Repository: kakaobrain/sparse-detr Branch: main Commit: f40632c3f467 Files: 64 Total size: 365.2 KB Directory structure: gitextract_r25kqp43/ ├── LICENSE ├── NOTICE ├── README.md ├── configs/ │ ├── r50_deformable_detr.sh │ ├── r50_efficient_detr.sh │ ├── r50_sparse_detr_rho_0.1.sh │ ├── r50_sparse_detr_rho_0.2.sh │ ├── r50_sparse_detr_rho_0.3.sh │ ├── swint_deformable_detr.sh │ ├── swint_efficient_detr.sh │ ├── swint_sparse_detr_rho_0.1.sh │ ├── swint_sparse_detr_rho_0.2.sh │ └── swint_sparse_detr_rho_0.3.sh ├── datasets/ │ ├── __init__.py │ ├── coco.py │ ├── coco_eval.py │ ├── coco_panoptic.py │ ├── data_prefetcher.py │ ├── panoptic_eval.py │ ├── samplers.py │ ├── torchvision_datasets/ │ │ ├── __init__.py │ │ └── coco.py │ └── transforms.py ├── engine.py ├── main.py ├── models/ │ ├── __init__.py │ ├── backbone.py │ ├── deformable_detr.py │ ├── deformable_transformer.py │ ├── matcher.py │ ├── ops/ │ │ ├── functions/ │ │ │ ├── __init__.py │ │ │ └── ms_deform_attn_func.py │ │ ├── make.sh │ │ ├── modules/ │ │ │ ├── __init__.py │ │ │ └── ms_deform_attn.py │ │ ├── setup.py │ │ ├── src/ │ │ │ ├── cpu/ │ │ │ │ ├── ms_deform_attn_cpu.cpp │ │ │ │ └── ms_deform_attn_cpu.h │ │ │ ├── cuda/ │ │ │ │ ├── ms_deform_attn_cuda.cu │ │ │ │ ├── ms_deform_attn_cuda.h │ │ │ │ └── ms_deform_im2col_cuda.cuh │ │ │ ├── ms_deform_attn.h │ │ │ └── vision.cpp │ │ └── test.py │ ├── position_encoding.py │ ├── segmentation.py │ └── swin_transformer/ │ ├── __init__.py │ ├── build.py │ ├── config.py │ ├── configs/ │ │ ├── default.yaml │ │ ├── swin_base_patch4_window7_224.yaml │ │ ├── swin_large_patch4_window7_224.yaml │ │ ├── swin_small_patch4_window7_224.yaml │ │ └── swin_tiny_patch4_window7_224.yaml │ └── swin_transformer.py ├── requirements.txt ├── tools/ │ ├── launch.py │ └── run_dist_launch.sh └── util/ ├── __init__.py ├── benchmark.py ├── box_ops.py ├── dam.py ├── misc.py └── plot_utils.py ================================================ FILE CONTENTS ================================================ ================================================ 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 2021 KAKAO BRAIN Corp. 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. Deformable DETR Copyright 2020 SenseTime 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. DETR Copyright 2020 - present, Facebook, Inc 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: NOTICE ================================================ =============================================================================== Deformable DETR's Apache License 2.0 =============================================================================== The overall structure of the code is based on the implementation in Deformable-DETR(https://github.com/fundamentalvision/Deformable-DETR). - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - Copyright (c) 2020 SenseTime 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. =============================================================================== DETR's Apache License 2.0 =============================================================================== Deformable DETR code is orginally built on the implementation in DETR (https://github.com/facebookresearch/detr). - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - Copyright (c) 2020 Facebook, Inc 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. =============================================================================== Swin Transformer' MIT License =============================================================================== The transformer backbone is based on the implementation in Swin Transformer (https://github.com/microsoft/Swin-Transformer). - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - Copyright (c) 2021 Microsoft Permission is hereby granted, free of charge, to any person obtaining a copy of this software and associated documentation files (the "Software"), to deal in the Software without restriction, including without limitation the rights to use, copy, modify, merge, publish, distribute, sublicense, and/or sell copies of the Software, and to permit persons to whom the Software is furnished to do so, subject to the following conditions: The above copyright notice and this permission notice shall be included in all copies or substantial portions of the Software. THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT. IN NO EVENT SHALL THE AUTHORS OR COPYRIGHT HOLDERS BE LIABLE FOR ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE. ================================================ FILE: README.md ================================================ [![KakaoBrain](https://img.shields.io/badge/kakao-brain-ffcd00.svg)](http://kakaobrain.com/) [![pytorch](https://img.shields.io/badge/pytorch-1.6.0-%2523ee4c2c.svg)](https://pytorch.org/) [![pytorch](https://img.shields.io/badge/pytorch-1.7.1-%2523ee4c2c.svg)](https://pytorch.org/) Sparse DETR (ICLR'22) ======== By [Byungseok Roh](https://scholar.google.com/citations?user=H4VWYHwAAAAJ)\*, [Jaewoong Shin](https://scholar.google.com/citations?user=i_o_95kAAAAJ)\*, [Wuhyun Shin](https://scholar.google.com/citations?user=bGwfkakAAAAJ)\*, and [Saehoon Kim](https://scholar.google.com/citations?user=_ZfueMIAAAAJ) at [Kakao Brain](https://www.kakaobrain.com). (*: Equal contribution) * This repository is an official implementation of the paper [Sparse DETR: Efficient End-to-End Object Detection with Learnable Sparsity](https://arxiv.org/abs/2111.14330). * The code and some instructions are built upon the official [Deformable DETR repository](https://github.com/fundamentalvision/Deformable-DETR). # Introduction **TL; DR.** Sparse DETR is an efficient end-to-end object detector that **sparsifies encoder tokens** by using the learnable DAM(Decoder Attention Map) predictor. It achieves better performance than Deformable DETR even with only 10% encoder queries on the COCO dataset.

**Abstract.** DETR is the first end-to-end object detector using a transformer encoder-decoder architecture and demonstrates competitive performance but low computational efficiency on high resolution feature maps. The subsequent work, Deformable DETR, enhances the efficiency of DETR by replacing dense attention with deformable attention, which achieves 10x faster convergence and improved performance. Deformable DETR uses the multiscale feature to ameliorate performance, however, the number of encoder tokens increases by 20x compared to DETR, and the computation cost of the encoder attention remains a bottleneck. In our preliminary experiment, we observe that the detection performance hardly deteriorates even if only a part of the encoder token is updated. Inspired by this observation, we propose *Sparse DETR* that selectively updates only the tokens expected to be referenced by the decoder, thus help the model effectively detect objects. In addition, we show that applying an auxiliary detection loss on the selected tokens in the encoder improves the performance while minimizing computational overhead. We validate that *Sparse DETR* achieves better performance than Deformable DETR even with only 10\% encoder tokens on the COCO dataset. Albeit only the encoder tokens are sparsified, the total computation cost decreases by 38\% and the frames per second (FPS) increases by 42\% compared to Deformable DETR. # Installation ## Requirements We have tested the code on the following environments: * Python 3.7.7 / Pytorch 1.6.0 / torchvisoin 0.7.0 / CUDA 10.1 / Ubuntu 18.04 * Python 3.8.3 / Pytorch 1.7.1 / torchvisoin 0.8.2 / CUDA 11.1 / Ubuntu 18.04 Run the following command to install dependencies: ```bash pip install -r requirements.txt ``` ## Compiling CUDA operators ```bash cd ./models/ops sh ./make.sh # unit test (should see all checking is True) python test.py ``` # Usage ## Dataset preparation Please download [COCO 2017 dataset](https://cocodataset.org/) and organize them as follows: ``` code_root/ └── data/ └── coco/ ├── train2017/ ├── val2017/ └── annotations/ ├── instances_train2017.json └── instances_val2017.json ``` ## Training ### Training on a single node For example, the command for training Sparse DETR with the keeping ratio of 10% on 8 GPUs is as follows: ```bash $ GPUS_PER_NODE=8 ./tools/run_dist_launch.sh 8 ./configs/swint_sparse_detr_rho_0.1.sh ``` ### Training on multiple nodes For example, the command Sparse DETR with the keeping ratio of 10% on 2 nodes of each with 8 GPUs is as follows: On node 1: ```bash $ MASTER_ADDR= NODE_RANK=0 GPUS_PER_NODE=8 ./tools/run_dist_launch.sh 16 ./configs/swint_sparse_detr_rho_0.1.sh ``` On node 2: ```bash $ MASTER_ADDR= NODE_RANK=1 GPUS_PER_NODE=8 ./tools/run_dist_launch.sh 16 ./configs/swint_sparse_detr_rho_0.1.sh ``` ### Direct argument control ```bash # Deformable DETR (with bounding-box-refinement and two-stage argument, if wanted) $ GPUS_PER_NODE=8 ./tools/run_dist_launch.sh 8 python main.py --with_box_refine --two_stage # Efficient DETR (with the class-specific head as describe in their paper) $ GPUS_PER_NODE=8 ./tools/run_dist_launch.sh 8 python main.py --with_box_refine --two_stage --eff_query_init --eff_specific_head # Sparse DETR (with the keeping ratio of 10% and encoder auxiliary loss) $ GPUS_PER_NODE=8 ./tools/run_dist_launch.sh 8 python main.py --with_box_refine --two_stage --eff_query_init --eff_specific_head --rho 0.1 --use_enc_aux_loss ``` ### Some tips to speed-up training * If your file system is slow to read images, you may consider enabling '--cache_mode' option to load the whole dataset into memory at the beginning of training. * You may increase the batch size to maximize the GPU utilization, according to GPU memory of yours, e.g., set '--batch_size 3' or '--batch_size 4'. ## Evaluation You can get the pre-trained model of Sparse DETR (the link is in "Main Results" session), then run the following command to evaluate it on COCO 2017 validation set: ```bash # Note that you should run the command with the corresponding configuration. $ ./configs/swint_sparse_detr_rho_0.1.sh --resume --eval ``` You can also run distributed evaluation by using ```./tools/run_dist_launch.sh```. # Main Results The tables below demonstrate the detection performance of Sparse DETR on the COCO 2017 validation set when using different backbones. * **Top-k** : sampling the top-k object queries instead of using the learned object queries(as in Efficient DETR). * **BBR** : performing bounding box refinement in the decoder block(as in Deformable DETR). * The **encoder auxiliary loss** proposed in our paper is only applied to Sparse DETR. * **FLOPs** and **FPS** are measured in the same way as used in Deformable DETR. * Refer to **Table 1** in the paper for more details. ## ResNet-50 backbone | Method | Epochs | ρ | Top-k & BBR | AP | #Params(M) | GFLOPs | B4FPS | Download | |:------------------:|:------:|:---:|:-----------:|:----:|:----------:|:------:|:-----:|:--------:| | Faster R-CNN + FPN | 109 | N/A | | 42.0 | 42M | 180G | 26 | | | DETR | 50 | N/A | | 35.0 | 41M | 86G | 28 | | | DETR | 500 | N/A | | 42.0 | 41M | 86G | 28 | | | DETR-DC5 | 500 | N/A | | 43.3 | 41M | 187G | 12 | | | PnP-DETR | 500 | 33% | | 41.1 | | | | | | | 500 | 50% | | 41.8 | | | | | | PnP-DETR-DC5 | 500 | 33% | | 42.7 | | | | | | | 500 | 50% | | 43.1 | | | | | | Deformable-DETR | 50 | N/A | | 43.9 | 39.8M | 172.9G | 19.1 | | | | 50 | N/A | o | 46.0 | 40.8M | 177.3G | 18.2 | | | Sparse-DETR | 50 | 10% | o | 45.3 | 40.9M | 105.4G | 26.5 | [link](https://twg.kakaocdn.net/brainrepo/sparse_detr/sparse_detr_r50_10.pth) | | | 50 | 20% | o | 45.6 | 40.9M | 112.9G | 24.8 | [link](https://twg.kakaocdn.net/brainrepo/sparse_detr/sparse_detr_r50_20.pth) | | | 50 | 30% | o | 46.0 | 40.9M | 120.5G | 23.2 | [link](https://twg.kakaocdn.net/brainrepo/sparse_detr/sparse_detr_r50_30.pth) | | | 50 | 40% | o | 46.2 | 40.9M | 128.0G | 21.8 | [link](https://twg.kakaocdn.net/brainrepo/sparse_detr/sparse_detr_r50_40.pth) | | | 50 | 50% | o | 46.3 | 40.9M | 135.6G | 20.5 | [link](https://twg.kakaocdn.net/brainrepo/sparse_detr/sparse_detr_r50_50.pth) | ## Swin-T backbone | Method | Epochs | ρ | Top-k & BBR | AP | #Params(M) | GFLOPs | B4FPS | Download | |:---------------:|:------:|:---:|:-----------:|:----:|:----------:|:------:|:-----:|:--------:| | DETR | 50 | N/A | | 35.9 | 45.0M | 91.6G | 26.8 | | | DETR | 500 | N/A | | 45.4 | 45.0M | 91.6G | 26.8 | | | Deformable-DETR | 50 | N/A | | 45.7 | 40.3M | 180.4G | 15.9 | | | | 50 | N/A | o | 48.0 | 41.3M | 184.8G | 15.4 | | | Sparse-DETR | 50 | 10% | o | 48.2 | 41.4M | 113.4G | 21.2 | [link](https://twg.kakaocdn.net/brainrepo/sparse_detr/sparse_detr_swint_10.pth) | | | 50 | 20% | o | 48.8 | 41.4M | 121.0G | 20 | [link](https://twg.kakaocdn.net/brainrepo/sparse_detr/sparse_detr_swint_20.pth) | | | 50 | 30% | o | 49.1 | 41.4M | 128.5G | 18.9 | [link](https://twg.kakaocdn.net/brainrepo/sparse_detr/sparse_detr_swint_30.pth) | | | 50 | 40% | o | 49.2 | 41.4M | 136.1G | 18 | [link](https://twg.kakaocdn.net/brainrepo/sparse_detr/sparse_detr_swint_40.pth) | | | 50 | 50% | o | 49.3 | 41.4M | 143.7G | 17.2 | [link](https://twg.kakaocdn.net/brainrepo/sparse_detr/sparse_detr_swint_50.pth) | ## Initializing ResNet-50 backbone with SCRL The performance of Sparse DETR can be further improved when the backbone network is initialized with the `SCRL`([Spatially Consistent Representation Learning](https://arxiv.org/abs/2103.06122)) that aims to learn dense representations in a self-supervised way, compared to the default initialization with the ImageNet pre-trained one, denoted as `IN-sup` in the table below. * We obtained pre-trained weights from [Torchvision](https://pytorch.org/tutorials/beginner/finetuning_torchvision_models_tutorial.html#sphx-glr-beginner-finetuning-torchvision-models-tutorial-py) for `IN-sup`, and the [SCRL GitHub repository](https://github.com/kakaobrain/scrl) for `SCRL`. * To reproduce the `SCRL` results, add `--scrl_pretrained_path ` to the training command. | Method | ρ | AP(IN-sup) | AP(SCRL) | AP(gain) | Download | |:-----------:|:---:|:-----------:|:--------:|:--------:|:--------:| | Sparse DETR | 10% | 45.3 | 46.9 | +1.6 | [link](https://twg.kakaocdn.net/brainrepo/sparse_detr/sparse_detr_r50_scrl_10.pth) | | | 20% | 45.6 | 47.2 | +1.7 | [link](https://twg.kakaocdn.net/brainrepo/sparse_detr/sparse_detr_r50_scrl_20.pth) | | | 30% | 46.0 | 47.4 | +1.4 | [link](https://twg.kakaocdn.net/brainrepo/sparse_detr/sparse_detr_r50_scrl_30.pth) | | | 40% | 46.2 | 47.7 | +1.5 | [link](https://twg.kakaocdn.net/brainrepo/sparse_detr/sparse_detr_r50_scrl_40.pth) | | | 50% | 46.3 | 47.9 | +1.6 | [link](https://twg.kakaocdn.net/brainrepo/sparse_detr/sparse_detr_r50_scrl_50.pth) | # Citation If you find Sparse DETR useful in your research, please consider citing: ```bibtex @inproceedings{roh2022sparse, title={Sparse DETR: Efficient End-to-End Object Detection with Learnable Sparsity}, author={Roh, Byungseok and Shin, JaeWoong and Shin, Wuhyun and Kim, Saehoon}, booktitle={ICLR}, year={2022} } ``` # License This project is released under the [Apache 2.0 license](./LICENSE). Copyright 2021 [Kakao Brain Corp](https://www.kakaobrain.com). All Rights Reserved. ================================================ FILE: configs/r50_deformable_detr.sh ================================================ #!/usr/bin/env bash set -x EXP_DIR=exps/r50_deformable_detr PY_ARGS=${@:1} python -u main.py \ --output_dir ${EXP_DIR} \ ${PY_ARGS} ================================================ FILE: configs/r50_efficient_detr.sh ================================================ #!/usr/bin/env bash set -x EXP_DIR=exps/r50_efficient_detr PY_ARGS=${@:1} python -u main.py \ --output_dir ${EXP_DIR} \ --with_box_refine \ --two_stage \ --eff_query_init \ --eff_specific_head \ ${PY_ARGS} ================================================ FILE: configs/r50_sparse_detr_rho_0.1.sh ================================================ #!/usr/bin/env bash set -x EXP_DIR=exps/r50_sparse_detr_0.1 PY_ARGS=${@:1} python -u main.py \ --output_dir ${EXP_DIR} \ --with_box_refine \ --two_stage \ --eff_query_init \ --eff_specific_head \ --rho 0.1 \ --use_enc_aux_loss \ ${PY_ARGS} ================================================ FILE: configs/r50_sparse_detr_rho_0.2.sh ================================================ #!/usr/bin/env bash set -x EXP_DIR=exps/r50_sparse_detr_0.2 PY_ARGS=${@:1} python -u main.py \ --output_dir ${EXP_DIR} \ --with_box_refine \ --two_stage \ --eff_query_init \ --eff_specific_head \ --rho 0.2 \ --use_enc_aux_loss \ ${PY_ARGS} ================================================ FILE: configs/r50_sparse_detr_rho_0.3.sh ================================================ #!/usr/bin/env bash set -x EXP_DIR=exps/r50_sparse_detr_0.3 PY_ARGS=${@:1} python -u main.py \ --output_dir ${EXP_DIR} \ --with_box_refine \ --two_stage \ --eff_query_init \ --eff_specific_head \ --rho 0.3 \ --use_enc_aux_loss \ ${PY_ARGS} ================================================ FILE: configs/swint_deformable_detr.sh ================================================ #!/usr/bin/env bash set -x EXP_DIR=exps/swint_deformable_detr PY_ARGS=${@:1} python -u main.py \ --output_dir ${EXP_DIR} \ --backbone swin-t \ ${PY_ARGS} ================================================ FILE: configs/swint_efficient_detr.sh ================================================ #!/usr/bin/env bash set -x EXP_DIR=exps/swint_efficient_detr PY_ARGS=${@:1} python -u main.py \ --output_dir ${EXP_DIR} \ --backbone swin-t \ --with_box_refine \ --two_stage \ --eff_query_init \ --eff_specific_head \ ${PY_ARGS} ================================================ FILE: configs/swint_sparse_detr_rho_0.1.sh ================================================ #!/usr/bin/env bash set -x EXP_DIR=exps/swint_sparse_detr_0.1 PY_ARGS=${@:1} python -u main.py \ --output_dir ${EXP_DIR} \ --backbone swin-t \ --with_box_refine \ --two_stage \ --eff_query_init \ --eff_specific_head \ --rho 0.1 \ --use_enc_aux_loss \ ${PY_ARGS} ================================================ FILE: configs/swint_sparse_detr_rho_0.2.sh ================================================ #!/usr/bin/env bash set -x EXP_DIR=exps/swint_sparse_detr_0.2 PY_ARGS=${@:1} python -u main.py \ --output_dir ${EXP_DIR} \ --backbone swin-t \ --with_box_refine \ --two_stage \ --eff_query_init \ --eff_specific_head \ --rho 0.2 \ --use_enc_aux_loss \ ${PY_ARGS} ================================================ FILE: configs/swint_sparse_detr_rho_0.3.sh ================================================ #!/usr/bin/env bash set -x EXP_DIR=exps/swint_sparse_detr_0.3 PY_ARGS=${@:1} python -u main.py \ --output_dir ${EXP_DIR} \ --backbone swin-t \ --with_box_refine \ --two_stage \ --eff_query_init \ --eff_specific_head \ --rho 0.3 \ --use_enc_aux_loss \ ${PY_ARGS} ================================================ FILE: datasets/__init__.py ================================================ # ------------------------------------------------------------------------ # Deformable DETR # Copyright (c) 2020 SenseTime. All Rights Reserved. # Licensed under the Apache License, Version 2.0 [see LICENSE for details] # ------------------------------------------------------------------------ # Modified from DETR (https://github.com/facebookresearch/detr) # Copyright (c) Facebook, Inc. and its affiliates. All Rights Reserved # ------------------------------------------------------------------------ import torch.utils.data from .torchvision_datasets import CocoDetection from .coco import build as build_coco def get_coco_api_from_dataset(dataset): for _ in range(10): # if isinstance(dataset, torchvision.datasets.CocoDetection): # break if isinstance(dataset, torch.utils.data.Subset): dataset = dataset.dataset if isinstance(dataset, CocoDetection): return dataset.coco def build_dataset(image_set, args): if args.dataset_file == 'coco': return build_coco(image_set, args) if args.dataset_file == 'coco_panoptic': # to avoid making panopticapi required for coco from .coco_panoptic import build as build_coco_panoptic return build_coco_panoptic(image_set, args) raise ValueError(f'dataset {args.dataset_file} not supported') ================================================ FILE: datasets/coco.py ================================================ # ------------------------------------------------------------------------ # Deformable DETR # Copyright (c) 2020 SenseTime. All Rights Reserved. # Licensed under the Apache License, Version 2.0 [see LICENSE for details] # ------------------------------------------------------------------------ # Modified from DETR (https://github.com/facebookresearch/detr) # Copyright (c) Facebook, Inc. and its affiliates. All Rights Reserved # ------------------------------------------------------------------------ """ COCO dataset which returns image_id for evaluation. Mostly copy-paste from https://github.com/pytorch/vision/blob/13b35ff/references/detection/coco_utils.py """ from pathlib import Path import torch import torch.utils.data from pycocotools import mask as coco_mask from .torchvision_datasets import CocoDetection as TvCocoDetection from util.misc import get_local_rank, get_local_size import datasets.transforms as T class CocoDetection(TvCocoDetection): def __init__(self, img_folder, ann_file, transforms, return_masks, cache_mode=False, local_rank=0, local_size=1): super(CocoDetection, self).__init__(img_folder, ann_file, cache_mode=cache_mode, local_rank=local_rank, local_size=local_size) self._transforms = transforms self.prepare = ConvertCocoPolysToMask(return_masks) def __getitem__(self, idx): img, target = super(CocoDetection, self).__getitem__(idx) image_id = self.ids[idx] target = {'image_id': image_id, 'annotations': target} img, target = self.prepare(img, target) if self._transforms is not None: img, target = self._transforms(img, target) return img, target def convert_coco_poly_to_mask(segmentations, height, width): masks = [] for polygons in segmentations: rles = coco_mask.frPyObjects(polygons, height, width) mask = coco_mask.decode(rles) if len(mask.shape) < 3: mask = mask[..., None] mask = torch.as_tensor(mask, dtype=torch.uint8) mask = mask.any(dim=2) masks.append(mask) if masks: masks = torch.stack(masks, dim=0) else: masks = torch.zeros((0, height, width), dtype=torch.uint8) return masks class ConvertCocoPolysToMask(object): def __init__(self, return_masks=False): self.return_masks = return_masks def __call__(self, image, target): w, h = image.size image_id = target["image_id"] image_id = torch.tensor([image_id]) anno = target["annotations"] anno = [obj for obj in anno if 'iscrowd' not in obj or obj['iscrowd'] == 0] boxes = [obj["bbox"] for obj in anno] # guard against no boxes via resizing boxes = torch.as_tensor(boxes, dtype=torch.float32).reshape(-1, 4) boxes[:, 2:] += boxes[:, :2] boxes[:, 0::2].clamp_(min=0, max=w) boxes[:, 1::2].clamp_(min=0, max=h) classes = [obj["category_id"] for obj in anno] classes = torch.tensor(classes, dtype=torch.int64) if self.return_masks: segmentations = [obj["segmentation"] for obj in anno] masks = convert_coco_poly_to_mask(segmentations, h, w) keypoints = None if anno and "keypoints" in anno[0]: keypoints = [obj["keypoints"] for obj in anno] keypoints = torch.as_tensor(keypoints, dtype=torch.float32) num_keypoints = keypoints.shape[0] if num_keypoints: keypoints = keypoints.view(num_keypoints, -1, 3) keep = (boxes[:, 3] > boxes[:, 1]) & (boxes[:, 2] > boxes[:, 0]) boxes = boxes[keep] classes = classes[keep] if self.return_masks: masks = masks[keep] if keypoints is not None: keypoints = keypoints[keep] target = {} target["boxes"] = boxes target["labels"] = classes if self.return_masks: target["masks"] = masks target["image_id"] = image_id if keypoints is not None: target["keypoints"] = keypoints # for conversion to coco api area = torch.tensor([obj["area"] for obj in anno]) iscrowd = torch.tensor([obj["iscrowd"] if "iscrowd" in obj else 0 for obj in anno]) target["area"] = area[keep] target["iscrowd"] = iscrowd[keep] target["orig_size"] = torch.as_tensor([int(h), int(w)]) target["size"] = torch.as_tensor([int(h), int(w)]) return image, target def make_coco_transforms(image_set): normalize = T.Compose([ T.ToTensor(), T.Normalize([0.485, 0.456, 0.406], [0.229, 0.224, 0.225]) ]) scales = [480, 512, 544, 576, 608, 640, 672, 704, 736, 768, 800] if image_set == 'train': return T.Compose([ T.RandomHorizontalFlip(), T.RandomSelect( T.RandomResize(scales, max_size=1333), T.Compose([ T.RandomResize([400, 500, 600]), T.RandomSizeCrop(384, 600), T.RandomResize(scales, max_size=1333), ]) ), normalize, ]) if image_set == 'val': return T.Compose([ T.RandomResize([800], max_size=1333), normalize, ]) raise ValueError(f'unknown {image_set}') def build(image_set, args): root = Path(args.coco_path) assert root.exists(), f'provided COCO path {root} does not exist' mode = 'instances' PATHS = { "train": (root / "train2017", root / "annotations" / f'{mode}_train2017.json'), "val": (root / "val2017", root / "annotations" / f'{mode}_val2017.json'), } img_folder, ann_file = PATHS[image_set] dataset = CocoDetection(img_folder, ann_file, transforms=make_coco_transforms(image_set), return_masks=args.masks, cache_mode=args.cache_mode, local_rank=get_local_rank(), local_size=get_local_size()) return dataset ================================================ FILE: datasets/coco_eval.py ================================================ # ------------------------------------------------------------------------ # Deformable DETR # Copyright (c) 2020 SenseTime. All Rights Reserved. # Licensed under the Apache License, Version 2.0 [see LICENSE for details] # ------------------------------------------------------------------------ # Modified from DETR (https://github.com/facebookresearch/detr) # Copyright (c) Facebook, Inc. and its affiliates. All Rights Reserved # ------------------------------------------------------------------------ """ COCO evaluator that works in distributed mode. Mostly copy-paste from https://github.com/pytorch/vision/blob/edfd5a7/references/detection/coco_eval.py The difference is that there is less copy-pasting from pycocotools in the end of the file, as python3 can suppress prints with contextlib """ import os import contextlib import copy import numpy as np import torch from pycocotools.cocoeval import COCOeval from pycocotools.coco import COCO import pycocotools.mask as mask_util from util.misc import all_gather class CocoEvaluator(object): def __init__(self, coco_gt, iou_types): assert isinstance(iou_types, (list, tuple)) coco_gt = copy.deepcopy(coco_gt) self.coco_gt = coco_gt self.iou_types = iou_types self.coco_eval = {} for iou_type in iou_types: self.coco_eval[iou_type] = COCOeval(coco_gt, iouType=iou_type) self.img_ids = [] self.eval_imgs = {k: [] for k in iou_types} def update(self, predictions): img_ids = list(np.unique(list(predictions.keys()))) self.img_ids.extend(img_ids) for iou_type in self.iou_types: results = self.prepare(predictions, iou_type) # suppress pycocotools prints with open(os.devnull, 'w') as devnull: with contextlib.redirect_stdout(devnull): coco_dt = COCO.loadRes(self.coco_gt, results) if results else COCO() coco_eval = self.coco_eval[iou_type] coco_eval.cocoDt = coco_dt coco_eval.params.imgIds = list(img_ids) img_ids, eval_imgs = evaluate(coco_eval) self.eval_imgs[iou_type].append(eval_imgs) def synchronize_between_processes(self): for iou_type in self.iou_types: self.eval_imgs[iou_type] = np.concatenate(self.eval_imgs[iou_type], 2) create_common_coco_eval(self.coco_eval[iou_type], self.img_ids, self.eval_imgs[iou_type]) def accumulate(self): for coco_eval in self.coco_eval.values(): coco_eval.accumulate() def summarize(self): for iou_type, coco_eval in self.coco_eval.items(): print("IoU metric: {}".format(iou_type)) coco_eval.summarize() def prepare(self, predictions, iou_type): if iou_type == "bbox": return self.prepare_for_coco_detection(predictions) elif iou_type == "segm": return self.prepare_for_coco_segmentation(predictions) elif iou_type == "keypoints": return self.prepare_for_coco_keypoint(predictions) else: raise ValueError("Unknown iou type {}".format(iou_type)) def prepare_for_coco_detection(self, predictions): coco_results = [] for original_id, prediction in predictions.items(): if len(prediction) == 0: continue boxes = prediction["boxes"] boxes = convert_to_xywh(boxes).tolist() scores = prediction["scores"].tolist() labels = prediction["labels"].tolist() coco_results.extend( [ { "image_id": original_id, "category_id": labels[k], "bbox": box, "score": scores[k], } for k, box in enumerate(boxes) ] ) return coco_results def prepare_for_coco_segmentation(self, predictions): coco_results = [] for original_id, prediction in predictions.items(): if len(prediction) == 0: continue scores = prediction["scores"] labels = prediction["labels"] masks = prediction["masks"] masks = masks > 0.5 scores = prediction["scores"].tolist() labels = prediction["labels"].tolist() rles = [ mask_util.encode(np.array(mask[0, :, :, np.newaxis], dtype=np.uint8, order="F"))[0] for mask in masks ] for rle in rles: rle["counts"] = rle["counts"].decode("utf-8") coco_results.extend( [ { "image_id": original_id, "category_id": labels[k], "segmentation": rle, "score": scores[k], } for k, rle in enumerate(rles) ] ) return coco_results def prepare_for_coco_keypoint(self, predictions): coco_results = [] for original_id, prediction in predictions.items(): if len(prediction) == 0: continue boxes = prediction["boxes"] boxes = convert_to_xywh(boxes).tolist() scores = prediction["scores"].tolist() labels = prediction["labels"].tolist() keypoints = prediction["keypoints"] keypoints = keypoints.flatten(start_dim=1).tolist() coco_results.extend( [ { "image_id": original_id, "category_id": labels[k], 'keypoints': keypoint, "score": scores[k], } for k, keypoint in enumerate(keypoints) ] ) return coco_results def convert_to_xywh(boxes): xmin, ymin, xmax, ymax = boxes.unbind(1) return torch.stack((xmin, ymin, xmax - xmin, ymax - ymin), dim=1) def merge(img_ids, eval_imgs): all_img_ids = all_gather(img_ids) all_eval_imgs = all_gather(eval_imgs) merged_img_ids = [] for p in all_img_ids: merged_img_ids.extend(p) merged_eval_imgs = [] for p in all_eval_imgs: merged_eval_imgs.append(p) merged_img_ids = np.array(merged_img_ids) merged_eval_imgs = np.concatenate(merged_eval_imgs, 2) # keep only unique (and in sorted order) images merged_img_ids, idx = np.unique(merged_img_ids, return_index=True) merged_eval_imgs = merged_eval_imgs[..., idx] return merged_img_ids, merged_eval_imgs def create_common_coco_eval(coco_eval, img_ids, eval_imgs): img_ids, eval_imgs = merge(img_ids, eval_imgs) img_ids = list(img_ids) eval_imgs = list(eval_imgs.flatten()) coco_eval.evalImgs = eval_imgs coco_eval.params.imgIds = img_ids coco_eval._paramsEval = copy.deepcopy(coco_eval.params) ################################################################# # From pycocotools, just removed the prints and fixed # a Python3 bug about unicode not defined ################################################################# def evaluate(self): ''' Run per image evaluation on given images and store results (a list of dict) in self.evalImgs :return: None ''' # tic = time.time() # print('Running per image evaluation...') p = self.params # add backward compatibility if useSegm is specified in params if p.useSegm is not None: p.iouType = 'segm' if p.useSegm == 1 else 'bbox' print('useSegm (deprecated) is not None. Running {} evaluation'.format(p.iouType)) # print('Evaluate annotation type *{}*'.format(p.iouType)) p.imgIds = list(np.unique(p.imgIds)) if p.useCats: p.catIds = list(np.unique(p.catIds)) p.maxDets = sorted(p.maxDets) self.params = p self._prepare() # loop through images, area range, max detection number catIds = p.catIds if p.useCats else [-1] if p.iouType == 'segm' or p.iouType == 'bbox': computeIoU = self.computeIoU elif p.iouType == 'keypoints': computeIoU = self.computeOks self.ious = { (imgId, catId): computeIoU(imgId, catId) for imgId in p.imgIds for catId in catIds} evaluateImg = self.evaluateImg maxDet = p.maxDets[-1] evalImgs = [ evaluateImg(imgId, catId, areaRng, maxDet) for catId in catIds for areaRng in p.areaRng for imgId in p.imgIds ] # this is NOT in the pycocotools code, but could be done outside evalImgs = np.asarray(evalImgs).reshape(len(catIds), len(p.areaRng), len(p.imgIds)) self._paramsEval = copy.deepcopy(self.params) # toc = time.time() # print('DONE (t={:0.2f}s).'