run_once=True fi done ================================================ FILE: youtube-8m-wangheda/eval_scripts/eval-video-distillchain-video-dcc.sh ================================================ start=$1 GPU_ID=0 MODEL=DistillchainDeepCombineChainModel MODEL_DIR="../model/distillchain_video_dcc" for checkpoint in $(for filename in $MODEL_DIR/model.ckpt-*.meta; do echo $filename | grep -o "ckpt-[0123456789]*.meta" | cut -d '-' -f 2 | cut -d '.' -f 1; done | sort -n); do if [ $checkpoint -gt $start ]; then echo $checkpoint; CUDA_VISIBLE_DEVICES=$GPU_ID python eval.py \ --train_dir="$MODEL_DIR" \ --model_checkpoint_path="${MODEL_DIR}/model.ckpt-${checkpoint}" \ --eval_data_pattern="/Youtube-8M/data/video/validate/validatea*" \ --distill_data_pattern="/Youtube-8M/model_predictions/validatea/distillation/ensemble_mean_model/*.tfrecord" \ --distillation_features=True \ --distillation_as_input=True \ --frame_features=False \ --feature_names="mean_rgb,mean_audio" \ --feature_sizes="1024,128" \ --model=$MODEL \ --moe_num_mixtures=4 \ --deep_chain_layers=4 \ --deep_chain_relu_cells=256 \ --batch_size=256 \ --run_once=True fi done ================================================ FILE: youtube-8m-wangheda/eval_scripts/eval-video-divergence-chain-model.sh ================================================ GPU_ID=0 EVERY=200 MODEL=MultiTaskDivergenceDeepCombineChainModel MODEL_DIR="../model/video_divergence_deep_combine_chain" start=$1 DIR="$(pwd)" for checkpoint in $(cd $MODEL_DIR && python ${DIR}/training_utils/select.py $EVERY); do echo $checkpoint; if [ $checkpoint -gt $start ]; then echo $checkpoint; CUDA_VISIBLE_DEVICES=$GPU_ID python eval.py \ --train_dir="$MODEL_DIR" \ --model_checkpoint_path="${MODEL_DIR}/model.ckpt-${checkpoint}" \ --eval_data_pattern="/Youtube-8M/data/video/validate/validatea*" \ --frame_features=False \ --feature_names="mean_rgb,mean_audio" \ --feature_sizes="1024,128" \ --batch_size=128 \ --model=$MODEL \ --lstm_cells="1024,128" \ --divergence_model_count=8 \ --deep_chain_layers=2 \ --deep_chain_relu_cells=128 \ --moe_num_mixtures=4 \ --multitask=True \ --label_loss=MultiTaskDivergenceCrossEntropyLoss \ --support_loss_percent=0.025 \ --num_readers=1 \ --run_once=True fi done ================================================ FILE: youtube-8m-wangheda/eval_scripts/eval-video-divergence-moe-model.sh ================================================ GPU_ID=1 EVERY=500 MODEL=MultiTaskDivergenceMoeModel MODEL_DIR="../model/video_divergence_moe" start=$1 DIR="$(pwd)" for checkpoint in $(cd $MODEL_DIR && python ${DIR}/training_utils/select.py $EVERY); do echo $checkpoint; if [ $checkpoint -gt $start ]; then echo $checkpoint; CUDA_VISIBLE_DEVICES=$GPU_ID python eval.py \ --train_dir="$MODEL_DIR" \ --model_checkpoint_path="${MODEL_DIR}/model.ckpt-${checkpoint}" \ --eval_data_pattern="/Youtube-8M/data/video/validate/validatea*" \ --frame_features=False \ --feature_names="mean_rgb,mean_audio" \ --feature_sizes="1024,128" \ --batch_size=128 \ --model=$MODEL \ --model=MultiTaskDivergenceMoeModel \ --lstm_cells="1024,128" \ --divergence_model_count=8 \ --moe_num_mixtures=4 \ --multitask=True \ --label_loss=MultiTaskDivergenceCrossEntropyLoss \ --support_loss_percent=0.025 \ --num_readers=1 \ --run_once=True fi done ================================================ FILE: youtube-8m-wangheda/eval_scripts/eval-video-hidden-chain.sh ================================================ GPU_ID=0 EVERY=100 MODEL=HiddenChainModel MODEL_DIR="../model/video_chain_moe16_hidden" start=$1 DIR="$(pwd)" for checkpoint in $(cd $MODEL_DIR && python ${DIR}/training_utils/select.py $EVERY); do echo $checkpoint; if [ $checkpoint -gt $start ]; then echo $checkpoint; CUDA_VISIBLE_DEVICES=$GPU_ID python eval.py \ --train_dir="$MODEL_DIR" \ --model_checkpoint_path="${MODEL_DIR}/model.ckpt-${checkpoint}" \ --eval_data_pattern="/Youtube-8M/data/video/validate/validatea*" \ --frame_features=False \ --feature_names="mean_rgb,mean_audio" \ --feature_sizes="1024,128" \ --multitask=True \ --support_type="label,label,label,label" \ --label_loss=MultiTaskCrossEntropyLoss \ --moe_num_mixtures=8 \ --batch_size=512 \ --model=$MODEL \ --run_once=True fi done ================================================ FILE: youtube-8m-wangheda/eval_scripts/eval-video-hidden-combine-chain.sh ================================================ GPU_ID=0 EVERY=100 MODEL=HiddenCombineChainModel MODEL_DIR="../model/video_chain_moe16_hidden_combine" start=$1 DIR="$(pwd)" for checkpoint in $(cd $MODEL_DIR && python ${DIR}/training_utils/select.py $EVERY); do echo $checkpoint; if [ $checkpoint -gt $start ]; then echo $checkpoint; CUDA_VISIBLE_DEVICES=$GPU_ID python eval.py \ --train_dir="$MODEL_DIR" \ --model_checkpoint_path="${MODEL_DIR}/model.ckpt-${checkpoint}" \ --eval_data_pattern="/Youtube-8M/data/video/validate/validatea*" \ --frame_features=False \ --feature_names="mean_rgb,mean_audio" \ --feature_sizes="1024,128" \ --moe_num_mixtures=8 \ --batch_size=512 \ --model=$MODEL \ --run_once=True fi done ================================================ FILE: youtube-8m-wangheda/eval_scripts/eval-video-logistic.sh ================================================ GPU_ID=1 EVERY=100 MODEL=LogisticModel MODEL_DIR="../model/video_logistic_model" start=$1 DIR="$(pwd)" for checkpoint in $(cd $MODEL_DIR && python ${DIR}/training_utils/select.py $EVERY); do echo $checkpoint; if [ $checkpoint -gt $start ]; then echo $checkpoint; CUDA_VISIBLE_DEVICES=$GPU_ID python eval.py \ --train_dir="$MODEL_DIR" \ --model_checkpoint_path="${MODEL_DIR}/model.ckpt-${checkpoint}" \ --eval_data_pattern="/Youtube-8M/data/video/validate/validatea*" \ --frame_features=False \ --feature_names="mean_rgb,mean_audio" \ --feature_sizes="1024,128" \ --model=$MODEL \ --batch_size=32 \ --run_once=True fi done ================================================ FILE: youtube-8m-wangheda/eval_scripts/eval-video-moe.sh ================================================ GPU_ID=1 EVERY=100 MODEL=MoeModel MODEL_DIR="/home/zhangt/yt8m/video_level_moe_model" start=$1 DIR="$(pwd)" for checkpoint in $(cd $MODEL_DIR && python ${DIR}/training_utils/select.py $EVERY); do echo $checkpoint; if [ $checkpoint -gt $start ]; then echo $checkpoint; CUDA_VISIBLE_DEVICES=$GPU_ID python eval.py \ --train_dir="$MODEL_DIR" \ --model_checkpoint_path="${MODEL_DIR}/model.ckpt-${checkpoint}" \ --eval_data_pattern="/Youtube-8M/data/video/validate/validatea*" \ --frame_features=False \ --feature_names="mean_rgb,mean_audio" \ --feature_sizes="1024,128" \ --batch_size=4096 \ --model=$MODEL \ --moe_num_mixtures=16 \ --run_once=True fi done ================================================ FILE: youtube-8m-wangheda/eval_scripts/eval-video-pairwise.sh ================================================ GPU_ID=1 EVERY=500 MODEL=MoeModel MODEL_DIR="../model/video_moe16_mixedloss" start=$1 DIR="$(pwd)" for checkpoint in $(cd $MODEL_DIR && python ${DIR}/training_utils/select.py $EVERY); do echo $checkpoint; if [ $checkpoint -gt $start ]; then echo $checkpoint; CUDA_VISIBLE_DEVICES=$GPU_ID python eval.py \ --train_dir="$MODEL_DIR" \ --model_checkpoint_path="${MODEL_DIR}/model.ckpt-${checkpoint}" \ --eval_data_pattern="/Youtube-8M/data/video/validate/validatea*" \ --frame_features=False \ --feature_names="mean_rgb,mean_audio" \ --feature_sizes="1024,128" \ --batch_size=256 \ --model=$MODEL \ --moe_num_mixtures=16 \ --run_once=True fi done ================================================ FILE: youtube-8m-wangheda/eval_scripts/eval-video-verydeep-combine-chain.sh ================================================ GPU_ID=1 EVERY=100 MODEL=DeepCombineChainModel MODEL_DIR="../model/video_chain_moe16_verydeep_combine" start=$1 DIR="$(pwd)" for checkpoint in $(cd $MODEL_DIR && python ${DIR}/training_utils/select.py $EVERY); do echo $checkpoint; if [ $checkpoint -gt $start ]; then echo $checkpoint; CUDA_VISIBLE_DEVICES=$GPU_ID python eval.py \ --train_dir="$MODEL_DIR" \ --model_checkpoint_path="${MODEL_DIR}/model.ckpt-${checkpoint}" \ --eval_data_pattern="/Youtube-8M/data/video/validate/validatea*" \ --frame_features=False \ --feature_names="mean_rgb,mean_audio" \ --feature_sizes="1024,128" \ --moe_num_mixtures=2 \ --deep_chain_layers=8 \ --deep_chain_relu_cells=128 \ --label_loss=MultiTaskCrossEntropyLoss \ --multitask=True \ --support_type="label,label,label,label,label,label,label,label" \ --support_loss_percent=0.025 \ --batch_size=128 \ --model=$MODEL \ --run_once=True fi done ================================================ FILE: youtube-8m-wangheda/eval_util.py ================================================ # Copyright 2016 Google Inc. All Rights Reserved. # # Licensed under the Apache License, Version 2.0 (the "License"); # you may not use this file except in compliance with the License. # You may obtain a copy of the License at # # http://www.apache.org/licenses/LICENSE-2.0 # # Unless required by applicable law or agreed to in writing, software # distributed under the License is distributed on an "AS-IS" BASIS, # WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied. # See the License for the specific language governing permissions and # limitations under the License. """Provides functions to help with evaluating models.""" import datetime import numpy from tensorflow.python.platform import gfile import mean_average_precision_calculator as map_calculator import average_precision_calculator as ap_calculator def flatten(l): """ Merges a list of lists into a single list. """ return [item for sublist in l for item in sublist] def calculate_hit_at_one(predictions, actuals): """Performs a local (numpy) calculation of the hit at one. Args: predictions: Matrix containing the outputs of the model. Dimensions are 'batch' x 'num_classes'. actuals: Matrix containing the ground truth labels. Dimensions are 'batch' x 'num_classes'. Returns: float: The average hit at one across the entire batch. """ top_prediction = numpy.argmax(predictions, 1) hits = actuals[numpy.arange(actuals.shape[0]), top_prediction] return numpy.average(hits) def calculate_recall_at_n(predictions, actuals, n): """Performs a local (numpy) calculation of the recall@n Args: predictions: Matrix containing the outputs of the model. Dimensions