.format(toc-tic)) return p.imgIds, evalImgs ################################################################# # end of straight copy from pycocotools, just removing the prints ################################################################# ================================================ FILE: datasets/coco_panoptic.py ================================================ # ------------------------------------------------------------------------ # Deformable DETR # Copyright (c) 2020 SenseTime. All Rights Reserved. # Licensed under the Apache License, Version 2.0 [see LICENSE for details] # ------------------------------------------------------------------------ # Modified from DETR (https://github.com/facebookresearch/detr) # Copyright (c) Facebook, Inc. and its affiliates. All Rights Reserved # ------------------------------------------------------------------------ import json from pathlib import Path import numpy as np import torch from PIL import Image from panopticapi.utils import rgb2id from util.box_ops import masks_to_boxes from .coco import make_coco_transforms class CocoPanoptic: def __init__(self, img_folder, ann_folder, ann_file, transforms=None, return_masks=True): with open(ann_file, 'r') as f: self.coco = json.load(f) # sort 'images' field so that they are aligned with 'annotations' # i.e., in alphabetical order self.coco['images'] = sorted(self.coco['images'], key=lambda x: x['id']) # sanity check if "annotations" in self.coco: for img, ann in zip(self.coco['images'], self.coco['annotations']): assert img['file_name'][:-4] == ann['file_name'][:-4] self.img_folder = img_folder self.ann_folder = ann_folder self.ann_file = ann_file self.transforms = transforms self.return_masks = return_masks def __getitem__(self, idx): ann_info = self.coco['annotations'][idx] if "annotations" in self.coco else self.coco['images'][idx] img_path = Path(self.img_folder) / ann_info['file_name'].replace('.png', '.jpg') ann_path = Path(self.ann_folder) / ann_info['file_name'] img = Image.open(img_path).convert('RGB') w, h = img.size if "segments_info" in ann_info: masks = np.asarray(Image.open(ann_path), dtype=np.uint32) masks = rgb2id(masks) ids = np.array([ann['id'] for ann in ann_info['segments_info']]) masks = masks == ids[:, None, None] masks = torch.as_tensor(masks, dtype=torch.uint8) labels = torch.tensor([ann['category_id'] for ann in ann_info['segments_info']], dtype=torch.int64) target = {} target['image_id'] = torch.tensor([ann_info['image_id'] if "image_id" in ann_info else ann_info["id"]]) if self.return_masks: target['masks'] = masks target['labels'] = labels target["boxes"] = masks_to_boxes(masks) target['size'] = torch.as_tensor([int(h), int(w)]) target['orig_size'] = torch.as_tensor([int(h), int(w)]) if "segments_info" in ann_info: for name in ['iscrowd', 'area']: target[name] = torch.tensor([ann[name] for ann in ann_info['segments_info']]) if self.transforms is not None: img, target = self.transforms(img, target) return img, target def __len__(self): return len(self.coco['images']) def get_height_and_width(self, idx): img_info = self.coco['images'][idx] height = img_info['height'] width = img_info['width'] return height, width def build(image_set, args): img_folder_root = Path(args.coco_path) ann_folder_root = Path(args.coco_panoptic_path) assert img_folder_root.exists(), f'provided COCO path {img_folder_root} does not exist' assert ann_folder_root.exists(), f'provided COCO path {ann_folder_root} does not exist' mode = 'panoptic' PATHS = { "train": ("train2017", Path("annotations") / f'{mode}_train2017.json'), "val": ("val2017", Path("annotations") / f'{mode}_val2017.json'), } img_folder, ann_file = PATHS[image_set] img_folder_path = img_folder_root / img_folder ann_folder = ann_folder_root / f'{mode}_{img_folder}' ann_file = ann_folder_root / ann_file dataset = CocoPanoptic(img_folder_path, ann_folder, ann_file, transforms=make_coco_transforms(image_set), return_masks=args.masks) return dataset ================================================ FILE: datasets/data_prefetcher.py ================================================ # ------------------------------------------------------------------------ # Deformable DETR # Copyright (c) 2020 SenseTime. All Rights Reserved. # Licensed under the Apache License, Version 2.0 [see LICENSE for details] # ------------------------------------------------------------------------ import torch def to_cuda(samples, targets, device): samples = samples.to(device, non_blocking=True) targets = [{k: v.to(device, non_blocking=True) for k, v in t.items()} for t in targets] return samples, targets class data_prefetcher(): def __init__(self, loader, device, prefetch=True): self.loader = iter(loader) self.prefetch = prefetch self.device = device if prefetch: self.stream = torch.cuda.Stream() self.preload() def preload(self): try: self.next_samples, self.next_targets = next(self.loader) except StopIteration: self.next_samples = None self.next_targets = None return # if record_stream() doesn't work, another option is to make sure device inputs are created # on the main stream. # self.next_input_gpu = torch.empty_like(self.next_input, device='cuda') # self.next_target_gpu = torch.empty_like(self.next_target, device='cuda') # Need to make sure the memory allocated for next_* is not still in use by the main stream # at the time we start copying to next_*: # self.stream.wait_stream(torch.cuda.current_stream()) with torch.cuda.stream(self.stream): self.next_samples, self.next_targets = to_cuda(self.next_samples, self.next_targets, self.device) # more code for the alternative if record_stream() doesn't work: # copy_ will record the use of the pinned source tensor in this side stream. # self.next_input_gpu.copy_(self.next_input, non_blocking=True) # self.next_target_gpu.copy_(self.next_target, non_blocking=True) # self.next_input = self.next_input_gpu # self.next_target = self.next_target_gpu # With Amp, it isn't necessary to manually convert data to half. # if args.fp16: # self.next_input = self.next_input.half() # else: def next(self): if self.prefetch: torch.cuda.current_stream().wait_stream(self.stream) samples = self.next_samples targets = self.next_targets if samples is not None: samples.record_stream(torch.cuda.current_stream()) if targets is not None: for t in targets: for k, v in t.items(): v.record_stream(torch.cuda.current_stream()) self.preload() else: try: samples, targets = next(self.loader) samples, targets = to_cuda(samples, targets, self.device) except StopIteration: samples = None targets = None return samples, targets ================================================ FILE: datasets/panoptic_eval.py ================================================ # ------------------------------------------------------------------------ # Deformable DETR # Copyright (c) 2020 SenseTime. All Rights Reserved. # Licensed under the Apache License, Version 2.0 [see LICENSE for details] # ------------------------------------------------------------------------ # Modified from DETR (https://github.com/facebookresearch/detr) # Copyright (c) Facebook, Inc. and its affiliates. All Rights Reserved # ------------------------------------------------------------------------ import json import os import util.misc as utils try: from panopticapi.evaluation import pq_compute except ImportError: pass class PanopticEvaluator(object): def __init__(self, ann_file, ann_folder, output_dir="panoptic_eval"): self.gt_json = ann_file self.gt_folder = ann_folder if utils.is_main_process(): if not os.path.exists(output_dir): os.mkdir(output_dir) self.output_dir = output_dir self.predictions = [] def update(self, predictions): for p in predictions: with open(os.path.join(self.output_dir, p["file_name"]), "wb") as f: f.write(p.pop("png_string")) self.predictions += predictions def synchronize_between_processes(self): all_predictions = utils.all_gather(self.predictions) merged_predictions = [] for p in all_predictions: merged_predictions += p self.predictions = merged_predictions def summarize(self): if utils.is_main_process(): json_data = {"annotations": self.predictions} predictions_json = os.path.join(self.output_dir, "predictions.json") with open(predictions_json, "w") as f: f.write(json.dumps(json_data)) return pq_compute(self.gt_json, predictions_json, gt_folder=self.gt_folder, pred_folder=self.output_dir) return None ================================================ FILE: datasets/samplers.py ================================================ # ------------------------------------------------------------------------ # Deformable DETR # Copyright (c) 2020 SenseTime. All Rights Reserved. # Licensed under the Apache License, Version 2.0 [see LICENSE for details] # ------------------------------------------------------------------------ # Modified from codes in torch.utils.data.distributed # ------------------------------------------------------------------------ import os import math import torch import torch.distributed as dist from torch.utils.data.sampler import Sampler class DistributedSampler(Sampler): """Sampler that restricts data loading to a subset of the dataset. It is especially useful in conjunction with :class:`torch.nn.parallel.DistributedDataParallel`. In such case, each process can pass a DistributedSampler instance as a DataLoader sampler, and load a subset of the original dataset that is exclusive to it. .. note:: Dataset is assumed to be of constant size. Arguments: dataset: Dataset used for sampling. num_replicas (optional): Number of processes participating in distributed training. rank (optional): Rank of the current process within num_replicas. """ def __init__(self, dataset, num_replicas=None, rank=None, local_rank=None, local_size=None, shuffle=True): if num_replicas is None: if not dist.is_available(): raise RuntimeError("Requires distributed package to be available") num_replicas = dist.get_world_size() if rank is None: if not dist.is_available(): raise RuntimeError("Requires distributed package to be available") rank = dist.get_rank() self.dataset = dataset self.num_replicas = num_replicas self.rank = rank self.epoch = 0 self.num_samples = int(math.ceil(len(self.dataset) * 1.0 / self.num_replicas)) self.total_size = self.num_samples * self.num_replicas self.shuffle = shuffle def __iter__(self): if self.shuffle: # deterministically shuffle based on epoch g = torch.Generator() g.manual_seed(self.epoch) indices = torch.randperm(len(self.dataset), generator=g).tolist() else: indices = torch.arange(len(self.dataset)).tolist() # add extra samples to make it evenly divisible indices += indices[: (self.total_size - len(indices))] assert len(indices) == self.total_size # subsample offset = self.num_samples * self.rank indices = indices[offset : offset + self.num_samples] assert len(indices) == self.num_samples return iter(indices) def __len__(self): return self.num_samples def set_epoch(self, epoch): self.epoch = epoch class NodeDistributedSampler(Sampler): """Sampler that restricts data loading to a subset of the dataset. It is especially useful in conjunction with :class:`torch.nn.parallel.DistributedDataParallel`. In such case, each process can pass a DistributedSampler instance as a DataLoader sampler, and load a subset of the original dataset that is exclusive to it. .. note:: Dataset is assumed to be of constant size. Arguments: dataset: Dataset used for sampling. num_replicas (optional): Number of processes participating in distributed training. rank (optional): Rank of the current process within num_replicas. """ def __init__(self, dataset, num_replicas=None, rank=None, local_rank=None, local_size=None, shuffle=True): if num_replicas is None: if not dist.is_available(): raise RuntimeError("Requires distributed package to be available") num_replicas = dist.get_world_size() if rank is None: if not dist.is_available(): raise RuntimeError("Requires distributed package to be available") rank = dist.get_rank() if local_rank is None: local_rank = int(os.environ.get('LOCAL_RANK', 0)) if local_size is None: local_size = int(os.environ.get('LOCAL_SIZE', 1)) self.dataset = dataset self.shuffle = shuffle self.num_replicas = num_replicas self.num_parts = local_size self.rank = rank self.local_rank = local_rank self.epoch = 0 self.num_samples = int(math.ceil(len(self.dataset) * 1.0 / self.num_replicas)) self.total_size = self.num_samples * self.num_replicas self.total_size_parts = self.num_samples * self.num_replicas // self.num_parts def __iter__(self): if self.shuffle: # deterministically shuffle based on epoch g = torch.Generator() g.manual_seed(self.epoch) indices = torch.randperm(len(self.dataset), generator=g).tolist() else: indices = torch.arange(len(self.dataset)).tolist() indices = [i for i in indices if i % self.num_parts == self.local_rank] # add extra samples to make it evenly divisible indices += indices[:(self.total_size_parts - len(indices))] assert len(indices) == self.total_size_parts # subsample indices = indices[self.rank // self.num_parts:self.total_size_parts:self.num_replicas // self.num_parts] assert len(indices) == self.num_samples return iter(indices) def __len__(self): return self.num_samples def set_epoch(self, epoch): self.epoch = epoch ================================================ FILE: datasets/torchvision_datasets/__init__.py ================================================ # ------------------------------------------------------------------------ # Deformable DETR # Copyright (c) 2020 SenseTime. All Rights Reserved. # Licensed under the Apache License, Version 2.0 [see LICENSE for details] # ------------------------------------------------------------------------ from .coco import CocoDetection ================================================ FILE: datasets/torchvision_datasets/coco.py ================================================ # ------------------------------------------------------------------------ # Deformable DETR # Copyright (c) 2020 SenseTime. All Rights Reserved. # Licensed under the Apache License, Version 2.0 [see LICENSE for details] # ------------------------------------------------------------------------ # Modified from torchvision # ------------------------------------------------------------------------ """ Copy-Paste from torchvision, but add utility of caching images on memory """ from torchvision.datasets.vision import VisionDataset from PIL import Image import os import os.path import tqdm from io import BytesIO class CocoDetection(VisionDataset): """`MS Coco Detection `_ Dataset. Args: root (string): Root directory where images are downloaded to. annFile (string): Path to json annotation file. transform (callable, optional): A function/transform that takes in an PIL image and returns a transformed version. E.g, ``transforms.ToTensor`` target_transform (callable, optional): A function/transform that takes in the target and transforms it. transforms (callable, optional): A function/transform that takes input sample and its target as entry and returns a transformed version. """ def __init__(self, root, annFile, transform=None, target_transform=None, transforms=None, cache_mode=False, local_rank=0, local_size=1): super(CocoDetection, self).__init__(root, transforms, transform, target_transform) from pycocotools.coco import COCO self.coco = COCO(annFile) self.ids = list(sorted(self.coco.imgs.keys())) self.cache_mode = cache_mode self.local_rank = local_rank self.local_size = local_size if cache_mode: self.cache = {} self.cache_images() def cache_images(self): self.cache = {} for index, img_id in zip(tqdm.trange(len(self.ids)), self.ids): if index % self.local_size != self.local_rank: continue path = self.coco.loadImgs(img_id)[0]['file_name'] with open(os.path.join(self.root, path), 'rb') as f: self.cache[path] = f.read() def get_image(self, path): if self.cache_mode: if path not in self.cache.keys(): with open(os.path.join(self.root, path), 'rb') as f: self.cache[path] = f.read() return Image.open(BytesIO(self.cache[path])).convert('RGB') return Image.open(os.path.join(self.root, path)).convert('RGB') def __getitem__(self, index): """ Args: index (int): Index Returns: tuple: Tuple (image, target). target is the object returned by ``coco.loadAnns``. """ coco = self.coco img_id = self.ids[index] ann_ids = coco.getAnnIds(imgIds=img_id) target = coco.loadAnns(ann_ids) path = coco.loadImgs(img_id)[0]['file_name'] img = self.get_image(path) if self.transforms is not None: img, target = self.transforms(img, target) return img, target def __len__(self): return len(self.ids) ================================================ FILE: datasets/transforms.py ================================================ # ------------------------------------------------------------------------ # Deformable DETR # Copyright (c) 2020 SenseTime. All Rights Reserved. # Licensed under the Apache License, Version 2.0 [see LICENSE for details] # ------------------------------------------------------------------------ # Modified from DETR (https://github.com/facebookresearch/detr) # Copyright (c) Facebook, Inc. and its affiliates. All Rights Reserved # ------------------------------------------------------------------------ """ Transforms and data augmentation for both image + bbox. """ import random import PIL import torch import torchvision.transforms as T import torchvision.transforms.functional as F from util.box_ops import box_xyxy_to_cxcywh from util.misc import interpolate def crop(image, target, region): cropped_image = F.crop(image, *region) target = target.copy() i, j, h, w = region # should we do something wrt the original size? target["size"] = torch.tensor([h, w]) fields = ["labels", "area", "iscrowd"] if "boxes" in target: boxes = target["boxes"] max_size = torch.as_tensor([w, h], dtype=torch.float32) cropped_boxes = boxes - torch.as_tensor([j, i, j, i]) cropped_boxes = torch.min(cropped_boxes.reshape(-1, 2, 2), max_size) cropped_boxes = cropped_boxes.clamp(min=0) area = (cropped_boxes[:, 1, :] - cropped_boxes[:, 0, :]).prod(dim=1) target["boxes"] = cropped_boxes.reshape(-1, 4) target["area"] = area fields.append("boxes") if "masks" in target: # FIXME should we update the area here if there are no boxes? target['masks'] = target['masks'][:, i:i + h, j:j + w] fields.append("masks") # remove elements for which the boxes or masks that have zero area if "boxes" in target or "masks" in target: # favor boxes selection when defining which elements to keep # this is compatible with previous implementation if "boxes" in target: cropped_boxes = target['boxes'].reshape(-1, 2, 2) keep = torch.all(cropped_boxes[:, 1, :] > cropped_boxes[:, 0, :], dim=1) else: keep = target['masks'].flatten(1).any(1) for field in fields: target[field] = target[field][keep] return cropped_image, target def hflip(image, target): flipped_image = F.hflip(image) w, h = image.size target = target.copy() if "boxes" in target: boxes = target["boxes"] boxes = boxes[:, [2, 1, 0, 3]] * torch.as_tensor([-1, 1, -1, 1]) + torch.as_tensor([w, 0, w, 0]) target["boxes"] = boxes if "masks" in target: target['masks'] = target['masks'].flip(-1) return flipped_image, target def resize(image, target, size, max_size=None): # size can be min_size (scalar) or (w, h) tuple def get_size_with_aspect_ratio(image_size, size, max_size=None): w, h = image_size if max_size is not None: min_original_size = float(min((w, h))) max_original_size = float(max((w, h))) if max_original_size / min_original_size * size > max_size: size = int(round(max_size * min_original_size / max_original_size)) if (w <= h and w == size) or (h <= w and h == size): return (h, w) if w < h: ow = size oh = int(size * h / w) else: oh = size ow = int(size * w / h) return (oh, ow) def get_size(image_size, size, max_size=None): if isinstance(size, (list, tuple)): return size[::-1] else: return get_size_with_aspect_ratio(image_size, size, max_size) size = get_size(image.size, size, max_size) rescaled_image = F.resize(image, size) if target is None: return rescaled_image, None ratios = tuple(float(s) / float(s_orig) for s, s_orig in zip(rescaled_image.size, image.size)) ratio_width, ratio_height = ratios target = target.copy() if "boxes" in target: boxes = target["boxes"] scaled_boxes = boxes * torch.as_tensor([ratio_width, ratio_height, ratio_width, ratio_height]) target["boxes"] = scaled_boxes if "area" in target: area = target["area"] scaled_area = area * (ratio_width * ratio_height) target["area"] = scaled_area h, w = size target["size"] = torch.tensor([h, w]) if "masks" in target: target['masks'] = interpolate( target['masks'][:, None].float(), size, mode="nearest")[:, 0] > 0.5 return rescaled_image, target def pad(image, target, padding): # assumes that we only pad on the bottom right corners padded_image = F.pad(image, (0, 0, padding[0], padding[1])) if target is None: return padded_image, None target = target.copy() # should we do something wrt the original size? target["size"] = torch.tensor(padded_image[::-1]) if "masks" in target: target['masks'] = torch.nn.functional.pad(target['masks'], (0, padding[0], 0, padding[1])) return padded_image, target class RandomCrop(object): def __init__(self, size): self.size = size def __call__(self, img, target): region = T.RandomCrop.get_params(img, self.size) return crop(img, target, region) class RandomSizeCrop(object): def __init__(self, min_size: int, max_size: int): self.min_size = min_size self.max_size = max_size def __call__(self, img: PIL.Image.Image, target: dict): w = random.randint(self.min_size, min(img.width, self.max_size)) h = random.randint(self.min_size, min(img.height, self.max_size)) region = T.RandomCrop.get_params(img, [h, w]) return crop(img, target, region) class CenterCrop(object): def __init__(self, size): self.size = size def __call__(self, img, target): image_width, image_height = img.size crop_height, crop_width = self.size crop_top = int(round((image_height - crop_height) / 2.)) crop_left = int(round((image_width - crop_width) / 2.)) return crop(img, target, (crop_top, crop_left, crop_height, crop_width)) class RandomHorizontalFlip(object): def __init__(self, p=0.5): self.p = p def __call__(self, img, target): if random.random() < self.p: return hflip(img, target) return img, target class RandomResize(object): def __init__(self, sizes, max_size=None): assert isinstance(sizes, (list, tuple)) self.sizes = sizes self.max_size = max_size def __call__(self, img, target=None): size = random.choice(self.sizes) return resize(img, target, size, self.max_size) class RandomPad(object): def __init__(self, max_pad): self.max_pad = max_pad def __call__(self, img, target): pad_x = random.randint(0, self.max_pad) pad_y = random.randint(0, self.max_pad) return pad(img, target, (pad_x, pad_y)) class RandomSelect(object): """ Randomly selects between transforms1 and transforms2, with probability p for transforms1 and (1 - p) for transforms2 """ def __init__(self, transforms1, transforms2, p=0.5): self.transforms1 = transforms1 self.transforms2 = transforms2 self.p = p def __call__(self, img, target): if random.random() < self.p: return self.transforms1(img, target) return self.transforms2(img, target) class ToTensor(object): def __call__(self, img, target): return F.to_tensor(img), target class RandomErasing(object): def __init__(self, *args, **kwargs): self.eraser = T.RandomErasing(*args, **kwargs) def __call__(self, img, target): return self.eraser(img), target class Normalize(object): def __init__(self, mean, std): self.mean = mean self.std = std def __call__(self, image, target=None): image = F.normalize(image, mean=self.mean, std=self.std) if target is None: return image, None target = target.copy() h, w = image.shape[-2:] if "boxes" in target: boxes = target["boxes"] boxes = box_xyxy_to_cxcywh(boxes) boxes = boxes / torch.tensor([w, h, w, h], dtype=torch.float32) target["boxes"] = boxes return image, target class Compose(object): def __init__(self, transforms): self.transforms = transforms def __call__(self, image, target): for t in self.transforms: image, target = t(image, target) return image, target def __repr__(self): format_string = self.__class__.__name__ + "(" for t in self.transforms: format_string += "\n" format_string += " {0}".format(t) format_string += "\n)" return format_string ================================================ FILE: engine.py ================================================ # ------------------------------------------------------------------------------------ # Sparse DETR # Copyright (c) 2021 KakaoBrain. All Rights Reserved. # Licensed under the Apache License, Version 2.0 [see LICENSE for details] # ------------------------------------------------------------------------------------ # Modified from Deformable DETR (https://github.com/fundamentalvision/Deformable-DETR) # Copyright (c) 2020 SenseTime. All Rights Reserved. # ------------------------------------------------------------------------------------ # Modified from DETR (https://github.com/facebookresearch/detr) # Copyright (c) Facebook, Inc. and its affiliates. All Rights Reserved # ------------------------------------------------------------------------------------ """ Train and eval functions used in main.py """ import math import os import sys from typing import Iterable import torch import util.misc as utils from datasets.coco_eval import CocoEvaluator from datasets.panoptic_eval import PanopticEvaluator from datasets.data_prefetcher import data_prefetcher from util.misc import check_unused_parameters def train_one_epoch(model: torch.nn.Module, criterion: torch.nn.Module, data_loader: Iterable, optimizer: torch.optim.Optimizer, device: torch.device, epoch: int, max_norm: float = 0, writer=None, total_iter=0): model.train() criterion.train() metric_logger = utils.MetricLogger(delimiter=" ") metric_logger.add_meter('lr', utils.SmoothedValue(window_size=1, fmt='{value:.6f}')) metric_logger.add_meter('class_error', utils.SmoothedValue(window_size=1, fmt='{value:.2f}')) metric_logger.add_meter('grad_norm', utils.SmoothedValue(window_size=1, fmt='{value:.2f}')) header = 'Epoch: [{}]'.format(epoch) print_freq = 10 prefetcher = data_prefetcher(data_loader, device, prefetch=True) samples, targets = prefetcher.next() for i in metric_logger.log_every(range(len(data_loader)), print_freq, header): outputs = model(samples) loss_dict = criterion(outputs, targets) weight_dict = criterion.weight_dict losses = sum(loss_dict[k] * weight_dict[k] for k in loss_dict.keys() if k in weight_dict) # reduce losses over all GPUs for logging purposes loss_dict_reduced = utils.reduce_dict(loss_dict) loss_dict_reduced_unscaled = {f'{k}_unscaled': v for k, v in loss_dict_reduced.items()} loss_dict_reduced_scaled = {k: v * weight_dict[k] for k, v in loss_dict_reduced.items() if k in weight_dict} losses_reduced_scaled = sum(loss_dict_reduced_scaled.values()) loss_value = losses_reduced_scaled.item() if not math.isfinite(loss_value): print("Loss is {}, stopping training".format(loss_value)) print(loss_dict_reduced) sys.exit(1) optimizer.zero_grad() losses.backward() if i == 0: check_unused_parameters(model, loss_dict, weight_dict) if max_norm > 0: grad_total_norm = torch.nn.utils.clip_grad_norm_(model.parameters(), max_norm) else: grad_total_norm = utils.get_total_grad_norm(model.parameters(), max_norm) metric_logger.update(loss=loss_value, **loss_dict_reduced_scaled, **loss_dict_reduced_unscaled) metric_logger.update(class_error=loss_dict_reduced['class_error']) metric_logger.update(lr=optimizer.param_groups[0]["lr"]) metric_logger.update(grad_norm=grad_total_norm) optimizer.step() if total_iter % (print_freq*10) == 0 and utils.is_main_process(): writer.add_scalar('train/loss', loss_value, total_iter) writer.add_scalar('train/class_error', loss_dict_reduced['class_error'], total_iter) writer.add_scalar('lr', optimizer.param_groups[0]["lr"], total_iter) writer.add_scalar('train/grad_norm', grad_total_norm, total_iter) for key, value in loss_dict_reduced_scaled.items(): writer.add_scalar('train/'+key, value, total_iter) for key, value in loss_dict_reduced_unscaled.items(): if "corr" in key: writer.add_scalar('train/'+key, value, total_iter) total_iter += 1 samples, targets = prefetcher.next() # gather the stats from all processes metric_logger.synchronize_between_processes() print("Averaged stats:", metric_logger) return {k: meter.global_avg for k, meter in metric_logger.meters.items()}, total_iter @torch.no_grad() def evaluate(model, criterion, postprocessors, data_loader, base_ds, device, args): model.eval() criterion.eval() metric_logger = utils.MetricLogger(delimiter=" ") metric_logger.add_meter('class_error', utils.SmoothedValue(window_size=1, fmt='{value:.2f}')) header = 'Test:' iou_types = tuple(k for k in ('segm', 'bbox') if k in postprocessors.keys()) coco_evaluator = CocoEvaluator(base_ds, iou_types) panoptic_evaluator = None if 'panoptic' in postprocessors.keys(): panoptic_evaluator = PanopticEvaluator( data_loader.dataset.ann_file, data_loader.dataset.ann_folder, output_dir=os.path.join(args.output_dir, "panoptic_eval"), ) for step, (samples, targets) in enumerate(metric_logger.log_every(data_loader, 10, header)): samples = samples.to(device) targets = [{k: v.to(device) for k, v in t.items()} for t in targets] outputs = model(samples) loss_dict = criterion(outputs, targets) weight_dict = criterion.weight_dict # reduce losses over all GPUs for logging purposes loss_dict_reduced = utils.reduce_dict(loss_dict) loss_dict_reduced_scaled = {k: v * weight_dict[k] for k, v in loss_dict_reduced.items() if k in weight_dict} loss_dict_reduced_unscaled = {f'{k}_unscaled': v for k, v in loss_dict_reduced.items()} metric_logger.update(loss=sum(loss_dict_reduced_scaled.values()), **loss_dict_reduced_scaled, **loss_dict_reduced_unscaled) metric_logger.update(class_error=loss_dict_reduced['class_error']) orig_target_sizes = torch.stack([t["orig_size"] for t in targets], dim=0) results = postprocessors['bbox'](outputs, orig_target_sizes) if 'segm' in postprocessors.keys(): target_sizes = torch.stack([t["size"] for t in targets], dim=0) results = postprocessors['segm'](results, outputs, orig_target_sizes, target_sizes) res = {target['image_id'].item(): output for target, output in zip(targets, results)} if coco_evaluator is not None: coco_evaluator.update(res) if panoptic_evaluator is not None: res_pano = postprocessors["panoptic"](outputs, target_sizes, orig_target_sizes) for i, target in enumerate(targets): image_id = target["image_id"].item() file_name = f"{image_id:012d}.png" res_pano[i]["image_id"] = image_id res_pano[i]["file_name"] = file_name panoptic_evaluator.update(res_pano) # gather the stats from all processes metric_logger.synchronize_between_processes() print("Averaged stats:", metric_logger) if coco_evaluator is not None: coco_evaluator.synchronize_between_processes() if panoptic_evaluator is not None: panoptic_evaluator.synchronize_between_processes() # accumulate predictions from all images if coco_evaluator is not None: coco_evaluator.accumulate() coco_evaluator.summarize() panoptic_res = None if panoptic_evaluator is not None: panoptic_res = panoptic_evaluator.summarize() stats = {k: meter.global_avg for k, meter in metric_logger.meters.items()} if coco_evaluator is not None: if 'bbox' in postprocessors.keys(): stats['coco_eval_bbox'] = coco_evaluator.coco_eval['bbox'].stats.tolist() if 'segm' in postprocessors.keys(): stats['coco_eval_masks'] = coco_evaluator.coco_eval['segm'].stats.tolist() if panoptic_res is not None: stats['PQ_all'] = panoptic_res["All"] stats['PQ_th'] = panoptic_res["Things"] stats['PQ_st'] = panoptic_res["Stuff"] return stats, coco_evaluator ================================================ FILE: main.py ================================================ # ------------------------------------------------------------------------------------ # Sparse DETR # Copyright (c) 2021 KakaoBrain. All Rights Reserved. # Licensed under the Apache License, Version 2.0 [see LICENSE for details] # ------------------------------------------------------------------------------------ # Modified from Deformable DETR (https://github.com/fundamentalvision/Deformable-DETR) # Copyright (c) 2020 SenseTime. All Rights Reserved. # ------------------------------------------------------------------------------------ # Modified from DETR (https://github.com/facebookresearch/detr) # Copyright (c) Facebook, Inc. and its affiliates. All Rights Reserved # ------------------------------------------------------------------------------------ import argparse import datetime import json import random import time from tabulate import tabulate from pathlib import Path import numpy as np import torch from torch.utils.data import DataLoader, Subset import datasets import util.misc as utils import datasets.samplers as samplers from datasets import build_dataset, get_coco_api_from_dataset from engine import evaluate, train_one_epoch from models import build_model from util.benchmark import compute_fps, compute_gflops from torch.utils.tensorboard import SummaryWriter def get_args_parser(): parser = argparse.ArgumentParser('Deformable DETR Detector', add_help=False) parser.add_argument('--lr', default=2e-4, type=float) parser.add_argument('--lr_backbone_names', default=["backbone.0"], type=str, nargs='+') parser.add_argument('--lr_backbone', default=2e-5, type=float) parser.add_argument('--lr_linear_proj_names', default=['reference_points', 'sampling_offsets'], type=str, nargs='+') parser.add_argument('--lr_linear_proj_mult', default=0.1, type=float) parser.add_argument('--batch_size', default=2, type=int) parser.add_argument('--weight_decay', default=1e-4, type=float) parser.add_argument('--epochs', default=50, type=int) parser.add_argument('--lr_drop', default=40, type=int) parser.add_argument('--lr_drop_epochs', default=None, type=int, nargs='+') parser.add_argument('--clip_max_norm', default=0.1, type=float, help='gradient clipping max norm') parser.add_argument('--sgd', action='store_true') # Variants of Deformable DETR parser.add_argument('--with_box_refine', default=False, action='store_true') parser.add_argument('--two_stage', default=False, action='store_true') # Model parameters parser.add_argument('--frozen_weights', type=str, default=None, help="Path to the pretrained model. If set, only the mask head will be trained") # * Backbone parser.add_argument('--backbone', default='resnet50', type=str, help="Name of the convolutional backbone to use") parser.add_argument('--dilation', action='store_true', help="If true, we replace stride with dilation in the last convolutional block (DC5)") parser.add_argument('--position_embedding', default='sine', type=str, choices=('sine', 'learned'), help="Type of positional embedding to use on top of the image features") parser.add_argument('--position_embedding_scale', default=2 * np.pi, type=float, help="position / size * scale") parser.add_argument('--num_feature_levels', default=4, type=int, help='number