are 'batch' x 'num_classes'. actuals: Matrix containing the ground truth labels. Dimensions are 'batch' x 'num_classes'. n: scalar of n Returns: float: The recall at n across the entire batch. """ aggregated_recall = 0.0 num_videos = actuals.shape[0] for row in numpy.arange(num_videos): num_labels = int(numpy.sum(actuals[row])) top_indices = numpy.argpartition(predictions[row], -n)[-n:] item_recall = 0.0 for label_index in top_indices: if predictions[row][label_index] > 0: item_recall += actuals[row][label_index] item_recall /= num_labels aggregated_recall += item_recall aggregated_recall /= num_videos return aggregated_recall def calculate_precision_at_equal_recall_rate(predictions, actuals): """Performs a local (numpy) calculation of the PERR. Args: predictions: Matrix containing the outputs of the model. Dimensions are 'batch' x 'num_classes'. actuals: Matrix containing the ground truth labels. Dimensions are 'batch' x 'num_classes'. Returns: float: The average precision at equal recall rate across the entire batch. """ aggregated_precision = 0.0 num_videos = actuals.shape[0] for row in numpy.arange(num_videos): num_labels = int(numpy.sum(actuals[row])) top_indices = numpy.argpartition(predictions[row], -num_labels)[-num_labels:] item_precision = 0.0 for label_index in top_indices: if predictions[row][label_index] > 0: item_precision += actuals[row][label_index] item_precision /= top_indices.size aggregated_precision += item_precision aggregated_precision /= num_videos return aggregated_precision def calculate_gap(predictions, actuals, top_k=20): """Performs a local (numpy) calculation of the global average precision. Only the top_k predictions are taken for each of the videos. Args: predictions: Matrix containing the outputs of the model. Dimensions are 'batch' x 'num_classes'. actuals: Matrix containing the ground truth labels. Dimensions are 'batch' x 'num_classes'. top_k: How many predictions to use per video. Returns: float: The global average precision. """ gap_calculator = ap_calculator.AveragePrecisionCalculator() sparse_predictions, sparse_labels, num_positives = top_k_by_class(predictions, actuals, top_k) gap_calculator.accumulate(flatten(sparse_predictions), flatten(sparse_labels), sum(num_positives)) return gap_calculator.peek_ap_at_n() def top_k_by_class(predictions, labels, k=20): """Extracts the top k predictions for each video, sorted by class. Args: predictions: A numpy matrix containing the outputs of the model. Dimensions are 'batch' x 'num_classes'. k: the top k non-zero entries to preserve in each prediction. Returns: A tuple (predictions,labels, true_positives). 'predictions' and 'labels' are lists of lists of floats. 'true_positives' is a list of scalars. The length of the lists are equal to the number of classes. The entries in the predictions variable are probability predictions, and the corresponding entries in the labels variable are the ground truth for those predictions. The entries in 'true_positives' are the number of true positives for each class in the ground truth. Raises: ValueError: An error occurred when the k is not a positive integer. """ if k <= 0: raise ValueError("k must be a positive integer.") k = min(k, predictions.shape[1]) num_classes = predictions.shape[1] prediction_triplets= [] for video_index in range(predictions.shape[0]): prediction_triplets.extend(top_k_triplets(predictions[video_index],labels[video_index], k)) out_predictions = [[] for v in range(num_classes)] out_labels = [[] for v in range(num_classes)] for triplet in prediction_triplets: out_predictions[triplet[0]].append(triplet[1]) out_labels[triplet[0]].append(triplet[2]) out_true_positives = [numpy.sum(labels[:,i]) for i in range(num_classes)] return out_predictions, out_labels, out_true_positives def top_k_triplets(predictions, labels, k=20): """Get the top_k for a 1-d numpy array. Returns a sparse list of tuples in (prediction, class) format""" m = len(predictions) k = min(k, m) indices = numpy.argpartition(predictions, -k)[-k:] return [(index, predictions[index], labels[index]) for index in indices] class EvaluationMetrics(object): """A class to store the evaluation metrics.""" def __init__(self, num_class, top_k): """Construct an EvaluationMetrics object to store the evaluation metrics. Args: num_class: A positive integer specifying the number of classes. top_k: A positive integer specifying how many predictions are considered per video. Raises: ValueError: An error occurred when MeanAveragePrecisionCalculator cannot not be constructed. """ self.sum_hit_at_one = 0.0 self.sum_perr = 0.0 self.sum_loss = 0.0 self.map_calculator = map_calculator.MeanAveragePrecisionCalculator(num_class) self.global_ap_calculator = ap_calculator.AveragePrecisionCalculator() self.top_k = top_k self.num_examples = 0 def accumulate(self, predictions, labels, loss): """Accumulate the metrics calculated locally for this mini-batch. Args: predictions: A numpy matrix containing the outputs of the model. Dimensions are 'batch' x 'num_classes'. labels: A numpy matrix containing the ground truth labels. Dimensions are 'batch' x 'num_classes'. loss: A numpy array containing the loss for each sample. Returns: dictionary: A dictionary storing the metrics for the mini-batch. Raises: ValueError: An error occurred when the shape of predictions and actuals does not match. """ batch_size = labels.shape[0] mean_hit_at_one = calculate_hit_at_one(predictions, labels) mean_perr = calculate_precision_at_equal_recall_rate(predictions, labels) mean_loss = numpy.mean(loss) # Take the top 20 predictions. sparse_predictions, sparse_labels, num_positives = top_k_by_class(predictions, labels, self.top_k) self.map_calculator.accumulate(sparse_predictions, sparse_labels, num_positives) self.global_ap_calculator.accumulate(flatten(sparse_predictions), flatten(sparse_labels), sum(num_positives)) self.num_examples += batch_size self.sum_hit_at_one += mean_hit_at_one * batch_size self.sum_perr += mean_perr * batch_size self.sum_loss += mean_loss * batch_size return {"hit_at_one": mean_hit_at_one, "perr": mean_perr, "loss": mean_loss} def get(self): """Calculate the evaluation metrics for the whole epoch. Raises: ValueError: If no examples were accumulated. Returns: dictionary: a dictionary storing the evaluation metrics for the epoch. The dictionary has the fields: avg_hit_at_one, avg_perr, avg_loss, and aps (default nan). """ if self.num_examples <= 0: raise ValueError("total_sample must be positive.") avg_hit_at_one = self.sum_hit_at_one / self.num_examples avg_perr = self.sum_perr / self.num_examples avg_loss = self.sum_loss / self.num_examples aps = self.map_calculator.peek_map_at_n() gap = self.global_ap_calculator.peek_ap_at_n() epoch_info_dict = {} return {"avg_hit_at_one": avg_hit_at_one, "avg_perr": avg_perr, "avg_loss": avg_loss, "aps": aps, "gap": gap} def clear(self): """Clear the evaluation metrics and reset the EvaluationMetrics object.""" self.sum_hit_at_one = 0.0 self.sum_perr = 0.0 self.sum_loss = 0.0 self.map_calculator.clear() self.global_ap_calculator.clear() self.num_examples = 0 ================================================ FILE: youtube-8m-wangheda/feature_transform.py ================================================ from tensorflow import flags flags.DEFINE_string("feature_transformer", "DefaultTransformer", "how to preprocess feature, defaults to identical, which means no transform") flags.DEFINE_string("engineer_types", "identical,avg,std,diff", "how to preprocess feature, defaults to identical, which means no transform") flags.DEFINE_integer("time_resolution", 8, "how many frames to mean pooling at a time") from all_feature_transform import * ================================================ FILE: youtube-8m-wangheda/frame_level_models.py ================================================ # Copyright 2016 Google Inc. All Rights Reserved. # # Licensed under the Apache License, Version 2.0 (the "License"); # you may not use this file except in compliance with the License. # You may obtain a copy of the License at # # http://www.apache.org/licenses/LICENSE-2.0 # # Unless required by applicable law or agreed to in writing, software # distributed under the License is distributed on an "AS-IS" BASIS, # WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied. # See the License for the specific language governing permissions and # limitations under the License. """Contains a collection of models which operate on variable-length sequences. """ from tensorflow import flags FLAGS = flags.FLAGS flags.DEFINE_integer("iterations", 30, "Number of frames per batch for DBoF.") flags.DEFINE_bool("dbof_add_batch_norm", True, "Adds batch normalization to the DBoF model.") flags.DEFINE_bool( "sample_random_frames", True, "If true samples random frames (for frame level models). If false, a random" "sequence of frames is sampled instead.") flags.DEFINE_integer("dbof_cluster_size", 8192, "Number of units in the DBoF cluster layer.") flags.DEFINE_integer("dbof_hidden_size", 1024, "Number of units in the DBoF hidden layer.") flags.DEFINE_string("dbof_pooling_method", "max", "The pooling method used in the DBoF cluster layer. " "Choices are 'average' and 'max'.") flags.DEFINE_string("video_level_classifier_model", "MoeModel", "Some Frame-Level models can be decomposed into a " "generalized pooling operation followed by a " "classifier layer") flags.DEFINE_string("video_level_classifier_support_model", "MoeModel", "Some Frame-Level models can be decomposed into a " "generalized pooling operation followed by a " "classifier layer") flags.DEFINE_bool("rnn_swap_memory", False, "If true, swap_memory = True.") flags.DEFINE_string("lstm_cells", "1024", "Number of LSTM cells.") flags.DEFINE_integer("lstm_layers", 2, "Number of LSTM layers.") flags.DEFINE_integer("gru_cells", 1024, "Number of GRU cells.") flags.DEFINE_integer("gru_layers", 2, "Number of GRU layers.") flags.DEFINE_integer("attention_size", 1, "Number of attention layers.") flags.DEFINE_integer("num_attentions", 5, "Number of attention cells per layer.") flags.DEFINE_string("cnn_filter_sizes", "1,2,3", "Sizes of cnn filters.") flags.DEFINE_string("cnn_filter_nums", "256,256,256", "Numbers of every cnn filters.") flags.DEFINE_integer("cnn_pooling_k", 4, "The k value for max-k pooling.") flags.DEFINE_string("lstm_normalization", "identical", "which normalization method") flags.DEFINE_integer("mm_label_embedding", 256, "size of label embedding vector") flags.DEFINE_string("wide_and_deep_models", "FrameLevelLogisticModel,LstmMemoryModel", "size of label embedding vector") flags.DEFINE_integer("deep_cnn_base_size", 128, "basic cnn size") flags.DEFINE_integer("lstm_look_back", 3, "how many adjacent input for a cell to look at") flags.DEFINE_integer("lstm_attentions", 8, "Attention size in lstm_attention_max_pooling_model.") flags.DEFINE_integer("positional_embedding_size", 32, "Positional embedding dimension use in lstm_positional_attention_max_pooling_model.") flags.DEFINE_bool("is_training", False, "used in batch normalization.") flags.DEFINE_integer("multiscale_cnn_lstm_layers", 1, "number of layers in multiscale cnn_lstm.") flags.DEFINE_integer("frame_seg_relu_cells", 256, "number of relu cells in frame-seg model.") flags.DEFINE_integer("distillchain_relu_cells", 256, "number of relu cells in distillchain model.") flags.DEFINE_integer("cnn_num_filters", 512, "number of filters in cnn conv layer.") import sys from os.path import dirname if dirname(__file__) not in sys.path: sys.path.append(dirname(__file__)) from all_frame_models import * ================================================ FILE: youtube-8m-wangheda/infer_scripts/infer-attentionlstm_moe4.sh ================================================ for part in test ensemble_train ensemble_validate train_samples; do CUDA_VISIBLE_DEVICES=1 python inference-pre-ensemble.py \ --output_dir="/Youtube-8M/model_predictions/${part}/attentionlstm_moe4" \ --model_checkpoint_path="../model/attentionlstm_moe4/model.ckpt-104135" \ --input_data_pattern="/Youtube-8M/data/frame/${part}/*.tfrecord" \ --frame_features=True \ --feature_names="rgb,audio" \ --feature_sizes="1024,128" \ --batch_size=64 \ --file_size=4096 done ================================================ FILE: youtube-8m-wangheda/infer_scripts/infer-biunilstm1024_moe8.sh ================================================ for part in test ensemble_train ensemble_validate train_samples; do CUDA_VISIBLE_DEVICES=0 python inference-pre-ensemble.py \ --output_dir="/Youtube-8M/model_predictions/${part}/biunilstm1024_moe4" \ --model_checkpoint_path="../model/biunilstm1024_moe4/model.ckpt-94882" \ --input_data_pattern="/Youtube-8M/data/frame/${part}/*.tfrecord" \ --frame_features=True \ --feature_names="rgb,audio" \ --feature_sizes="1024,128" \ --batch_size=32 \ --file_size=4096 done ================================================ FILE: youtube-8m-wangheda/infer_scripts/infer-cnn_deep_combine_chain.sh ================================================ #for part in ensemble_train ensemble_validate test; do for part in train_samples; do CUDA_VISIBLE_DEVICES=0 python inference-pre-ensemble.py \ --output_dir="/Youtube-8M/model_predictions/${part}/cnn_deep_combine_chain" \ --model_checkpoint_path="../model/cnn_deep_combine_chain/model.ckpt-228332" \ --input_data_pattern="/Youtube-8M/data/frame/${part}/*.tfrecord" \ --frame_features=True \ --feature_names="rgb,audio" \ --feature_sizes="1024,128" \ --batch_size=32 \ --file_size=4096 done ================================================ FILE: youtube-8m-wangheda/infer_scripts/infer-cnn_lstmmemory1024_moe8.sh ================================================ for part in test ensemble_train ensemble_validate train_samples; do CUDA_VISIBLE_DEVICES=0 python inference-pre-ensemble.py \ --output_dir="/Youtube-8M/model_predictions/${part}/cnnlstmmemory1024_moe8" \ --model_checkpoint_path="../model/cnnlstmmemory1024_moe8/model.ckpt-149354" \ --input_data_pattern="/Youtube-8M/data/frame/${part}/*.tfrecord" \ --frame_features=True \ --feature_names="rgb,audio" \ --feature_sizes="1024,128" \ --batch_size=64 \ --file_size=4096 done ================================================ FILE: youtube-8m-wangheda/infer_scripts/infer-cnn_model.sh ================================================ #for part in ensemble_train ensemble_validate test; do for part in ensemble_validate; do CUDA_VISIBLE_DEVICES=0 python inference-pre-ensemble.py \ --output_dir="/Youtube-8M/model_predictions/${part}/cnn_model" \ --model_checkpoint_path="../model/cnn_model/model.ckpt-374098" \ --input_data_pattern="/Youtube-8M/data/frame/${part}/*.tfrecord" \ --frame_features=True \ --feature_names="rgb,audio" \ --feature_sizes="1024,128" \ --model=CnnModel \ --cnn_num_filters=512 \ --moe_num_mixtures=4 \ --batch_size=128 \ --file_size=4096 done ================================================ FILE: youtube-8m-wangheda/infer_scripts/infer-dbof.sh ================================================ for part in test ensemble_train ensemble_validate; do CUDA_VISIBLE_DEVICES=1 python inference-pre-ensemble.py \ --output_dir="/Youtube-8M/model_predictions/${part}/dbof_model" \ --model_checkpoint_path="../model/dbof_model/model.ckpt-184058" \ --input_data_pattern="/Youtube-8M/data/frame/${part}/*.tfrecord" \ --frame_features=True \ --feature_names="rgb,audio" \ --feature_sizes="1024,128" \ --model=DbofModel \ --batch_size=64 \ --num_readers=1 \ --file_size=4096 done ================================================ FILE: youtube-8m-wangheda/infer_scripts/infer-deep_cnn_deep_combine.sh ================================================ #for part in ensemble_train ensemble_validate test; do for part in train_samples; do CUDA_VISIBLE_DEVICES=0 python inference-pre-ensemble.py \ --output_dir="/Youtube-8M/model_predictions/${part}/deep_cnn_deep_combine_chain" \ --model_checkpoint_path="../model/deep_cnn_deep_combine_chain/model.ckpt-303289" \ --input_data_pattern="/Youtube-8M/data/frame/${part}/*.tfrecord" \ --frame_features=True \ --feature_names="rgb,audio" \ --feature_sizes="1024,128" \ --batch_size=32 \ --file_size=4096 done ================================================ FILE: youtube-8m-wangheda/infer_scripts/infer-deeplstm1024_layer6_moe4.sh ================================================ for part in test ensemble_train ensemble_validate train_samples; do CUDA_VISIBLE_DEVICES=0 python inference-pre-ensemble.py \ --output_dir="/Youtube-8M/model_predictions/${part}/deeplstm1024_layer6_moe4" \ --model_checkpoint_path="../model/deeplstm1024_layer6_moe4/model.ckpt-175048" \ --input_data_pattern="/Youtube-8M/data/frame/${part}/*.tfrecord" \ --frame_features=True \ --feature_names="rgb,audio" \ --feature_sizes="1024,128" \ --batch_size=128 \ --file_size=4096 done ================================================ FILE: youtube-8m-wangheda/infer_scripts/infer-distill_video_dcc.sh ================================================ for part in ensemble_train ensemble_validate test; do CUDA_VISIBLE_DEVICES=1 python inference-pre-ensemble.py \ --output_dir="/Youtube-8M/model_predictions/${part}/distillation_video_dcc_noise" \ --model_checkpoint_path="../model/distillation_video_dcc_noise/scene1_percent_0.4/model.ckpt-47842" \ --input_data_pattern="/Youtube-8M/data/video/${part}/*.tfrecord" \ --frame_features=False \ --feature_names="mean_rgb,mean_audio" \ --feature_sizes="1024,128" \ --model=DeepCombineChainModel \ --deep_chain_relu_cells=256 \ --deep_chain_layers=4 \ --moe_num_mixtures=4 \ --batch_size=32 \ --file_size=4096 done ================================================ FILE: youtube-8m-wangheda/infer_scripts/infer-distillation-cnn-dcc.sh ================================================ for part in train; do CUDA_VISIBLE_DEVICES=0 python inference-pre-ensemble.py \ --output_dir="/Youtube-8M/model_predictions_local/${part}/distillation/distillchain_cnn_dcc" \ --model_checkpoint_path="../model/distillation/distillchain_cnn_dcc/model.ckpt-74488" \ --input_data_pattern="/Youtube-8M/data/frame/${part}/*.tfrecord" \ --frame_features=True \ --feature_names="rgb,audio" \ --feature_sizes="1024,128" \ --distill_data_pattern="/Youtube-8M/model_predictions/${part}/distillation/ensemble_mean_model/*.tfrecord" \ --distillation_features=True \ --distillation_as_input=True \ --model=DistillchainCnnDeepCombineChainModel \ --deep_chain_layers=3 \ --deep_chain_relu_cells=256 \ --moe_num_mixtures=4 \ --batch_size=128 \ --file_size=4096 done ================================================ FILE: youtube-8m-wangheda/infer_scripts/infer-distillation-lstmattention8max.sh ================================================ # for distillation version 2 data cd ../youtube-8m-zhangteng #for part in ensemble_train ensemble_validate test; do for part in train; do output_dir="/Youtube-8M/model_predictions_x32/${part}/distillation/distillchain_lstm_attention8max" if [ ! -d $output_dir ]; then CUDA_VISIBLE_DEVICES=1 python inference-pre-ensemble-distill.py \ --output_dir="$output_dir" \ --model_checkpoint_path="../model/distillation/frame_level_lstm_extend_distillchain_model/model.ckpt-74796" \ --input_data_pattern="/Youtube-8M/data/frame/${part}/*.tfrecord" \ --distill_data_pattern="/Youtube-8M/model_predictions/${part}/distillation/ensemble_mean_model/*.tfrecord" \ --frame_features=True \ --feature_names="rgb,audio" \ --distill_names="predictions" \ --feature_sizes="1024,128" \ --distill_sizes="4716" \ --model=LstmExtendModel \ --video_level_classifier_model=MoeExtendDistillChainModel \ --moe_num_extend=8 \ --moe_num_mixtures=8 \ --train=False \ --batch_size=32 \ --file_size=4096 fi done ================================================ FILE: youtube-8m-wangheda/infer_scripts/infer-distillation-lstmgate.sh ================================================ # for distillation version 2 data cd ../youtube-8m-zhangteng #for part in ensemble_train ensemble_validate test; do for part in train; do output_dir="/Youtube-8M/model_predictions_x32/${part}/distillation/distillchain_lstm_gate_moe8" if [ ! -d $output_dir ]; then CUDA_VISIBLE_DEVICES=1 python inference-pre-ensemble-distill.py \ --output_dir="$output_dir" \ --model_checkpoint_path="../model/distillation/frame_level_lstm_gate_distillchain_model/model.ckpt-63344" \ --input_data_pattern="/Youtube-8M/data/frame/${part}/*.tfrecord" \ --distill_data_pattern="/Youtube-8M/model_predictions/${part}/distillation/ensemble_mean_model/*.tfrecord" \ --frame_features=True \ --feature_names="rgb,audio" \ --distill_names="predictions" \ --feature_sizes="1024,128" \ --distill_sizes="4716" \ --model=LstmGateModel \ --video_level_classifier_model=MoeDistillChainModel \ --moe_num_extend=4 \ --moe_method=None \ --lstm_cells=1024 \ --moe_num_mixtures=8 \ --train=False \ --batch_size=32 \ --file_size=4096 fi done ================================================ FILE: youtube-8m-wangheda/infer_scripts/infer-distillation-video-dcc.sh ================================================ #for part in ensemble_train ensemble_validate test; do for part in train; do CUDA_VISIBLE_DEVICES=1 python inference-pre-ensemble.py \ --output_dir="/Youtube-8M/model_predictions_x32/${part}/distillation/distillchain_video_dcc" \ --model_checkpoint_path="../model/distillation/distillchain_video_dcc/model.ckpt-19296" \ --input_data_pattern="/Youtube-8M/data/video/${part}/*.tfrecord" \ --frame_features=False \ --feature_names="mean_rgb,mean_audio" \ --feature_sizes="1024,128" \ --distill_data_pattern="/Youtube-8M/model_predictions/${part}/distillation/ensemble_mean_model/*.tfrecord" \ --distillation_features=True \ --distillation_as_input=True \ --model=DistillchainDeepCombineChainModel \ --moe_num_mixtures=4 \ --deep_chain_layers=4 \ --deep_chain_relu_cells=256 \ --batch_size=32 \ --file_size=4096 done ================================================ FILE: youtube-8m-wangheda/infer_scripts/infer-distillation.sh ================================================ #!