of feature levels') # * Modified architecture parser.add_argument('--backbone_from_scratch', default=False, action='store_true') parser.add_argument('--finetune_early_layers', default=False, action='store_true') parser.add_argument('--scrl_pretrained_path', default='', type=str) # * Transformer parser.add_argument('--enc_layers', default=6, type=int, help="Number of encoding layers in the transformer") parser.add_argument('--dec_layers', default=6, type=int, help="Number of decoding layers in the transformer") parser.add_argument('--dim_feedforward', default=1024, type=int, help="Intermediate size of the feedforward layers in the transformer blocks") parser.add_argument('--hidden_dim', default=256, type=int, help="Size of the embeddings (dimension of the transformer)") parser.add_argument('--dropout', default=0.1, type=float, help="Dropout applied in the transformer") parser.add_argument('--nheads', default=8, type=int, help="Number of attention heads inside the transformer's attentions") parser.add_argument('--num_queries', default=300, type=int, help="Number of query slots") parser.add_argument('--dec_n_points', default=4, type=int) parser.add_argument('--enc_n_points', default=4, type=int) # * Efficient DETR parser.add_argument('--eff_query_init', default=False, action='store_true') parser.add_argument('--eff_specific_head', default=False, action='store_true') # * Sparse DETR parser.add_argument('--use_enc_aux_loss', default=False, action='store_true') parser.add_argument('--rho', default=0., type=float) # * Segmentation parser.add_argument('--masks', action='store_true', help="Train segmentation head if the flag is provided") # Loss parser.add_argument('--no_aux_loss', dest='aux_loss', action='store_false', help="Disables auxiliary decoding losses (loss at each layer)") # * Matcher parser.add_argument('--set_cost_class', default=2, type=float, help="Class coefficient in the matching cost") parser.add_argument('--set_cost_bbox', default=5, type=float, help="L1 box coefficient in the matching cost") parser.add_argument('--set_cost_giou', default=2, type=float, help="giou box coefficient in the matching cost") # * Loss coefficients parser.add_argument('--mask_loss_coef', default=1, type=float) parser.add_argument('--dice_loss_coef', default=1, type=float) parser.add_argument('--cls_loss_coef', default=2, type=float) parser.add_argument('--bbox_loss_coef', default=5, type=float) parser.add_argument('--giou_loss_coef', default=2, type=float) parser.add_argument('--mask_prediction_coef', default=1, type=float) parser.add_argument('--focal_alpha', default=0.25, type=float) # * dataset parameters parser.add_argument('--dataset_file', default='coco') parser.add_argument('--coco_path', default='./data/coco', type=str) parser.add_argument('--coco_panoptic_path', type=str) parser.add_argument('--remove_difficult', action='store_true') parser.add_argument('--output_dir', default='', help='path where to save, empty for no saving') parser.add_argument('--device', default='cuda', help='device to use for training / testing') parser.add_argument('--seed', default=42, type=int) parser.add_argument('--resume', default='', help='resume from checkpoint') parser.add_argument('--start_epoch', default=0, type=int, metavar='N', help='start epoch') parser.add_argument('--eval', action='store_true') parser.add_argument('--num_workers', default=2, type=int) parser.add_argument('--cache_mode', default=False, action='store_true', help='whether to cache images on memory') # * benchmark parser.add_argument('--approx_benchmark_only', default=False, action='store_true') parser.add_argument('--benchmark_only', default=False, action='store_true') parser.add_argument('--no_benchmark', dest='benchmark', action='store_false') return parser def main(args): utils.init_distributed_mode(args) print("git:\n {}\n".format(utils.get_sha())) if args.frozen_weights is not None: assert args.masks, "Frozen training is meant for segmentation only" print(args) device = torch.device(args.device) # fix the seed for reproducibility seed = args.seed + utils.get_rank() torch.manual_seed(seed) np.random.seed(seed) random.seed(seed) model, criterion, postprocessors = build_model(args) model.to(device) model_without_ddp = model dataset_val_org = build_dataset(image_set='val', args=args) if args.approx_benchmark_only or args.benchmark_only: assert not args.distributed and args.benchmark if utils.is_main_process() and args.benchmark: n_params = sum(p.numel() for p in model.parameters() if p.requires_grad) if args.benchmark_only: gflops = compute_gflops(model, dataset_val_org, approximated=False) else: gflops = compute_gflops(model, dataset_val_org, approximated=True) fps = compute_fps(model, dataset_val_org, num_iters=20, batch_size=1) bfps = compute_fps(model, dataset_val_org, num_iters=20, batch_size=4) tab_keys = ["#Params(M)", "GFLOPs", "FPS", "B4FPS"] tab_vals = [n_params / 10 ** 6, gflops, fps, bfps] table = tabulate([tab_vals], headers=tab_keys, tablefmt="pipe", floatfmt=".3f", stralign="center", numalign="center") print("===== Benchmark (Crude Approx.) =====\n" + table) if args.approx_benchmark_only or args.benchmark_only: import sys; sys.exit() if args.distributed: # wait for benchmark in the main process torch.distributed.barrier() dataset_train = build_dataset(image_set='train', args=args) dataset_val = dataset_val_org if args.distributed: if args.cache_mode: sampler_train = samplers.NodeDistributedSampler(dataset_train) sampler_val = samplers.NodeDistributedSampler(dataset_val, shuffle=False) else: sampler_train = samplers.DistributedSampler(dataset_train) sampler_val = samplers.DistributedSampler(dataset_val, shuffle=False) else: sampler_train = torch.utils.data.RandomSampler(dataset_train) sampler_val = torch.utils.data.SequentialSampler(dataset_val) batch_sampler_train = torch.utils.data.BatchSampler( sampler_train, args.batch_size, drop_last=True) data_loader_train = DataLoader(dataset_train, batch_sampler=batch_sampler_train, collate_fn=utils.collate_fn, num_workers=args.num_workers, pin_memory=True) data_loader_val = DataLoader(dataset_val, args.batch_size, sampler=sampler_val, drop_last=False, collate_fn=utils.collate_fn, num_workers=args.num_workers, pin_memory=True) args = utils.scale_learning_rate(args) def match_name_keywords(n, name_keywords): out = False for b in name_keywords: if b in n: out = True break return out param_dicts = [ { "params": [p for n, p in model_without_ddp.named_parameters() if (not match_name_keywords(n, args.lr_backbone_names) and not match_name_keywords(n, args.lr_linear_proj_names) and p.requires_grad)], "lr": args.lr, }, { "params": [p for n, p in model_without_ddp.named_parameters() if (match_name_keywords(n, args.lr_backbone_names) and not match_name_keywords(n, args.lr_linear_proj_names) and p.requires_grad)], "lr": args.lr_backbone, }, { "params": [p for n, p in model_without_ddp.named_parameters() if match_name_keywords(n, args.lr_linear_proj_names) and p.requires_grad], "lr": args.lr * args.lr_linear_proj_mult, } ] if args.sgd: optimizer = torch.optim.SGD(param_dicts, lr=args.lr, momentum=0.9, weight_decay=args.weight_decay) else: optimizer = torch.optim.AdamW(param_dicts, lr=args.lr, weight_decay=args.weight_decay) lr_scheduler = torch.optim.lr_scheduler.StepLR(optimizer, args.lr_drop) if args.distributed: model = torch.nn.parallel.DistributedDataParallel(model, device_ids=[args.gpu], find_unused_parameters=True) model_without_ddp = model.module if args.dataset_file == "coco_panoptic": # We also evaluate AP during panoptic training, on original coco DS coco_val = datasets.coco.build("val", args) base_ds = get_coco_api_from_dataset(coco_val) else: base_ds = get_coco_api_from_dataset(dataset_val) if args.frozen_weights is not None: checkpoint = torch.load(args.frozen_weights, map_location='cpu') model_without_ddp.detr.load_state_dict(checkpoint['model']) output_dir = Path(args.output_dir) if args.resume: if args.resume.startswith('https'): checkpoint = torch.hub.load_state_dict_from_url( args.resume, map_location='cpu', check_hash=True) else: checkpoint = torch.load(args.resume, map_location='cpu') missing_keys, unexpected_keys = model_without_ddp.load_state_dict(checkpoint['model'], strict=False) unexpected_keys = [k for k in unexpected_keys if not (k.endswith('total_params') or k.endswith('total_ops'))] if len(missing_keys) > 0: print('Missing Keys: {}'.format(missing_keys)) if len(unexpected_keys) > 0: print('Unexpected Keys: {}'.format(unexpected_keys)) if not args.eval and 'optimizer' in checkpoint and 'lr_scheduler' in checkpoint and 'epoch' in checkpoint: import copy p_groups = copy.deepcopy(optimizer.param_groups) optimizer.load_state_dict(checkpoint['optimizer']) for pg, pg_old in zip(optimizer.param_groups, p_groups): pg['lr'] = pg_old['lr'] pg['initial_lr'] = pg_old['initial_lr'] print(optimizer.param_groups) lr_scheduler.load_state_dict(checkpoint['lr_scheduler']) # todo: this is a hack for doing experiment that resume from checkpoint # and also modify lr scheduler (e.g., decrease lr in advance). args.override_resumed_lr_drop = True if args.override_resumed_lr_drop: print('Warning: (hack) args.override_resumed_lr_drop is set to True, ' 'so args.lr_drop would override lr_drop in resumed lr_scheduler.') lr_scheduler.step_size = args.lr_drop lr_scheduler.base_lrs = list(map(lambda group: group['initial_lr'], optimizer.param_groups)) lr_scheduler.step(lr_scheduler.last_epoch) args.start_epoch = checkpoint['epoch'] + 1 # check the resumed model if not args.eval: test_stats, coco_evaluator = evaluate( model, criterion, postprocessors, data_loader_val, base_ds, device, args ) if args.eval: print("Start evaluation") start_time = time.time() test_stats, coco_evaluator = evaluate(model, criterion, postprocessors, data_loader_val, base_ds, device, args) if args.output_dir: utils.save_on_master(coco_evaluator.coco_eval["bbox"].eval, output_dir / "eval.pth") print_final_result_on_master(model, dataset_val_org, args, test_stats, start_time) return if utils.is_main_process(): writer = SummaryWriter(output_dir) else: writer = None total_iter = 0 print("Start training") start_time = time.time() for epoch in range(args.start_epoch, args.epochs): if args.distributed: sampler_train.set_epoch(epoch) train_stats, total_iter = train_one_epoch( model, criterion, data_loader_train, optimizer, device, epoch, args.clip_max_norm, writer, total_iter) lr_scheduler.step() if args.output_dir: checkpoint_paths = [output_dir / 'checkpoint.pth'] # extra checkpoint before LR drop and every 5 epochs if (epoch + 1) % args.lr_drop == 0 or (epoch + 1) % 5 == 0: checkpoint_paths.append(output_dir / f'checkpoint{epoch:04}.pth') for checkpoint_path in checkpoint_paths: utils.save_on_master({ 'model': model_without_ddp.state_dict(), 'optimizer': optimizer.state_dict(), 'lr_scheduler': lr_scheduler.state_dict(), 'epoch': epoch, 'args': args, }, checkpoint_path) test_stats, coco_evaluator = evaluate( model, criterion, postprocessors, data_loader_val, base_ds, device, args ) # write test status if utils.is_main_process(): writer.add_scalar('test/AP', test_stats['coco_eval_bbox'][0], epoch) writer.add_scalar('test/AP50', test_stats['coco_eval_bbox'][1], epoch) writer.add_scalar('test/AP75', test_stats['coco_eval_bbox'][2], epoch) writer.add_scalar('test/APs', test_stats['coco_eval_bbox'][3], epoch) writer.add_scalar('test/APm', test_stats['coco_eval_bbox'][4], epoch) writer.add_scalar('test/APl', test_stats['coco_eval_bbox'][5], epoch) writer.add_scalar('test/class_error', test_stats['class_error'], epoch) writer.add_scalar('test/loss', test_stats['loss'], epoch) writer.add_scalar('test/loss_ce', test_stats['loss_ce'], epoch) writer.add_scalar('test/loss_bbox', test_stats['loss_bbox'], epoch) writer.add_scalar('test/loss_giou', test_stats['loss_giou'], epoch) for key, value in test_stats.items(): if "corr" in key: writer.add_scalar('test/'+key, value, epoch) log_stats = {**{f'train_{k}': v for k, v in train_stats.items()}, **{f'test_{k}': v for k, v in test_stats.items()}, 'epoch': epoch} if args.output_dir and utils.is_main_process(): if args.benchmark: log_stats.update({'params': n_params, 'gflops': gflops, 'fps': fps, 'bfps': bfps}) with (output_dir / "log.txt").open("a") as f: f.write(json.dumps(log_stats) + "\n") # for evaluation logs if coco_evaluator is not None: (output_dir / 'eval').mkdir(exist_ok=True) if "bbox" in coco_evaluator.coco_eval: filenames = ['latest.pth'] if epoch % 50 == 0: filenames.append(f'{epoch:03}.pth') for name in filenames: torch.save(coco_evaluator.coco_eval["bbox"].eval, output_dir / "eval" / name) print_final_result_on_master(model, dataset_val_org, args, test_stats, start_time) def print_final_result_on_master(model, dataset_val, args, test_stats, start_time=None): if not utils.is_main_process(): return False # training wallclock-time / gpus-hours num_gpus = args.world_size if args.distributed else 1 if start_time is not None: total_time = time.time() - start_time gpu_hours = total_time / 3600 * num_gpus gpu_hours_per_epoch = gpu_hours / args.epochs total_time_str = str(datetime.timedelta(seconds=int(total_time))) else: total_time_str, gpu_hours, gpu_hours_per_epoch = ["N/A"] * 3 # make result table now = datetime.datetime.now().strftime("%h%d %H:%M") tab_keys = ["Time", "output_dir", "epochs", "bsz", "#GPUs"] tab_vals = [now, Path(args.output_dir), args.epochs, int(args.batch_size * num_gpus), num_gpus] tab_keys += ["AP", "AP50", "AP75", "APs", "APm", "APl"] tab_vals += [v * 100 for v in test_stats['coco_eval_bbox'][:6]] tab_keys += ["E/T", "GPU*hrs", "GPU*hrs/ep"] tab_vals += [total_time_str, gpu_hours, gpu_hours_per_epoch] # add benchmark if args.benchmark: gflops = compute_gflops(model, dataset_val, approximated=False) fps = compute_fps(model, dataset_val, num_iters=300, batch_size=1) bfps = compute_fps(model, dataset_val, num_iters=300, batch_size=4) tab_keys += ['GFLOPs', 'FPS', 'B4FPS'] tab_vals += [gflops, fps, bfps] table = tabulate([tab_vals], headers=tab_keys, tablefmt="pipe", floatfmt=".3f", stralign="center", numalign="center") # dump to the file with open("log_result.txt", "a") as f: f.write("\n" + table + "\n") print(f"Save the final result to ./log_result.txt\n{table}") if __name__ == '__main__': parser = argparse.ArgumentParser('Sparse DETR training and evaluation script', parents=[get_args_parser()]) args = parser.parse_args() if args.output_dir: Path(args.output_dir).mkdir(parents=True, exist_ok=True) main(args) ================================================ FILE: models/__init__.py ================================================ # ------------------------------------------------------------------------------------ # Sparse DETR # Copyright (c) 2021 KakaoBrain. All Rights Reserved. # Licensed under the Apache License, Version 2.0 [see LICENSE for details] # ------------------------------------------------------------------------------------ # Modified from Deformable DETR (https://github.com/fundamentalvision/Deformable-DETR) # Copyright (c) 2020 SenseTime. All Rights Reserved. # ------------------------------------------------------------------------------------ # Modified from DETR (https://github.com/facebookresearch/detr) # Copyright (c) Facebook, Inc. and its affiliates. All Rights Reserved # ------------------------------------------------------------------------------------ from .deformable_detr import build def build_model(args): return build(args) ================================================ FILE: models/backbone.py ================================================ # ------------------------------------------------------------------------------------ # Sparse DETR # Copyright (c) 2021 KakaoBrain. All Rights Reserved. # Licensed under the Apache License, Version 2.0 [see LICENSE for details] # ------------------------------------------------------------------------------------ # Modified from Deformable DETR (https://github.com/fundamentalvision/Deformable-DETR) # Copyright (c) 2020 SenseTime. All Rights Reserved. # ------------------------------------------------------------------------------------ # Modified from DETR (https://github.com/facebookresearch/detr) # Copyright (c) Facebook, Inc. and its affiliates. All Rights Reserved # ------------------------------------------------------------------------------------ """ Backbone modules. """ from collections import OrderedDict import torch import torch.nn.functional as F import torchvision from torch import nn from torchvision.models._utils import IntermediateLayerGetter from typing import Dict, List from models import swin_transformer from util.misc import NestedTensor, is_main_process from .position_encoding import build_position_encoding class FrozenBatchNorm2d(torch.nn.Module): """ BatchNorm2d where the batch statistics and the affine parameters are fixed. Copy-paste from torchvision.misc.ops with added eps before rsqrt, without which any other models than torchvision.models.resnet[18,34,50,101] produce nans. """ def __init__(self, n, eps=1e-5): super(FrozenBatchNorm2d, self).__init__() self.register_buffer("weight", torch.ones(n)) self.register_buffer("bias", torch.zeros(n)) self.register_buffer("running_mean", torch.zeros(n)) self.register_buffer("running_var", torch.ones(n)) self.eps = eps def _load_from_state_dict(self, state_dict, prefix, local_metadata, strict, missing_keys, unexpected_keys, error_msgs): num_batches_tracked_key = prefix + 'num_batches_tracked' if num_batches_tracked_key in state_dict: del state_dict[num_batches_tracked_key] super(FrozenBatchNorm2d, self)._load_from_state_dict( state_dict, prefix, local_metadata, strict, missing_keys, unexpected_keys, error_msgs) def forward(self, x): # move reshapes to the beginning # to make it fuser-friendly w = self.weight.reshape(1, -1, 1, 1) b = self.bias.reshape(1, -1, 1, 1) rv = self.running_var.reshape(1, -1, 1, 1) rm = self.running_mean.reshape(1, -1, 1, 1) eps = self.eps scale = w * (rv + eps).rsqrt() bias = b - rm * scale return x * scale + bias class BackboneBase(nn.Module): def __init__(self, backbone: nn.Module, train_backbone: bool, return_interm_layers: bool, args): # TODO: args -> duplicated args super().__init__() if 'none' in args.backbone: self.strides = [1] # not used, actually (length only matters) self.num_channels = [3] return_layers = self.get_return_layers('identity', (0,)) self.body = IntermediateLayerGetter(backbone, return_layers=return_layers) elif 'resnet' in args.backbone: if not args.backbone_from_scratch and not args.finetune_early_layers: print("Freeze early layers.") for name, parameter in backbone.named_parameters(): if not train_backbone or all([k not in name for k in ['layer2', 'layer3', 'layer4']]): parameter.requires_grad_(False) else: print('Finetune early layers as well.') layer_name = "layer" if return_interm_layers: return_layers = self.get_return_layers(layer_name, (2, 3, 4)) self.strides = [8, 16, 32] self.num_channels = [512, 1024, 2048] else: return_layers = self.get_return_layers(layer_name, (4,)) self.strides = [32] self.num_channels = [2048] self.body = IntermediateLayerGetter(backbone, return_layers=return_layers) elif 'swin' in args.backbone: if return_interm_layers: num_channels = [int(backbone.embed_dim * 2 ** i) for i in range(backbone.num_layers)] return_layers = [2, 3, 4] self.strides = [8, 16, 32] self.num_channels = num_channels[1:] else: return_layers = [4] self.strides = [32] self.num_channels = num_channels[-1] self.body = backbone else: raise ValueError(f"Unknown backbone name: {args.backbone}") @staticmethod def get_return_layers(name: str, layer_ids): return {name + str(n): str(i) for i, n in enumerate(layer_ids)} def forward(self, tensor_list: NestedTensor): xs = self.body(tensor_list.tensors) out: Dict[str, NestedTensor] = {} for name, x in xs.items(): m = tensor_list.mask assert m is not None mask = F.interpolate(m[None].float(), size=x.shape[-2:]).to(torch.bool)[0] out[name] = NestedTensor(x, mask) return out class DummyBackbone(torch.nn.Module): def __init__(self): super().__init__() self.identity0 = torch.nn.Identity() class Backbone(BackboneBase): """ResNet backbone with frozen BatchNorm.""" def __init__(self, name: str, train_backbone: bool, return_interm_layers: bool, dilation: bool, args): print(f"Backbone: {name}") pretrained = is_main_process() and not args.backbone_from_scratch and not args.scrl_pretrained_path if not pretrained: print("Train backbone from scratch.") else: print("Load pretrained weights") if "none" in name: backbone = DummyBackbone() elif "resnet" in name: assert name not in ("resnet18", "resnet34"), "number of channels are hard coded" backbone = getattr(torchvision.models, name)( replace_stride_with_dilation=[False, False, dilation], pretrained=pretrained, norm_layer=FrozenBatchNorm2d) elif "swin" in name: assert not dilation, "not supported" if not args.backbone_from_scratch and not args.finetune_early_layers: print("Freeze early layers.") frozen_stages = 2 else: print('Finetune early layers as well.') frozen_stages = -1 if return_interm_layers: out_indices = [1, 2, 3] else: out_indices = [3] backbone = swin_transformer.build_model( name, out_indices=out_indices, frozen_stages=frozen_stages, pretrained=pretrained) else: raise ValueError(f"Unknown backbone name: {args.backbone}") if args.scrl_pretrained_path: assert "resnet" in name, "Currently only resnet50 is available." ckpt = torch.load(args.scrl_pretrained_path, map_location="cpu") translate_map = { "encoder.0" : "conv1", "encoder.1" : "bn1", "encoder.4" : "layer1", "encoder.5" : "layer2", "encoder.6" : "layer3", "encoder.7" : "layer4", } state_dict = { translate_map[k[:9]] + k[9:] : v for k, v in ckpt["online_network_state_dict"].items() if "encoder" in k } backbone.load_state_dict(state_dict, strict=False) super().__init__(backbone, train_backbone, return_interm_layers, args) if dilation and "resnet" in name: self.strides[-1] = self.strides[-1] // 2 class Joiner(nn.Sequential): def __init__(self, backbone, position_embedding): super().__init__(backbone, position_embedding) self.strides = backbone.strides self.num_channels = backbone.num_channels def forward(self, tensor_list: NestedTensor): xs = self[0](tensor_list) out: List[NestedTensor] = [] pos = [] for name, x in sorted(xs.items()): out.append(x) # position encoding for x in out: pos.append(self[1](x).to(x.tensors.dtype)) return out, pos def test_backbone(backbone): imgs = [ torch.randn(2, 3, 633, 122), torch.randn(2, 3, 322, 532), torch.randn(2, 3, 236, 42), ] return [backbone(img).shape for img in imgs] def build_backbone(args): # test_backbone(torchvision.models.resnet50()) position_embedding = build_position_encoding(args) train_backbone = args.lr_backbone > 0 return_interm_layers = args.masks or (args.num_feature_levels > 1) backbone = Backbone(args.backbone, train_backbone, return_interm_layers, args.dilation, args) model = Joiner(backbone, position_embedding) return model ================================================ FILE: models/deformable_detr.py ================================================ # ------------------------------------------------------------------------------------ # Sparse DETR # Copyright (c) 2021 KakaoBrain. All Rights Reserved. # Licensed under the Apache License, Version 2.0 [see LICENSE for details] # ------------------------------------------------------------------------------------ # Modified from Deformable DETR (https://github.com/fundamentalvision/Deformable-DETR) # Copyright (c) 2020 SenseTime. All Rights Reserved. # ------------------------------------------------------------------------------------ # Modified from DETR (https://github.com/facebookresearch/detr) # Copyright (c) Facebook, Inc. and its affiliates. All Rights Reserved # ------------------------------------------------------------------------------------ """ Deformable DETR model and criterion classes. """ import torch import torch.nn.functional as F from torch import nn import math from util import box_ops from util.misc import (NestedTensor, nested_tensor_from_tensor_list, accuracy, get_world_size, interpolate, is_dist_avail_and_initialized, inverse_sigmoid) from util.dam import idx_to_flat_grid, attn_map_to_flat_grid, compute_corr from .backbone import build_backbone from .matcher import build_matcher from .segmentation import (DETRsegm, PostProcessPanoptic, PostProcessSegm, dice_loss, sigmoid_focal_loss) from .deformable_transformer import build_deforamble_transformer import copy def _get_clones(module, N): return nn.ModuleList([copy.deepcopy(module) for i in range(N)]) class DeformableDETR(nn.Module): """ This is the Deformable DETR module that performs object detection """ def __init__(self, backbone, transformer, num_classes, num_queries, num_feature_levels, aux_loss=True, with_box_refine=False, two_stage=False, args=None): """ Initializes the model. Parameters: backbone: torch module of the backbone to be used. See backbone.py transformer: torch module of the transformer architecture. See transformer.py num_classes: number of object classes num_queries: number of object queries, ie detection slot. This is the maximal number of objects DETR can detect in a single image. For COCO, we recommend 100 queries. aux_loss: True if auxiliary decoding losses (loss at each decoder layer) are to be used. with_box_refine: iterative bounding box refinement two_stage: two-stage Deformable DETR """ super().__init__() self.num_queries = num_queries self.transformer = transformer hidden_dim = transformer.d_model self.class_embed = nn.Linear(hidden_dim, num_classes) self.bbox_embed = MLP(hidden_dim, hidden_dim, output_dim=4, num_layers=3) self.num_feature_levels = num_feature_levels if not two_stage: self.query_embed = nn.Embedding(num_queries, hidden_dim * 2) # will be splited into query_embed(query_pos) & tgt later if num_feature_levels > 1: num_backbone_outs = len(backbone.strides) input_proj_list = [] for _ in range(num_backbone_outs): in_channels = backbone.num_channels[_] input_proj_list.append(nn.Sequential( nn.Conv2d(in_channels, hidden_dim, kernel_size=1), nn.GroupNorm(32, hidden_dim), )) for _ in range(num_feature_levels - num_backbone_outs): input_proj_list.append(nn.Sequential( nn.Conv2d(in_channels, hidden_dim, kernel_size=3, stride=2, padding=1), nn.GroupNorm(32, hidden_dim), )) in_channels = hidden_dim self.input_proj = nn.ModuleList(input_proj_list) else: self.input_proj = nn.ModuleList([ nn.Sequential( nn.Conv2d(backbone.num_channels[0], hidden_dim, kernel_size=1), nn.GroupNorm(32, hidden_dim), )]) self.backbone = backbone self.aux_loss = aux_loss self.with_box_refine = with_box_refine self.two_stage = two_stage self.use_enc_aux_loss = args.use_enc_aux_loss self.rho = args.rho prior_prob = 0.01 bias_value = -math.log((1 - prior_prob) / prior_prob) self.class_embed.bias.data = torch.ones(num_classes) * bias_value nn.init.constant_(self.bbox_embed.layers[-1].weight.data, 0) nn.init.constant_(self.bbox_embed.layers[-1].bias.data, 0) for proj in self.input_proj: nn.init.xavier_uniform_(proj[0].weight, gain=1) nn.init.constant_(proj[0].bias, 0) # hack implementation: a list of embedding heads (see the order) # n: dec_layers / m: enc_layers # [dec_0, dec_1, ..., dec_n-1, encoder, backbone, enc_0, enc_1, ..., enc_m-2] # at each layer of decoder (by default) num_pred = transformer.decoder.num_layers if self.two_stage: # at the end of encoder num_pred += 1 if self.use_enc_aux_loss: # at each layer of encoder (excl. the last) num_pred += transformer.encoder.num_layers - 1 if with_box_refine or self.use_enc_aux_loss: # individual heads with the same initialization self.class_embed = _get_clones(self.class_embed, num_pred) self.bbox_embed = _get_clones(self.bbox_embed, num_pred) nn.init.constant_(self.bbox_embed[0].layers[-1].bias.data[2:], -2.0) else: # shared heads nn.init.constant_(self.bbox_embed.layers[-1].bias.data[2:], -2.0) self.class_embed = nn.ModuleList([self.class_embed for _ in range(num_pred)]) self.bbox_embed = nn.ModuleList([self.bbox_embed for _ in range(num_pred)]) if two_stage: # hack implementation self.transformer.decoder.class_embed = self.class_embed self.transformer.decoder.bbox_embed = self.bbox_embed for box_embed in self.transformer.decoder.bbox_embed: nn.init.constant_(box_embed.layers[-1].bias.data[2:], 0.0) if self.use_enc_aux_loss: # the output from the last layer should be specially treated as an input of decoder num_layers_excluding_the_last = transformer.encoder.num_layers - 1 self.transformer.encoder.aux_heads = True self.transformer.encoder.class_embed = self.class_embed[-num_layers_excluding_the_last:] self.transformer.encoder.bbox_embed = self.bbox_embed[-num_layers_excluding_the_last:] for box_embed in self.transformer.encoder.bbox_embed: nn.init.constant_(box_embed.layers[-1].bias.data[2:], 0.0) def forward(self, samples: NestedTensor): """ The forward expects a NestedTensor, which consists of: - samples.tensor: batched images, of shape [batch_size x 3 x H x W] - samples.mask: a binary mask of shape [batch_size x H x W], containing 1 on padded pixels It returns a dict with the following elements: - "pred_logits": the classification logits (including no-object) for all queries. Shape= [batch_size x num_queries x (num_classes + 1)] - "pred_boxes": The normalized boxes coordinates for all queries, represented as (center_x, center_y, height, width). These values are normalized in [0, 1], relative to the size of each individual image (disregarding possible padding). See PostProcess for information on how to retrieve the unnormalized bounding box. - "aux_outputs": Optional, only returned when auxilary losses are activated. It is a list of dictionnaries containing the two above keys for each decoder layer. """ ########### # Backbone if not isinstance(samples, NestedTensor): samples = nested_tensor_from_tensor_list(samples) features, pos = self.backbone(samples) srcs = [] masks = [] # multi-scale features projected from ~C5 with 1x1 conv for l, feat in enumerate(features): src, mask = feat.decompose() srcs.append(self.input_proj[l](src)) masks.append(mask) assert mask is not None # multi-scale features smaller than C5 projected with 2 strided 3x3 conv if self.num_feature_levels > len(srcs): _len_srcs = len(srcs) for l in range(_len_srcs, self.num_feature_levels): if l == _len_srcs: # feature scale 1/32 src = self.input_proj[l](features[-1].tensors) else: # feature scale <1/64: recursively downsample the last projection src = self.input_proj[l](srcs[-1]) m = samples.mask mask = F.interpolate(m[None].float(), size=src.shape[-2:]).to(torch.bool)[0] pos_l = self.backbone[1](NestedTensor(src, mask)).to(src.dtype) srcs.append(src) masks.append(mask) pos.append(pos_l) ########### # Transformer encoder & decoder query_embeds = None if not self.two_stage: query_embeds = self.query_embed.weight (hs, init_reference, inter_references, enc_outputs_class, enc_outputs_coord_unact, backbone_mask_prediction, enc_inter_outputs_class, enc_inter_outputs_coord, sampling_locations_enc, attn_weights_enc, sampling_locations_dec, attn_weights_dec, backbone_topk_proposals, spatial_shapes, level_start_index) = \ self.transformer(srcs, masks, pos, query_embeds) ########### # Detection heads outputs_classes = [] outputs_coords = [] for lvl in range(len(hs)): # lvl: level of decoding layer outputs_class = self.class_embed[lvl](hs[lvl]) outputs_coord = self.bbox_embed[lvl](hs[lvl]) assert init_reference is not None and inter_references is not None if lvl == 0: reference = init_reference else: reference = inter_references[lvl - 1] reference = inverse_sigmoid(reference) if reference.shape[-1] == 4: outputs_coord += reference else: assert reference.shape[-1] == 2 outputs_coord[..., :2] += reference outputs_coord = outputs_coord.sigmoid() outputs_classes.append(outputs_class) outputs_coords.append(outputs_coord) outputs_class = torch.stack(outputs_classes) outputs_coord = torch.stack(outputs_coords) # the topmost layer output out = { "pred_logits": outputs_class[-1], "pred_boxes": outputs_coord[-1], "sampling_locations_enc": sampling_locations_enc, "attn_weights_enc": attn_weights_enc, "sampling_locations_dec": sampling_locations_dec, "attn_weights_dec": attn_weights_dec, "spatial_shapes": spatial_shapes, "level_start_index": level_start_index, } if backbone_topk_proposals is not None: out["backbone_topk_proposals"] = backbone_topk_proposals if self.aux_loss: # make loss from every intermediate layers (excluding the last one) out['aux_outputs'] = self._set_aux_loss(outputs_class[:-1], outputs_coord[:-1]) if self.two_stage: enc_outputs_coord = enc_outputs_coord_unact.sigmoid() out['enc_outputs'] = {'pred_logits': enc_outputs_class, 'pred_boxes': enc_outputs_coord} if self.rho: out["backbone_mask_prediction"] = backbone_mask_prediction if self.use_enc_aux_loss: out['aux_outputs_enc'] = self._set_aux_loss(enc_inter_outputs_class, enc_inter_outputs_coord) if self.rho: out["sparse_token_nums"] = self.transformer.sparse_token_nums out['mask_flatten'] = torch.cat([m.flatten(1) for m in masks], 1) return out @torch.jit.unused def _set_aux_loss(self, outputs_class, outputs_coord): # this is a workaround to make torchscript happy, as torchscript # doesn't support dictionary with non-homogeneous values, such # as a dict having both a Tensor and a list. return [{'pred_logits': a, 'pred_boxes': b} for a, b in zip(outputs_class, outputs_coord)] class SetCriterion(nn.Module): """ This class computes the loss for DETR. The process happens in two steps: 1) we compute hungarian assignment between ground truth boxes and the outputs of the model 2) we supervise each pair of matched ground-truth / prediction (supervise class and box) """ def __init__(self, num_classes, matcher, weight_dict, losses, args): """ Create the criterion. Parameters: num_classes: number of object categories, omitting the special no-object category matcher: module able to compute a matching between targets and proposals weight_dict: dict containing as key the names of the losses and as values their relative weight. losses: list of all the losses to be applied. See get_loss for list of available losses. focal_alpha: alpha in Focal Loss """ super().