/bin/bash # the script directory DIR="$( cd "$( dirname "${BASH_SOURCE[0]}" )" && pwd )" cd ${DIR}/../ model_name="distillation/video_chain_moe2_verydeep_combine" checkpoint="28403" for part in ensemble_validate test ensemble_train; do output_dir="/Youtube-8M/model_predictions/${part}/${model_name}" if [ ! -d $output_dir ]; then CUDA_VISIBLE_DEVICES=0 python inference-pre-ensemble.py \ --output_dir="$output_dir" \ --model_checkpoint_path="../model/${model_name}/model.ckpt-${checkpoint}" \ --input_data_pattern="/Youtube-8M/data/video/${part}/*.tfrecord" \ --frame_features=False \ --feature_names="mean_rgb,mean_audio" \ --feature_sizes="1024,128" \ --model=DeepCombineChainModel \ --deep_chain_relu_cells=128 \ --deep_chain_layers=8 \ --moe_num_mixtures=2 \ --batch_size=1024 \ --file_size=4096 fi done model_name="distillation/cnn_deep_combine_chain" checkpoint="228332" for part in ensemble_validate test ensemble_train; do output_dir="/Youtube-8M/model_predictions/${part}/${model_name}" if [ ! -d $output_dir ]; then CUDA_VISIBLE_DEVICES=0 python inference-pre-ensemble.py \ --output_dir="$output_dir" \ --model_checkpoint_path="../model/${model_name}/model.ckpt-${checkpoint}" \ --input_data_pattern="/Youtube-8M/data/frame/${part}/*.tfrecord" \ --frame_features=True \ --feature_names="rgb,audio" \ --feature_sizes="1024,128" \ --model=CnnDeepCombineChainModel \ --deep_chain_layers=4 \ --deep_chain_relu_cells=128 \ --moe_num_mixtures=4 \ --batch_size=128 \ --file_size=4096 fi done model_name="distillation/lstmparallelfinaloutput1024_moe8" checkpoint="144351" for part in ensemble_validate test ensemble_train; do output_dir="/Youtube-8M/model_predictions/${part}/${model_name}" if [ ! -d $output_dir ]; then CUDA_VISIBLE_DEVICES=0 python inference-pre-ensemble.py \ --output_dir="$output_dir" \ --model_checkpoint_path="../model/${model_name}/model.ckpt-${checkpoint}" \ --input_data_pattern="/Youtube-8M/data/frame/${part}/*.tfrecord" \ --frame_features=True \ --feature_names="rgb,audio" \ --feature_sizes="1024,128" \ --model=LstmParallelFinaloutputModel \ --lstm_cells="1024,128" \ --moe_num_mixtures=8 \ --rnn_swap_memory=True \ --batch_size=128 \ --file_size=4096 fi done # The Lstm-Attention-Max-pooling model cd ${DIR}/../../youtube-8m-zhangteng/ model_name="distillation/frame_level_lstm_extend8_model" checkpoint="181785" for part in ensemble_validate test ensemble_train; do output_dir="/Youtube-8M/model_predictions/${part}/${model_name}" if [ ! -d $output_dir ]; then CUDA_VISIBLE_DEVICES=0 python inference_with_rebuild.py \ --output_dir="$output_dir" \ --model_checkpoint_path="../model/${model_name}/model.ckpt-${checkpoint}" \ --input_data_pattern="/Youtube-8M/data/frame/${part}/*.tfrecord" \ --frame_features=True \ --feature_names="rgb,audio" \ --feature_sizes="1024,128" \ --model=LstmExtendModel \ --video_level_classifier_model=MoeExtendModel \ --moe_num_mixtures=8 \ --moe_num_extend=8 \ --batch_size=128 \ --file_size=4096 fi done ================================================ FILE: youtube-8m-wangheda/infer_scripts/infer-distillchain-cnn-dcc.sh ================================================ for part in ensemble_train test ensemble_validate; do CUDA_VISIBLE_DEVICES=1 python inference-pre-ensemble.py \ --output_dir="/Youtube-8M/model_predictions/${part}/distillchain_cnn_dcc" \ --model_checkpoint_path="../model/distillchain_cnn_dcc/model.ckpt-113446" \ --input_data_pattern="/Youtube-8M/data/frame/${part}/*.tfrecord" \ --frame_features=True \ --feature_names="rgb,audio" \ --feature_sizes="1024,128" \ --distill_data_pattern="/Youtube-8M/model_predictions/${part}/distillation/ensemble_mean_model/*.tfrecord" \ --distillation_features=True \ --distillation_as_input=True \ --model=DistillchainCnnDeepCombineChainModel \ --deep_chain_layers=3 \ --deep_chain_relu_cells=256 \ --moe_num_mixtures=4 \ --batch_size=32 \ --file_size=4096 done ================================================ FILE: youtube-8m-wangheda/infer_scripts/infer-distillchain-lstmcnn.sh ================================================ for part in ensemble_train ensemble_validate test; do output_dir="/Youtube-8M/model_predictions/${part}/distillchain_lstmcnn_dcc" if [ ! -d $output_dir ]; then CUDA_VISIBLE_DEVICES=1 python inference-pre-ensemble.py \ --output_dir="$output_dir" \ --model_checkpoint_path="../model/distillchain_lstmcnn_dcc/model.ckpt-79669" \ --input_data_pattern="/Youtube-8M/data/frame/${part}/*.tfrecord" \ --distill_data_pattern="/Youtube-8M/model_predictions/${part}/distillation/ensemble_mean_model/*.tfrecord" \ --frame_features=True \ --feature_names="rgb,audio" \ --feature_sizes="1024,128" \ --distillation_features=True \ --distillation_as_input=True \ --model=DistillchainLstmCnnDeepCombineChainModel \ --deep_chain_layers=2 \ --deep_chain_relu_cells=128 \ --distillchain_relu_cells=256 \ --moe_num_mixtures=4 \ --lstm_layers=1 \ --lstm_cells="1024,128" \ --rnn_swap_memory=True \ --batch_size=32 \ --file_size=4096 fi done ================================================ FILE: youtube-8m-wangheda/infer_scripts/infer-distillchain-lstmparalleloutput.sh ================================================ for part in ensemble_train ensemble_validate test; do CUDA_VISIBLE_DEVICES=1 python inference-pre-ensemble.py \ --output_dir="/Youtube-8M/model_predictions/${part}/distillchain_lstmparalleloutput" \ --model_checkpoint_path="../model/distillchain_lstmparalleloutput/model.ckpt-75261" \ --input_data_pattern="/Youtube-8M/data/frame/${part}/*.tfrecord" \ --frame_features=True \ --feature_names="rgb,audio" \ --feature_sizes="1024,128" \ --distill_data_pattern="/Youtube-8M/model_predictions/${part}/distillation/ensemble_mean_model/*.tfrecord" \ --distillation_features=True \ --distillation_as_input=True \ --model=DistillchainLstmParallelFinaloutputModel \ --rnn_swap_memory=True \ --lstm_cells="1024,128" \ --moe_num_mixtures=8 \ --batch_size=128 \ --file_size=4096 done ================================================ FILE: youtube-8m-wangheda/infer_scripts/infer-distillchain-v2-boost-lstmparalleloutput.sh ================================================ for part in ensemble_train ensemble_validate test; do output_dir="/Youtube-8M/model_predictions/${part}/distillchain_v2_boost_lstmparalleloutput" if [ ! -d $output_dir ]; then CUDA_VISIBLE_DEVICES=1 python inference-pre-ensemble.py \ --output_dir="$output_dir" \ --model_checkpoint_path="../model/distillchain_v2_boost_lstmparalleloutput/model.ckpt-86757" \ --input_data_pattern="/Youtube-8M/data/frame/${part}/*.tfrecord" \ --distill_data_pattern="/Youtube-8M/model_predictions/${part}/distillation/ensemble_v2_matrix_model/*.tfrecord" \ --frame_features=True \ --feature_names="rgb,audio" \ --feature_sizes="1024,128" \ --distillation_features=False \ --distillation_as_input=True \ --model=DistillchainLstmParallelFinaloutputModel \ --rnn_swap_memory=True \ --lstm_layers=1 \ --lstm_cells="1024,128" \ --moe_num_mixtures=8 \ --batch_size=32 \ --file_size=4096 fi done ================================================ FILE: youtube-8m-wangheda/infer_scripts/infer-distillchain-v2-lstmattention8max.sh ================================================ for part in ensemble_train ensemble_validate test; do output_dir="/Youtube-8M/model_predictions/${part}/distillchain_v2_lstmattention8max" if [ ! -d $output_dir ]; then CUDA_VISIBLE_DEVICES=0 python inference-pre-ensemble.py \ --output_dir="$output_dir" \ --model_checkpoint_path="../model/distillchain_v2_lstmattention8max/model.ckpt-" \ --input_data_pattern="/Youtube-8M/data/frame/${part}/*.tfrecord" \ --distill_data_pattern="/Youtube-8M/model_predictions/${part}/distillation/ensemble_v2_matrix_model/*.tfrecord" \ --frame_features=True \ --feature_names="rgb,audio" \ --feature_sizes="1024,128" \ --distillation_features=False \ --distillation_as_input=True \ --model=DistillchainLstmAttentionMaxPoolingModel \ --moe_num_mixtures=8 \ --lstm_attentions=8 \ --lstm_cells=1024 \ --rnn_swap_memory=True \ --moe_num_mixtures=8 \ --batch_size=128 \ --file_size=4096 fi done ================================================ FILE: youtube-8m-wangheda/infer_scripts/infer-distillchain-v2-lstmcnn.sh ================================================ for part in ensemble_train ensemble_validate test; do output_dir="/Youtube-8M/model_predictions/${part}/distillchain_v2_lstmcnn_dcc" if [ ! -d $output_dir ]; then CUDA_VISIBLE_DEVICES=0 python inference-pre-ensemble.py \ --output_dir="$output_dir" \ --model_checkpoint_path="../model/distillchain_v2_lstmcnn_dcc/model.ckpt-77277" \ --input_data_pattern="/Youtube-8M/data/frame/${part}/*.tfrecord" \ --distill_data_pattern="/Youtube-8M/model_predictions/${part}/distillation/ensemble_v2_matrix_model/*.tfrecord" \ --frame_features=True \ --feature_names="rgb,audio" \ --feature_sizes="1024,128" \ --distillation_features=False \ --distillation_as_input=True \ --model=DistillchainLstmCnnDeepCombineChainModel \ --deep_chain_layers=2 \ --deep_chain_relu_cells=128 \ --distillchain_relu_cells=256 \ --moe_num_mixtures=4 \ --lstm_layers=1 \ --lstm_cells="1024,128" \ --rnn_swap_memory=True \ --batch_size=128 \ --file_size=4096 fi done ================================================ FILE: youtube-8m-wangheda/infer_scripts/infer-distillchain-v2-lstmparalleloutput.sh ================================================ for part in ensemble_train ensemble_validate test; do CUDA_VISIBLE_DEVICES=0 python inference-pre-ensemble.py \ --output_dir="/Youtube-8M/model_predictions/${part}/distillchain_v2_lstmparalleloutput" \ --model_checkpoint_path="../model/distillchain_v2_lstmparalleloutput/model.ckpt-74190" \ --input_data_pattern="/Youtube-8M/data/frame/${part}/*.tfrecord" \ --frame_features=True \ --feature_names="rgb,audio" \ --feature_sizes="1024,128" \ --distill_data_pattern="/Youtube-8M/model_predictions/${part}/distillation/ensemble_v2_matrix_model/*.tfrecord" \ --distillation_features=False \ --distillation_as_input=True \ --model=DistillchainLstmParallelFinaloutputModel \ --rnn_swap_memory=True \ --lstm_cells="1024,128" \ --moe_num_mixtures=4 \ --batch_size=128 \ --file_size=4096 done ================================================ FILE: youtube-8m-wangheda/infer_scripts/infer-distillchain-v2-multilstm.sh ================================================ for part in ensemble_train ensemble_validate test; do CUDA_VISIBLE_DEVICES=0 python inference-pre-ensemble.py \ --output_dir="/Youtube-8M/model_predictions/${part}/distillchain_v2_multilstm_dcc" \ --model_checkpoint_path="../model/distillchain_v2_multilstm_dcc/model.ckpt-74272" \ --input_data_pattern="/Youtube-8M/data/frame/${part}/*.tfrecord" \ --distill_data_pattern="/Youtube-8M/model_predictions/${part}/distillation/ensemble_v2_matrix_model/*.tfrecord" \ --frame_features=True \ --feature_names="rgb,audio" \ --feature_sizes="1024,128" \ --distillation_features=False \ --distillation_as_input=True \ --model=DistillchainLstmMemoryDeepCombineChainModel \ --lstm_layers=1 \ --lstm_cells=1024 \ --moe_num_mixtures=4 \ --distillation_relu_cells=256 \ --deep_chain_relu_cells=256 \ --deep_chain_layers=2 \ --rnn_swap_memory=True \ --batch_size=128 \ --file_size=4096 done ================================================ FILE: youtube-8m-wangheda/infer_scripts/infer-distillchain-v2-multiscal-cnnlstm.sh ================================================ for part in ensemble_train ensemble_validate test; do CUDA_VISIBLE_DEVICES=0 python inference-pre-ensemble.py \ --output_dir="/Youtube-8M/model_predictions/${part}/distillchain_v2_multiscale_cnnlstm" \ --model_checkpoint_path="../model/distillchain_v2_multiscale_cnnlstm/model.ckpt-72344" \ --input_data_pattern="/Youtube-8M/data/frame/${part}/*.tfrecord" \ --distill_data_pattern="/Youtube-8M/model_predictions/${part}/distillation/ensemble_v2_matrix_model/*.tfrecord" \ --frame_features=True \ --feature_names="rgb,audio" \ --feature_sizes="1024,128" \ --distillation_features=False \ --distillation_as_input=True \ --model=DistillchainMultiscaleCnnLstmModel \ --multiscale_cnn_lstm_layers=3 \ --moe_num_mixtures=4 \ --rnn_swap_memory=True \ --is_training=False \ --batch_size=128 \ --file_size=4096 done ================================================ FILE: youtube-8m-wangheda/infer_scripts/infer-distillchain-v2-video-dcc.sh ================================================ for part in ensemble_train ensemble_validate test; do CUDA_VISIBLE_DEVICES=0 python inference-pre-ensemble.py \ --output_dir="/Youtube-8M/model_predictions/${part}/distillchain_v2_video_dcc" \ --model_checkpoint_path="../model/distillchain_v2_video_dcc/model.ckpt-17137" \ --input_data_pattern="/Youtube-8M/data/video/${part}/*.tfrecord" \ --distill_data_pattern="/Youtube-8M/model_predictions/${part}/distillation/ensemble_v2_matrix_model/*.tfrecord" \ --frame_features=False \ --feature_names="mean_rgb,mean_audio" \ --feature_sizes="1024,128" \ --distillation_features=True \ --distillation_as_input=True \ --model=DistillchainDeepCombineChainModel \ --moe_num_mixtures=4 \ --deep_chain_layers=4 \ --deep_chain_relu_cells=256 \ --batch_size=1024 \ --file_size=4096 done ================================================ FILE: youtube-8m-wangheda/infer_scripts/infer-frame_seg.sh ================================================ for part in ensemble_train ensemble_validate test; do CUDA_VISIBLE_DEVICES=1 python inference-pre-ensemble.py \ --output_dir="/Youtube-8M/model_predictions_local/${part}/frame_seg_model" \ --model_checkpoint_path="../model/frame_seg_model/model.ckpt-27945" \ --input_data_pattern="/Youtube-8M/data/frame/${part}/*.tfrecord" \ --frame_features=True \ --feature_names="rgb,audio" \ --feature_sizes="1024,128" \ --feature_transformer=IdenticalTransformer \ --model=FrameSegModel \ --moe_num_mixtures=16 \ --batch_size=64 \ --file_size=4096 done ================================================ FILE: youtube-8m-wangheda/infer_scripts/infer-framehop_lstm.sh ================================================ for part in test ensemble_train ensemble_validate train_samples; do CUDA_VISIBLE_DEVICES=1 python inference-pre-ensemble.py \ --output_dir="/Youtube-8M/model_predictions/${part}/framehop_lstm" \ --model_checkpoint_path="../model/framehop_lstm/model.ckpt-143746" \ --input_data_pattern="/Youtube-8M/data/frame/${part}/*.tfrecord" \ --frame_features=True \ --feature_names="rgb,audio" \ --feature_sizes="1024,128" \ --feature_transformer=IdenticalTransformer \ --batch_size=64 \ --file_size=4096 done ================================================ FILE: youtube-8m-wangheda/infer_scripts/infer-lstm_attention8_max.sh ================================================ #for part in ensemble_train ensemble_validate test; do for part in train_samples; do CUDA_VISIBLE_DEVICES=0 python inference-pre-ensemble.py \ --output_dir="/Youtube-8M/model_predictions/${part}/lstm_attention8_max" \ --model_checkpoint_path="/home/zhangt/yt8m/frame_level_lstm_extend8_model/model.ckpt-181785" \ --input_data_pattern="/Youtube-8M/data/frame/${part}/*.tfrecord" \ --frame_features=True \ --feature_names="rgb,audio" \ --feature_sizes="1024,128" \ --batch_size=32 \ --file_size=4096 done ================================================ FILE: youtube-8m-wangheda/infer_scripts/infer-lstm_cnn_deep_combine_chain.sh ================================================ for part in test ensemble_validate ensemble_train; do #for part in train_samples; do CUDA_VISIBLE_DEVICES=0 python inference-pre-ensemble.py \ --output_dir="/Youtube-8M/model_predictions/${part}/lstm_cnn_deep_combine_chain" \ --model_checkpoint_path="../model/lstm_cnn_deep_combine_chain/model.ckpt-259179" \ --input_data_pattern="/Youtube-8M/data/frame/${part}/*.tfrecord" \ --frame_features=True \ --feature_names="rgb,audio" \ --feature_sizes="1024,128" \ --batch_size=32 \ --file_size=4096 done ================================================ FILE: youtube-8m-wangheda/infer_scripts/infer-lstmattlstm1024_moe8.sh ================================================ for part in test ensemble_train ensemble_validate train_samples; do CUDA_VISIBLE_DEVICES=1 python inference-pre-ensemble.py \ --output_dir="/Youtube-8M/model_predictions/${part}/lstmattlstm1024_moe8" \ --model_checkpoint_path="../model/lstmattlstm1024_moe8/model.ckpt-86809" \ --input_data_pattern="/Youtube-8M/data/frame/${part}/*.tfrecord" \ --frame_features=True \ --feature_names="rgb,audio" \ --feature_sizes="1024,128" \ --batch_size=64 \ --file_size=4096 done ================================================ FILE: youtube-8m-wangheda/infer_scripts/infer-lstmmemory-audio.sh ================================================ CUDA_VISIBLE_DEVICES=1 python inference.py \ --output_file="../model/audio_lstmmemory1024_layer1_moe8/error_analysis.train_samples.tsv" \ --model_checkpoint_path="../model/audio_lstmmemory1024_layer1_moe8/model.ckpt-187979" \ --input_data_pattern="/Youtube-8M/data/frame/train/train1r.tfrecord" \ --frame_features=True \ --feature_names="audio" \ --feature_sizes="128" \ --batch_size=128 \ --file_size=4096 ================================================ FILE: youtube-8m-wangheda/infer_scripts/infer-lstmmemory-layer1.sh ================================================ #for part in ensemble_train ensemble_validate test; do for part in ensemble_validate; do CUDA_VISIBLE_DEVICES=1 python inference-pre-ensemble.py \ --output_dir="/Youtube-8M/model_predictions/${part}/lstmmemory1024_layer1_moe8" \ --model_checkpoint_path="../model/lstmmemory1024_layer1_moe8/model.ckpt-149022" \ --input_data_pattern="/Youtube-8M/data/frame/${part}/*.tfrecord" \ --frame_features=True \ --feature_names="rgb,audio" \ --feature_sizes="1024,128" \ --model=LstmMemoryModel \ --lstm_cells=1024 \ --lstm_layers=1 \ --moe_num_mixtures=8 \ --rnn_swap_memory=False \ --batch_size=128 \ --file_size=4096 done ================================================ FILE: youtube-8m-wangheda/infer_scripts/infer-lstmmemory1024_deep_combine_chain_add_length.sh ================================================ #for part in ensemble_train ensemble_validate test; do for part in train_samples; do CUDA_VISIBLE_DEVICES=1 python inference-pre-ensemble.py \ --output_dir="/Youtube-8M/model_predictions/${part}/lstmmem1024_layer2_moe4_deep_combine_chain_add_length" \ --model_checkpoint_path="../model/lstmmem1024_deep_combine_chain_length/model.ckpt-148035" \ --input_data_pattern="/Youtube-8M/data/frame/${part}/*.tfrecord" \ --frame_features=True \ --feature_names="rgb,audio" \ --feature_sizes="1024,128" \ --batch_size=32 \ --file_size=4096 done ================================================ FILE: youtube-8m-wangheda/infer_scripts/infer-lstmmemory1024_moe8.sh ================================================ #for part in ensemble_train ensemble_validate test; do for part in train_samples; do CUDA_VISIBLE_DEVICES=1 python inference-pre-ensemble.py \ --output_dir="/Youtube-8M/model_predictions/${part}/lstmmemory_cell1024_layer2_moe8" \ --model_checkpoint_path="../model/lstmmemory1024_moe8/model.ckpt-123144" \ --input_data_pattern="/Youtube-8M/data/frame/${part}/*.tfrecord" \ --frame_features=True \ --feature_names="rgb,audio" \ --feature_sizes="1024,128" \ --batch_size=32 \ --file_size=4096 done ================================================ FILE: youtube-8m-wangheda/infer_scripts/infer-lstmmemory2048_moe4.sh ================================================ #for part in ensemble_train ensemble_validate test; do for part in train_samples; do CUDA_VISIBLE_DEVICES=0 python inference-pre-ensemble.py \ --output_dir="/Youtube-8M/model_predictions/${part}/lstmmemory_cell2048_layer2_moe4" \ --model_checkpoint_path="../model/lstmmemory2048_moe4_batch128/model.ckpt-109434" \ --input_data_pattern="/Youtube-8M/data/frame/${part}/*.tfrecord" \ --frame_features=True \ --feature_names="rgb,audio" \ --feature_sizes="1024,128" \ --batch_size=32 \ --file_size=4096 done ================================================ FILE: youtube-8m-wangheda/infer_scripts/infer-lstmparallelmemory1024_moe8.sh ================================================ #for part in ensemble_train ensemble_validate test; do for part in train_samples; do CUDA_VISIBLE_DEVICES=0 python inference-pre-ensemble.py \ --output_dir="/Youtube-8M/model_predictions/${part}/lstmparallelmemory1024_moe8" \ --model_checkpoint_path="../model/lstmparallelmemory1024_moe8/model.ckpt-111155" \ --input_data_pattern="/Youtube-8M/data/frame/${part}/*.tfrecord" \ --frame_features=True \ --feature_names="rgb,audio" \ --feature_sizes="1024,128" \ --batch_size=32 \ --file_size=4096 done ================================================ FILE: youtube-8m-wangheda/infer_scripts/infer-lstmparalleloutput1024_moe8.sh ================================================ for part in test ensemble_train ensemble_validate train_samples; do CUDA_VISIBLE_DEVICES=0 python inference-pre-ensemble.py \ --output_dir="/Youtube-8M/model_predictions/${part}/lstmparallelfinaloutput1024_moe8" \ --model_checkpoint_path="../model/lstmparallelfinaloutput1024_moe8/model.ckpt-144351" \ --input_data_pattern="/Youtube-8M/data/frame/${part}/*.tfrecord" \ --frame_features=True \ --feature_names="rgb,audio" \ --feature_sizes="1024,128" \ --batch_size=64 \ --file_size=4096 done ================================================ FILE: youtube-8m-wangheda/infer_scripts/infer-model_input.sh ================================================ for part in test ensemble_train ensemble_validate; do CUDA_VISIBLE_DEVICES=1 python inference-pre-ensemble-get-input.py \ --output_dir="/Youtube-8M/model_predictions/${part}/model_input" \ --input_data_pattern="/Youtube-8M/data/video/${part}/*.tfrecord" \ --frame_features=False \ --feature_names="mean_rgb,mean_audio" \ --feature_sizes="1024,128" \ --batch_size=128 \ --file_size=4096 done ================================================ FILE: youtube-8m-wangheda/infer_scripts/infer-multilstmmemory1024_moe4_deep_chain.sh ================================================ #for part in test ensemble_validate ensemble_train; do for part in train_samples; do CUDA_VISIBLE_DEVICES=1 python inference-pre-ensemble.py \ --output_dir="/Youtube-8M/model_predictions/${part}/multilstmmemory1024_moe4_deep_chain" \ --model_checkpoint_path="../model/multilstmmemory1024_moe4_deep_chain/model.ckpt-133508" \ --input_data_pattern="/Youtube-8M/data/frame/${part}/*.tfrecord" \ --frame_features=True \ --feature_names="rgb,audio" \ --feature_sizes="1024,128" \ --batch_size=32 \ --file_size=4096 done ================================================ FILE: youtube-8m-wangheda/infer_scripts/infer-multires_lstm_deep_combine_chain.sh ================================================ for part in test ensemble_validate ensemble_train; do #for part in train_samples; do CUDA_VISIBLE_DEVICES=0 python inference-pre-ensemble.py \ --output_dir="/Youtube-8M/model_predictions/${part}/multires_lstm_deep_combine_chain" \ --model_checkpoint_path="../model/multires_lstm_deep_combine_chain/model.ckpt-108969" \ --input_data_pattern="/Youtube-8M/data/frame/${part}/*.tfrecord" \ --frame_features=True \ --feature_names="rgb,audio" \ --feature_sizes="1024,128" \ --feature_transformer=IdenticalTransformer \ --batch_size=32 \ --file_size=4096 done ================================================ FILE: youtube-8m-wangheda/infer_scripts/infer-positional-lstmattention8max.sh ================================================ for part in test ensemble_train ensemble_validate; do CUDA_VISIBLE_DEVICES=1 python inference-pre-ensemble.py \ --output_dir="/Youtube-8M/model_predictions/${part}/lstm_positional_attention8max" \ --model_checkpoint_path="../model/lstm_positional_attention8max/model.ckpt-198407" \ --input_data_pattern="/Youtube-8M/data/frame/${part}/*.tfrecord" \ --frame_features=True \ --feature_names="rgb,audio" \ --feature_sizes="1024,128" \ --model=LstmPositionalAttentionMaxPoolingModel \ --moe_num_mixtures=8 \ --lstm_attentions=8 \ --positional_embedding_size=32 \ --rnn_swap_memory=True \ --batch_size=32 \ --file_size=4096 done ================================================ FILE: youtube-8m-wangheda/infer_scripts/infer-video-distillchain-video-dcc.sh ================================================ for part in ensemble_train ensemble_validate test; do CUDA_VISIBLE_DEVICES=0 python inference-pre-ensemble.py \ --output_dir="/Youtube-8M/model_predictions/${part}/distillchain_video_dcc" \ --model_checkpoint_path="../model/distillchain_video_dcc/model.ckpt-19296" \ --input_data_pattern="/Youtube-8M/data/video/${part}/*.tfrecord" \ --frame_features=False \ --feature_names="mean_rgb,mean_audio" \ --feature_sizes="1024,128" \ --distill_data_pattern="/Youtube-8M/model_predictions/${part}/distillation/ensemble_mean_model/*.tfrecord" \ --distillation_features=True \ --distillation_as_input=True \ --model=DistillchainDeepCombineChainModel \ --moe_num_mixtures=4 \ --deep_chain_layers=4 \ --deep_chain_relu_cells=256 \ --batch_size=1024 \ --file_size=4096 done ================================================ FILE: youtube-8m-wangheda/infer_scripts/infer-video_group_moe4_noise0.2_layer4_elu.sh ================================================ #for part in ensemble_train ensemble_validate test; do for part in train_samples; do CUDA_VISIBLE_DEVICES=0 python inference-pre-ensemble.py \ --output_dir="/Youtube-8M/model_predictions/${part}/video_moe4_noise0.2_layer4_elu" \ --model_checkpoint_path="/home/zhangt/yt8m/video_level_moenoise0.2_model/moe_4layers_elu/model.ckpt-22845" \ --input_data_pattern="/Youtube-8M/data/video/${part}/*.tfrecord" \ --frame_features=False \ --feature_names="mean_rgb,mean_audio" \ --feature_sizes="1024,128" \ --batch_size=128 \ --file_size=4096 done ================================================ FILE: youtube-8m-wangheda/infer_scripts/infer-video_logistic.sh ================================================ for part in ensemble_train ensemble_validate test; do CUDA_VISIBLE_DEVICES=1 python inference-pre-ensemble.py \ --output_dir="/Youtube-8M/model_predictions/${part}/video_logistic_model" \ --model_checkpoint_path="../model/video_logistic_model/model.ckpt-23581" \ --input_data_pattern="/Youtube-8M/data/video/${part}/*.tfrecord" \ --frame_features=False \ --feature_names="mean_rgb,mean_audio" \ --feature_sizes="1024,128" \ --model=LogisticModel \ --batch_size=32 \ --file_size=4096 done ================================================ FILE: youtube-8m-wangheda/infer_scripts/infer-video_moe16_model.sh ================================================ for part in ensemble_train ensemble_validate test; do CUDA_VISIBLE_DEVICES=0 python inference-pre-ensemble.py \ --output_dir="/Youtube-8M/model_predictions/${part}/video_moe16_model" \ --model_checkpoint_path="../model/video_moe16_model/model.ckpt-19058" \ --input_data_pattern="/Youtube-8M/data/video/${part}/*.tfrecord" \ --frame_features=False \ --feature_names="mean_rgb,mean_audio" \ --feature_sizes="1024,128" \ --model=MoeModel \ --moe_num_mixtures=16 \ --batch_size=128 \ --file_size=4096 done ================================================ FILE: youtube-8m-wangheda/infer_scripts/infer-video_very_deep_combine_chain.sh ================================================ #for part in ensemble_train ensemble_validate test; do for part in train_samples; do CUDA_VISIBLE_DEVICES=0 python inference-pre-ensemble.py \ --output_dir="/Youtube-8M/model_predictions/${part}/video_very_deep_combine_chain" \ --model_checkpoint_path="../model/video_chain_moe2_verydeep_combine/model.ckpt-28403" \ --input_data_pattern="/Youtube-8M/data/video/${part}/*.tfrecord" \ --frame_features=False \ --feature_names="mean_rgb,mean_audio" \ --feature_sizes="1024,128" \ --batch_size=128 \ --file_size=4096 done ================================================ FILE: youtube-8m-wangheda/inference-layer.py ================================================ # Copyright 2016 Google Inc. All Rights Reserved. # # Licensed under the Apache License, Version 2.0 (the "License"); # you may not use this file except in compliance with the License. # You may obtain a copy of the License at # # http://www.apache.org/licenses/LICENSE-2.0 # # Unless required by applicable law or agreed to in writing, software # distributed under the License is distributed on an "AS-IS" BASIS, # WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied. # See the License for the specific language governing permissions and # limitations under the License. """Binary for generating predictions over a set of videos.""" import os import time import numpy import numpy as np import tensorflow as tf from tensorflow import app from tensorflow import flags from tensorflow import gfile from tensorflow import logging import eval_util import losses import readers import utils FLAGS = flags.FLAGS if __name__ == '__main__': flags.DEFINE_string("train_dir", "/tmp/yt8m_model/", "The directory to load the model files from.") flags.DEFINE_string("model_checkpoint_path", "", "The file path to load the model from.") flags.DEFINE_string("output_file", "", "The file to save the predictions to.") flags.DEFINE_integer("file_size", 4096, "Number of examples put into a file.") flags.DEFINE_string( "input_data_pattern", "", "File glob defining the evaluation dataset in tensorflow.SequenceExample " "format. The SequenceExamples are expected to have an 'rgb' byte array " "sequence feature as well as a 'labels' int64 context feature.") # Model flags. flags.DEFINE_bool( "frame_features", False, "If set, then --eval_data_pattern must be frame-level features. " "Otherwise, --eval_data_pattern must be aggregated video-level " "features. The model must also be set appropriately (i.e. to read 3D " "batches VS 4D batches.") flags.DEFINE_integer( "batch_size", 8192, "How many examples to process per batch.") flags.DEFINE_string("feature_names", "mean_rgb", "Name of the feature " "to use for training.") flags.DEFINE_string("feature_sizes", "1024", "Length of the feature vectors.") flags.DEFINE_string("layer_name", "model/RNN/concat:0", "Name of the layer" "to extract value from.") # Other flags. flags.DEFINE_integer("num_readers", 1, "How many threads to use for reading input files.") flags.DEFINE_integer("top_k", 20, "How many predictions to output per video.") def format_lines(video_ids, predictions, top_k): batch_size = len(video_ids) for video_index in range(batch_size): top_indices = numpy.argpartition(predictions[video_index], -top_k)[-top_k:] line = [(class_index, predictions[video_index][class_index]) for class_index in top_indices] # print("Type - Test :") # print(type(video_ids[video_index])) # print(video_ids[video_index].decode('utf-8')) line = sorted(line, key=lambda p: -p[1]) yield video_ids[video_index].decode('utf-8') + "," + " ".join("%i %f" % pair for pair in line) + "\n" def get_input_data_tensors(reader, data_pattern, batch_size, num_readers=1): """Creates the section of the graph which reads the input data. Args: reader: A class which parses the input data. data_pattern: A 'glob' style path to the data files. batch_size: How many examples to process at a time. num_readers: How many I/O threads to use. Returns: A tuple containing the features tensor, labels tensor, and optionally a tensor containing the number of frames per video. The exact dimensions depend on the reader being used. Raises: IOError: If no files matching the given pattern were found. """ with tf.name_scope("input"): files = gfile.Glob(data_pattern) if not files: raise IOError("Unable to find input files. data_pattern='" + data_pattern + "'") logging.info("number of input files: " + str(len(files))) filename_queue = tf.train.string_input_producer( files, num_epochs=1, shuffle=False) examples_and_labels = [reader.prepare_reader(filename_queue) for _ in range(num_readers)] video_id_batch, video_batch, video_label_batch, num_frames_batch = ( tf.train.batch_join(examples_and_labels, batch_size=batch_size, allow_smaller_final_batch = True, enqueue_many=True)) return video_id_batch, video_batch, video_label_batch, num_frames_batch def get_output_feature(video_id, labels, feature_dict): feature_maps = {'video_id': tf.train.Feature(bytes_list=tf.train.BytesList(value=[video_id])), 'labels': tf.train.Feature(int64_list=tf.train.Int64List(value=labels))} for feature_name in feature_dict: feature_maps[feature_name] = tf.train.Feature( float_list=tf.train.FloatList(value=feature_dict[feature_name])) example = tf.train.Example(features=tf.train.Features(feature=feature_maps)) return example def write_to_record(id_batch, label_batch, feature_dict, filenum, num_examples_processed): writer = tf.python_io.TFRecordWriter(FLAGS.output_file+str(filenum)+'.tfrecord') for i in range(num_examples_processed): slice_dict = lambda d, j: dict([(k,d[k][j,:]) for k in d]) video_id = id_batch[i] label = np.nonzero(label_batch[i,:])[0] example = get_output_feature(video_id, label, slice_dict(feature_dict, i)) serialized = example.SerializeToString() writer.write(serialized) writer.close() def inference(reader, train_dir, data_pattern, out_file_location, batch_size, top_k): with tf.Session() as sess: video_id_batch, video_batch, video_label_batch, num_frames_batch = get_input_data_tensors(reader, data_pattern, batch_size) if FLAGS.model_checkpoint_path: latest_checkpoint = FLAGS.model_checkpoint_path else: latest_checkpoint = tf.train.latest_checkpoint(train_dir) if latest_checkpoint is None: raise Exception("unable to find a checkpoint at location: %s" % train_dir) else: meta_graph_location = latest_checkpoint + ".meta" logging.info("loading meta-graph: " + meta_graph_location) saver = tf.train.import_meta_graph(meta_graph_location, clear_devices=True) logging.info("restoring variables from " + latest_checkpoint) saver.restore(sess, latest_checkpoint) input_tensor = tf.get_collection("input_batch_raw")[0] num_frames_tensor = tf.get_collection("num_frames")[0] layer_tensor = tf.get_default_graph().get_tensor_by_name(FLAGS.layer_name) print "layer", layer_tensor # Workaround for num_epochs issue. def set_up_init_ops(variables): init_op_list = [] for variable in list(variables): if "train_input" in variable.name: init_op_list.append(tf.assign(variable, 1)) variables.remove(variable) init_op_list.append(tf.variables_initializer(variables)) return init_op_list sess.run(set_up_init_ops(tf.get_collection_ref( tf.GraphKeys.LOCAL_VARIABLES))) coord = tf.train.Coordinator() threads = tf.train.start_queue_runners(sess=sess, coord=coord) num_examples_processed = 0 start_time = time.time() try: file_num = 0 video_id_list = [] video_label_list = [] video_layer_list = [] def write_record(video_id_list, video_label_list, video_layer_list, file_num, num_examples_processed): video_id_list = np.concatenate(video_id_list,axis=0) video_label_list = np.concatenate(video_label_list,axis=0) video_layer_list = np.concatenate(video_layer_list,axis=0) feature_dict = {"layer": video_layer_list} write_to_record(video_id_list, video_label_list, feature_dict, file_num, num_examples_processed) while not coord.should_stop(): video_id_batch_val, video_batch_val, video_label_batch_val, num_frames_batch_val = sess.run([video_id_batch, video_batch, video_label_batch, num_frames_batch]) layer_val = sess.run(layer_tensor, feed_dict={input_tensor: video_batch_val, num_frames_tensor: num_frames_batch_val}) now = time.time() num_examples_processed += len(video_batch_val) video_id_list.append(video_id_batch_val) video_label_list.append(video_label_batch_val) video_layer_list.append(layer_val) if num_examples_processed>=FLAGS.file_size: write_record(video_id_list, video_label_list, video_layer_list, file_num, num_examples_processed) file_num += 1 video_id_list = [] video_label_list = [] video_layer_list = [] num_examples_processed = 0 logging.info("num examples processed: " + str(num_examples_processed) + " elapsed seconds: " + "{0:.2f}".format(now-start_time)) if num_examples_processed>0: write_record(video_id_list, video_label_list, video_layer_list, file_num, num_examples_processed) except tf.errors.OutOfRangeError: logging.info('Done with inference. The output file was written to ' + out_file_location) finally: coord.request_stop() coord.join(threads) sess.close() def main(unused_argv): logging.set_verbosity(tf.logging.INFO) # convert feature_names and feature_sizes to lists of values feature_names, feature_sizes = utils.GetListOfFeatureNamesAndSizes( FLAGS.feature_names, FLAGS.feature_sizes) if FLAGS.frame_features: reader = readers.YT8MFrameFeatureReader(feature_names=feature_names, feature_sizes=feature_sizes) else: reader = readers.YT8MAggregatedFeatureReader(feature_names=feature_names, feature_sizes=feature_sizes) if FLAGS.output_file is "": raise ValueError("'output_file' was not specified. " "Unable to continue with inference.") if FLAGS.input_data_pattern is "": raise ValueError("'input_data_pattern' was not specified. " "Unable to continue with inference.") inference(reader, FLAGS.train_dir, FLAGS.input_data_pattern, FLAGS.output_file, FLAGS.batch_size, FLAGS.top_k) if __name__ == "__main__": app.run() ================================================ FILE: youtube-8m-wangheda/inference-pre-ensemble-get-input.py ================================================ # WARNING: This is a temporary script for writing the input directly into files, do not edit base on the file # WARNING: This is a temporary script for writing the input directly into files, do not edit base on the file # WARNING: This is a temporary script for writing the input directly into files, do not edit base on the file # WARNING: This is a temporary script for writing the input directly into files, do not edit base on the file # WARNING: This is a temporary script for writing the input directly into files, do not edit base on the file # WARNING: This is a temporary script for writing the input directly into files, do not edit base on the file # WARNING: This is a temporary script for writing the input directly into files, do not edit base on the file # WARNING: This is a temporary script for writing the input directly into files, do not edit base on the file # WARNING: This is a temporary script for writing the input directly into files, do not edit base on the file # WARNING: This is a temporary script for writing the input directly into files, do not edit base on the file # WARNING: This is a temporary script for writing the input directly into files, do not edit base on the file # WARNING: This is a temporary script for writing the input directly into files, do not edit base on the file # WARNING: This is a temporary script for writing the input directly into files, do not edit base on the file # WARNING: This is a temporary script for writing the input directly into files, do not edit base on the file # WARNING: This is a temporary script for writing the input directly into files, do not edit base on the file # WARNING: This is a temporary script for writing the input directly into files, do not edit base on the file # WARNING: This is a temporary script for writing the input directly into files, do not edit base on the file # WARNING: This is a temporary script for writing the input directly into files, do not edit base on the file # WARNING: This is a temporary script for writing the input directly into files, do not edit base on the file # WARNING: This is a temporary script for writing the input directly into files, do not edit base on the file # WARNING: This is a temporary script for writing the input directly into files, do not edit base on the file # WARNING: This is a temporary script for writing the input directly into files, do not edit base on the file # WARNING: This is a temporary script for writing the input directly into files, do not edit base on the file # WARNING: This is a temporary script for writing the input directly into files, do not edit base on the file # WARNING: This is a temporary script for writing the input directly into files, do not edit base on the file # WARNING: This is a temporary script for writing the input directly into files, do not edit base on the file # WARNING: This is a temporary script for writing the input directly into files, do not edit base on the file # WARNING: This is a temporary script for writing the input directly into files, do not edit base on the file # WARNING: This is a temporary script for writing the input directly into files, do not edit base on the file # WARNING: This is a temporary script for writing the input directly into files, do not edit base on the file # WARNING: This is a temporary script for writing the input directly into files, do not edit base on the file # WARNING: This is a temporary script for writing the input directly into files, do not edit base on the file # WARNING: This is a temporary script for writing the input directly into files, do not edit base on the file # WARNING: This is a temporary script for writing the input directly into files, do not edit base on the file # WARNING: This is a temporary script for writing the input directly into files, do not edit base on the file # WARNING: This is a temporary script for writing the input directly into files, do not edit base on the file # WARNING: This is a temporary script for writing the input directly into files, do not edit base on the file # WARNING: This is a temporary script for writing the input directly into files, do not edit base on the file # WARNING: This is a temporary script for writing the input directly into files, do not edit base on the file # WARNING: This is a temporary script for writing the input directly into files, do not edit base on the file # WARNING: This is a temporary script for writing the input directly into files, do not edit base on the file # WARNING: This is a temporary script for writing the input directly into files, do not edit base on the file # WARNING: This is a temporary script for writing the input directly into files, do not edit base on the file # WARNING: This is a temporary script for writing the input directly into files, do not edit base on the file # WARNING: This is a temporary script for writing the input directly into files, do not edit base on the file # WARNING: This is a temporary script for writing the input directly into files, do not edit base on the file # WARNING: This is a temporary script for writing the input directly into files, do not edit base on the file # WARNING: This is a temporary script for writing the input directly into files, do not edit base on the file # WARNING: This is a temporary script for writing the input directly into files, do not edit base on the file # WARNING: This is a temporary script for writing the input directly into files, do not edit base on the file # WARNING: This is a temporary script for writing the input directly into files, do not edit base on the file # WARNING: This is a temporary script for writing the input directly into files, do not edit base on the file # WARNING: This is a temporary script for writing the input directly into files, do not edit base on the file # WARNING: This is a temporary script for writing the input directly into files, do not edit base on the file # WARNING: This is a temporary script for writing the input directly into files, do not edit base on the file # WARNING: This is a temporary script for writing the input directly into files, do not edit base on the file # WARNING: This is a temporary script for writing the input directly into files, do not edit base on the file # WARNING: This is a temporary script for writing the input directly into files, do not edit base on the file # WARNING: This is a temporary script for writing the input directly into files, do not edit base on the file # WARNING: This is a temporary script for writing the input directly into files, do not edit base on the file # WARNING: This is a temporary script for writing the input directly into files, do not edit base on the file # WARNING: This is a temporary script for writing the input directly into files, do not edit base on the file # WARNING: This is a temporary script for writing the input directly into files, do not edit base on the file # WARNING: This is a temporary script for writing the input directly into files, do not edit base on the file # Copyright 2016 Google Inc. All Rights Reserved. # # Licensed under the Apache License, Version 2.0 (the "License"); # you may not use this file except in compliance with the License. # You may obtain a copy of the License at # # http://www.apache.org/licenses/LICENSE-2.0 # # Unless required by applicable law or agreed to in writing, software # distributed under the License is distributed on an "AS-IS" BASIS, # WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied. # See the License for the specific language governing permissions and # limitations under the License. """Binary for generating predictions over a set of videos.""" import os import time import numpy import tensorflow as tf from tensorflow import app from tensorflow import flags from tensorflow import gfile from tensorflow import logging import eval_util import losses import readers import utils import numpy as np FLAGS = flags.FLAGS if __name__ == '__main__': flags.DEFINE_string("model_checkpoint_path", "", "The file path to load the model from.") flags.DEFINE_string("output_dir", "", "The file to save the predictions to.") flags.DEFINE_string( "input_data_pattern", "", "File glob defining the evaluation dataset in tensorflow.SequenceExample " "format. The SequenceExamples are expected to have an 'rgb' byte array " "sequence feature as well as a 'labels' int64 context feature.") # Model flags. flags.DEFINE_bool( "frame_features", False, "If set, then --eval_data_pattern must be frame-level features. " "Otherwise, --eval_data_pattern must be aggregated video-level " "features. The model must also be set appropriately (i.e. to read 3D " "batches VS 4D batches.") flags.DEFINE_integer( "batch_size", 8192, "How many examples to process per batch.") flags.DEFINE_string("feature_names", "mean_rgb", "Name of the feature " "to use for training.") flags.DEFINE_string("feature_sizes", "1024", "Length of the feature vectors.") flags.DEFINE_integer("file_size", 4096, "Number of frames per batch for DBoF.") # Other flags. flags.DEFINE_integer("num_readers", 1, "How many threads to use for reading input files.") flags.DEFINE_integer("top_k", 20, "How many predictions to output per video.") def get_input_data_tensors(reader, data_pattern, batch_size, num_readers=1): """Creates the section of the graph which reads the input data. Args: reader: A class which parses the input data. data_pattern: A 'glob' style path to the data files. batch_size: How many examples to process at a time. num_readers: How many I/O threads to use. Returns: A tuple containing the features tensor, labels tensor, and optionally a tensor containing the number of frames per video. The exact dimensions depend on the reader being used. Raises: IOError: If no files matching the given pattern were found. """ with tf.name_scope("input"): files = gfile.Glob(data_pattern) files.sort() if not files: raise IOError("Unable to find input files. data_pattern='" + data_pattern + "'") logging.info("number of input files: " + str(len(files))) filename_queue = tf.train.string_input_producer( files, num_epochs=1, shuffle=False) examples_and_labels = [reader.prepare_reader(filename_queue) for _ in range(num_readers)] video_id_batch, video_batch, unused_labels, num_frames_batch = ( tf.train.batch_join(examples_and_labels, batch_size=batch_size, allow_smaller_final_batch=True, enqueue_many=True)) return video_id_batch, video_batch, unused_labels, num_frames_batch def inference(reader, model_checkpoint_path, data_pattern, out_file_location, batch_size, top_k): with tf.Session() as sess: video_id_batch, video_batch, video_label_batch, num_frames_batch = get_input_data_tensors(reader, data_pattern, batch_size) # Workaround for num_epochs issue. def set_up_init_ops(variables): init_op_list = [] for variable in list(variables): if "train_input" in variable.name: init_op_list.append(tf.assign(variable, 1)) variables.remove(variable) init_op_list.append(tf.variables_initializer(variables)) return init_op_list sess.run(set_up_init_ops(tf.get_collection_ref( tf.GraphKeys.LOCAL_VARIABLES))) coord = tf.train.Coordinator() threads = tf.train.start_queue_runners(sess=sess, coord=coord) num_examples_processed = 0 start_time = time.time() video_id = [] video_label = [] video_inputs = [] video_features = [] filenum = 0 directory = FLAGS.output_dir if not os.path.exists(directory): os.makedirs(directory) else: raise IOError("Output path exists! path='" + directory + "'") try: while not coord.should_stop(): video_id_batch_val, video_batch_val, video_label_batch_val, num_frames_batch_val = sess.run([video_id_batch, video_batch, video_label_batch, num_frames_batch]) now = time.time() num_examples_processed += len(video_batch_val) video_id.append(video_id_batch_val) video_label.append(video_label_batch_val) video_inputs.append(video_batch_val) if num_examples_processed>=FLAGS.file_size: assert num_examples_processed==FLAGS.file_size, "num_examples_processed should be equal to file_size" video_id = np.concatenate(video_id,axis=0) video_label = np.concatenate(video_label,axis=0) video_inputs = np.concatenate(video_inputs,axis=0) write_to_record(video_id, video_label, video_inputs, filenum, num_examples_processed) filenum += 1 video_id = [] video_label = [] video_inputs = [] num_examples_processed = 0 logging.info("num examples processed: " + str(num_examples_processed) + " elapsed seconds: " + "{0:.2f}".format(now-start_time)) except tf.errors.OutOfRangeError: logging.info('Done with inference. The output file was written to ' + out_file_location) finally: coord.request_stop() if num_examples_processed