__init__() self.num_classes = num_classes self.matcher = matcher self.weight_dict = weight_dict self.losses = losses self.focal_alpha = args.focal_alpha self.eff_specific_head = args.eff_specific_head def loss_labels(self, outputs, targets, indices, num_boxes, log=True): """Classification loss (NLL) targets dicts must contain the key "labels" containing a tensor of dim [nb_target_boxes] """ assert 'pred_logits' in outputs src_logits = outputs['pred_logits'] idx = self._get_src_permutation_idx(indices) target_classes_o = torch.cat([t["labels"][J] for t, (_, J) in zip(targets, indices)]) target_classes = torch.full(src_logits.shape[:2], self.num_classes, dtype=torch.int64, device=src_logits.device) target_classes[idx] = target_classes_o target_classes_onehot = torch.zeros([src_logits.shape[0], src_logits.shape[1], src_logits.shape[2] + 1], dtype=src_logits.dtype, layout=src_logits.layout, device=src_logits.device) target_classes_onehot.scatter_(2, target_classes.unsqueeze(-1), 1) target_classes_onehot = target_classes_onehot[:,:,:-1] loss_ce = sigmoid_focal_loss(src_logits, target_classes_onehot, num_boxes, alpha=self.focal_alpha, gamma=2) loss_ce = loss_ce * src_logits.shape[1] losses = {'loss_ce': loss_ce} if log: # TODO this should probably be a separate loss, not hacked in this one here losses['class_error'] = 100 - accuracy(src_logits[idx], target_classes_o)[0] return losses @torch.no_grad() def loss_cardinality(self, outputs, targets, indices, num_boxes): """ Compute the cardinality error, ie the absolute error in the number of predicted non-empty boxes This is not really a loss, it is intended for logging purposes only. It doesn't propagate gradients """ pred_logits = outputs['pred_logits'] device = pred_logits.device tgt_lengths = torch.as_tensor([len(v["labels"]) for v in targets], device=device) # Count the number of predictions that are NOT "no-object" (which is the last class) card_pred = (pred_logits.argmax(-1) != pred_logits.shape[-1] - 1).sum(1) card_err = F.l1_loss(card_pred.float(), tgt_lengths.float()) losses = {'cardinality_error': card_err} return losses def loss_boxes(self, outputs, targets, indices, num_boxes): """Compute the losses related to the bounding boxes, the L1 regression loss and the GIoU loss targets dicts must contain the key "boxes" containing a tensor of dim [nb_target_boxes, 4] The target boxes are expected in format (center_x, center_y, h, w), normalized by the image size. """ assert 'pred_boxes' in outputs idx = self._get_src_permutation_idx(indices) src_boxes = outputs['pred_boxes'][idx] target_boxes = torch.cat([t['boxes'][i] for t, (_, i) in zip(targets, indices)], dim=0) loss_bbox = F.l1_loss(src_boxes, target_boxes, reduction='none') losses = {} losses['loss_bbox'] = loss_bbox.sum() / num_boxes loss_giou = 1 - torch.diag(box_ops.generalized_box_iou( box_ops.box_cxcywh_to_xyxy(src_boxes), box_ops.box_cxcywh_to_xyxy(target_boxes))) losses['loss_giou'] = loss_giou.sum() / num_boxes return losses def loss_masks(self, outputs, targets, indices, num_boxes): """Compute the losses related to the masks: the focal loss and the dice loss. targets dicts must contain the key "masks" containing a tensor of dim [nb_target_boxes, h, w] """ assert "pred_masks" in outputs src_idx = self._get_src_permutation_idx(indices) tgt_idx = self._get_tgt_permutation_idx(indices) src_masks = outputs["pred_masks"] # TODO use valid to mask invalid areas due to padding in loss target_masks, valid = nested_tensor_from_tensor_list([t["masks"] for t in targets]).decompose() target_masks = target_masks.to(src_masks) src_masks = src_masks[src_idx] # upsample predictions to the target size src_masks = interpolate(src_masks[:, None], size=target_masks.shape[-2:], mode="bilinear", align_corners=False) src_masks = src_masks[:, 0].flatten(1) target_masks = target_masks[tgt_idx].flatten(1) losses = { "loss_mask": sigmoid_focal_loss(src_masks, target_masks, num_boxes), "loss_dice": dice_loss(src_masks, target_masks, num_boxes), } return losses def loss_mask_prediction(self, outputs, targets, indices, num_boxes, layer=None): assert "backbone_mask_prediction" in outputs assert "sampling_locations_dec" in outputs assert "attn_weights_dec" in outputs assert "spatial_shapes" in outputs assert "level_start_index" in outputs mask_prediction = outputs["backbone_mask_prediction"] loss_key = "loss_mask_prediction" sampling_locations_dec = outputs["sampling_locations_dec"] attn_weights_dec = outputs["attn_weights_dec"] spatial_shapes = outputs["spatial_shapes"] level_start_index = outputs["level_start_index"] flat_grid_attn_map_dec = attn_map_to_flat_grid( spatial_shapes, level_start_index, sampling_locations_dec, attn_weights_dec).sum(dim=(1,2)) losses = {} if 'mask_flatten' in outputs: flat_grid_attn_map_dec = flat_grid_attn_map_dec.masked_fill( outputs['mask_flatten'], flat_grid_attn_map_dec.min()-1) sparse_token_nums = outputs["sparse_token_nums"] num_topk = sparse_token_nums.max() topk_idx_tgt = torch.topk(flat_grid_attn_map_dec, num_topk)[1] target = torch.zeros_like(mask_prediction) for i in range(target.shape[0]): target[i].scatter_(0, topk_idx_tgt[i][:sparse_token_nums[i]], 1) losses.update({loss_key: F.multilabel_soft_margin_loss(mask_prediction, target)}) return losses @torch.no_grad() def corr(self, outputs, targets, indices, num_boxes): if "backbone_topk_proposals" not in outputs.keys(): return {} assert "backbone_topk_proposals" in outputs assert "sampling_locations_dec" in outputs assert "attn_weights_dec" in outputs assert "spatial_shapes" in outputs assert "level_start_index" in outputs backbone_topk_proposals = outputs["backbone_topk_proposals"] sampling_locations_dec = outputs["sampling_locations_dec"] attn_weights_dec = outputs["attn_weights_dec"] spatial_shapes = outputs["spatial_shapes"] level_start_index = outputs["level_start_index"] flat_grid_topk = idx_to_flat_grid(spatial_shapes, backbone_topk_proposals) flat_grid_attn_map_dec = attn_map_to_flat_grid( spatial_shapes, level_start_index, sampling_locations_dec, attn_weights_dec).sum(dim=(1,2)) corr = compute_corr(flat_grid_topk, flat_grid_attn_map_dec, spatial_shapes) losses = {} losses["corr_mask_attn_map_dec_all"] = corr[0].mean() for i, _corr in enumerate(corr[1:]): losses[f"corr_mask_attn_map_dec_{i}"] = _corr.mean() return losses def _get_src_permutation_idx(self, indices): # permute predictions following indices batch_idx = torch.cat([torch.full_like(src, i) for i, (src, _) in enumerate(indices)]) src_idx = torch.cat([src for (src, _) in indices]) return batch_idx, src_idx def _get_tgt_permutation_idx(self, indices): # permute targets following indices batch_idx = torch.cat([torch.full_like(tgt, i) for i, (_, tgt) in enumerate(indices)]) tgt_idx = torch.cat([tgt for (_, tgt) in indices]) return batch_idx, tgt_idx def get_loss(self, loss, outputs, targets, indices, num_boxes, **kwargs): loss_map = { 'labels': self.loss_labels, 'cardinality': self.loss_cardinality, 'boxes': self.loss_boxes, 'masks': self.loss_masks, "mask_prediction": self.loss_mask_prediction, "corr": self.corr, } assert loss in loss_map, f'do you really want to compute {loss} loss?' return loss_map[loss](outputs, targets, indices, num_boxes, **kwargs) def forward(self, outputs, targets): """ This performs the loss computation. Parameters: outputs: dict of tensors, see the output specification of the model for the format targets: list of dicts, such that len(targets) == batch_size. The expected keys in each dict depends on the losses applied, see each loss' doc """ outputs_without_aux = {k: v for k, v in outputs.items() if k not in ['aux_outputs', 'enc_outputs', 'backbone_outputs', 'mask_flatten']} # Retrieve the matching between the outputs of the last layer and the targets indices = self.matcher(outputs_without_aux, targets) # Compute the average number of target boxes accross all nodes, for normalization purposes num_boxes = sum(len(t["labels"]) for t in targets) num_boxes = torch.as_tensor([num_boxes], dtype=torch.float, device=next(iter(outputs.values())).device) if is_dist_avail_and_initialized(): torch.distributed.all_reduce(num_boxes) num_boxes = torch.clamp(num_boxes / get_world_size(), min=1).item() # Compute all the requested losses losses = {} for loss in self.losses: kwargs = {} losses.update(self.get_loss(loss, outputs, targets, indices, num_boxes, **kwargs)) # In case of auxiliary losses, we repeat this process with the output of each intermediate layer. if 'aux_outputs' in outputs: for i, aux_outputs in enumerate(outputs['aux_outputs']): indices = self.matcher(aux_outputs, targets) for loss in self.losses: if loss in ['masks', "mask_prediction", "corr"]: # Intermediate masks losses are too costly to compute, we ignore them. continue kwargs = {} if loss == 'labels': # Logging is enabled only for the last layer kwargs['log'] = False l_dict = self.get_loss(loss, aux_outputs, targets, indices, num_boxes, **kwargs) l_dict = {k + f'_{i}': v for k, v in l_dict.items()} losses.update(l_dict) if 'enc_outputs' in outputs: enc_outputs = outputs['enc_outputs'] bin_targets = copy.deepcopy(targets) if not self.eff_specific_head: for bt in bin_targets: bt['labels'] = torch.zeros_like(bt['labels']) # all labels are zero (meaning foreground) indices = self.matcher(enc_outputs, bin_targets) for loss in self.losses: if loss in ['masks', "mask_prediction", "corr"]: # Intermediate masks losses are too costly to compute, we ignore them. continue kwargs = {} if loss == 'labels': # Logging is enabled only for the last layer kwargs['log'] = False l_dict = self.get_loss(loss, enc_outputs, bin_targets, indices, num_boxes, **kwargs) l_dict = {k + f'_enc': v for k, v in l_dict.items()} losses.update(l_dict) if 'backbone_outputs' in outputs: backbone_outputs = outputs['backbone_outputs'] bin_targets = copy.deepcopy(targets) if not self.eff_specific_head: for bt in bin_targets: bt['labels'] = torch.zeros_like(bt['labels']) # all labels are zero (meaning foreground) indices = self.matcher(backbone_outputs, bin_targets) for loss in self.losses: if loss in ['masks', "mask_prediction", "corr"]: # Intermediate masks losses are too costly to compute, we ignore them. continue kwargs = {} if loss == 'labels': # Logging is enabled only for the last layer kwargs['log'] = False l_dict = self.get_loss(loss, backbone_outputs, bin_targets, indices, num_boxes, **kwargs) l_dict = {k + f'_backbone': v for k, v in l_dict.items()} losses.update(l_dict) if 'aux_outputs_enc' in outputs: for i, aux_outputs in enumerate(outputs['aux_outputs_enc']): indices = self.matcher(aux_outputs, targets) for loss in self.losses: if loss in ['masks', "mask_prediction", "corr"]: # Intermediate masks losses are too costly to compute, we ignore them. continue kwargs = {} if loss == 'labels': # Logging is enabled only for the last layer kwargs['log'] = False l_dict = self.get_loss(loss, aux_outputs, targets, indices, num_boxes, **kwargs) l_dict = {k + f'_enc_{i}': v for k, v in l_dict.items()} losses.update(l_dict) return losses class PostProcess(nn.Module): """ This module converts the model's output into the format expected by the coco api""" @torch.no_grad() def forward(self, outputs, target_sizes): """ Perform the computation Parameters: outputs: raw outputs of the model target_sizes: tensor of dimension [batch_size x 2] containing the size of each images of the batch For evaluation, this must be the original image size (before any data augmentation) For visualization, this should be the image size after data augment, but before padding """ out_logits, out_bbox = outputs['pred_logits'], outputs['pred_boxes'] assert len(out_logits) == len(target_sizes) assert target_sizes.shape[1] == 2 prob = out_logits.sigmoid() topk_values, topk_indexes = torch.topk(prob.view(out_logits.shape[0], -1), 100, dim=1) scores = topk_values topk_boxes = topk_indexes // out_logits.shape[2] labels = topk_indexes % out_logits.shape[2] boxes = box_ops.box_cxcywh_to_xyxy(out_bbox) boxes = torch.gather(boxes, 1, topk_boxes.unsqueeze(-1).repeat(1,1,4)) # and from relative [0, 1] to absolute [0, height] coordinates img_h, img_w = target_sizes.unbind(1) scale_fct = torch.stack([img_w, img_h, img_w, img_h], dim=1) boxes = boxes * scale_fct[:, None, :] results = [{'scores': s, 'labels': l, 'boxes': b} for s, l, b in zip(scores, labels, boxes)] return results class MLP(nn.Module): """ Very simple multi-layer perceptron (also called FFN)""" def __init__(self, input_dim, hidden_dim, output_dim, num_layers): super().__init__() self.num_layers = num_layers h = [hidden_dim] * (num_layers - 1) self.layers = nn.ModuleList(nn.Linear(n, k) for n, k in zip([input_dim] + h, h + [output_dim])) def forward(self, x): for i, layer in enumerate(self.layers): x = F.relu(layer(x)) if i < self.num_layers - 1 else layer(x) return x def build(args): num_classes = 20 if args.dataset_file != 'coco' else 91 if args.dataset_file == "coco_panoptic": num_classes = 250 device = torch.device(args.device) backbone = build_backbone(args) transformer = build_deforamble_transformer(args) model = DeformableDETR( backbone, transformer, num_classes=num_classes, num_queries=args.num_queries, num_feature_levels=args.num_feature_levels, aux_loss=args.aux_loss, with_box_refine=args.with_box_refine, two_stage=args.two_stage, args=args, ) if args.masks: model = DETRsegm(model, freeze_detr=(args.frozen_weights is not None)) matcher = build_matcher(args) weight_dict = {'loss_ce': args.cls_loss_coef, 'loss_bbox': args.bbox_loss_coef} weight_dict['loss_giou'] = args.giou_loss_coef if args.masks: weight_dict["loss_mask"] = args.mask_loss_coef weight_dict["loss_dice"] = args.dice_loss_coef # TODO this is a hack aux_weight_dict = {} if args.aux_loss: for i in range(args.dec_layers - 1): aux_weight_dict.update({k + f'_{i}': v for k, v in weight_dict.items()}) if args.two_stage: aux_weight_dict.update({k + f'_enc': v for k, v in weight_dict.items()}) if args.use_enc_aux_loss: for i in range(args.enc_layers - 1): aux_weight_dict.update({k + f'_enc_{i}': v for k, v in weight_dict.items()}) if args.rho: aux_weight_dict.update({k + f'_backbone': v for k, v in weight_dict.items()}) if aux_weight_dict: weight_dict.update(aux_weight_dict) weight_dict['loss_mask_prediction'] = args.mask_prediction_coef losses = ['labels', 'boxes', 'cardinality', "corr"] if args.masks: losses += ["masks"] if args.rho: losses += ["mask_prediction"] # num_classes, matcher, weight_dict, losses, focal_alpha=0.25 criterion = SetCriterion(num_classes, matcher, weight_dict, losses, args) criterion.to(device) postprocessors = {'bbox': PostProcess()} if args.masks: postprocessors['segm'] = PostProcessSegm() if args.dataset_file == "coco_panoptic": is_thing_map = {i: i <= 90 for i in range(201)} postprocessors["panoptic"] = PostProcessPanoptic(is_thing_map, threshold=0.85) return model, criterion, postprocessors ================================================ FILE: models/deformable_transformer.py ================================================ # ------------------------------------------------------------------------------------ # Sparse DETR # Copyright (c) 2021 KakaoBrain. All Rights Reserved. # Licensed under the Apache License, Version 2.0 [see LICENSE for details] # ------------------------------------------------------------------------------------ # Modified from Deformable DETR (https://github.com/fundamentalvision/Deformable-DETR) # Copyright (c) 2020 SenseTime. All Rights Reserved. # ------------------------------------------------------------------------------------ # Modified from DETR (https://github.com/facebookresearch/detr) # Copyright (c) Facebook, Inc. and its affiliates. All Rights Reserved # ------------------------------------------------------------------------------------ import copy from typing import Optional, List import math import torch import torch.nn.functional as F from torch import nn, Tensor from torch.nn.init import xavier_uniform_, constant_, uniform_, normal_ from util.misc import inverse_sigmoid from models.ops.modules import MSDeformAttn class DeformableTransformer(nn.Module): def __init__(self, d_model=256, nhead=8, num_encoder_layers=6, num_decoder_layers=6, dim_feedforward=1024, dropout=0.1, activation="relu", return_intermediate_dec=False, num_feature_levels=4, dec_n_points=4, enc_n_points=4, two_stage=False, two_stage_num_proposals=300, args=None): super().__init__() self.d_model = d_model self.nhead = nhead self.two_stage = two_stage self.two_stage_num_proposals = two_stage_num_proposals self.eff_query_init = args.eff_query_init self.eff_specific_head = args.eff_specific_head # there's no need to compute reference points if above 2 conditions meet simultaneously self._log_args('eff_query_init', 'eff_specific_head') self.rho = args.rho self.use_enc_aux_loss = args.use_enc_aux_loss self.sparse_enc_head = 1 if self.two_stage and self.rho else 0 if self.rho: self.enc_mask_predictor = MaskPredictor(self.d_model, self.d_model) else: self.enc_mask_predictor = None encoder_layer = DeformableTransformerEncoderLayer(d_model, dim_feedforward, dropout, activation, num_feature_levels, nhead, enc_n_points) self.encoder = DeformableTransformerEncoder(encoder_layer, num_encoder_layers, self.d_model) decoder_layer = DeformableTransformerDecoderLayer(d_model, dim_feedforward, dropout, activation, num_feature_levels, nhead, dec_n_points) self.decoder = DeformableTransformerDecoder(decoder_layer, num_decoder_layers, return_intermediate_dec) self.level_embed = nn.Parameter(torch.Tensor(num_feature_levels, d_model)) if self.two_stage: self.enc_output = nn.Linear(d_model, d_model) self.enc_output_norm = nn.LayerNorm(d_model) if self.two_stage: self.pos_trans = nn.Linear(d_model * 2, d_model * (1 if self.eff_query_init else 2)) self.pos_trans_norm = nn.LayerNorm(d_model * (1 if self.eff_query_init else 2)) if not self.two_stage: self.reference_points = nn.Linear(d_model, 2) self._reset_parameters() def _log_args(self, *names): print('==============') print("\n".join([f"{name}: {getattr(self, name)}" for name in names])) print('==============') def _reset_parameters(self): for p in self.parameters(): if p.dim() > 1: nn.init.xavier_uniform_(p) for m in self.modules(): if isinstance(m, MSDeformAttn): m._reset_parameters() if hasattr(self, 'reference_points'): xavier_uniform_(self.reference_points.weight.data, gain=1.0) constant_(self.reference_points.bias.data, 0.) normal_(self.level_embed) def get_proposal_pos_embed(self, proposals): # proposals: N, L(top_k), 4(bbox coords.) num_pos_feats = 128 temperature = 10000 scale = 2 * math.pi dim_t = torch.arange(num_pos_feats, dtype=torch.float32, device=proposals.device) # 128 dim_t = temperature ** (2 * (dim_t // 2) / num_pos_feats) proposals = proposals.sigmoid() * scale # N, L, 4 pos = proposals[:, :, :, None] / dim_t # N, L, 4, 128 # apply sin/cos alternatively pos = torch.stack((pos[:, :, :, 0::2].sin(), pos[:, :, :, 1::2].cos()), dim=4) # N, L, 4, 64, 2 pos = pos.flatten(2) # N, L, 512 (4 x 128) return pos def gen_encoder_output_proposals(self, memory, memory_padding_mask, spatial_shapes, process_output=True): """Make region proposals for each multi-scale features considering their shapes and padding masks, and project & normalize the encoder outputs corresponding to these proposals. - center points: relative grid coordinates in the range of [0.01, 0.99] (additional mask) - width/height: 2^(layer_id) * s (s=0.05) / see the appendix A.4 Tensor shape example: Args: memory: torch.Size([2, 15060, 256]) memory_padding_mask: torch.Size([2, 15060]) spatial_shape: torch.Size([4, 2]) Returns: output_memory: torch.Size([2, 15060, 256]) - same shape with memory ( + additional mask + linear layer + layer norm ) output_proposals: torch.Size([2, 15060, 4]) - x, y, w, h """ N_, S_, C_ = memory.shape proposals = [] _cur = 0 for lvl, (H_, W_) in enumerate(spatial_shapes): # level of encoded feature scale mask_flatten_ = memory_padding_mask[:, _cur:(_cur + H_ * W_)].view(N_, H_, W_, 1) valid_H = torch.sum(~mask_flatten_[:, :, 0, 0], 1) valid_W = torch.sum(~mask_flatten_[:, 0, :, 0], 1) grid_y, grid_x = torch.meshgrid(torch.linspace(0, H_ - 1, H_, dtype=torch.float32, device=memory.device), torch.linspace(0, W_ - 1, W_, dtype=torch.float32, device=memory.device)) grid = torch.cat([grid_x.unsqueeze(-1), grid_y.unsqueeze(-1)], -1) scale = torch.cat([valid_W.unsqueeze(-1), valid_H.unsqueeze(-1)], 1).view(N_, 1, 1, 2) grid = (grid.unsqueeze(0).expand(N_, -1, -1, -1) + 0.5) / scale wh = torch.ones_like(grid) * 0.05 * (2.0 ** lvl) proposal = torch.cat((grid, wh), -1).view(N_, -1, 4) proposals.append(proposal) _cur += (H_ * W_) output_proposals = torch.cat(proposals, 1) output_proposals_valid = ((output_proposals > 0.01) & (output_proposals < 0.99)).all(-1, keepdim=True) output_proposals = torch.log(output_proposals / (1 - output_proposals)) # inverse of sigmoid output_proposals = output_proposals.masked_fill(memory_padding_mask.unsqueeze(-1), float('inf')) output_proposals = output_proposals.masked_fill(~output_proposals_valid, float('inf')) # sigmoid(inf) = 1 output_memory = memory if process_output: output_memory = output_memory.masked_fill(memory_padding_mask.unsqueeze(-1), float(0)) output_memory = output_memory.masked_fill(~output_proposals_valid, float(0)) output_memory = self.enc_output_norm(self.enc_output(output_memory)) return output_memory, output_proposals, (~memory_padding_mask).sum(axis=-1) def get_valid_ratio(self, mask): _, H, W = mask.shape valid_H = torch.sum(~mask[:, :, 0], 1) valid_W = torch.sum(~mask[:, 0, :], 1) valid_ratio_h = valid_H.float() / H valid_ratio_w = valid_W.float() / W valid_ratio = torch.stack([valid_ratio_w, valid_ratio_h], -1) return valid_ratio def forward(self, srcs, masks, pos_embeds, query_embed=None): assert self.two_stage or query_embed is not None ########### # prepare input for encoder src_flatten = [] mask_flatten = [] lvl_pos_embed_flatten = [] spatial_shapes = [] for lvl, (src, mask, pos_embed) in enumerate(zip(srcs, masks, pos_embeds)): bs, c, h, w = src.shape spatial_shape = (h, w) spatial_shapes.append(spatial_shape) src = src.flatten(2).transpose(1, 2) mask = mask.flatten(1) pos_embed = pos_embed.flatten(2).transpose(1, 2) lvl_pos_embed = pos_embed + self.level_embed[lvl].view(1, 1, -1) lvl_pos_embed_flatten.append(lvl_pos_embed) src_flatten.append(src) mask_flatten.append(mask) src_flatten = torch.cat(src_flatten, 1) mask_flatten = torch.cat(mask_flatten, 1) lvl_pos_embed_flatten = torch.cat(lvl_pos_embed_flatten, 1) spatial_shapes = torch.as_tensor(spatial_shapes, dtype=torch.long, device=src_flatten.device) level_start_index = torch.cat((spatial_shapes.new_zeros((1, )), spatial_shapes.prod(1).cumsum(0)[:-1])) # valid ratios across multi-scale features of the same image can be varied, # while they are interpolated and binarized on different resolutions. valid_ratios = torch.stack([self.get_valid_ratio(m) for m in masks], 1) ########### # prepare for sparse encoder if self.rho or self.use_enc_aux_loss: backbone_output_memory, backbone_output_proposals, valid_token_nums = self.gen_encoder_output_proposals( src_flatten+lvl_pos_embed_flatten, mask_flatten, spatial_shapes, process_output=bool(self.rho)) self.valid_token_nums = valid_token_nums if self.rho: sparse_token_nums = (valid_token_nums * self.rho).int() + 1 backbone_topk = int(max(sparse_token_nums)) self.sparse_token_nums = sparse_token_nums backbone_topk = min(backbone_topk, backbone_output_memory.shape[1]) backbone_mask_prediction = self.enc_mask_predictor(backbone_output_memory).squeeze(-1) # excluding pad area backbone_mask_prediction = backbone_mask_prediction.masked_fill(mask_flatten, backbone_mask_prediction.min()) backbone_topk_proposals = torch.topk(backbone_mask_prediction, backbone_topk, dim=1)[1] else: backbone_topk_proposals = None backbone_outputs_class = None backbone_outputs_coord_unact = None sparse_token_nums= None ########### # encoder if self.encoder: output_proposals = backbone_output_proposals if self.use_enc_aux_loss else None encoder_output = self.encoder(src_flatten, spatial_shapes, level_start_index, valid_ratios, pos=lvl_pos_embed_flatten, padding_mask=mask_flatten, topk_inds=backbone_topk_proposals, output_proposals=output_proposals, sparse_token_nums=sparse_token_nums) memory, sampling_locations_enc, attn_weights_enc = encoder_output[:3] if self.use_enc_aux_loss: enc_inter_outputs_class, enc_inter_outputs_coord_unact = encoder_output[3:5] else: memory = src_flatten + lvl_pos_embed_flatten ########### # prepare input for decoder bs, _, c = memory.shape # torch.Size([N, L, 256]) topk_proposals = None if self.two_stage: # finalize the first stage output # project & normalize the memory and make proposal bounding boxes on them output_memory, output_proposals, _ = self.gen_encoder_output_proposals(memory, mask_flatten, spatial_shapes) # hack implementation for two-stage Deformable DETR (using the last layer registered in class/bbox_embed) # 1) a linear projection for bounding box binary classification (fore/background) enc_outputs_class = self.decoder.class_embed[self.decoder.num_layers](output_memory) # 2) 3-layer FFN for bounding box regression enc_outputs_coord_offset = self.decoder.bbox_embed[self.decoder.num_layers](output_memory) enc_outputs_coord_unact = output_proposals + enc_outputs_coord_offset # appendix A.4 # top scoring bounding boxes are picked as the final region proposals. # these proposals are fed into the decoder as initial boxes for the iterative bounding box refinement. topk = self.two_stage_num_proposals # enc_outputs_class: torch.Size([N, L, 91]) if self.eff_specific_head: # take the best score for judging objectness with class specific head enc_outputs_fg_class = enc_outputs_class.topk(1, dim=2).values[... , 0] else: # take the score from the binary(fore/background) classfier # though outputs have 91 output dim, the 1st dim. alone will be used for the loss computation. enc_outputs_fg_class = enc_outputs_class[..., 0] topk_proposals = torch.topk(enc_outputs_fg_class, topk, dim=1)[1] topk_coords_unact = torch.gather(enc_outputs_coord_unact, 1, topk_proposals.unsqueeze(-1).repeat(1, 1, 4)) topk_coords_unact = topk_coords_unact.detach() reference_points = topk_coords_unact.sigmoid() init_reference_out = reference_points # pos_embed -> linear layer -> layer norm pos_trans_out = self.pos_trans_norm(self.pos_trans(self.get_proposal_pos_embed(topk_coords_unact))) if self.eff_query_init: # Efficient-DETR uses top-k memory as the initialization of `tgt` (query vectors) tgt = torch.gather(memory, 1, topk_proposals.unsqueeze(-1).repeat(1, 1, memory.size(-1))) query_embed = pos_trans_out else: query_embed, tgt = torch.split(pos_trans_out, c, dim=2) else: query_embed, tgt = torch.split(query_embed, c, dim=1) query_embed = query_embed.unsqueeze(0).expand(bs, -1, -1) tgt = tgt.unsqueeze(0).expand(bs, -1, -1) reference_points = self.reference_points(query_embed).sigmoid() init_reference_out = reference_points ########### # decoder hs, inter_references, sampling_locations_dec, attn_weights_dec = self.decoder(tgt, reference_points, src=memory, src_spatial_shapes=spatial_shapes, src_level_start_index=level_start_index, src_valid_ratios=valid_ratios, query_pos=query_embed, src_padding_mask=mask_flatten, topk_inds=topk_proposals) inter_references_out = inter_references ret = [] ret += [hs, init_reference_out, inter_references_out] ret += [enc_outputs_class, enc_outputs_coord_unact] if self.two_stage else [None] * 2 if self.rho: ret += [backbone_mask_prediction] else: ret += [None] ret += [enc_inter_outputs_class, enc_inter_outputs_coord_unact] if self.use_enc_aux_loss else [None] * 2 ret += [sampling_locations_enc, attn_weights_enc, sampling_locations_dec, attn_weights_dec] ret += [backbone_topk_proposals, spatial_shapes, level_start_index] return ret class DeformableTransformerEncoderLayer(nn.Module): def __init__(self, d_model=256, d_ffn=1024, dropout=0.1, activation="relu", n_levels=4, n_heads=8, n_points=4): super().__init__() # self attention self.self_attn = MSDeformAttn(d_model, n_levels, n_heads, n_points) self.dropout1 = nn.Dropout(dropout) self.norm1 = nn.LayerNorm(d_model) # ffn self.linear1 = nn.Linear(d_model, d_ffn) self.activation = _get_activation_fn(activation) self.dropout2 = nn.Dropout(dropout) self.linear2 = nn.Linear(d_ffn, d_model) self.dropout3 = nn.Dropout(dropout) self.norm2 = nn.LayerNorm(d_model) @staticmethod def with_pos_embed(tensor, pos): return tensor if pos is None else tensor + pos def forward_ffn(self, src): src2 = self.linear2(self.dropout2(self.activation(self.linear1(src)))) src = src + self.dropout3(src2) src = self.norm2(src) return src def forward(self, src, pos, reference_points, spatial_shapes, level_start_index, padding_mask=None, tgt=None): if tgt is None: # self attention src2, sampling_locations, attn_weights = self.self_attn(self.with_pos_embed(src, pos), reference_points, src, spatial_shapes, level_start_index, padding_mask) src = src + self.dropout1(src2) src = self.norm1(src) # torch.Size([2, 13101, 256]) # ffn src = self.forward_ffn(src) return src, sampling_locations, attn_weights else: # self attention tgt2, sampling_locations, attn_weights = self.self_attn(self.with_pos_embed(tgt, pos), reference_points, src, spatial_shapes, level_start_index, padding_mask) tgt = tgt + self.dropout1(tgt2) tgt = self.norm1(tgt) # ffn tgt = self.forward_ffn(tgt) return tgt, sampling_locations, attn_weights class DeformableTransformerEncoder(nn.Module): def __init__(self, encoder_layer, num_layers, mask_predictor_dim=256): super().__init__() self.layers = _get_clones(encoder_layer, num_layers) self.num_layers = num_layers # hack implementation self.aux_heads = False self.class_embed = None self.bbox_embed = None @staticmethod def get_reference_points(spatial_shapes, valid_ratios, device): """Make reference points for every single point on the multi-scale feature maps. Each point has K reference points on every the multi-scale features. """ reference_points_list = [] for lvl, (H_, W_) in enumerate(spatial_shapes): ref_y, ref_x = torch.meshgrid(torch.linspace(0.5, H_ - 0.5, H_, dtype=torch.float32, device=device), torch.linspace(0.5, W_ - 0.5, W_, dtype=torch.float32, device=device)) ref_y = ref_y.reshape(-1)[None] / (valid_ratios[:, None, lvl, 1] * H_) ref_x = ref_x.reshape(-1)[None] / (valid_ratios[:, None, lvl, 0] * W_) # out-of-reference points have relative coords. larger than 1 ref = torch.stack((ref_x, ref_y), -1) reference_points_list.append(ref) reference_points = torch.cat(reference_points_list, 1) reference_points = reference_points[:, :, None] * valid_ratios[:, None] # >>> reference_points[:, :, None].shape # torch.Size([2, 13101, 1, 2]) # >>> valid_ratios[:, None].shape # torch.Size([2, 1, 4, 2]) return reference_points def forward(self, src, spatial_shapes, level_start_index, valid_ratios, pos=None, padding_mask=None, topk_inds=None, output_proposals=None, sparse_token_nums=None): if self.aux_heads: assert output_proposals is not None else: assert output_proposals is None output = src sparsified_keys = False if topk_inds is None else True reference_points = self.get_reference_points(spatial_shapes, valid_ratios, device=src.device) reference_points_orig = reference_points pos_orig = pos output_proposals_orig = output_proposals sampling_locations_all = [] attn_weights_all = [] if self.aux_heads: enc_inter_outputs_class = [] enc_inter_outputs_coords = [] if sparsified_keys: assert topk_inds is not None B_, N_, S_, P_ = reference_points.shape reference_points = torch.gather(reference_points.view(B_, N_, -1), 1, topk_inds.unsqueeze(-1).repeat(1, 1, S_*P_)).view(B_, -1, S_, P_) tgt = torch.gather(output, 1, topk_inds.unsqueeze(-1).repeat(1, 1, output.size(-1))) pos = torch.gather(pos, 1, topk_inds.unsqueeze(-1).repeat(1, 1, pos.size(-1))) if output_proposals is not None: output_proposals = output_proposals.gather(1, topk_inds.unsqueeze(-1).repeat(1, 1, output_proposals.size(-1))) else: tgt = None for lid, layer in enumerate(self.layers): # if tgt is None: self-attention / if tgt is not None: cross-attention w.r.t. the target queries tgt, sampling_locations, attn_weights = layer(output, pos, reference_points, spatial_shapes, level_start_index, padding_mask, tgt=tgt if sparsified_keys else None) sampling_locations_all.append(sampling_locations) attn_weights_all.append(attn_weights) if sparsified_keys: if sparse_token_nums is None: output = output.scatter(1, topk_inds.unsqueeze(-1).repeat(1, 1, tgt.size(-1)), tgt) else: outputs = [] for i in range(topk_inds.shape[0]): outputs.append(output[i].scatter(0, topk_inds[i][:sparse_token_nums[i]].unsqueeze(-1).repeat(1, tgt.size(-1)), tgt[i][:sparse_token_nums[i]])) output = torch.stack(outputs) else: output = tgt if self.aux_heads and lid < self.num_layers - 1: # feed outputs to aux. heads output_class = self.class_embed[lid](tgt) output_offset = self.bbox_embed[lid](tgt) output_coords_unact = output_proposals + output_offset # values to be used for loss compuation enc_inter_outputs_class.append(output_class) enc_inter_outputs_coords.append(output_coords_unact.sigmoid()) # Change dimension from [num_layer, batch_size, ...] to [batch_size, num_layer, ...] sampling_locations_all = torch.stack(sampling_locations_all, dim=1) attn_weights_all = torch.stack(attn_weights_all, dim=1) ret = [output, sampling_locations_all, attn_weights_all] if self.aux_heads: ret += [enc_inter_outputs_class, enc_inter_outputs_coords] return ret class DeformableTransformerDecoderLayer(nn.Module): def __init__(self, d_model=256, d_ffn=1024, dropout=0.1, activation="relu", n_levels=4, n_heads=8, n_points=4): super().__init__() # cross attention self.cross_attn = MSDeformAttn(d_model, n_levels, n_heads, n_points) self.dropout1 = nn.Dropout(dropout) self.norm1 = nn.LayerNorm(d_model) # self attention self.self_attn = nn.MultiheadAttention(d_model, n_heads, dropout=dropout) self.dropout2 = nn.Dropout(dropout) self.norm2 = nn.LayerNorm(d_model) # ffn self.linear1 = nn.Linear(d_model, d_ffn) self.activation = _get_activation_fn(activation) self.dropout3 = nn.Dropout(dropout) self.linear2 = nn.Linear(d_ffn, d_model) self.dropout4 = nn.Dropout(dropout) self.norm3 = nn.LayerNorm(d_model) @staticmethod def with_pos_embed(tensor, pos): return tensor if pos is None else tensor + pos def forward_ffn(self, tgt): tgt2 = self.linear2(self.dropout3(self.activation(self.linear1(tgt)))) tgt = tgt + self.dropout4(tgt2) tgt = self.norm3(tgt) return tgt def forward(self, tgt, query_pos, reference_points, src, src_spatial_shapes, level_start_index, src_padding_mask=None): # self attention q = k = self.with_pos_embed(tgt, query_pos) tgt2 = self.self_attn(q.transpose(0, 1), k.transpose(0, 1), tgt.transpose(0, 1))[0].transpose(0, 1) tgt = tgt + self.dropout2(tgt2) tgt = self.norm2(tgt) # cross attention assert reference_points is not None, "deformable attention needs reference points!" tgt2, sampling_locations, attn_weights = self.cross_attn(self.with_pos_embed(tgt, query_pos), reference_points, src, src_spatial_shapes, level_start_index, src_padding_mask) tgt = tgt + self.dropout1(tgt2) tgt = self.norm1(tgt) # ffn tgt = self.forward_ffn(tgt) # torch.Size([2, 300, 256]) return tgt, sampling_locations, attn_weights class DeformableTransformerDecoder(nn.Module): def __init__(self, decoder_layer, num_layers, return_intermediate=False): super().__init__() self.layers = _get_clones(decoder_layer, num_layers) self.num_layers = num_layers self.return_intermediate = return_intermediate # hack implementation for iterative bounding box refinement and two-stage Deformable DETR self.bbox_embed = None self.class_embed = None def forward(self, tgt, reference_points, src, src_spatial_shapes, src_level_start_index, src_valid_ratios, query_pos=None, src_padding_mask=None, topk_inds=None): """ Args: tgt: torch.Size([2, 300, 256]) (query vectors) reference_points: torch.Size([2, 300, 2]) src: torch.Size([2, 13101, 256]) (last MS feature map from the encoder) query_pos: torch.Size([2, 300, 256]) (learned positional embedding of query vectors) - `tgt` and `query_pos` are originated from the same query embedding. - `tgt` changes through the forward pass as object query vector while `query_pos` does not and is added as positional embedding. Returns: (when return_intermediate=True) output: torch.Size([6, 2, 300, 256]) reference_points: torch.Size([6, 2, 300, 2]) """ output = tgt intermediate = [] intermediate_reference_points = [] sampling_locations_all = [] attn_weights_all = [] for lid, layer in enumerate(self.layers): if reference_points is None: reference_points_input = None elif reference_points.shape[-1] == 4: # output from iterative bounding box refinement # reference_points: N, top_k, 4(x/y/w/h) # src_valid_ratios: N, num_feature_levels, 2(w/h) # reference_points_input: N, top_k, num_feature_levels, 4(x/y/w/h) reference_points_input = reference_points[:, :, None] \ * torch.cat([src_valid_ratios, src_valid_ratios], -1)[:, None] else: assert reference_points.shape[-1] == 2 reference_points_input = reference_points[:, :, None] * src_valid_ratios[:, None] output, sampling_locations, attn_weights = layer(output, query_pos, reference_points_input, src, src_spatial_shapes, src_level_start_index, src_padding_mask) sampling_locations_all.append(sampling_locations) attn_weights_all.append(attn_weights) # hack implementation for iterative bounding box refinement if self.bbox_embed is not None: assert reference_points is not None, "box refinement needs reference points!" tmp = self.bbox_embed[lid](output) if reference_points.shape[-1] == 4: new_reference_points = tmp + inverse_sigmoid(reference_points) new_reference_points = new_reference_points.sigmoid() else: assert reference_points.shape[-1] == 2 new_reference_points = tmp new_reference_points[..., :2] = tmp[..., :2] + inverse_sigmoid(reference_points) new_reference_points = new_reference_points.sigmoid() reference_points = new_reference_points.detach() if self.return_intermediate: intermediate.append(output) intermediate_reference_points.append(reference_points) # Change dimension from [num_layer, batch_size, ...] to [batch_size, num_layer, ...] sampling_locations_all = torch.stack(sampling_locations_all, dim=1) attn_weights_all = torch.stack(attn_weights_all, dim=1) if self.return_intermediate: intermediate_outputs = torch.stack(intermediate) if intermediate_reference_points[0] is None: intermediate_reference_points = None else: intermediate_reference_points = torch.stack(intermediate_reference_points) return intermediate_outputs, intermediate_reference_points, sampling_locations_all, attn_weights_all return output, reference_points, sampling_locations_all, attn_weights_all class MaskPredictor(nn.Module): def __init__(self, in_dim, h_dim): super().__init__() self.h_dim = h_dim self.layer1 = nn.Sequential( nn.LayerNorm(in_dim), nn.Linear(in_dim, h_dim), nn.GELU() ) self.layer2 = nn.Sequential( nn.Linear(h_dim, h_dim // 2), nn.GELU(), nn.Linear(h_dim // 2, h_dim // 4), nn.GELU(), nn.Linear(h_dim // 4, 1) ) def forward(self, x): z = self.layer1(x) z_local, z_global = torch.split(z, self.h_dim // 2, dim=-1) z_global = z_global.mean(dim=1, keepdim=True).expand(-1, z_local.shape[1], -1) z = torch.cat([z_local, z_global], dim=-1) out = self.layer2(z) return out def _get_clones(module, N): return nn.ModuleList([copy.deepcopy(module) for i in range(N)]) def _get_activation_fn(activation): """Return an activation function given a string""" if activation == "relu": return F.relu if activation == "gelu": return F.gelu if activation == "glu": return F.glu raise RuntimeError(F"activation should be relu/gelu, not {activation}.") def build_deforamble_transformer(args): return DeformableTransformer( d_model=args.hidden_dim, nhead=args.nheads, num_encoder_layers=args.enc_layers, num_decoder_layers=args.dec_layers, dim_feedforward=args.dim_feedforward, dropout=args.dropout, activation="relu", return_intermediate_dec=True, num_feature_levels=args.num_feature_levels, dec_n_points=args.dec_n_points, enc_n_points=args.enc_n_points, two_stage=args.two_stage, two_stage_num_proposals=args.num_queries, args=args) ================================================ FILE: models/matcher.py ================================================ # ------------------------------------------------------------------------------------ # Sparse DETR # Copyright (c) 2021 KakaoBrain. All Rights Reserved. # Licensed under the Apache License, Version 2.0 [see LICENSE for details] # ------------------------------------------------------------------------------------ # Modified from Deformable DETR (https://github.com/fundamentalvision/Deformable-DETR) # Copyright (c) 2020 SenseTime. All Rights Reserved. # ------------------------------------------------------------------------------------ # Modified from DETR (https://github.com/facebookresearch/detr) # Copyright (c) Facebook, Inc. and its affiliates. All Rights Reserved # ------------------------------------------------------------------------------------ """ Modules to compute the matching cost and solve the corresponding LSAP. """ import torch from scipy.optimize import linear_sum_assignment from torch import nn from util.box_ops import box_cxcywh_to_xyxy, generalized_box_iou class HungarianMatcher(nn.Module): """This class computes an assignment between the targets and the predictions of the network For efficiency reasons, the targets don't include the no_object. Because of this, in general, there are more predictions than targets. In this case, we do a 1-to-1 matching of the best predictions, while the others are un-matched (and thus treated as non-objects). """ def __init__(self, cost_class: float = 1, cost_bbox: float = 1, cost_giou: float = 1): """Creates the matcher Params: cost_class: This is the relative weight of the classification error in the matching cost cost_bbox: This is the relative weight of the L1 error of the bounding box coordinates in the matching cost cost_giou: This is the relative weight of the giou loss of the bounding box in the matching cost """ super().__init__() self.cost_class = cost_class self.cost_bbox = cost_bbox self.cost_giou = cost_giou assert cost_class != 0 or cost_bbox != 0 or cost_giou != 0, "all costs cant be 0" def forward(self, outputs, targets): """ Performs the matching Params: outputs: This is a dict that contains at least these entries: "pred_logits": Tensor of dim [batch_size, num_queries, num_classes] with the classification logits "pred_boxes": Tensor of dim [batch_size, num_queries, 4] with the predicted box coordinates targets: This is a list of targets (len(targets) = batch_size), where each target is a dict containing: "labels": Tensor of dim [num_target_boxes] (where num_target_boxes is the number of ground-truth objects in the target) containing the class labels "boxes": Tensor of dim [num_target_boxes, 4] containing the target box coordinates Returns: A list of size batch_size, containing tuples of (index_i, index_j) where: - index_i is the indices of the selected predictions (in order) - index_j is the indices of the corresponding selected targets (in order) For each batch element, it holds: len(index_i) = len(index_j) = min(num_queries, num_target_boxes) """ with torch.no_grad(): bs, num_queries = outputs["pred_logits"].shape[:2] # We flatten to compute the cost matrices in a batch out_prob = outputs["pred_logits"].flatten(0, 1).sigmoid() out_bbox = outputs["pred_boxes"].flatten(0, 1) # [batch_size * num_queries, 4] # Also concat the target labels and boxes tgt_ids = torch.cat([v["labels"] for v in targets]) tgt_bbox = torch.cat([v["boxes"] for v in targets]) # Compute the classification cost. alpha = 0.25 gamma = 2.0 neg_cost_class = (1 - alpha) * (out_prob ** gamma) * (-(1 - out_prob + 1e-8).log()) pos_cost_class = alpha * ((1 - out_prob) ** gamma) * (-(out_prob + 1e-8).log()) cost_class = pos_cost_class[:, tgt_ids] - neg_cost_class[:, tgt_ids] # Compute the L1 cost between boxes cost_bbox = torch.cdist(out_bbox, tgt_bbox, p=1) # Compute the giou cost betwen boxes cost_giou = -generalized_box_iou(box_cxcywh_to_xyxy(out_bbox), box_cxcywh_to_xyxy(tgt_bbox)) # Final cost matrix C = self.cost_bbox * cost_bbox + self.cost_class * cost_class + self.cost_giou * cost_giou C = C.view(bs, num_queries, -1).cpu() sizes = [len(v["boxes"]) for v in targets] indices = [linear_sum_assignment(c[i]) for i, c in enumerate(C.split(sizes, -1))] return [(torch.as_tensor(i, dtype=torch.int64), torch.as_tensor([_j % size for _j in j], dtype=torch.int64)) for (i, j), size in zip(indices, sizes)] def build_matcher(args): return HungarianMatcher(cost_class=args.set_cost_class, cost_bbox=args.set_cost_bbox, cost_giou=args.set_cost_giou) ================================================ FILE: models/ops/functions/__init__.py ================================================ # ------------------------------------------------------------------------------------------------ # Deformable DETR # Copyright (c) 2020 SenseTime. All Rights Reserved. # Licensed under the Apache License, Version 2.0 [see LICENSE for details] # ------------------------------------------------------------------------------------------------ # Modified from https://github.com/chengdazhi/Deformable-Convolution-V2-PyTorch/tree/pytorch_1.0.0 # ------------------------------------------------------------------------------------------------ from .ms_deform_attn_func import MSDeformAttnFunction ================================================ FILE: models/ops/functions/ms_deform_attn_func.py ================================================ # ------------------------------------------------------------------------------------------------ # Deformable DETR # Copyright (c) 2020 SenseTime. All Rights Reserved. # Licensed under the Apache License, Version 2.0 [see LICENSE for details] # ------------------------------------------------------------------------------------------------ # Modified from https://github.com/chengdazhi/Deformable-Convolution-V2-PyTorch/tree/pytorch_1.0.0 # ------------------------------------------------------------------------------------------------ from __future__ import absolute_import from __future__ import print_function from __future__ import division import torch import torch.nn.functional as F from torch.autograd import Function from torch.autograd.function import once_differentiable import MultiScaleDeformableAttention as MSDA class MSDeformAttnFunction(Function): @staticmethod def forward(ctx, value, value_spatial_shapes, value_level_start_index, sampling_locations, attention_weights, im2col_step): ctx.im2col_step = im2col_step output = MSDA.ms_deform_attn_forward( value, value_spatial_shapes, value_level_start_index, sampling_locations, attention_weights, ctx.im2col_step) ctx.save_for_backward(value, value_spatial_shapes, value_level_start_index, sampling_locations, attention_weights) return output @staticmethod @once_differentiable def backward(ctx, grad_output): value, value_spatial_shapes, value_level_start_index, sampling_locations, attention_weights = ctx.saved_tensors grad_value, grad_sampling_loc, grad_attn_weight = \ MSDA.ms_deform_attn_backward( value, value_spatial_shapes, value_level_start_index, sampling_locations, attention_weights, grad_output, ctx.im2col_step) return grad_value, None, None, grad_sampling_loc, grad_attn_weight, None def ms_deform_attn_core_pytorch(value, value_spatial_shapes, sampling_locations, attention_weights): # for debug and test only, # need to use cuda version instead N_, S_, M_, D_ = value.shape _, Lq_, M_, L_, P_, _ = sampling_locations.shape value_list = value.split([H_ * W_ for H_, W_ in value_spatial_shapes], dim=1) sampling_grids = 2 * sampling_locations - 1 sampling_value_list = [] for lid_, (H_, W_) in enumerate(value_spatial_shapes): # N_, H_*W_, M_, D_ -> N_, H_*W_, M_*D_ -> N_, M_*D_, H_*W_ -> N_*M_, D_, H_, W_ value_l_ = value_list[lid_].flatten(2).transpose(1, 2).reshape(N_*M_, D_, H_, W_) # N_, Lq_, M_, P_, 2 -> N_, M_, Lq_, P_, 2 -> N_*M_, Lq_, P_, 2 sampling_grid_l_ = sampling_grids[:, :, :, lid_].transpose(1, 2).flatten(0, 1) # N_*M_, D_, Lq_, P_ sampling_value_l_ = F.grid_sample(value_l_, sampling_grid_l_, mode='bilinear', padding_mode='zeros', align_corners=False) sampling_value_list.append(sampling_value_l_) # (N_, Lq_, M_, L_, P_) -> (N_, M_, Lq_, L_, P_) -> (N_, M_, 1, Lq_, L_*P_) attention_weights = attention_weights.transpose(1, 2).reshape(N_*M_, 1, Lq_, L_*P_) output = (torch.stack(sampling_value_list, dim=-2).flatten(-2) * attention_weights).sum(-1).view(N_, M_*D_, Lq_) return output.transpose(1, 2).contiguous() ================================================ FILE: models/ops/make.sh ================================================ #!/usr/bin/env bash # ------------------------------------------------------------------------------------------------ # Deformable DETR # Copyright (c) 2020 SenseTime. All Rights Reserved. # Licensed under the Apache License, Version 2.0 [see LICENSE for details] # ------------------------------------------------------------------------------------------------ # Modified from https://github.com/chengdazhi/Deformable-Convolution-V2-PyTorch/tree/pytorch_1.0.0 # ------------------------------------------------------------------------------------------------ python setup.py build install ================================================ FILE: models/ops/modules/__init__.py ================================================ # ------------------------------------------------------------------------------------ # Sparse DETR # Copyright (c) 2021 KakaoBrain. All Rights Reserved. # Licensed under the Apache License, Version 2.0 [see LICENSE for details] # ------------------------------------------------------------------------------------------------ # Deformable DETR # Copyright (c) 2020 SenseTime. All Rights Reserved. # ------------------------------------------------------------------------------------------------ # Modified from https://github.com/chengdazhi/Deformable-Convolution-V2-PyTorch/tree/pytorch_1.0.0 # ------------------------------------------------------------------------------------------------ from .ms_deform_attn import MSDeformAttn ================================================ FILE: models/ops/modules/ms_deform_attn.py ================================================ # ------------------------------------------------------------------------------------ # Sparse DETR # Copyright (c) 2021 KakaoBrain. All Rights Reserved. # Licensed under the Apache License, Version 2.0 [see LICENSE for details] # ------------------------------------------------------------------------------------------------ # Deformable DETR # Copyright (c) 2020 SenseTime. All Rights Reserved. # ------------------------------------------------------------------------------------------------ # Modified from https://github.com/chengdazhi/Deformable-Convolution-V2-PyTorch/tree/pytorch_1.0.0 # ------------------------------------------------------------------------------------------------ from __future__ import absolute_import from __future__ import print_function from __future__ import division import warnings import math import torch from torch import nn import torch.nn.functional as F from torch.nn.init import xavier_uniform_, constant_ from ..functions import MSDeformAttnFunction def _is_power_of_2(n): if (not isinstance(n, int)) or (n < 0): raise ValueError("invalid input for _is_power_of_2: {} (type: {})".format(n, type(n))) return (n & (n-1) == 0) and n != 0 class MSDeformAttn(nn.Module): def __init__(self, d_model=256, n_levels=4, n_heads=8, n_points=4): """ Multi-Scale Deformable Attention Module :param d_model hidden dimension :param n_levels number of feature levels :param n_heads number of attention heads :param n_points number of sampling points per attention head per feature level """ super().__init__() if d_model % n_heads != 0: raise ValueError('d_model must be divisible by n_heads, but got {} and {}'.format(d_model, n_heads)) _d_per_head = d_model // n_heads # you'd better set _d_per_head to a power of 2 which is more efficient in our CUDA implementation if not _is_power_of_2(_d_per_head): warnings.warn("You'd better set d_model in MSDeformAttn to make the dimension of each attention head a power of 2 " "which is more efficient in our CUDA implementation.") self.im2col_step = 64 self.d_model = d_model self.n_levels = n_levels self.n_heads = n_heads self.n_points = n_points self.sampling_offsets = nn.Linear(d_model, n_heads * n_levels * n_points * 2) self.attention_weights = nn.Linear(d_model, n_heads * n_levels * n_points) self.value_proj = nn.Linear(d_model, d_model) self.output_proj = nn.Linear(d_model, d_model) self.python_ops_for_test = False self._reset_parameters() def _reset_parameters(self): constant_(self.sampling_offsets.weight.data, 0.) thetas = torch.arange(self.n_heads, dtype=torch.float32) * (2.0 * math.pi / self.n_heads) grid_init = torch.stack([thetas.cos(), thetas.sin()], -1) grid_init = (grid_init / grid_init.abs().max(-1, keepdim=True)[0]).view(self.n_heads, 1, 1, 2).repeat(1, self.n_levels, self.n_points, 1) for i in range(self.n_points): grid_init[:, :, i, :] *= i + 1 with torch.no_grad(): self.sampling_offsets.bias = nn.Parameter(grid_init.view(-1)) constant_(self.attention_weights.weight.data, 0.) constant_(self.attention_weights.bias.data, 0.) xavier_uniform_(self.value_proj.weight.data) constant_(self.value_proj.bias.data, 0.) xavier_uniform_(self.output_proj.weight.data) constant_(self.output_proj.bias.data, 0.) def forward(self, query, reference_points, input_flatten, input_spatial_shapes, input_level_start_index, input_padding_mask=None): """ :param query (N, Length_{query}, C) :param reference_points (N, Length_{query}, n_levels, 2), range in [0, 1], top-left (0,0), bottom-right (1, 1), including padding area or (N, Length_{query}, n_levels, 4), add additional (w, h) to form reference boxes :param input_flatten (N, \sum_{l=0}^{L-1} H_l \cdot W_l, C) :param input_spatial_shapes (n_levels, 2), [(H_0, W_0), (H_1, W_1), ..., (H_{L-1}, W_{L-1})] :param input_level_start_index (n_levels, ), [0, H_0*W_0, H_0*W_0+H_1*W_1, H_0*W_0+H_1*W_1+H_2*W_2, ..., H_0*W_0+H_1*W_1+...+H_{L-1}*W_{L-1}] :param input_padding_mask (N, \sum_{l=0}^{L-1} H_l \cdot W_l), True for padding elements, False for non-padding elements :return output (N, Length_{query}, C) """ N, Len_q, _ = query.shape N, Len_in, _ = input_flatten.shape assert (input_spatial_shapes[:, 0] * input_spatial_shapes[:, 1]).sum() == Len_in value = self.value_proj(input_flatten) if input_padding_mask is not None: value = value.masked_fill(input_padding_mask[..., None], float(0)) value = value.view(N, Len_in, self.n_heads, self.d_model // self.n_heads) sampling_offsets = self.sampling_offsets(query).view(N, Len_q, self.n_heads, self.n_levels, self.n_points, 2) attention_weights = self.attention_weights(query).view(N, Len_q, self.n_heads, self.n_levels * self.n_points) attention_weights = F.softmax(attention_weights, -1).view(N, Len_q, self.n_heads, self.n_levels, self.n_points) # N, Len_q, n_heads, n_levels, n_points, 2 if reference_points.shape[-1] == 2: offset_normalizer = torch.stack([input_spatial_shapes[..., 1], input_spatial_shapes[..., 0]], -1) sampling_locations = reference_points[:, :, None, :, None, :] \ + sampling_offsets / offset_normalizer[None, None, None, :, None, :] elif reference_points.shape[-1] == 4: sampling_locations = reference_points[:, :, None, :, None, :2] \ + sampling_offsets / self.n_points * reference_points[:, :, None, :, None, 2:] * 0.5 else: raise ValueError( 'Last dim of reference_points must be 2 or 4, but get {} instead.'.format(reference_points.shape[-1])) if not self.python_ops_for_test: output = MSDeformAttnFunction.apply( value, input_spatial_shapes, input_level_start_index, sampling_locations, attention_weights, self.im2col_step) else: output = ms_deform_attn_core_pytorch(value, input_spatial_shapes, sampling_locations, attention_weights) output = self.output_proj(output) return output, sampling_locations, attention_weights def ms_deform_attn_core_pytorch(value, value_spatial_shapes, sampling_locations, attention_weights): # for debug and test only, # need to use cuda version instead N_, S_, M_, D_ = value.shape _, Lq_, M_, L_, P_, _ = sampling_locations.shape value_list = value.split([H_ * W_ for H_, W_ in value_spatial_shapes], dim=1) sampling_grids = 2 * sampling_locations - 1 sampling_value_list = [] for lid_, (H_, W_) in enumerate(value_spatial_shapes): # N_, H_*W_, M_, D_ -> N_, H_*W_, M_*D_ -> N_, M_*D_, H_*W_ -> N_*M_, D_, H_, W_ value_l_ = value_list[lid_].flatten(2).transpose(1, 2).reshape(N_*M_, D_, H_, W_) # N_, Lq_, M_, P_, 2 -> N_, M_, Lq_, P_, 2 -> N_*M_, Lq_, P_, 2 sampling_grid_l_ = sampling_grids[:, :, :, lid_].transpose(1, 2).flatten(0, 1) # N_*M_, D_, Lq_, P_ sampling_value_l_ = F.grid_sample(value_l_, sampling_grid_l_, mode='bilinear', padding_mode='zeros', align_corners=False) sampling_value_list.append(sampling_value_l_) # (N_, Lq_, M_, L_, P_) -> (N_, M_, Lq_, L_, P_) -> (N_, M_, 1, Lq_, L_*P_) attention_weights = attention_weights.transpose(1, 2).reshape(N_*M_, 1, Lq_, L_*P_) output = (torch.stack(sampling_value_list, dim=-2).flatten(-2) * attention_weights).sum(-1).view(N_, M_*D_, Lq_) return output.transpose(1, 2).contiguous() ================================================ FILE: models/ops/setup.py ================================================ # ------------------------------------------------------------------------------------------------ # Deformable DETR # Copyright (c) 2020 SenseTime. All Rights Reserved. # Licensed under the Apache License, Version 2.0 [see LICENSE for details] # ------------------------------------------------------------------------------------------------ # Modified from https://github.com/chengdazhi/Deformable-Convolution-V2-PyTorch/tree/pytorch_1.0.0 # ------------------------------------------------------------------------------------------------ import os import glob import torch from torch.utils.cpp_extension import CUDA_HOME from torch.utils.cpp_extension import CppExtension from torch.utils.cpp_extension import CUDAExtension from setuptools import find_packages from setuptools import setup requirements = ["torch", "torchvision"] def get_extensions(): this_dir = os.path.dirname(os.path.abspath(__file__)) extensions_dir = os.path.join(this_dir, "src") main_file = glob.glob(os.path.join(extensions_dir, "*.cpp")) source_cpu = glob.glob(os.path.join(extensions_dir, "cpu", "*.cpp")) source_cuda = glob.glob(os.path.join(extensions_dir, "cuda", "*.cu")) sources = main_file + source_cpu extension = CppExtension extra_compile_args = {"cxx": []} define_macros = [] if torch.cuda.is_available() and CUDA_HOME is not None: extension = CUDAExtension sources += source_cuda define_macros += [("WITH_CUDA", None)] extra_compile_args["nvcc"] = [ "-DCUDA_HAS_FP16=1", "-D__CUDA_NO_HALF_OPERATORS__", "-D__CUDA_NO_HALF_CONVERSIONS__", "-D__CUDA_NO_HALF2_OPERATORS__", ] else: raise NotImplementedError('Cuda is not availabel') sources = [os.path.join(extensions_dir, s) for s in sources] include_dirs = [extensions_dir] ext_modules = [ extension( "MultiScaleDeformableAttention", sources, include_dirs=include_dirs, define_macros=define_macros, extra_compile_args=extra_compile_args, ) ] return ext_modules setup( name="MultiScaleDeformableAttention", version="1.0", author="Weijie Su", url="https://github.com/fundamentalvision/Deformable-DETR", description="PyTorch Wrapper for CUDA Functions of Multi-Scale Deformable Attention", packages=find_packages(exclude=("configs", "tests",)), ext_modules=get_extensions(), cmdclass={"build_ext": torch.utils.cpp_extension.BuildExtension}, ) ================================================ FILE: models/ops/src/cpu/ms_deform_attn_cpu.cpp ================================================ /*! ************************************************************************************************** * Deformable DETR * Copyright (c) 2020 SenseTime. All Rights Reserved. * Licensed under the Apache License, Version 2.0 [see LICENSE for details] ************************************************************************************************** * Modified from https://github.com/chengdazhi/Deformable-Convolution-V2-PyTorch/tree/pytorch_1.0.0 ************************************************************************************************** */ #include #include #include at::Tensor ms_deform_attn_cpu_forward( const at::Tensor &value, const at::Tensor &spatial_shapes, const at::Tensor &level_start_index, const at::Tensor &sampling_loc, const at::Tensor &attn_weight, const int im2col_step) { AT_ERROR("Not implement on cpu"); } std::vector ms_deform_attn_cpu_backward( const at::Tensor &value, const at::Tensor &spatial_shapes, const at::Tensor &level_start_index, const at::Tensor &sampling_loc, const at::Tensor &attn_weight, const at::Tensor &grad_output, const int im2col_step) { AT_ERROR("Not implement on cpu"); } ================================================ FILE: models/ops/src/cpu/ms_deform_attn_cpu.h ================================================ /*! ************************************************************************************************** * Deformable DETR * Copyright (c) 2020 SenseTime. All Rights Reserved. * Licensed under the Apache License, Version 2.0 [see LICENSE for details] ************************************************************************************************** * Modified from https://github.com/chengdazhi/Deformable-Convolution-V2-PyTorch/tree/pytorch_1.0.0 ************************************************************************************************** */ #pragma once #include at::Tensor ms_deform_attn_cpu_forward( const at::Tensor &value, const at::Tensor &spatial_shapes, const at::Tensor &level_start_index, const at::Tensor &sampling_loc, const at::Tensor &attn_weight, const int im2col_step); std::vector ms_deform_attn_cpu_backward( const at::Tensor &value, const at::Tensor &spatial_shapes, const at::Tensor &level_start_index, const at::Tensor &sampling_loc, const at::Tensor &attn_weight, const at::Tensor &grad_output, const int im2col_step); ================================================ FILE: models/ops/src/cuda/ms_deform_attn_cuda.cu ================================================ /*! ************************************************************************************************** * Deformable DETR * Copyright (c) 2020 SenseTime. All Rights Reserved. * Licensed under the Apache License, Version 2.0 [see LICENSE for details] ************************************************************************************************** * Modified from https://github.com/chengdazhi/Deformable-Convolution-V2-PyTorch/tree/pytorch_1.0.0 ************************************************************************************************** */ #include #include "cuda/ms_deform_im2col_cuda.cuh" #include #include #include #include at::Tensor ms_deform_attn_cuda_forward( const at::Tensor &value, const at::Tensor &spatial_shapes, const at::Tensor &level_start_index, const at::Tensor &sampling_loc, const at::Tensor &attn_weight, const int im2col_step) { AT_ASSERTM(value.is_contiguous(), "value tensor has to be contiguous"); AT_ASSERTM(spatial_shapes.is_contiguous(), "spatial_shapes tensor has to be contiguous"); AT_ASSERTM(level_start_index.is_contiguous(), "level_start_index tensor has to be contiguous"); AT_ASSERTM(sampling_loc.is_contiguous(), "sampling_loc tensor has to be contiguous"); AT_ASSERTM(attn_weight.is_contiguous(), "attn_weight tensor has to be contiguous"); AT_ASSERTM(value.type().is_cuda(), "value must be a CUDA tensor"); AT_ASSERTM(spatial_shapes.type().is_cuda(), "spatial_shapes must be a CUDA tensor"); AT_ASSERTM(level_start_index.type().is_cuda(), "level_start_index must be a CUDA tensor"); AT_ASSERTM(sampling_loc.type().is_cuda(), "sampling_loc must be a CUDA tensor"); AT_ASSERTM(attn_weight.type().is_cuda(), "attn_weight must be a CUDA tensor"); const int batch = value.size(0); const int spatial_size = value.size(1); const int num_heads = value.size(2); const int channels = value.size(3); const int num_levels = spatial_shapes.size(0); const int num_query = sampling_loc.size(1); const int num_point = sampling_loc.size(4); const int im2col_step_ = std::min(batch, im2col_step); AT_ASSERTM(batch % im2col_step_ == 0, "batch(%d) must divide im2col_step(%d)", batch, im2col_step_); auto output = at::zeros({batch, num_query, num_heads, channels}, value.options()); const int batch_n = im2col_step_; auto output_n = output.view({batch/im2col_step_, batch_n, num_query, num_heads, channels}); auto per_value_size = spatial_size * num_heads * channels; auto per_sample_loc_size = num_query * num_heads * num_levels * num_point * 2; auto per_attn_weight_size = num_query * num_heads * num_levels * num_point; for (int n = 0; n < batch/im2col_step_; ++n) { auto columns = output_n.select(0, n); AT_DISPATCH_FLOATING_TYPES(value.type(), "ms_deform_attn_forward_cuda", ([&] { ms_deformable_im2col_cuda(at::cuda::getCurrentCUDAStream(), value.data() + n * im2col_step_ * per_value_size, spatial_shapes.data(), level_start_index.data(), sampling_loc.data() + n * im2col_step_ * per_sample_loc_size, attn_weight.data() + n * im2col_step_ * per_attn_weight_size, batch_n, spatial_size, num_heads, channels, num_levels, num_query, num_point, columns.data()); })); } output = output.view({batch, num_query, num_heads*channels}); return output; } std::vector ms_deform_attn_cuda_backward( const at::Tensor &value, const at::Tensor &spatial_shapes, const at::Tensor &level_start_index, const at::Tensor &sampling_loc, const at::Tensor &attn_weight, const at::Tensor &grad_output, const int im2col_step) { AT_ASSERTM(value.is_contiguous(), "value tensor has to be contiguous"); AT_ASSERTM(spatial_shapes.is_contiguous(), "spatial_shapes tensor has to be contiguous"); AT_ASSERTM(level_start_index.is_contiguous(), "level_start_index tensor has to be contiguous"); AT_ASSERTM(sampling_loc.is_contiguous(), "sampling_loc tensor has to be contiguous"); AT_ASSERTM(attn_weight.is_contiguous(), "attn_weight tensor has to be contiguous"); AT_ASSERTM(grad_output.is_contiguous(), "grad_output tensor has to be contiguous"); AT_ASSERTM(value.type().is_cuda(), "value must be a CUDA tensor"); AT_ASSERTM(spatial_shapes.type().is_cuda(), "spatial_shapes must be a CUDA tensor"); AT_ASSERTM(level_start_index.type().is_cuda(), "level_start_index must be a CUDA tensor"); AT_ASSERTM(sampling_loc.type().is_cuda(), "sampling_loc must be a CUDA tensor"); AT_ASSERTM(attn_weight.type().is_cuda(), "attn_weight must be a CUDA tensor"); AT_ASSERTM(grad_output.type().is_cuda(), "grad_output must be a CUDA tensor"); const int batch = value.size(0); const int spatial_size = value.size(1); const int num_heads = value.size(2); const int channels = value.size(3); const int num_levels = spatial_shapes.size(0); const int num_query = sampling_loc.size(1); const int num_point = sampling_loc.size(4); const int im2col_step_ = std::min(batch, im2col_step); AT_ASSERTM(batch % im2col_step_ == 0, "batch(%d) must divide im2col_step(%d)", batch, im2col_step_); auto grad_value = at::zeros_like(value); auto grad_sampling_loc = at::zeros_like(sampling_loc); auto grad_attn_weight = at::zeros_like(attn_weight); const int batch_n = im2col_step_; auto per_value_size = spatial_size * num_heads * channels; auto per_sample_loc_size = num_query * num_heads * num_levels * num_point * 2; auto per_attn_weight_size = num_query * num_heads * num_levels * num_point; auto grad_output_n = grad_output.view({batch/im2col_step_, batch_n, num_query, num_heads, channels}); for (int n = 0; n < batch/im2col_step_; ++n) { auto grad_output_g = grad_output_n.select(0, n); AT_DISPATCH_FLOATING_TYPES(value.type(), "ms_deform_attn_backward_cuda", ([&] { ms_deformable_col2im_cuda(at::cuda::getCurrentCUDAStream(), grad_output_g.data(), value.data() + n * im2col_step_ * per_value_size, spatial_shapes.data(), level_start_index.data(), sampling_loc.data() + n * im2col_step_ * per_sample_loc_size, attn_weight.data() + n * im2col_step_ * per_attn_weight_size, batch_n, spatial_size, num_heads, channels, num_levels, num_query, num_point, grad_value.data() + n * im2col_step_ * per_value_size, grad_sampling_loc.data() + n * im2col_step_ * per_sample_loc_size, grad_attn_weight.data() + n * im2col_step_ * per_attn_weight_size); })); } return { grad_value, grad_sampling_loc, grad_attn_weight }; } ================================================ FILE: models/ops/src/cuda/ms_deform_attn_cuda.h ================================================ /*! ************************************************************************************************** * Deformable DETR * Copyright (c) 2020 SenseTime. All Rights Reserved. * Licensed under the Apache License, Version 2.0 [see LICENSE for details] ************************************************************************************************** * Modified from https://github.com/chengdazhi/Deformable-Convolution-V2-PyTorch/tree/pytorch_1.0.0 ************************************************************************************************** */ #pragma once #include at::Tensor ms_deform_attn_cuda_forward( const at::Tensor &value, const at::Tensor &spatial_shapes, const at::Tensor &level_start_index, const at::Tensor &sampling_loc, const at::Tensor &attn_weight, const int im2col_step); std::vector ms_deform_attn_cuda_backward( const at::Tensor &value, const at::Tensor &spatial_shapes, const at::Tensor &level_start_index, const at::Tensor &sampling_loc, const at::Tensor &attn_weight, const at::Tensor &grad_output, const int im2col_step); ================================================ FILE: models/ops/src/cuda/ms_deform_im2col_cuda.cuh ================================================ /*! ************************************************************************** * Deformable DETR * Copyright (c) 2020 SenseTime. All Rights Reserved. * Licensed under the Apache License, Version 2.0 [see LICENSE for details] ************************************************************************** * Modified from DCN (https://github.com/msracver/Deformable-ConvNets) * Copyright (c) 2018 Microsoft ************************************************************************** */ #include #include #include #include #include #include #define CUDA_KERNEL_LOOP(i, n) \ for (int i = blockIdx.x * blockDim.x + threadIdx.x; \ i < (n); \ i += blockDim.x * gridDim.x) const int CUDA_NUM_THREADS = 1024; inline int GET_BLOCKS(const int N, const int num_threads) { return (N + num_threads - 1) / num_threads; } template __device__ scalar_t ms_deform_attn_im2col_bilinear(const scalar_t* &bottom_data, const int &height, const int &width, const int &nheads, const int &channels, const scalar_t &h, const scalar_t &w, const int &m, const int &c) { const int h_low = floor(h); const int w_low = floor(w); const int h_high = h_low + 1; const int w_high = w_low + 1; const scalar_t lh = h - h_low; const scalar_t lw = w - w_low; const scalar_t hh = 1 - lh, hw = 1 - lw; const int w_stride = nheads * channels; const int h_stride = width * w_stride; const int h_low_ptr_offset = h_low * h_stride; const int h_high_ptr_offset = h_low_ptr_offset + h_stride; const int w_low_ptr_offset = w_low * w_stride; const int w_high_ptr_offset = w_low_ptr_offset + w_stride; const int base_ptr = m * channels + c; scalar_t v1 = 0; if (h_low >= 0 && w_low >= 0) { const int ptr1 = h_low_ptr_offset + w_low_ptr_offset + base_ptr; v1 = bottom_data[ptr1]; } scalar_t v2 = 0; if (h_low >= 0 && w_high <= width - 1) { const int ptr2 = h_low_ptr_offset + w_high_ptr_offset + base_ptr; v2 = bottom_data[ptr2]; } scalar_t v3 = 0; if (h_high <= height - 1 && w_low >= 0) { const int ptr3 = h_high_ptr_offset + w_low_ptr_offset + base_ptr; v3 = bottom_data[ptr3]; } scalar_t v4 = 0; if (h_high <= height - 1 && w_high <= width - 1) { const int ptr4 = h_high_ptr_offset + w_high_ptr_offset + base_ptr; v4 = bottom_data[ptr4]; } const scalar_t w1 = hh * hw, w2 = hh * lw, w3 = lh * hw, w4 = lh * lw; const scalar_t val = (w1 * v1 + w2 * v2 + w3 * v3 + w4 * v4); return val; } template __device__ void ms_deform_attn_col2im_bilinear(const scalar_t* &bottom_data, const int &height, const int &width, const int &nheads, const int &channels, const scalar_t &h, const scalar_t &w, const int &m, const int &c, const scalar_t &top_grad, const scalar_t &attn_weight, scalar_t* &grad_value, scalar_t* grad_sampling_loc, scalar_t* grad_attn_weight) { const int h_low = floor(h); const int w_low = floor(w); const int h_high = h_low + 1; const int w_high = w_low + 1; const scalar_t lh = h - h_low; const scalar_t lw = w - w_low; const scalar_t hh = 1 - lh, hw = 1 - lw; const int w_stride = nheads * channels; const int h_stride = width * w_stride; const int h_low_ptr_offset = h_low * h_stride; const int h_high_ptr_offset = h_low_ptr_offset + h_stride; const int w_low_ptr_offset = w_low * w_stride; const int w_high_ptr_offset = w_low_ptr_offset + w_stride; const int base_ptr = m * channels + c; const scalar_t w1 = hh * hw, w2 = hh * lw, w3 = lh * hw, w4 = lh * lw; const scalar_t top_grad_value = top_grad * attn_weight; scalar_t grad_h_weight = 0, grad_w_weight = 0; scalar_t v1 = 0; if (h_low >= 0 && w_low >= 0) { const int ptr1 = h_low_ptr_offset + w_low_ptr_offset + base_ptr; v1 = bottom_data[ptr1]; grad_h_weight -= hw * v1; grad_w_weight -= hh * v1; atomicAdd(grad_value+ptr1, w1*top_grad_value); } scalar_t v2 = 0; if (h_low >= 0 && w_high <= width - 1) { const int ptr2 = h_low_ptr_offset + w_high_ptr_offset + base_ptr; v2 = bottom_data[ptr2]; grad_h_weight -= lw * v2; grad_w_weight += hh * v2; atomicAdd(grad_value+ptr2, w2*top_grad_value); } scalar_t v3 = 0; if (h_high <= height - 1 && w_low >= 0) { const int ptr3 = h_high_ptr_offset + w_low_ptr_offset + base_ptr; v3 = bottom_data[ptr3]; grad_h_weight += hw * v3; grad_w_weight -= lh * v3; atomicAdd(grad_value+ptr3, w3*top_grad_value); } scalar_t v4 = 0; if (h_high <= height - 1 && w_high <= width - 1) { const int ptr4 = h_high_ptr_offset + w_high_ptr_offset + base_ptr; v4 = bottom_data[ptr4]; grad_h_weight += lw * v4; grad_w_weight += lh * v4; atomicAdd(grad_value+ptr4, w4*top_grad_value); } const scalar_t val = (w1 * v1 + w2 * v2 + w3 * v3 + w4 * v4); *grad_attn_weight = top_grad * val; *grad_sampling_loc = width * grad_w_weight * top_grad_value; *(grad_sampling_loc + 1) = height * grad_h_weight * top_grad_value; } template __device__ void ms_deform_attn_col2im_bilinear_gm(const scalar_t* &bottom_data, const int &height, const int &width, const int &nheads, const int &channels, const scalar_t &h, const scalar_t &w, const int &m, const int &c, const scalar_t &top_grad, const scalar_t &attn_weight, scalar_t* &grad_value, scalar_t* grad_sampling_loc, scalar_t* grad_attn_weight) { const int h_low = floor(h); const int w_low = floor(w); const int h_high = h_low + 1; const int w_high = w_low + 1; const scalar_t lh = h - h_low; const scalar_t lw = w - w_low; const scalar_t hh = 1 - lh, hw = 1 - lw; const int w_stride = nheads * channels; const int h_stride = width * w_stride; const int h_low_ptr_offset = h_low * h_stride; const int h_high_ptr_offset = h_low_ptr_offset + h_stride; const int w_low_ptr_offset = w_low * w_stride; const int w_high_ptr_offset = w_low_ptr_offset + w_stride; const int base_ptr = m * channels + c; const scalar_t w1 = hh * hw, w2 = hh * lw, w3 = lh * hw, w4 = lh * lw; const scalar_t top_grad_value = top_grad * attn_weight; scalar_t grad_h_weight = 0, grad_w_weight = 0; scalar_t v1 = 0; if (h_low >= 0 && w_low >= 0) { const int ptr1 = h_low_ptr_offset + w_low_ptr_offset + base_ptr; v1 = bottom_data[ptr1]; grad_h_weight -= hw * v1; grad_w_weight -= hh * v1; atomicAdd(grad_value+ptr1, w1*top_grad_value); } scalar_t v2 = 0; if (h_low >= 0 && w_high <= width - 1) { const int ptr2 = h_low_ptr_offset + w_high_ptr_offset + base_ptr; v2 = bottom_data[ptr2]; grad_h_weight -= lw * v2; grad_w_weight += hh * v2; atomicAdd(grad_value+ptr2, w2*top_grad_value); } scalar_t v3 = 0; if (h_high <= height - 1 && w_low >= 0) { const int ptr3 = h_high_ptr_offset + w_low_ptr_offset + base_ptr; v3 = bottom_data[ptr3]; grad_h_weight += hw * v3; grad_w_weight -= lh * v3; atomicAdd(grad_value+ptr3, w3*top_grad_value); } scalar_t v4 = 0; if (h_high <= height - 1 && w_high <= width - 1) { const int ptr4 = h_high_ptr_offset + w_high_ptr_offset + base_ptr; v4 = bottom_data[ptr4]; grad_h_weight += lw * v4; grad_w_weight += lh * v4; atomicAdd(grad_value+ptr4, w4*top_grad_value); } const scalar_t val = (w1 * v1 + w2 * v2 + w3 * v3 + w4 * v4); atomicAdd(grad_attn_weight, top_grad * val); atomicAdd(grad_sampling_loc, width * grad_w_weight * top_grad_value); atomicAdd(grad_sampling_loc + 1, height * grad_h_weight * top_grad_value); } template __global__ void ms_deformable_im2col_gpu_kernel(const int n, const scalar_t *data_value, const int64_t *data_spatial_shapes, const int64_t *data_level_start_index, const scalar_t *data_sampling_loc, const scalar_t *data_attn_weight, const int batch_size, const int spatial_size, const int num_heads, const int channels, const int num_levels, const int num_query, const int num_point, scalar_t *data_col) { CUDA_KERNEL_LOOP(index, n) { int _temp = index; const int c_col = _temp % channels; _temp /= channels; const int sampling_index = _temp; const int m_col = _temp % num_heads; _temp /= num_heads; const int q_col = _temp % num_query; _temp /= num_query; const int b_col = _temp; scalar_t *data_col_ptr = data_col + index; int data_weight_ptr = sampling_index * num_levels * num_point; int data_loc_w_ptr = data_weight_ptr << 1; const int qid_stride = num_heads * channels; const int data_value_ptr_init_offset = b_col * spatial_size * qid_stride; scalar_t col = 0; for (int l_col=0; l_col < num_levels; ++l_col) { const int level_start_id = data_level_start_index[l_col]; const int spatial_h_ptr = l_col << 1; const int spatial_h = data_spatial_shapes[spatial_h_ptr]; const int spatial_w = data_spatial_shapes[spatial_h_ptr + 1]; const scalar_t *data_value_ptr = data_value + (data_value_ptr_init_offset + level_start_id * qid_stride); for (int p_col=0; p_col < num_point; ++p_col) { const scalar_t loc_w = data_sampling_loc[data_loc_w_ptr]; const scalar_t loc_h = data_sampling_loc[data_loc_w_ptr + 1]; const scalar_t weight = data_attn_weight[data_weight_ptr]; const scalar_t h_im = loc_h * spatial_h - 0.5; const scalar_t w_im = loc_w * spatial_w - 0.5; if (h_im > -1 && w_im > -1 && h_im < spatial_h && w_im < spatial_w) { col += ms_deform_attn_im2col_bilinear(data_value_ptr, spatial_h, spatial_w, num_heads, channels, h_im, w_im, m_col, c_col) * weight; } data_weight_ptr += 1; data_loc_w_ptr += 2; } } *data_col_ptr = col; } } template __global__ void ms_deformable_col2im_gpu_kernel_shm_blocksize_aware_reduce_v1(const int n, const scalar_t *grad_col, const scalar_t *data_value, const int64_t *data_spatial_shapes, const int64_t *data_level_start_index, const scalar_t *data_sampling_loc, const scalar_t *data_attn_weight, const int batch_size, const int spatial_size, const int num_heads, const int channels, const int num_levels, const int num_query, const int num_point, scalar_t *grad_value, scalar_t *grad_sampling_loc, scalar_t *grad_attn_weight) { CUDA_KERNEL_LOOP(index, n) { __shared__ scalar_t cache_grad_sampling_loc[blockSize * 2]; __shared__ scalar_t cache_grad_attn_weight[blockSize]; unsigned int tid = threadIdx.x; int _temp = index; const int c_col = _temp % channels; _temp /= channels; const int sampling_index = _temp; const int m_col = _temp % num_heads; _temp /= num_heads; const int q_col = _temp % num_query; _temp /= num_query; const int b_col = _temp; const scalar_t top_grad = grad_col[index]; int data_weight_ptr = sampling_index * num_levels * num_point; int data_loc_w_ptr = data_weight_ptr << 1; const int grad_sampling_ptr = data_weight_ptr; grad_sampling_loc += grad_sampling_ptr << 1; grad_attn_weight += grad_sampling_ptr; const int grad_weight_stride = 1; const int grad_loc_stride = 2; const int qid_stride = num_heads * channels; const int data_value_ptr_init_offset = b_col * spatial_size * qid_stride; for (int l_col=0; l_col < num_levels; ++l_col) { const int level_start_id = data_level_start_index[l_col]; const int spatial_h_ptr = l_col << 1; const int spatial_h = data_spatial_shapes[spatial_h_ptr]; const int spatial_w = data_spatial_shapes[spatial_h_ptr + 1]; const int value_ptr_offset = data_value_ptr_init_offset + level_start_id * qid_stride; const scalar_t *data_value_ptr = data_value + value_ptr_offset; scalar_t *grad_value_ptr = grad_value + value_ptr_offset; for (int p_col=0; p_col < num_point; ++p_col) { const scalar_t loc_w = data_sampling_loc[data_loc_w_ptr]; const scalar_t loc_h = data_sampling_loc[data_loc_w_ptr + 1]; const scalar_t weight = data_attn_weight[data_weight_ptr]; const scalar_t h_im = loc_h * spatial_h - 0.5; const scalar_t w_im = loc_w * spatial_w - 0.5; *(cache_grad_sampling_loc+(threadIdx.x << 1)) = 0; *(cache_grad_sampling_loc+((threadIdx.x << 1) + 1)) = 0; *(cache_grad_attn_weight+threadIdx.x)=0; if (h_im > -1 && w_im > -1 && h_im < spatial_h && w_im < spatial_w) { ms_deform_attn_col2im_bilinear( data_value_ptr, spatial_h, spatial_w, num_heads, channels, h_im, w_im, m_col, c_col, top_grad, weight, grad_value_ptr, cache_grad_sampling_loc+(threadIdx.x << 1), cache_grad_attn_weight+threadIdx.x); } __syncthreads(); if (tid == 0) { scalar_t _grad_w=cache_grad_sampling_loc[0], _grad_h=cache_grad_sampling_loc[1], _grad_a=cache_grad_attn_weight[0]; int sid=2; for (unsigned int tid = 1; tid < blockSize; ++tid) { _grad_w += cache_grad_sampling_loc[sid]; _grad_h += cache_grad_sampling_loc[sid + 1]; _grad_a += cache_grad_attn_weight[tid]; sid += 2; } *grad_sampling_loc = _grad_w; *(grad_sampling_loc + 1) = _grad_h; *grad_attn_weight = _grad_a; } __syncthreads(); data_weight_ptr += 1; data_loc_w_ptr += 2; grad_attn_weight += grad_weight_stride; grad_sampling_loc += grad_loc_stride; } } } } template __global__ void ms_deformable_col2im_gpu_kernel_shm_blocksize_aware_reduce_v2(const int n, const scalar_t *grad_col, const scalar_t *data_value, const int64_t *data_spatial_shapes, const int64_t *data_level_start_index, const scalar_t *data_sampling_loc, const scalar_t *data_attn_weight, const int batch_size, const int spatial_size, const int num_heads, const int channels, const int num_levels, const int num_query, const int num_point, scalar_t *grad_value, scalar_t *grad_sampling_loc, scalar_t *grad_attn_weight) { CUDA_KERNEL_LOOP(index, n) { __shared__ scalar_t cache_grad_sampling_loc[blockSize * 2]; __shared__ scalar_t cache_grad_attn_weight[blockSize]; unsigned int tid = threadIdx.x; int _temp = index; const int c_col = _temp % channels; _temp /= channels; const int sampling_index = _temp; const int m_col = _temp % num_heads; _temp /= num_heads; const int q_col = _temp % num_query; _temp /= num_query; const int b_col = _temp; const scalar_t top_grad = grad_col[index]; int data_weight_ptr = sampling_index * num_levels * num_point; int data_loc_w_ptr = data_weight_ptr << 1; const int grad_sampling_ptr = data_weight_ptr; grad_sampling_loc += grad_sampling_ptr << 1; grad_attn_weight += grad_sampling_ptr; const int grad_weight_stride = 1; const int grad_loc_stride = 2; const int qid_stride = num_heads * channels; const int data_value_ptr_init_offset = b_col * spatial_size * qid_stride; for (int l_col=0; l_col < num_levels; ++l_col) { const int level_start_id = data_level_start_index[l_col]; const int spatial_h_ptr = l_col << 1; const int spatial_h = data_spatial_shapes[spatial_h_ptr]; const int spatial_w = data_spatial_shapes[spatial_h_ptr + 1]; const int value_ptr_offset = data_value_ptr_init_offset + level_start_id * qid_stride; const scalar_t *data_value_ptr = data_value + value_ptr_offset; scalar_t *grad_value_ptr = grad_value + value_ptr_offset; for (int p_col=0; p_col < num_point; ++p_col) { const scalar_t loc_w = data_sampling_loc[data_loc_w_ptr]; const scalar_t loc_h = data_sampling_loc[data_loc_w_ptr + 1]; const scalar_t weight = data_attn_weight[data_weight_ptr]; const scalar_t h_im = loc_h * spatial_h - 0.5; const scalar_t w_im = loc_w * spatial_w - 0.5; *(cache_grad_sampling_loc+(threadIdx.x << 1)) = 0; *(cache_grad_sampling_loc+((threadIdx.x << 1) + 1)) = 0; *(cache_grad_attn_weight+threadIdx.x)=0; if (h_im > -1 && w_im > -1 && h_im < spatial_h && w_im < spatial_w) { ms_deform_attn_col2im_bilinear( data_value_ptr, spatial_h, spatial_w, num_heads, channels, h_im, w_im, m_col, c_col, top_grad, weight, grad_value_ptr, cache_grad_sampling_loc+(threadIdx.x << 1), cache_grad_attn_weight+threadIdx.x); } __syncthreads(); for (unsigned int s=blockSize/2; s>0; s>>=1) { if (tid < s) { const unsigned int xid1 = tid << 1; const unsigned int xid2 = (tid + s) << 1; cache_grad_attn_weight[tid] += cache_grad_attn_weight[tid + s]; cache_grad_sampling_loc[xid1] += cache_grad_sampling_loc[xid2]; cache_grad_sampling_loc[xid1 + 1] += cache_grad_sampling_loc[xid2 + 1]; } __syncthreads(); } if (tid == 0) { *grad_sampling_loc = cache_grad_sampling_loc[0]; *(grad_sampling_loc + 1) = cache_grad_sampling_loc[1]; *grad_attn_weight = cache_grad_attn_weight[0]; } __syncthreads(); data_weight_ptr += 1; data_loc_w_ptr += 2; grad_attn_weight += grad_weight_stride; grad_sampling_loc += grad_loc_stride; } } } } template __global__ void ms_deformable_col2im_gpu_kernel_shm_reduce_v1(const int n, const scalar_t *grad_col, const scalar_t *data_value, const int64_t *data_spatial_shapes, const int64_t *data_level_start_index, const scalar_t *data_sampling_loc, const scalar_t *data_attn_weight, const int batch_size, const int spatial_size, const int num_heads, const int channels, const int num_levels, const int num_query, const int num_point, scalar_t *grad_value, scalar_t *grad_sampling_loc, scalar_t *grad_attn_weight) { CUDA_KERNEL_LOOP(index, n) { extern __shared__ int _s[]; scalar_t* cache_grad_sampling_loc = (scalar_t*)_s; scalar_t* cache_grad_attn_weight = cache_grad_sampling_loc + 2 * blockDim.x; unsigned int tid = threadIdx.x; int _temp = index; const int c_col = _temp % channels; _temp /= channels; const int sampling_index = _temp; const int m_col = _temp % num_heads; _temp /= num_heads; const int q_col = _temp % num_query; _temp /= num_query; const int b_col = _temp; const scalar_t top_grad = grad_col[index]; int data_weight_ptr = sampling_index * num_levels * num_point; int data_loc_w_ptr = data_weight_ptr << 1; const int grad_sampling_ptr = data_weight_ptr; grad_sampling_loc += grad_sampling_ptr << 1; grad_attn_weight += grad_sampling_ptr; const int grad_weight_stride = 1; const int grad_loc_stride = 2; const int qid_stride = num_heads * channels; const int data_value_ptr_init_offset = b_col * spatial_size * qid_stride; for (int l_col=0; l_col < num_levels; ++l_col) { const int level_start_id = data_level_start_index[l_col]; const int spatial_h_ptr = l_col << 1; const int spatial_h = data_spatial_shapes[spatial_h_ptr]; const int spatial_w = data_spatial_shapes[spatial_h_ptr + 1]; const int value_ptr_offset = data_value_ptr_init_offset + level_start_id * qid_stride; const scalar_t *data_value_ptr = data_value + value_ptr_offset; scalar_t *grad_value_ptr = grad_value + value_ptr_offset; for (int p_col=0; p_col < num_point; ++p_col) { const scalar_t loc_w = data_sampling_loc[data_loc_w_ptr]; const scalar_t loc_h = data_sampling_loc[data_loc_w_ptr + 1]; const scalar_t weight = data_attn_weight[data_weight_ptr]; const scalar_t h_im = loc_h * spatial_h - 0.5; const scalar_t w_im = loc_w * spatial_w - 0.5; *(cache_grad_sampling_loc+(threadIdx.x << 1)) = 0; *(cache_grad_sampling_loc+((threadIdx.x << 1) + 1)) = 0; *(cache_grad_attn_weight+threadIdx.x)=0; if (h_im > -1 && w_im > -1 && h_im < spatial_h && w_im < spatial_w) { ms_deform_attn_col2im_bilinear( data_value_ptr, spatial_h, spatial_w, num_heads, channels, h_im, w_im, m_col, c_col, top_grad, weight, grad_value_ptr, cache_grad_sampling_loc+(threadIdx.x << 1), cache_grad_attn_weight+threadIdx.x); } __syncthreads(); if (tid == 0) { scalar_t _grad_w=cache_grad_sampling_loc[0], _grad_h=cache_grad_sampling_loc[1], _grad_a=cache_grad_attn_weight[0]; int sid=2; for (unsigned int tid = 1; tid < blockDim.x; ++tid) { _grad_w += cache_grad_sampling_loc[sid]; _grad_h += cache_grad_sampling_loc[sid + 1]; _grad_a += cache_grad_attn_weight[tid]; sid += 2; } *grad_sampling_loc = _grad_w; *(grad_sampling_loc + 1) = _grad_h; *grad_attn_weight = _grad_a; } __syncthreads(); data_weight_ptr += 1; data_loc_w_ptr += 2; grad_attn_weight += grad_weight_stride; grad_sampling_loc += grad_loc_stride; } } } } template __global__ void ms_deformable_col2im_gpu_kernel_shm_reduce_v2(const int n, const scalar_t *grad_col, const scalar_t *data_value, const int64_t *data_spatial_shapes, const int64_t *data_level_start_index, const scalar_t *data_sampling_loc, const scalar_t *data_attn_weight, const int batch_size, const int spatial_size, const int num_heads, const int channels, const int num_levels, const int num_query, const int num_point, scalar_t *grad_value, scalar_t *grad_sampling_loc, scalar_t *grad_attn_weight) { CUDA_KERNEL_LOOP(index, n) { extern __shared__ int _s[]; scalar_t* cache_grad_sampling_loc = (scalar_t*)_s; scalar_t* cache_grad_attn_weight = cache_grad_sampling_loc + 2 * blockDim.x; unsigned int tid = threadIdx.x; int _temp = index; const int c_col = _temp % channels; _temp /= channels; const int sampling_index = _temp; const int m_col = _temp % num_heads; _temp /= num_heads; const int q_col = _temp % num_query; _temp /= num_query; const int b_col = _temp; const scalar_t top_grad = grad_col[index]; int data_weight_ptr = sampling_index * num_levels * num_point; int data_loc_w_ptr = data_weight_ptr << 1; const int grad_sampling_ptr = data_weight_ptr; grad_sampling_loc += grad_sampling_ptr << 1; grad_attn_weight += grad_sampling_ptr; const int grad_weight_stride = 1; const int grad_loc_stride = 2; const int qid_stride = num_heads * channels; const int data_value_ptr_init_offset = b_col * spatial_size * qid_stride; for (int l_col=0; l_col < num_levels; ++l_col) { const int level_start_id = data_level_start_index[l_col]; const int spatial_h_ptr = l_col << 1; const int spatial_h = data_spatial_shapes[spatial_h_ptr]; const int spatial_w = data_spatial_shapes[spatial_h_ptr + 1]; const int value_ptr_offset = data_value_ptr_init_offset + level_start_id * qid_stride; const scalar_t *data_value_ptr = data_value + value_ptr_offset; scalar_t *grad_value_ptr = grad_value + value_ptr_offset; for (int p_col=0; p_col < num_point; ++p_col) { const scalar_t loc_w = data_sampling_loc[data_loc_w_ptr]; const scalar_t loc_h = data_sampling_loc[data_loc_w_ptr + 1]; const scalar_t weight = data_attn_weight[data_weight_ptr]; const scalar_t h_im = loc_h * spatial_h - 0.5; const scalar_t w_im = loc_w * spatial_w - 0.5; *(cache_grad_sampling_loc+(threadIdx.x << 1)) = 0; *(cache_grad_sampling_loc+((threadIdx.x << 1) + 1)) = 0; *(cache_grad_attn_weight+threadIdx.x)=0; if (h_im > -1 && w_im > -1 && h_im < spatial_h && w_im < spatial_w) { ms_deform_attn_col2im_bilinear( data_value_ptr, spatial_h, spatial_w, num_heads, channels, h_im, w_im, m_col, c_col, top_grad, weight, grad_value_ptr, cache_grad_sampling_loc+(threadIdx.x << 1), cache_grad_attn_weight+threadIdx.x); } __syncthreads(); for (unsigned int s=blockDim.x/2, spre=blockDim.x; s>0; s>>=1, spre>>=1) { if (tid < s) { const unsigned int xid1 = tid << 1; const unsigned int xid2 = (tid + s) << 1; cache_grad_attn_weight[tid] += cache_grad_attn_weight[tid + s]; cache_grad_sampling_loc[xid1] += cache_grad_sampling_loc[xid2]; cache_grad_sampling_loc[xid1 + 1] += cache_grad_sampling_loc[xid2 + 1]; if (tid + (s << 1) < spre) { cache_grad_attn_weight[tid] += cache_grad_attn_weight[tid + (s << 1)]; cache_grad_sampling_loc[xid1] += cache_grad_sampling_loc[xid2 + (s << 1)]; cache_grad_sampling_loc[xid1 + 1] += cache_grad_sampling_loc[xid2 + 1 + (s << 1)]; } } __syncthreads(); } if (tid == 0) { *grad_sampling_loc = cache_grad_sampling_loc[0]; *(grad_sampling_loc + 1) = cache_grad_sampling_loc[1]; *grad_attn_weight = cache_grad_attn_weight[0]; } __syncthreads(); data_weight_ptr += 1; data_loc_w_ptr += 2; grad_attn_weight += grad_weight_stride; grad_sampling_loc += grad_loc_stride; } } } } template __global__ void ms_deformable_col2im_gpu_kernel_shm_reduce_v2_multi_blocks(const int n, const scalar_t *grad_col, const scalar_t *data_value, const int64_t *data_spatial_shapes, const int64_t *data_level_start_index, const scalar_t *data_sampling_loc, const scalar_t *data_attn_weight, const int batch_size, const int spatial_size, const int num_heads, const int channels, const int num_levels, const int num_query, const int num_point, scalar_t *grad_value, scalar_t *grad_sampling_loc, scalar_t *grad_attn_weight) { CUDA_KERNEL_LOOP(index, n) { extern __shared__ int _s[]; scalar_t* cache_grad_sampling_loc = (scalar_t*)_s; scalar_t* cache_grad_attn_weight = cache_grad_sampling_loc + 2 * blockDim.x; unsigned int tid = threadIdx.x; int _temp = index; const int c_col = _temp % channels; _temp /= channels; const int sampling_index = _temp; const int m_col = _temp % num_heads; _temp /= num_heads; const int q_col = _temp % num_query; _temp /= num_query; const int b_col = _temp; const scalar_t top_grad = grad_col[index]; int data_weight_ptr = sampling_index * num_levels * num_point; int data_loc_w_ptr = data_weight_ptr << 1; const int grad_sampling_ptr = data_weight_ptr; grad_sampling_loc += grad_sampling_ptr << 1; grad_attn_weight += grad_sampling_ptr; const int grad_weight_stride = 1; const int grad_loc_stride = 2; const int qid_stride = num_heads * channels; const int data_value_ptr_init_offset = b_col * spatial_size * qid_stride; for (int l_col=0; l_col < num_levels; ++l_col) { const int level_start_id = data_level_start_index[l_col]; const int spatial_h_ptr = l_col << 1; const int spatial_h = data_spatial_shapes[spatial_h_ptr]; const int spatial_w = data_spatial_shapes[spatial_h_ptr + 1]; const int value_ptr_offset = data_value_ptr_init_offset + level_start_id * qid_stride; const scalar_t *data_value_ptr = data_value + value_ptr_offset; scalar_t *grad_value_ptr = grad_value + value_ptr_offset; for (int p_col=0; p_col < num_point; ++p_col) { const scalar_t loc_w = data_sampling_loc[data_loc_w_ptr]; const scalar_t loc_h = data_sampling_loc[data_loc_w_ptr + 1]; const scalar_t weight = data_attn_weight[data_weight_ptr]; const scalar_t h_im = loc_h * spatial_h - 0.5; const scalar_t w_im = loc_w * spatial_w - 0.5; *(cache_grad_sampling_loc+(threadIdx.x << 1)) = 0; *(cache_grad_sampling_loc+((threadIdx.x << 1) + 1)) = 0; *(cache_grad_attn_weight+threadIdx.x)=0; if (h_im > -1 && w_im > -1 && h_im < spatial_h && w_im < spatial_w) { ms_deform_attn_col2im_bilinear( data_value_ptr, spatial_h, spatial_w, num_heads, channels, h_im, w_im, m_col, c_col, top_grad, weight, grad_value_ptr, cache_grad_sampling_loc+(threadIdx.x << 1), cache_grad_attn_weight+threadIdx.x); } __syncthreads(); for (unsigned int s=blockDim.x/2, spre=blockDim.x; s>0; s>>=1, spre>>=1) { if (tid < s) { const unsigned int xid1 = tid << 1; const unsigned int xid2 = (tid + s) << 1; cache_grad_attn_weight[tid] += cache_grad_attn_weight[tid + s]; cache_grad_sampling_loc[xid1] += cache_grad_sampling_loc[xid2]; cache_grad_sampling_loc[xid1 + 1] += cache_grad_sampling_loc[xid2 + 1]; if (tid + (s << 1) < spre) { cache_grad_attn_weight[tid] += cache_grad_attn_weight[tid + (s << 1)]; cache_grad_sampling_loc[xid1] += cache_grad_sampling_loc[xid2 + (s << 1)]; cache_grad_sampling_loc[xid1 + 1] += cache_grad_sampling_loc[xid2 + 1 + (s << 1)]; } } __syncthreads(); } if (tid == 0) { atomicAdd(grad_sampling_loc, cache_grad_sampling_loc[0]); atomicAdd(grad_sampling_loc + 1, cache_grad_sampling_loc[1]); atomicAdd(grad_attn_weight, cache_grad_attn_weight[0]); } __syncthreads(); data_weight_ptr += 1; data_loc_w_ptr += 2; grad_attn_weight += grad_weight_stride; grad_sampling_loc += grad_loc_stride; } } } } template __global__ void ms_deformable_col2im_gpu_kernel_gm(const int n, const scalar_t *grad_col, const scalar_t *data_value, const int64_t *data_spatial_shapes, const int64_t *data_level_start_index, const scalar_t *data_sampling_loc, const scalar_t *data_attn_weight, const int batch_size, const int spatial_size, const int num_heads, const int channels, const int num_levels, const int num_query, const int num_point, scalar_t *grad_value, scalar_t *grad_sampling_loc, scalar_t *grad_attn_weight) { CUDA_KERNEL_LOOP(index, n) { int _temp = index; const int c_col = _temp % channels; _temp /= channels; const int sampling_index = _temp; const int m_col = _temp % num_heads; _temp /= num_heads; const int q_col = _temp % num_query; _temp /= num_query; const int b_col = _temp; const scalar_t top_grad = grad_col[index]; int data_weight_ptr = sampling_index * num_levels * num_point; int data_loc_w_ptr = data_weight_ptr << 1; const int grad_sampling_ptr = data_weight_ptr; grad_sampling_loc += grad_sampling_ptr << 1; grad_attn_weight += grad_sampling_ptr; const int grad_weight_stride = 1; const int grad_loc_stride = 2; const int qid_stride = num_heads * channels; const int data_value_ptr_init_offset = b_col * spatial_size * qid_stride; for (int l_col=0; l_col < num_levels; ++l_col) { const int level_start_id = data_level_start_index[l_col]; const int spatial_h_ptr = l_col << 1; const int spatial_h = data_spatial_shapes[spatial_h_ptr]; const int spatial_w = data_spatial_shapes[spatial_h_ptr + 1]; const int value_ptr_offset = data_value_ptr_init_offset + level_start_id * qid_stride; const scalar_t *data_value_ptr = data_value + value_ptr_offset; scalar_t *grad_value_ptr = grad_value + value_ptr_offset; for (int p_col=0; p_col < num_point; ++p_col) { const scalar_t loc_w = data_sampling_loc[data_loc_w_ptr]; const scalar_t loc_h = data_sampling_loc[data_loc_w_ptr + 1]; const scalar_t weight = data_attn_weight[data_weight_ptr]; const scalar_t h_im = loc_h * spatial_h - 0.5; const scalar_t w_im = loc_w * spatial_w - 0.5; if (h_im > -1 && w_im > -1 && h_im < spatial_h && w_im < spatial_w) { ms_deform_attn_col2im_bilinear_gm( data_value_ptr, spatial_h, spatial_w, num_heads, channels, h_im, w_im, m_col, c_col, top_grad, weight, grad_value_ptr, grad_sampling_loc, grad_attn_weight); } data_weight_ptr += 1; data_loc_w_ptr += 2; grad_attn_weight += grad_weight_stride; grad_sampling_loc += grad_loc_stride; } } } } template void ms_deformable_im2col_cuda(cudaStream_t stream, const scalar_t* data_value, const int64_t* data_spatial_shapes, const int64_t* data_level_start_index, const scalar_t* data_sampling_loc, const scalar_t* data_attn_weight, const int batch_size, const int spatial_size, const int num_heads, const int channels, const int num_levels, const int num_query, const int num_point, scalar_t* data_col) { const int num_kernels = batch_size * num_query * num_heads * channels; const int num_actual_kernels = batch_size * num_query * num_heads * channels; const int num_threads = CUDA_NUM_THREADS; ms_deformable_im2col_gpu_kernel <<>>( num_kernels, data_value, data_spatial_shapes, data_level_start_index, data_sampling_loc, data_attn_weight, batch_size, spatial_size, num_heads, channels, num_levels, num_query, num_point, data_col); cudaError_t err = cudaGetLastError(); if (err != cudaSuccess) { printf("error in ms_deformable_im2col_cuda: %s\n", cudaGetErrorString(err)); } } template void ms_deformable_col2im_cuda(cudaStream_t stream, const scalar_t* grad_col, const scalar_t* data_value, const int64_t * data_spatial_shapes, const int64_t * data_level_start_index, const scalar_t * data_sampling_loc, const scalar_t * data_attn_weight, const int batch_size, const int spatial_size, const int num_heads, const int channels, const int num_levels, const int num_query, const int num_point, scalar_t* grad_value, scalar_t* grad_sampling_loc, scalar_t* grad_attn_weight) { const int num_threads = (channels > CUDA_NUM_THREADS)?CUDA_NUM_THREADS:channels; const int num_kernels = batch_size * num_query * num_heads * channels; const int num_actual_kernels = batch_size * num_query * num_heads * channels; if (channels > 1024) { if ((channels & 1023) == 0) { ms_deformable_col2im_gpu_kernel_shm_reduce_v2_multi_blocks <<>>( num_kernels, grad_col, data_value, data_spatial_shapes, data_level_start_index, data_sampling_loc, data_attn_weight, batch_size, spatial_size, num_heads, channels, num_levels, num_query, num_point, grad_value, grad_sampling_loc, grad_attn_weight); } else { ms_deformable_col2im_gpu_kernel_gm <<>>( num_kernels, grad_col, data_value, data_spatial_shapes, data_level_start_index, data_sampling_loc, data_attn_weight, batch_size, spatial_size, num_heads, channels, num_levels, num_query, num_point, grad_value, grad_sampling_loc, grad_attn_weight); } } else{ switch(channels) { case 1: ms_deformable_col2im_gpu_kernel_shm_blocksize_aware_reduce_v1 <<>>( num_kernels, grad_col, data_value, data_spatial_shapes, data_level_start_index, data_sampling_loc, data_attn_weight, batch_size, spatial_size, num_heads, channels, num_levels, num_query, num_point, grad_value, grad_sampling_loc, grad_attn_weight); break; case 2: ms_deformable_col2im_gpu_kernel_shm_blocksize_aware_reduce_v1 <<>>( num_kernels, grad_col, data_value, data_spatial_shapes, data_level_start_index, data_sampling_loc, data_attn_weight, batch_size, spatial_size, num_heads, channels, num_levels, num_query, num_point, grad_value, grad_sampling_loc, grad_attn_weight); break; case 4: ms_deformable_col2im_gpu_kernel_shm_blocksize_aware_reduce_v1 <<>>( num_kernels, grad_col, data_value, data_spatial_shapes, data_level_start_index, data_sampling_loc, data_attn_weight, batch_size, spatial_size, num_heads, channels, num_levels, num_query, num_point, grad_value, grad_sampling_loc, grad_attn_weight); break; case 8: ms_deformable_col2im_gpu_kernel_shm_blocksize_aware_reduce_v1 <<>>( num_kernels, grad_col, data_value, data_spatial_shapes, data_level_start_index, data_sampling_loc, data_attn_weight, batch_size, spatial_size, num_heads, channels, num_levels, num_query, num_point, grad_value, grad_sampling_loc, grad_attn_weight); break; case 16: ms_deformable_col2im_gpu_kernel_shm_blocksize_aware_reduce_v1 <<>>( num_kernels, grad_col, data_value, data_spatial_shapes, data_level_start_index, data_sampling_loc, data_attn_weight, batch_size, spatial_size, num_heads, channels, num_levels, num_query, num_point, grad_value, grad_sampling_loc, grad_attn_weight); break; case 32: ms_deformable_col2im_gpu_kernel_shm_blocksize_aware_reduce_v1 <<>>( num_kernels, grad_col, data_value, data_spatial_shapes, data_level_start_index, data_sampling_loc, data_attn_weight, batch_size, spatial_size, num_heads, channels, num_levels, num_query, num_point, grad_value, grad_sampling_loc, grad_attn_weight); break; case 64: ms_deformable_col2im_gpu_kernel_shm_blocksize_aware_reduce_v2 <<>>( num_kernels, grad_col, data_value, data_spatial_shapes, data_level_start_index, data_sampling_loc, data_attn_weight, batch_size, spatial_size, num_heads, channels, num_levels, num_query, num_point, grad_value, grad_sampling_loc, grad_attn_weight); break; case 128: ms_deformable_col2im_gpu_kernel_shm_blocksize_aware_reduce_v2 <<>>( num_kernels, grad_col, data_value, data_spatial_shapes, data_level_start_index, data_sampling_loc, data_attn_weight, batch_size, spatial_size, num_heads, channels, num_levels, num_query, num_point, grad_value, grad_sampling_loc, grad_attn_weight); break; case 256: ms_deformable_col2im_gpu_kernel_shm_blocksize_aware_reduce_v2 <<>>( num_kernels, grad_col, data_value, data_spatial_shapes, data_level_start_index, data_sampling_loc, data_attn_weight, batch_size, spatial_size, num_heads, channels, num_levels, num_query, num_point, grad_value, grad_sampling_loc, grad_attn_weight); break; case 512: ms_deformable_col2im_gpu_kernel_shm_blocksize_aware_reduce_v2 <<>>( num_kernels, grad_col, data_value, data_spatial_shapes, data_level_start_index, data_sampling_loc, data_attn_weight, batch_size, spatial_size, num_heads, channels, num_levels, num_query, num_point, grad_value, grad_sampling_loc, grad_attn_weight); break; case 1024: ms_deformable_col2im_gpu_kernel_shm_blocksize_aware_reduce_v2 <<>>( num_kernels, grad_col, data_value, data_spatial_shapes, data_level_start_index, data_sampling_loc, data_attn_weight, batch_size, spatial_size, num_heads, channels, num_levels, num_query, num_point, grad_value, grad_sampling_loc, grad_attn_weight); break; default: if (channels < 64) { ms_deformable_col2im_gpu_kernel_shm_reduce_v1 <<>>( num_kernels, grad_col, data_value, data_spatial_shapes, data_level_start_index, data_sampling_loc, data_attn_weight, batch_size, spatial_size, num_heads, channels, num_levels, num_query, num_point, grad_value, grad_sampling_loc, grad_attn_weight); } else { ms_deformable_col2im_gpu_kernel_shm_reduce_v2 <<>>( num_kernels, grad_col, data_value, data_spatial_shapes, data_level_start_index, data_sampling_loc, data_attn_weight, batch_size, spatial_size, num_heads, channels, num_levels, num_query, num_point, grad_value, grad_sampling_loc, grad_attn_weight); } } } cudaError_t err = cudaGetLastError(); if (err != cudaSuccess) { printf("error in ms_deformable_col2im_cuda: %s\n", cudaGetErrorString(err)); } } ================================================ FILE: models/ops/src/ms_deform_attn.h ================================================ /*! ************************************************************************************************** * Deformable DETR * Copyright (c) 2020 SenseTime. All Rights Reserved. * Licensed under the Apache License, Version 2.0 [see LICENSE for details] ************************************************************************************************** * Modified from https://github.com/chengdazhi/Deformable-Convolution-V2-PyTorch/tree/pytorch_1.0.0 ************************************************************************************************** */ #pragma once #include "cpu/ms_deform_attn_cpu.h" #ifdef WITH_CUDA #include "cuda/ms_deform_attn_cuda.h" #endif at::Tensor ms_deform_attn_forward( const at::Tensor &value, const at::Tensor &spatial_shapes, const at::Tensor &level_start_index, const at::Tensor &sampling_loc, const at::Tensor &attn_weight, const int im2col_step) { if (value.type().is_cuda()) { #ifdef WITH_CUDA return ms_deform_attn_cuda_forward( value, spatial_shapes, level_start_index, sampling_loc, attn_weight, im2col_step); #else AT_ERROR("Not compiled with GPU support"); #endif } AT_ERROR("Not implemented on the CPU"); } std::vector ms_deform_attn_backward( const at::Tensor &value, const at::Tensor &spatial_shapes, const at::Tensor &level_start_index, const at::Tensor &sampling_loc, const at::Tensor &attn_weight, const at::Tensor &grad_output, const int im2col_step) { if (value.type().is_cuda()) { #ifdef WITH_CUDA return ms_deform_attn_cuda_backward( value, spatial_shapes, level_start_index, sampling_loc, attn_weight, grad_output, im2col_step); #else AT_ERROR("Not compiled with GPU support"); #endif } AT_ERROR("Not implemented on the CPU"); } ================================================ FILE: models/ops/src/vision.cpp ================================================ /*! ************************************************************************************************** * Deformable DETR * Copyright (c) 2020 SenseTime. All Rights Reserved. * Licensed under the Apache License, Version 2.0 [see LICENSE for details] ************************************************************************************************** * Modified from https://github.com/chengdazhi/Deformable-Convolution-V2-PyTorch/tree/pytorch_1.0.0 ************************************************************************************************** */ #include "ms_deform_attn.h" PYBIND11_MODULE(TORCH_EXTENSION_NAME, m) { m.def("ms_deform_attn_forward", &ms_deform_attn_forward, "ms_deform_attn_forward"); m.def("ms_deform_attn_backward", &ms_deform_attn_backward, "ms_deform_attn_backward"); } ================================================ FILE: models/ops/test.py ================================================ # ------------------------------------------------------------------------------------------------ # Deformable DETR # Copyright (c) 2020 SenseTime. All Rights Reserved. # Licensed under the Apache License, Version 2.0 [see LICENSE for details] # ------------------------------------------------------------------------------------------------ # Modified from https://github.com/chengdazhi/Deformable-Convolution-V2-PyTorch/tree/pytorch_1.0.0 # ------------------------------------------------------------------------------------------------ from __future__ import absolute_import from __future__ import print_function from __future__ import division import time import torch import torch.nn as nn from torch.autograd import gradcheck from functions.ms_deform_attn_func import MSDeformAttnFunction, ms_deform_attn_core_pytorch N, M, D = 1, 2, 2 Lq, L, P = 2, 2, 2 shapes = torch.as_tensor([(6, 4), (3, 2)], dtype=torch.long).cuda() level_start_index = torch.cat((shapes.new_zeros((1, )), shapes.prod(1).cumsum(0)[:-1])) S = sum([(H*W).item() for H, W in shapes]) torch.manual_seed(3) @torch.no_grad() def check_forward_equal_with_pytorch_double(): value = torch.rand(N, S, M, D).cuda() * 0.01 sampling_locations = torch.rand(N, Lq, M, L, P, 2).cuda() attention_weights = torch.rand(N, Lq, M, L, P).cuda() + 1e-5 attention_weights /= attention_weights.sum(-1, keepdim=True).sum(-2, keepdim=True) im2col_step = 2 output_pytorch = ms_deform_attn_core_pytorch(value.double(), shapes, sampling_locations.double(), attention_weights.double()).detach().cpu() output_cuda = MSDeformAttnFunction.apply(value.double(), shapes, level_start_index, sampling_locations.double(), attention_weights.double(), im2col_step).detach().cpu() fwdok = torch.allclose(output_cuda, output_pytorch) max_abs_err = (output_cuda - output_pytorch).abs().max() max_rel_err = ((output_cuda - output_pytorch).abs() / output_pytorch.abs()).max() print(f'* {fwdok} check_forward_equal_with_pytorch_double: max_abs_err {max_abs_err:.2e} max_rel_err {max_rel_err:.2e}') @torch.no_grad() def check_forward_equal_with_pytorch_float(): value = torch.rand(N, S, M, D).cuda() * 0.01 sampling_locations = torch.rand(N, Lq, M, L, P, 2).cuda() attention_weights = torch.rand(N, Lq, M, L, P).cuda() + 1e-5 attention_weights /= attention_weights.sum(-1, keepdim=True).sum(-2, keepdim=True) im2col_step = 2 output_pytorch = ms_deform_attn_core_pytorch(value, shapes, sampling_locations, attention_weights).detach().cpu() output_cuda = MSDeformAttnFunction.apply(value, shapes, level_start_index, sampling_locations, attention_weights, im2col_step).detach().cpu() fwdok = torch.allclose(output_cuda, output_pytorch, rtol=1e-2, atol=1e-3) max_abs_err = (output_cuda - output_pytorch).abs().max() max_rel_err = ((output_cuda - output_pytorch).abs() / output_pytorch.abs()).max() print(f'* {fwdok} check_forward_equal_with_pytorch_float: max_abs_err {max_abs_err:.2e} max_rel_err {max_rel_err:.2e}') def check_gradient_numerical(channels=4, grad_value=True, grad_sampling_loc=True, grad_attn_weight=True): value = torch.rand(N, S, M, channels).cuda() * 0.01 sampling_locations = torch.rand(N, Lq, M, L, P, 2).cuda() attention_weights = torch.rand(N, Lq, M, L, P).cuda() + 1e-5 attention_weights /= attention_weights.sum(-1, keepdim=True).sum(-2, keepdim=True) im2col_step = 2 func = MSDeformAttnFunction.apply value.requires_grad = grad_value sampling_locations.requires_grad = grad_sampling_loc attention_weights.requires_grad = grad_attn_weight gradok = gradcheck(func, (value.double(), shapes, level_start_index, sampling_locations.double(), attention_weights.double(), im2col_step)) print(f'* {gradok} check_gradient_numerical(D={channels})') if __name__ == '__main__': check_forward_equal_with_pytorch_double() check_forward_equal_with_pytorch_float() for channels in [30, 32, 64, 71, 1025, 2048, 3096]: check_gradient_numerical(channels, True, True, True) ================================================ FILE: models/position_encoding.py ================================================ # ------------------------------------------------------------------------------------ # Sparse DETR # Copyright (c) 2021 KakaoBrain. All Rights Reserved. # Licensed under the Apache License, Version 2.0 [see LICENSE for details] # ------------------------------------------------------------------------------------ # Modified from Deformable DETR (https://github.com/fundamentalvision/Deformable-DETR) # Copyright (c) 2020 SenseTime. All Rights Reserved. # ------------------------------------------------------------------------------------ # Modified from DETR (https://github.com/facebookresearch/detr) # Copyright (c) Facebook, Inc. and its affiliates. All Rights Reserved # ------------------------------------------------------------------------------------ """ Various positional encodings for the transformer. """ import math import torch from torch import nn from util.misc import NestedTensor class PositionEmbeddingSine(nn.Module): """ This is a more standard version of the position embedding, very similar to the one used by the Attention is all you need paper, generalized to work on images. """ def __init__(self, num_pos_feats=64, temperature=10000, normalize=False, scale=None): super().__init__() self.num_pos_feats = num_pos_feats self.temperature = temperature self.normalize = normalize if scale is not None and normalize is False: raise ValueError("normalize should be True if scale is passed") if scale is None: scale = 2 * math.pi self.scale = scale def forward(self, tensor_list: NestedTensor): x = tensor_list.tensors mask = tensor_list.mask assert mask is not None not_mask = ~mask y_embed = not_mask.cumsum(1, dtype=torch.float32) x_embed = not_mask.cumsum(2, dtype=torch.float32) if self.normalize: eps = 1e-6 y_embed = (y_embed - 0.5) / (y_embed[:, -1:, :] + eps) * self.scale x_embed = (x_embed - 0.5) / (x_embed[:, :, -1:] + eps) * self.scale dim_t = torch.arange(self.num_pos_feats, dtype=torch.float32, device=x.device) dim_t = self.temperature ** (2 * (dim_t // 2) / self.num_pos_feats) pos_x = x_embed[:, :, :, None] / dim_t pos_y = y_embed[:, :, :, None] / dim_t pos_x = torch.stack((pos_x[:, :, :, 0::2].sin(), pos_x[:, :, :, 1::2].cos()), dim=4).flatten(3) pos_y = torch.stack((pos_y[:, :, :, 0::2].sin(), pos_y[:, :, :, 1::2].cos()), dim=4).flatten(3) pos = torch.cat((pos_y, pos_x), dim=3).permute(0, 3, 1, 2) return pos class PositionEmbeddingLearned(nn.Module): """ Absolute pos embedding, learned. """ def __init__(self, num_pos_feats=256): super().__init__() self.row_embed = nn.Embedding(50, num_pos_feats) self.col_embed = nn.Embedding(50, num_pos_feats) self.reset_parameters() def reset_parameters(self): nn.init.uniform_(self.row_embed.weight) nn.init.uniform_(self.col_embed.weight) def forward(self, tensor_list: NestedTensor): x = tensor_list.tensors h, w = x.shape[-2:] i = torch.arange(w, device=x.device) j = torch.arange(h, device=x.device) x_emb = self.col_embed(i) y_emb = self.row_embed(j) pos = torch.cat([ x_emb.unsqueeze(0).repeat(h, 1, 1), y_emb.unsqueeze(1).repeat(1, w, 1), ], dim=-1).permute(2, 0, 1).unsqueeze(0).repeat(x.shape[0], 1, 1, 1) return pos def build_position_encoding(args): N_steps = args.hidden_dim // 2 if args.position_embedding in ('v2', 'sine'): # TODO find a better way of exposing other arguments position_embedding = PositionEmbeddingSine(N_steps, normalize=True) elif args.position_embedding in ('v3', 'learned'): position_embedding = PositionEmbeddingLearned(N_steps) else: raise ValueError(f"not supported {args.position_embedding}") return position_embedding ================================================ FILE: models/segmentation.py ================================================ # ------------------------------------------------------------------------------------ # Sparse DETR # Copyright (c) 2021 KakaoBrain. All Rights Reserved. # Licensed under the Apache License, Version 2.0 [see LICENSE for details] # ------------------------------------------------------------------------------------ # Modified from Deformable DETR (https://github.com/fundamentalvision/Deformable-DETR) # Copyright (c) 2020 SenseTime. All Rights Reserved. # ------------------------------------------------------------------------------------ # Modified from DETR (https://github.com/facebookresearch/detr) # Copyright (c) Facebook, Inc. and its affiliates. All Rights Reserved # ------------------------------------------------------------------------------------ """ This file provides the definition of the convolutional heads used to predict masks, as well as the losses """ import io from collections import defaultdict import torch import torch.nn as nn import torch.nn.functional as F from PIL import Image import util.box_ops as box_ops from util.misc import NestedTensor, interpolate, nested_tensor_from_tensor_list try: from panopticapi.utils import id2rgb, rgb2id except ImportError: pass class DETRsegm(nn.Module): def __init__(self, detr, freeze_detr=False): super().__init__() self.detr = detr if freeze_detr: for p in self.parameters(): p.requires_grad_(False) hidden_dim, nheads = detr.transformer.d_model, detr.transformer.nhead self.bbox_attention = MHAttentionMap(hidden_dim, hidden_dim, nheads, dropout=0) self.mask_head = MaskHeadSmallConv(hidden_dim + nheads, [1024, 512, 256], hidden_dim) def forward(self, samples: NestedTensor): if not isinstance(samples, NestedTensor): samples = nested_tensor_from_tensor_list(samples) features, pos = self.detr.backbone(samples) bs = features[-1].tensors.shape[0] src, mask = features[-1].decompose() src_proj = self.detr.input_proj(src) hs, memory = self.detr.transformer(src_proj, mask, self.detr.query_embed.weight, pos[-1]) outputs_class = self.detr.class_embed(hs) outputs_coord = self.detr.bbox_embed(hs).sigmoid() out = {"pred_logits": outputs_class[-1], "pred_boxes": outputs_coord[-1]} if self.detr.aux_loss: out["aux_outputs"] = [ {"pred_logits": a, "pred_boxes": b} for a, b in zip(outputs_class[:-1], outputs_coord[:-1]) ] # FIXME h_boxes takes the last one computed, keep this in mind bbox_mask = self.bbox_attention(hs[-1], memory, mask=mask) seg_masks = self.mask_head(src_proj, bbox_mask, [features[2].tensors, features[1].tensors, features[0].tensors]) outputs_seg_masks = seg_masks.view(bs, self.detr.num_queries, seg_masks.shape[-2], seg_masks.shape[-1]) out["pred_masks"] = outputs_seg_masks return out class MaskHeadSmallConv(nn.Module): """ Simple convolutional head, using group norm. Upsampling is done using a FPN approach """ def __init__(self, dim, fpn_dims, context_dim): super().__init__() inter_dims = [dim, context_dim // 2, context_dim // 4, context_dim // 8, context_dim // 16, context_dim // 64] self.lay1 = torch.nn.Conv2d(dim, dim, 3, padding=1) self.gn1 = torch.nn.GroupNorm(8, dim) self.lay2 = torch.nn.Conv2d(dim, inter_dims[1], 3, padding=1) self.gn2 = torch.nn.GroupNorm(8, inter_dims[1]) self.lay3 = torch.nn.Conv2d(inter_dims[1], inter_dims[2], 3, padding=1) self.gn3 = torch.nn.GroupNorm(8, inter_dims[2]) self.lay4 = torch.nn.Conv2d(inter_dims[2], inter_dims[3], 3, padding=1) self.gn4 = torch.nn.GroupNorm(8, inter_dims[3]) self.lay5 = torch.nn.Conv2d(inter_dims[3], inter_dims[4], 3, padding=1) self.gn5 = torch.nn.GroupNorm(8, inter_dims[4]) self.out_lay = torch.nn.Conv2d(inter_dims[4], 1, 3, padding=1) self.dim = dim self.adapter1 = torch.nn.Conv2d(fpn_dims[0], inter_dims[1], 1) self.adapter2 = torch.nn.Conv2d(fpn_dims[1], inter_dims[2], 1) self.adapter3 = torch.nn.Conv2d(fpn_dims[2], inter_dims[3], 1) for m in self.modules(): if isinstance(m, nn.Conv2d): nn.init.kaiming_uniform_(m.weight, a=1) nn.init.constant_(m.bias, 0) def forward(self, x, bbox_mask, fpns): def expand(tensor, length): return tensor.unsqueeze(1).repeat(1, int(length), 1, 1, 1).flatten(0, 1) x = torch.cat([expand(x, bbox_mask.shape[1]), bbox_mask.flatten(0, 1)], 1) x = self.lay1(x) x = self.gn1(x) x = F.relu(x) x = self.lay2(x) x = self.gn2(x) x = F.relu(x) cur_fpn = self.adapter1(fpns[0]) if cur_fpn.size(0) != x.size(0): cur_fpn = expand(cur_fpn, x.size(0) / cur_fpn.size(0)) x = cur_fpn + F.interpolate(x, size=cur_fpn.shape[-2:], mode="nearest") x = self.lay3(x) x = self.gn3(x) x = F.relu(x) cur_fpn = self.adapter2(fpns[1]) if cur_fpn.size(0) != x.size(0): cur_fpn = expand(cur_fpn, x.size(0) / cur_fpn.size(0)) x = cur_fpn + F.interpolate(x, size=cur_fpn.shape[-2:], mode="nearest") x = self.lay4(x) x = self.gn4(x) x = F.relu(x) cur_fpn = self.adapter3(fpns[2]) if cur_fpn.size(0) != x.size(0): cur_fpn = expand(cur_fpn, x.size(0) / cur_fpn.size(0)) x = cur_fpn + F.interpolate(x, size=cur_fpn.shape[-2:], mode="nearest") x = self.lay5(x) x = self.gn5(x) x = F.relu(x) x = self.out_lay(x) return x class MHAttentionMap(nn.Module): """This is a 2D attention module, which only returns the attention softmax (no multiplication by value)""" def __init__(self, query_dim, hidden_dim, num_heads, dropout=0, bias=True): super().__init__() self.num_heads = num_heads self.hidden_dim = hidden_dim self.dropout = nn.Dropout(dropout) self.q_linear = nn.Linear(query_dim, hidden_dim, bias=bias) self.k_linear = nn.Linear(query_dim, hidden_dim, bias=bias) nn.init.zeros_(self.k_linear.bias) nn.init.zeros_(self.q_linear.bias) nn.init.xavier_uniform_(self.k_linear.weight) nn.init.xavier_uniform_(self.q_linear.weight) self.normalize_fact = float(hidden_dim / self.num_heads) ** -0.5 def forward(self, q, k, mask=None): q = self.q_linear(q) k = F.conv2d(k, self.k_linear.weight.unsqueeze(-1).unsqueeze(-1), self.k_linear.bias) qh = q.view(q.shape[0], q.shape[1], self.num_heads, self.hidden_dim // self.num_heads) kh = k.view(k.shape[0], self.num_heads, self.hidden_dim // self.num_heads, k.shape[-2], k.shape[-1]) weights = torch.einsum("bqnc,bnchw->bqnhw", qh * self.normalize_fact, kh) if mask is not None: weights.masked_fill_(mask.unsqueeze(1).unsqueeze(1), float("-inf")) weights = F.softmax(weights.flatten(2), dim=-1).view_as(weights) weights = self.dropout(weights) return weights def dice_loss(inputs, targets, num_boxes): """ Compute the DICE loss, similar to generalized IOU for masks Args: inputs: A float tensor of arbitrary shape. The predictions for each example. targets: A float tensor with the same shape as inputs. Stores the binary classification label for each element in inputs (0 for the negative class and 1 for the positive class). """ inputs = inputs.sigmoid() inputs = inputs.flatten(1) numerator = 2 * (inputs * targets).sum(1) denominator = inputs.sum(-1) + targets.sum(-1) loss = 1 - (numerator + 1) / (denominator + 1) return loss.sum() / num_boxes def sigmoid_focal_loss(inputs, targets, num_boxes, alpha: float = 0.25, gamma: float = 2, idx=None): """ Loss used in RetinaNet for dense detection: https://arxiv.org/abs/1708.02002. Args: inputs: A float tensor of arbitrary shape. The predictions for each example. targets: A float tensor with the same shape as inputs. Stores the binary classification label for each element in inputs (0 for the negative class and 1 for the positive class). alpha: (optional) Weighting factor in range (0,1) to balance positive vs negative examples. Default = -1 (no weighting). gamma: Exponent of the modulating factor (1 - p_t) to balance easy vs hard examples. Returns: Loss tensor """ prob = inputs.sigmoid() ce_loss = F.binary_cross_entropy_with_logits(inputs, targets, reduction="none") p_t = prob * targets + (1 - prob) * (1 - targets) loss = ce_loss * ((1 - p_t) ** gamma) if alpha >= 0: alpha_t = alpha * targets + (1 - alpha) * (1 - targets) loss = alpha_t * loss if idx is not None: return loss[idx].mean(1).sum() / num_boxes return loss.mean(1).sum() / num_boxes class PostProcessSegm(nn.Module): def __init__(self, threshold=0.5): super().__init__() self.threshold = threshold @torch.no_grad() def forward(self, results, outputs, orig_target_sizes, max_target_sizes): assert len(orig_target_sizes) == len(max_target_sizes) max_h, max_w = max_target_sizes.max(0)[0].tolist() outputs_masks = outputs["pred_masks"].squeeze(2) outputs_masks = F.interpolate(outputs_masks, size=(max_h, max_w), mode="bilinear", align_corners=False) outputs_masks = (outputs_masks.sigmoid() > self.threshold).cpu() for i, (cur_mask, t, tt) in enumerate(zip(outputs_masks, max_target_sizes, orig_target_sizes)): img_h, img_w = t[0], t[1] results[i]["masks"] = cur_mask[:, :img_h, :img_w].unsqueeze(1) results[i]["masks"] = F.interpolate( results[i]["masks"].float(), size=tuple(tt.tolist()), mode="nearest" ).byte() return results class PostProcessPanoptic(nn.Module): """This class converts the output of the model to the final panoptic result, in the format expected by the coco panoptic API """ def __init__(self, is_thing_map, threshold=0.85): """ Parameters: is_thing_map: This is a whose keys are the class ids, and the values a boolean indicating whether the class is a thing (True) or a stuff (False) class threshold: confidence threshold: segments with confidence lower than this will be deleted """ super().__init__() self.threshold = threshold self.is_thing_map = is_thing_map def forward(self, outputs, processed_sizes, target_sizes=None): """ This function computes the panoptic prediction from the model's predictions. Parameters: outputs: This is a dict coming directly from the model. See the model doc for the content. processed_sizes: This is a list of tuples (or torch tensors) of sizes of the images that were passed to the model, ie the size after data augmentation but before batching. target_sizes: This is a list of tuples (or torch tensors) corresponding to the requested final size of each prediction. If left to None, it will default to the processed_sizes """ if target_sizes is None: target_sizes = processed_sizes assert len(processed_sizes) == len(target_sizes) out_logits, raw_masks, raw_boxes = outputs["pred_logits"], outputs["pred_masks"], outputs["pred_boxes"] assert len(out_logits) == len(raw_masks) == len(target_sizes) preds = [] def to_tuple(tup): if isinstance(tup, tuple): return tup return tuple(tup.cpu().tolist()) for cur_logits, cur_masks, cur_boxes, size, target_size in zip( out_logits, raw_masks, raw_boxes, processed_sizes, target_sizes ): # we filter empty queries and detection below threshold scores, labels = cur_logits.softmax(-1).max(-1) keep = labels.ne(outputs["pred_logits"].shape[-1] - 1) & (scores > self.threshold) cur_scores, cur_classes = cur_logits.softmax(-1).max(-1) cur_scores = cur_scores[keep] cur_classes = cur_classes[keep] cur_masks = cur_masks[keep] cur_masks = interpolate(cur_masks[None], to_tuple(size), mode="bilinear").squeeze(0) cur_boxes = box_ops.box_cxcywh_to_xyxy(cur_boxes[keep]) h, w = cur_masks.shape[-2:] assert len(cur_boxes) == len(cur_classes) # It may be that we have several predicted masks for the same stuff class. # In the following, we track the list of masks ids for each stuff class (they are merged later on) cur_masks = cur_masks.flatten(1) stuff_equiv_classes = defaultdict(lambda: []) for k, label in enumerate(cur_classes): if not self.is_thing_map[label.item()]: stuff_equiv_classes[label.item()].append(k) def get_ids_area(masks, scores, dedup=False): # This helper function creates the final panoptic segmentation image # It also returns the area of the masks that appears on the image m_id = masks.transpose(0, 1).softmax(-1) if m_id.shape[-1] == 0: # We didn't detect any mask :( m_id = torch.zeros((h, w), dtype=torch.long, device=m_id.device) else: m_id = m_id.argmax(-1).view(h, w) if dedup: # Merge the masks corresponding to the same stuff class for equiv in stuff_equiv_classes.values(): if len(equiv) > 1: for eq_id in equiv: m_id.masked_fill_(m_id.eq(eq_id), equiv[0]) final_h, final_w = to_tuple(target_size) seg_img = Image.fromarray(id2rgb(m_id.view(h, w).cpu().numpy())) seg_img = seg_img.resize(size=(final_w, final_h), resample=Image.NEAREST) np_seg_img = ( torch.ByteTensor(torch.ByteStorage.from_buffer(seg_img.tobytes())).view(final_h, final_w, 3).numpy() ) m_id = torch.from_numpy(rgb2id(np_seg_img)) area = [] for i in range(len(scores)): area.append(m_id.eq(i).sum().item()) return area, seg_img area, seg_img = get_ids_area(cur_masks, cur_scores, dedup=True) if cur_classes.numel() > 0: # We know filter empty masks as long as we find some while True: filtered_small = torch.as_tensor( [area[i] <= 4 for i, c in enumerate(cur_classes)], dtype=torch.bool, device=keep.device ) if filtered_small.any().item(): cur_scores = cur_scores[~filtered_small] cur_classes = cur_classes[~filtered_small] cur_masks = cur_masks[~filtered_small] area, seg_img = get_ids_area(cur_masks, cur_scores) else: break else: cur_classes = torch.ones(1, dtype=torch.long, device=cur_classes.device) segments_info = [] for i, a in enumerate(area): cat = cur_classes[i].item() segments_info.append({"id": i, "isthing": self.is_thing_map[cat], "category_id": cat, "area": a}) del cur_classes with io.BytesIO() as out: seg_img.save(out, format="PNG") predictions = {"png_string": out.getvalue(), "segments_info": segments_info} preds.append(predictions) return preds ================================================ FILE: models/swin_transformer/__init__.py ================================================ # ------------------------------------------------------------------------------------ # Sparse DETR # Copyright (c) 2021 KakaoBrain. All Rights Reserved. # Licensed under the Apache License, Version 2.0 [see LICENSE for details] # ------------------------------------------------------------------------------------ from .build import build_model ================================================ FILE: models/swin_transformer/build.py ================================================ # ------------------------------------------------------------------------------ # Sparse DETR # Copyright (c) 2021 KakaoBrain. All Rights Reserved. # Licensed under the Apache License, Version 2.0 [see LICENSE for details] # ------------------------------------------------------------------------------ from collections import abc, OrderedDict import os import yaml from .swin_transformer import SwinTransformer from .config import Config import torch CONFIG_MAP = { "swin-t": "models/swin_transformer/configs/swin_tiny_patch4_window7_224.yaml", "swin-s": "models/swin_transformer/configs/swin_small_patch4_window7_224.yaml", "swin-b": "models/swin_transformer/configs/swin_base_patch4_window7_224.yaml", "swin-l": "models/swin_transformer/configs/swin_large_patch4_window7_224.yaml", } CHECKPOINT_MAP = { "swin-t": "/data/public/rw/team-autolearn/pretrainedmodels/swin/swin_tiny_patch4_window7_224.pth", } def build_model(name, out_indices, frozen_stages, pretrained): config_file = CONFIG_MAP[name] config = load_config_yaml(config_file) config = Config(config) config.freeze() model_type = config.MODEL.TYPE if model_type == 'swin': model = SwinTransformer(pretrain_img_size=config.DATA.IMG_SIZE, patch_size=config.MODEL.SWIN.PATCH_SIZE, in_chans=config.MODEL.SWIN.IN_CHANS, embed_dim=config.MODEL.SWIN.EMBED_DIM, depths=config.MODEL.SWIN.DEPTHS, num_heads=config.MODEL.SWIN.NUM_HEADS, window_size=config.MODEL.SWIN.WINDOW_SIZE, mlp_ratio=config.MODEL.SWIN.MLP_RATIO, qkv_bias=config.MODEL.SWIN.QKV_BIAS, qk_scale=config.MODEL.SWIN.QK_SCALE, drop_rate=config.MODEL.DROP_RATE, drop_path_rate=config.MODEL.DROP_PATH_RATE, ape=config.MODEL.SWIN.APE, patch_norm=config.MODEL.SWIN.PATCH_NORM, use_checkpoint=config.TRAIN.USE_CHECKPOINT, out_indices=out_indices, frozen_stages=frozen_stages) else: raise NotImplementedError(f"Unkown model: {model_type}") if pretrained: ckpt_path = CHECKPOINT_MAP[name] state_dict = torch.load(ckpt_path) model.load_state_dict(state_dict['model'], strict=False) return model def _update_dict(tar, src): """recursive dict update.""" for k, v in src.items(): if isinstance(v, abc.Mapping): tar[k] = _update_dict(tar.get(k, {}), v) else: tar[k] = v return tar def load_config_yaml(cfg_file, config=None): if config is None: config = OrderedDict() with open(cfg_file, 'r') as f: config_src = yaml.load(f, Loader=yaml.FullLoader) for cfg in config_src.setdefault('BASE', ['']): if cfg: load_config_yaml( os.path.join(os.path.dirname(cfg_file), cfg), config ) print('=> merge config from {}'.format(cfg_file)) _update_dict(config, config_src) return config ================================================ FILE: models/swin_transformer/config.py ================================================ # ------------------------------------------------------------------------------ # Sparse DETR # Copyright (c) 2021 KakaoBrain. All Rights Reserved. # Licensed under the Apache License, Version 2.0 [see LICENSE for details] # ------------------------------------------------------------------------------ import collections from collections import OrderedDict from copy import deepcopy import logging from os.path import basename, splitext from pprint import pformat from types import SimpleNamespace import yaml class Config(SimpleNamespace): """Dictionary-based but also dot-accessible configuration object, which will rescue you from the messy brackets and quotation marks while accessing nested dictionaries. As the usage example below, a value can be easily assigned to a new field with hierarchies by using Python's usual assignment syntax. Due to the side effects of this feature, it is safe that the user call '.freeze()' before using the Config instance as a fixed configuration. Otherwise, even when a wanted attribute is called with an incorrect name, AttributeError will be silently ignored and returns an empty config, which could be resulting in unwanted consequences. Usage: >>> cfg = Config() >>> cfg.foo = 1 >>> cfg.bar.baz = 2 >>> cfg['bar']['baz'] == cfg.bar.baz True >>> cfg.pprint() --- foo: 1 bar: baz: 2 ... >>> cfg.freeze() >>> cfg.new = 3 RuntimeError: Can't set new attribute after being freezed! """ def __init__(self, _dict=None, **kwargs): super().__init__(**kwargs) self._freezed = False self._order = list() if _dict is not None: self._set_with_nested_dict(_dict) def _set_with_nested_dict(self, _dict): for key, value in _dict.items(): if isinstance(value, dict): self.__setattr__(key, Config(value)) else: self.__setattr__(key, value) self._order.append(key) @property def freezed(self): return self._freezed @classmethod def from_yaml(cls, yaml_file): """Initialize configuration with a YAML file.""" return cls(OrderedDict(yaml.load(open(yaml_file, "r"), Loader=yaml.FullLoader))) def __repr__(self): return 'Config' + self.to_dict().__repr__() def __getitem__(self, item): return self.__getattr__(item) def __getattr__(self, item): try: return self.__getattribute__(item) except AttributeError as e: if self._freezed: raise AttributeError(f"Can't find the field: {item}") from e else: # if there's no attribute with the given name, # make new one and assign an empty config. self.__setattr__(item, Config()) return self.__getattribute__(item) def __setattr__(self, item, value): if item != '_freezed' and self.__dict__['_freezed']: raise RuntimeError("Can't set new attribute after being freezed!") super().__setattr__(item, value) def __bool__(self): return len([k for k in self.to_dict().keys() if not k.startswith('_')]) > 0 def __len__(self): return len(self.to_dict()) def __getstate__(self): return self.to_dict() def __setstate__(self, state): self._set_with_nested_dict(state) def __contains__(self, item): return self.to_dict().__contains__(item) def __deepcopy__(self, memodict={}): return Config(_dict=deepcopy(self.to_dict())) def __iter__(self): # for iterable unpacking return self.to_dict().__iter__() def pformat(self): return yaml.dump(self.to_dict(), indent=4, sort_keys=False, explicit_start=True, explicit_end=True) def pprint(self): return print(self.pformat()) def freeze(self): self._freezed = True for value in self.__dict__.values(): if isinstance(value, Config): value.freeze() return self def defrost(self): self._freezed = False for value in self.__dict__.values(): if isinstance(value, Config): value.defrost() return self def get(self, *args, **kwargs): return self.to_dict().get(*args, **kwargs) def keys(self): return self.to_dict().keys() def values(self): return self.to_dict().values() def items(self): return self.to_dict().items() def clone(self): return self.__deepcopy__() def update(self, dict_, delimiter='/'): for k, v in dict_.items(): self._update(k, v, delimiter) def _update(self, key, value, delimiter='/'): obj = self keys = key.split(delimiter) for k in keys[:-1]: obj = obj.__getattr__(k) obj.__setattr__(keys[-1], value) def to_dict(self): out_dict = OrderedDict() for key, value in self.__dict__.items(): if isinstance(value, Config): out_dict[key] = value.to_dict() else: if not key.startswith('_'): out_dict[key] = value return dict(out_dict) ================================================ FILE: models/swin_transformer/configs/default.yaml ================================================ DATA: IMG_SIZE: 224 TRAIN: USE_CHECKPOINT: false MODEL: SWIN: APE: false DEPTHS: [2, 2, 6, 2] EMBED_DIM: 96 IN_CHANS: 3 MLP_RATIO: 4.0 NUM_HEADS: [3, 6, 12, 24] PATCH_NORM: true PATCH_SIZE: 4 QKV_BIAS: true QK_SCALE: null WINDOW_SIZE: 7 DROP_RATE: 0.0 DROP_PATH_RATE: 0.1 NUM_CLASSES: 1000 ================================================ FILE: models/swin_transformer/configs/swin_base_patch4_window7_224.yaml ================================================ BASE: ['default.yaml'] MODEL: TYPE: swin NAME: swin_base_patch4_window7_224 DROP_PATH_RATE: 0.5 SWIN: EMBED_DIM: 128 DEPTHS: [ 2, 2, 18, 2 ] NUM_HEADS: [ 4, 8, 16, 32 ] WINDOW_SIZE: 7 ================================================ FILE: models/swin_transformer/configs/swin_large_patch4_window7_224.yaml ================================================ BASE: ['default.yaml'] MODEL: TYPE: swin NAME: swin_large_patch4_window7_224 SWIN: EMBED_DIM: 192 DEPTHS: [ 2, 2, 18, 2 ] NUM_HEADS: [ 6, 12, 24, 48 ] WINDOW_SIZE: 7 ================================================ FILE: models/swin_transformer/configs/swin_small_patch4_window7_224.yaml ================================================ BASE: ['default.yaml'] MODEL: TYPE: swin NAME: swin_small_patch4_window7_224 DROP_PATH_RATE: 0.3 SWIN: EMBED_DIM: 96 DEPTHS: [ 2, 2, 18, 2 ] NUM_HEADS: [ 3, 6, 12, 24 ] WINDOW_SIZE: 7 ================================================ FILE: models/swin_transformer/configs/swin_tiny_patch4_window7_224.yaml ================================================ BASE: ['default.yaml'] MODEL: TYPE: swin NAME: swin_tiny_patch4_window7_224 DROP_PATH_RATE: 0.2 SWIN: EMBED_DIM: 96 DEPTHS: [ 2, 2, 6, 2 ] NUM_HEADS: [ 3, 6, 12, 24 ] WINDOW_SIZE: 7 ================================================ FILE: models/swin_transformer/swin_transformer.py ================================================ # ------------------------------------------------------------------------------ # Sparse DETR # Copyright (c) 2021 KakaoBrain. All Rights Reserved. # Licensed under the Apache License, Version 2.0 [see LICENSE for details] # ------------------------------------------------------------------------------ # Modified from Swin Transformer (https://github.com/microsoft/Swin-Transformer) # Copyright (c) 2021 Microsoft. All Rights Reserved. # Written by Ze Liu # ------------------------------------------------------------------------------ import numpy as np import torch import torch.nn as nn import torch.utils.checkpoint as checkpoint import torch.nn.functional as F from timm.models.layers import DropPath, to_2tuple, trunc_normal_ class Mlp(nn.Module): def __init__(self, in_features, hidden_features=None, out_features=None, act_layer=nn.GELU, drop=0.): super().__init__() out_features = out_features or in_features hidden_features = hidden_features or in_features self.fc1 = nn.Linear(in_features, hidden_features) self.act = act_layer() self.fc2 = nn.Linear(hidden_features, out_features) self.drop = nn.Dropout(drop) def forward(self, x): x = self.fc1(x) x = self.act(x) x = self.drop(x) x = self.fc2(x) x = self.drop(x) return x def window_partition(x, window_size): """ Args: x: (B, H, W, C) window_size (int): window size Returns: windows: (num_windows*B, window_size, window_size, C) """ B, H, W, C = x.shape x = x.view(B, H // window_size, window_size, W // window_size, window_size, C) windows = x.permute(0, 1, 3, 2, 4, 5).contiguous().view(-1, window_size, window_size, C) return windows def window_reverse(windows, window_size, H, W): """ Args: windows: (num_windows*B, window_size, window_size, C) window_size (int): Window size H (int): Height of image W (int): Width of image Returns: x: (B, H, W, C) """ B = int(windows.shape[0] / (float(H) * float(W) / window_size / window_size)) x = windows.view(B, H // window_size, W // window_size, window_size, window_size, -1) x = x.permute(0, 1, 3, 2, 4, 5).contiguous().view(B, H, W, -1) return x class WindowAttention(nn.Module): """ Window based multi-head self attention (W-MSA) module with relative position bias. It supports both of shifted and non-shifted window. Args: dim (int): Number of input channels. window_size (tuple[int]): The height and width of the window. num_heads (int): Number of attention heads. qkv_bias (bool, optional): If True, add a learnable bias to query, key, value. Default: True qk_scale (float | None, optional): Override default qk scale of head_dim ** -0.5 if set attn_drop (float, optional): Dropout ratio of attention weight. Default: 0.0 proj_drop (float, optional): Dropout ratio of output. Default: 0.0 """ def __init__(self, dim, window_size, num_heads, qkv_bias=True, qk_scale=None, attn_drop=0., proj_drop=0.): super().__init__() self.dim = dim self.window_size = window_size # Wh, Ww self.num_heads = num_heads head_dim = dim // num_heads self.scale = qk_scale or head_dim ** -0.5 # define a parameter table of relative position bias self.relative_position_bias_table = nn.Parameter( torch.zeros((2 * window_size[0] - 1) * (2 * window_size[1] - 1), num_heads)) # 2*Wh-1 * 2*Ww-1, nH # get pair-wise relative position index for each token inside the window coords_h = torch.arange(self.window_size[0]) coords_w = torch.arange(self.window_size[1]) coords = torch.stack(torch.meshgrid([coords_h, coords_w])) # 2, Wh, Ww coords_flatten = torch.flatten(coords, 1) # 2, Wh*Ww relative_coords = coords_flatten[:, :, None] - coords_flatten[:, None, :] # 2, Wh*Ww, Wh*Ww relative_coords = relative_coords.permute(1, 2, 0).contiguous() # Wh*Ww, Wh*Ww, 2 relative_coords[:, :, 0] += self.window_size[0] - 1 # shift to start from 0 relative_coords[:, :, 1] += self.window_size[1] - 1 relative_coords[:, :, 0] *= 2 * self.window_size[1] - 1 relative_position_index = relative_coords.sum(-1) # Wh*Ww, Wh*Ww self.register_buffer("relative_position_index", relative_position_index) self.qkv = nn.Linear(dim, dim * 3, bias=qkv_bias) self.attn_drop = nn.Dropout(attn_drop) self.proj = nn.Linear(dim, dim) self.proj_drop = nn.Dropout(proj_drop) trunc_normal_(self.relative_position_bias_table, std=.02) self.softmax = nn.Softmax(dim=-1) def forward(self, x, mask=None): """ Forward function. Args: x: input features with shape of (num_windows*B, N, C) mask: (0/-inf) mask with shape of (num_windows, Wh*Ww, Wh*Ww) or None """ B_, N, C = x.shape qkv = self.qkv(x).reshape(B_, N, 3, self.num_heads, C // self.num_heads).permute(2, 0, 3, 1, 4) q, k, v = qkv[0], qkv[1], qkv[2] # make torchscript happy (cannot use tensor as tuple) q = q * self.scale attn = (q @ k.transpose(-2, -1)) relative_position_bias = self.relative_position_bias_table[self.relative_position_index.view(-1)].view( self.window_size[0] * self.window_size[1], self.window_size[0] * self.window_size[1], -1) # Wh*Ww,Wh*Ww,nH relative_position_bias = relative_position_bias.permute(2, 0, 1).contiguous() # nH, Wh*Ww, Wh*Ww attn = attn + relative_position_bias.unsqueeze(0) if mask is not None: nW = mask.shape[0] attn = attn.view(B_ // nW, nW, self.num_heads, N, N) + mask.unsqueeze(1).unsqueeze(0) attn = attn.view(-1, self.num_heads, N, N) attn = self.softmax(attn) else: attn = self.softmax(attn) attn = self.attn_drop(attn) x = (attn @ v).transpose(1, 2).reshape(B_, N, C) x = self.proj(x) x = self.proj_drop(x) return x class SwinTransformerBlock(nn.Module): """ Swin Transformer Block. Args: dim (int): Number of input channels. num_heads (int): Number of attention heads. window_size (int): Window size. shift_size (int): Shift size for SW-MSA. mlp_ratio (float): Ratio of mlp hidden dim to embedding dim. qkv_bias (bool, optional): If True, add a learnable bias to query, key, value. Default: True qk_scale (float | None, optional): Override default qk scale of head_dim ** -0.5 if set. drop (float, optional): Dropout rate. Default: 0.0 attn_drop (float, optional): Attention dropout rate. Default: 0.0 drop_path (float, optional): Stochastic depth rate. Default: 0.0 act_layer (nn.Module, optional): Activation layer. Default: nn.GELU norm_layer (nn.Module, optional): Normalization layer. Default: nn.LayerNorm """ def __init__(self, dim, num_heads, window_size=7, shift_size=0, mlp_ratio=4., qkv_bias=True, qk_scale=None, drop=0., attn_drop=0., drop_path=0., act_layer=nn.GELU, norm_layer=nn.LayerNorm): super().__init__() self.dim = dim self.num_heads = num_heads self.window_size = window_size self.shift_size = shift_size self.mlp_ratio = mlp_ratio assert 0 <= self.shift_size < self.window_size, "shift_size must in 0-window_size" self.norm1 = norm_layer(dim) self.attn = WindowAttention( dim, window_size=to_2tuple(self.window_size), num_heads=num_heads, qkv_bias=qkv_bias, qk_scale=qk_scale, attn_drop=attn_drop, proj_drop=drop) self.drop_path = DropPath(drop_path) if drop_path > 0. else nn.Identity() self.norm2 = norm_layer(dim) mlp_hidden_dim = int(dim * mlp_ratio) self.mlp = Mlp(in_features=dim, hidden_features=mlp_hidden_dim, act_layer=act_layer, drop=drop) self.H = None self.W = None def forward(self, x, mask_matrix): """ Forward function. Args: x: Input feature, tensor size (B, H*W, C). H, W: Spatial resolution of the input feature. mask_matrix: Attention mask for cyclic shift. """ B, L, C = x.shape H, W = self.H, self.W assert L == H * W, "input feature has wrong size" shortcut = x x = self.norm1(x) x = x.view(B, H, W, C) # pad feature maps to multiples of window size pad_l = pad_t = 0 pad_r = (self.window_size - W % self.window_size) % self.window_size pad_b = (self.window_size - H % self.window_size) % self.window_size x = F.pad(x, (0, 0, pad_l, pad_r, pad_t, pad_b)) _, Hp, Wp, _ = x.shape # cyclic shift if self.shift_size > 0: shifted_x = torch.roll(x, shifts=(-self.shift_size, -self.shift_size), dims=(1, 2)) attn_mask = mask_matrix else: shifted_x = x attn_mask = None # partition windows x_windows = window_partition(shifted_x, self.window_size) # nW*B, window_size, window_size, C x_windows = x_windows.view(-1, self.window_size * self.window_size, C) # nW*B, window_size*window_size, C # W-MSA/SW-MSA attn_windows = self.attn(x_windows, mask=attn_mask) # nW*B, window_size*window_size, C # merge windows attn_windows = attn_windows.view(-1, self.window_size, self.window_size, C) shifted_x = window_reverse(attn_windows, self.window_size, Hp, Wp) # B H' W' C # reverse cyclic shift if self.shift_size > 0: x = torch.roll(shifted_x, shifts=(self.shift_size, self.shift_size), dims=(1, 2)) else: x = shifted_x if pad_r > 0 or pad_b > 0: x = x[:, :H, :W, :].contiguous() x = x.view(B, H * W, C) # FFN x = shortcut + self.drop_path(x) x = x + self.drop_path(self.mlp(self.norm2(x))) return x class PatchMerging(nn.Module): """ Patch Merging Layer Args: dim (int): Number of input channels. norm_layer (nn.Module, optional): Normalization layer. Default: nn.LayerNorm """ def __init__(self, dim, norm_layer=nn.LayerNorm): super().__init__() self.dim = dim self.reduction = nn.Linear(4 * dim, 2 * dim, bias=False) self.norm = norm_layer(4 * dim) def forward(self, x, H, W): """ Forward function. Args: x: Input feature, tensor size (B, H*W, C). H, W: Spatial resolution of the input feature. """ B, L, C = x.shape assert L == H * W, "input feature has wrong size" x = x.view(B, H, W, C) # padding pad_input = (H % 2 == 1) or (W % 2 == 1) if pad_input: x = F.pad(x, (0, 0, 0, W % 2, 0, H % 2)) x0 = x[:, 0::2, 0::2, :] # B H/2 W/2 C x1 = x[:, 1::2, 0::2, :] # B H/2 W/2 C x2 = x[:, 0::2, 1::2, :] # B H/2 W/2 C x3 = x[:, 1::2, 1::2, :] # B H/2 W/2 C x = torch.cat([x0, x1, x2, x3], -1) # B H/2 W/2 4*C x = x.view(B, -1, 4 * C) # B H/2*W/2 4*C x = self.norm(x) x = self.reduction(x) return x class BasicLayer(nn.Module): """ A basic Swin Transformer layer for one stage. Args: dim (int): Number of feature channels depth (int): Depths of this stage. num_heads (int): Number of attention head. window_size (int): Local window size. Default: 7. mlp_ratio (float): Ratio of mlp hidden dim to embedding dim. Default: 4. qkv_bias (bool, optional): If True, add a learnable bias to query, key, value. Default: True qk_scale (float | None, optional): Override default qk scale of head_dim ** -0.5 if set. drop (float, optional): Dropout rate. Default: 0.0 attn_drop (float, optional): Attention dropout rate. Default: 0.0 drop_path (float | tuple[float], optional): Stochastic depth rate. Default: 0.0 norm_layer (nn.Module, optional): Normalization layer. Default: nn.LayerNorm downsample (nn.Module | None, optional): Downsample layer at the end of the layer. Default: None use_checkpoint (bool): Whether to use checkpointing to save memory. Default: False. """ def __init__(self, dim, depth, num_heads, window_size=7, mlp_ratio=4., qkv_bias=True, qk_scale=None, drop=0., attn_drop=0., drop_path=0., norm_layer=nn.LayerNorm, downsample=None, use_checkpoint=False): super().__init__() self.window_size = window_size self.shift_size = window_size // 2 self.depth = depth self.use_checkpoint = use_checkpoint # build blocks self.blocks = nn.ModuleList([ SwinTransformerBlock( dim=dim, num_heads=num_heads, window_size=window_size, shift_size=0 if (i % 2 == 0) else window_size // 2, mlp_ratio=mlp_ratio, qkv_bias=qkv_bias, qk_scale=qk_scale, drop=drop, attn_drop=attn_drop, drop_path=drop_path[i] if isinstance(drop_path, list) else drop_path, norm_layer=norm_layer) for i in range(depth)]) # patch merging layer if downsample is not None: self.downsample = downsample(dim=dim, norm_layer=norm_layer) else: self.downsample = None def forward(self, x, H, W): """ Forward function. Args: x: Input feature, tensor size (B, H*W, C). H, W: Spatial resolution of the input feature. """ # calculate attention mask for SW-MSA Hp = int(np.ceil(float(H) / self.window_size)) * self.window_size Wp = int(np.ceil(float(W) / self.window_size)) * self.window_size img_mask = torch.zeros((1, Hp, Wp, 1), device=x.device) # 1 Hp Wp 1 h_slices = (slice(0, -self.window_size), slice(-self.window_size, -self.shift_size), slice(-self.shift_size, None)) w_slices = (slice(0, -self.window_size), slice(-self.window_size, -self.shift_size), slice(-self.shift_size, None)) cnt = 0 for h in h_slices: for w in w_slices: img_mask[:, h, w, :] = cnt cnt += 1 mask_windows = window_partition(img_mask, self.window_size) # nW, window_size, window_size, 1 mask_windows = mask_windows.view(-1, self.window_size * self.window_size) attn_mask = mask_windows.unsqueeze(1) - mask_windows.unsqueeze(2) attn_mask = attn_mask.masked_fill(attn_mask != 0, float(-100.0)).masked_fill(attn_mask == 0, float(0.0)) for blk in self.blocks: blk.H, blk.W = H, W if self.use_checkpoint: x = checkpoint.checkpoint(blk, x, attn_mask) else: x = blk(x, attn_mask) if self.downsample is not None: x_down = self.downsample(x, H, W) Wh, Ww = (H + 1) // 2, (W + 1) // 2 return x, H, W, x_down, Wh, Ww else: return x, H, W, x, H, W class PatchEmbed(nn.Module): """ Image to Patch Embedding Args: patch_size (int): Patch token size. Default: 4. in_chans (int): Number of input image channels. Default: 3. embed_dim (int): Number of linear projection output channels. Default: 96. norm_layer (nn.Module, optional): Normalization layer. Default: None """ def __init__(self, patch_size=4, in_chans=3, embed_dim=96, norm_layer=None): super().__init__() patch_size = to_2tuple(patch_size) self.patch_size = patch_size self.in_chans = in_chans self.embed_dim = embed_dim self.proj = nn.Conv2d(in_chans, embed_dim, kernel_size=patch_size, stride=patch_size) if norm_layer is not None: self.norm = norm_layer(embed_dim) else: self.norm = None def forward(self, x): """Forward function.""" # padding _, _, H, W = x.size() if W % self.patch_size[1] != 0: x = F.pad(x, (0, self.patch_size[1] - W % self.patch_size[1])) if H % self.patch_size[0] != 0: x = F.pad(x, (0, 0, 0, self.patch_size[0] - H % self.patch_size[0])) x = self.proj(x) # B C Wh Ww if self.norm is not None: Wh, Ww = x.size(2), x.size(3) x = x.flatten(2).transpose(1, 2) x = self.norm(x) x = x.transpose(1, 2).view(-1, self.embed_dim, Wh, Ww) return x class SwinTransformer(nn.Module): """ Swin Transformer backbone. A PyTorch impl of : `Swin Transformer: Hierarchical Vision Transformer using Shifted Windows` - https://arxiv.org/pdf/2103.14030 Args: pretrain_img_size (int): Input image size for training the pretrained model, used in absolute postion embedding. Default 224. patch_size (int | tuple(int)): Patch size. Default: 4. in_chans (int): Number of input image channels. Default: 3. embed_dim (int): Number of linear projection output channels. Default: 96. depths (tuple[int]): Depths of each Swin Transformer stage. num_heads (tuple[int]): Number of attention head of each stage. window_size (int): Window size. Default: 7. mlp_ratio (float): Ratio of mlp hidden dim to embedding dim. Default: 4. qkv_bias (bool): If True, add a learnable bias to query, key, value. Default: True qk_scale (float): Override default qk scale of head_dim ** -0.5 if set. drop_rate (float): Dropout rate. attn_drop_rate (float): Attention dropout rate. Default: 0. drop_path_rate (float): Stochastic depth rate. Default: 0.2. norm_layer (nn.Module): Normalization layer. Default: nn.LayerNorm. ape (bool): If True, add absolute position embedding to the patch embedding. Default: False. patch_norm (bool): If True, add normalization after patch embedding. Default: True. out_indices (Sequence[int]): Output from which stages. frozen_stages (int): Stages to be frozen (stop grad and set eval mode). -1 means not freezing any parameters. use_checkpoint (bool): Whether to use checkpointing to save memory. Default: False. """ def __init__(self, pretrain_img_size=224, patch_size=4, in_chans=3, embed_dim=96, depths=[2, 2, 6, 2], num_heads=[3, 6, 12, 24], window_size=7, mlp_ratio=4., qkv_bias=True, qk_scale=None, drop_rate=0., attn_drop_rate=0., drop_path_rate=0.2, norm_layer=nn.LayerNorm, ape=False, patch_norm=True, out_indices=(0, 1, 2, 3), frozen_stages=-1, use_checkpoint=False): super().__init__() self.pretrain_img_size = pretrain_img_size self.num_layers = len(depths) self.embed_dim = embed_dim self.ape = ape self.patch_norm = patch_norm self.out_indices = out_indices self.frozen_stages = frozen_stages # split image into non-overlapping patches self.patch_embed = PatchEmbed( patch_size=patch_size, in_chans=in_chans, embed_dim=embed_dim, norm_layer=norm_layer if self.patch_norm else None) # absolute position embedding if self.ape: pretrain_img_size = to_2tuple(pretrain_img_size) patch_size = to_2tuple(patch_size) patches_resolution = [pretrain_img_size[0] // patch_size[0], pretrain_img_size[1] // patch_size[1]] self.absolute_pos_embed = nn.Parameter(torch.zeros(1, embed_dim, patches_resolution[0], patches_resolution[1])) trunc_normal_(self.absolute_pos_embed, std=.02) self.pos_drop = nn.Dropout(p=drop_rate) # stochastic depth dpr = [x.item() for x in torch.linspace(0, drop_path_rate, sum(depths))] # stochastic depth decay rule # build layers self.layers = nn.ModuleList() for i_layer in range(self.num_layers): layer = BasicLayer( dim=int(embed_dim * 2 ** i_layer), depth=depths[i_layer], num_heads=num_heads[i_layer], window_size=window_size, mlp_ratio=mlp_ratio, qkv_bias=qkv_bias, qk_scale=qk_scale, drop=drop_rate, attn_drop=attn_drop_rate, drop_path=dpr[sum(depths[:i_layer]):sum(depths[:i_layer + 1])], norm_layer=norm_layer, downsample=PatchMerging if (i_layer < self.num_layers - 1) else None, use_checkpoint=use_checkpoint) self.layers.append(layer) num_features = [int(embed_dim * 2 ** i) for i in range(self.num_layers)] self.num_features = num_features # add a norm layer for each output for i_layer in out_indices: layer = norm_layer(num_features[i_layer]) layer_name = f'norm{i_layer}' self.add_module(layer_name, layer) self.apply(self._init_weights) self._freeze_stages() def _freeze_stages(self): if self.frozen_stages >= 0: self.patch_embed.eval() for param in self.patch_embed.parameters(): param.requires_grad = False if self.frozen_stages >= 1 and self.ape: self.absolute_pos_embed.requires_grad = False if self.frozen_stages >= 2: self.pos_drop.eval() for i in range(0, self.frozen_stages - 1): m = self.layers[i] m.eval() for param in m.parameters(): param.requires_grad = False def _init_weights(self, m): if isinstance(m, nn.Linear): trunc_normal_(m.weight, std=.02) if isinstance(m, nn.Linear) and m.bias is not None: nn.init.constant_(m.bias, 0) elif isinstance(m, nn.LayerNorm): nn.init.constant_(m.bias, 0) nn.init.constant_(m.weight, 1.0) def forward(self, x): """Forward function.""" x = self.patch_embed(x) Wh, Ww = x.size(2), x.size(3) if self.ape: # interpolate the position embedding to the corresponding size absolute_pos_embed = F.interpolate(self.absolute_pos_embed, size=(Wh, Ww), mode='bicubic') x = (x + absolute_pos_embed).flatten(2).transpose(1, 2) # B Wh*Ww C else: x = x.flatten(2).transpose(1, 2) x = self.pos_drop(x) outs = {} for i in range(self.num_layers): layer = self.layers[i] x_out, H, W, x, Wh, Ww = layer(x, Wh, Ww) if i in self.out_indices: norm_layer = getattr(self, f'norm{i}') x_out = norm_layer(x_out) out = x_out.view(-1, H, W, self.num_features[i]).permute(0, 3, 1, 2).contiguous() outs[str(len(outs))] = out return outs ================================================ FILE: requirements.txt ================================================ pycocotools tqdm scipy timm fvcore tensorboard ================================================ FILE: tools/launch.py ================================================ # -------------------------------------------------------------------------------------------------------------------------- # Deformable DETR # Copyright (c) 2020 SenseTime. All Rights Reserved. # Licensed under the Apache License, Version 2.0 [see LICENSE for details] # -------------------------------------------------------------------------------------------------------------------------- # Modified from https://github.com/pytorch/pytorch/blob/173f224570017b4b1a3a1a13d0bff280a54d9cd9/torch/distributed/launch.py # -------------------------------------------------------------------------------------------------------------------------- r""" `torch.distributed.launch` is a module that spawns up multiple distributed training processes on each of the training nodes. The utility can be used for single-node distributed training, in which one or more processes per node will be spawned. The utility can be used for either CPU training or GPU training. If the utility is used for GPU training, each distributed process will be operating on a single GPU. This can achieve well-improved single-node training performance. It can also be used in multi-node distributed training, by spawning up multiple processes on each node for well-improved multi-node distributed training performance as well. This will especially be benefitial for systems with multiple Infiniband interfaces that have direct-GPU support, since all of them can be utilized for aggregated communication bandwidth. In both cases of single-node distributed training or multi-node distributed training, this utility will launch the given number of processes per node (``--nproc_per_node``). If used for GPU training, this number needs to be less or euqal to the number of GPUs on the current system (``nproc_per_node``), and each process will be operating on a single GPU from *GPU 0 to GPU (nproc_per_node - 1)*. **How to use this module:** 1. Single-Node multi-process distributed training :: >>> python -m torch.distributed.launch --nproc_per_node=NUM_GPUS_YOU_HAVE YOUR_TRAINING_SCRIPT.py (--arg1 --arg2 --arg3 and all other arguments of your training script) 2. Multi-Node multi-process distributed training: (e.g. two nodes) Node 1: *(IP: 192.168.1.1, and has a free port: 1234)* :: >>> python -m torch.distributed.launch --nproc_per_node=NUM_GPUS_YOU_HAVE --nnodes=2 --node_rank=0 --master_addr="192.168.1.1" --master_port=1234 YOUR_TRAINING_SCRIPT.py (--arg1 --arg2 --arg3 and all other arguments of your training script) Node 2: :: >>> python -m torch.distributed.launch --nproc_per_node=NUM_GPUS_YOU_HAVE --nnodes=2 --node_rank=1 --master_addr="192.168.1.1" --master_port=1234 YOUR_TRAINING_SCRIPT.py (--arg1 --arg2 --arg3 and all other arguments of your training script) 3. To look up what optional arguments this module offers: :: >>> python -m torch.distributed.launch --help **Important Notices:** 1. This utilty and multi-process distributed (single-node or multi-node) GPU training currently only achieves the best performance using the NCCL distributed backend. Thus NCCL backend is the recommended backend to use for GPU training. 2. In your training program, you must parse the command-line argument: ``--local_rank=LOCAL_PROCESS_RANK``, which will be provided by this module. If your training program uses GPUs, you should ensure that your code only runs on the GPU device of LOCAL_PROCESS_RANK. This can be done by: Parsing the local_rank argument :: >>> import argparse >>> parser = argparse.ArgumentParser() >>> parser.add_argument("--local_rank", type=int) >>> args = parser.parse_args() Set your device to local rank using either :: >>> torch.cuda.set_device(arg.local_rank) # before your code runs or :: >>> with torch.cuda.device(arg.local_rank): >>> # your code to run 3. In your training program, you are supposed to call the following function at the beginning to start the distributed backend. You need to make sure that the init_method uses ``env://``, which is the only supported ``init_method`` by this module. :: torch.distributed.init_process_group(backend='YOUR BACKEND', init_method='env://') 4. In your training program, you can either use regular distributed functions or use :func:`torch.nn.parallel.DistributedDataParallel` module. If your training program uses GPUs for training and you would like to use :func:`torch.nn.parallel.DistributedDataParallel` module, here is how to configure it. :: model = torch.nn.parallel.DistributedDataParallel(model, device_ids=[arg.local_rank], output_device=arg.local_rank) Please ensure that ``device_ids`` argument is set to be the only GPU device id that your code will be operating on. This is generally the local rank of the process. In other words, the ``device_ids`` needs to be ``[args.local_rank]``, and ``output_device`` needs to be ``args.local_rank`` in order to use this utility 5. Another way to pass ``local_rank`` to the subprocesses via environment variable ``LOCAL_RANK``. This behavior is enabled when you launch the script with ``--use_env=True``. You must adjust the subprocess example above to replace ``args.local_rank`` with ``os.environ['LOCAL_RANK']``; the launcher will not pass ``--local_rank`` when you specify this flag. .. warning:: ``local_rank`` is NOT globally unique: it is only unique per process on a machine. Thus, don't use it to decide if you should, e.g., write to a networked filesystem. See https://github.com/pytorch/pytorch/issues/12042 for an example of how things can go wrong if you don't do this correctly. """ import sys import subprocess import os import socket from argparse import ArgumentParser, REMAINDER import torch def parse_args(): """ Helper function parsing the command line options @retval ArgumentParser """ parser = ArgumentParser(description="PyTorch distributed training launch " "helper utilty that will spawn up " "multiple distributed processes") # Optional arguments for the launch helper parser.add_argument("--nnodes", type=int, default=1, help="The number of nodes to use for distributed " "training") parser.add_argument("--node_rank", type=int, default=0, help="The rank of the node for multi-node distributed " "training") parser.add_argument("--nproc_per_node", type=int, default=1, help="The number of processes to launch on each node, " "for GPU training, this is recommended to be set " "to the number of GPUs in your system so that " "each process can be bound to a single GPU.") parser.add_argument("--master_addr", default="127.0.0.1", type=str, help="Master node (rank 0)'s address, should be either " "the IP address or the hostname of node 0, for " "single node multi-proc training, the " "--master_addr can simply be 127.0.0.1") parser.add_argument("--master_port", default=29500, type=int, help="Master node (rank 0)'s free port that needs to " "be used for communciation during distributed " "training") # positional parser.add_argument("training_script", type=str, help="The full path to the single GPU training " "program/script to be launched in parallel, " "followed by all the arguments for the " "training script") # rest from the training program parser.add_argument('training_script_args', nargs=REMAINDER) return parser.parse_args() def main(): args = parse_args() # world size in terms of number of processes dist_world_size = args.nproc_per_node * args.nnodes # set PyTorch distributed related environmental variables current_env = os.environ.copy() current_env["MASTER_ADDR"] = args.master_addr current_env["MASTER_PORT"] = str(args.master_port) current_env["WORLD_SIZE"] = str(dist_world_size) processes = [] for local_rank in range(0, args.nproc_per_node): # each process's rank dist_rank = args.nproc_per_node * args.node_rank + local_rank current_env["RANK"] = str(dist_rank) current_env["LOCAL_RANK"] = str(local_rank) cmd = [args.training_script] + args.training_script_args process = subprocess.Popen(cmd, env=current_env) processes.append(process) for process in processes: process.wait() if process.returncode != 0: raise subprocess.CalledProcessError(returncode=process.returncode, cmd=process.args) if __name__ == "__main__": main() ================================================ FILE: tools/run_dist_launch.sh ================================================ #!/usr/bin/env bash # ------------------------------------------------------------------------ # Deformable DETR # Copyright (c) 2020 SenseTime. All Rights Reserved. # Licensed under the Apache License, Version 2.0 [see LICENSE for details] # ------------------------------------------------------------------------ set -x GPUS=$1 RUN_COMMAND=${@:2} if [ $GPUS -lt 8 ]; then GPUS_PER_NODE=${GPUS_PER_NODE:-$GPUS} else GPUS_PER_NODE=${GPUS_PER_NODE:-8} fi MASTER_ADDR=${MASTER_ADDR:-"127.0.0.1"} MASTER_PORT=${MASTER_PORT:-"29500"} NODE_RANK=${NODE_RANK:-0} let "NNODES=GPUS/GPUS_PER_NODE" python ./tools/launch.py \ --nnodes ${NNODES} \ --node_rank ${NODE_RANK} \ --master_addr ${MASTER_ADDR} \ --master_port ${MASTER_PORT} \ --nproc_per_node ${GPUS_PER_NODE} \ ${RUN_COMMAND} ================================================ FILE: util/__init__.py ================================================ # ------------------------------------------------------------------------ # Deformable DETR # Copyright (c) 2020 SenseTime. All Rights Reserved. # Licensed under the Apache License, Version 2.0 [see LICENSE for details] # ------------------------------------------------------------------------ # Modified from DETR (https://github.com/facebookresearch/detr) # Copyright (c) Facebook, Inc. and its affiliates. All Rights Reserved # ------------------------------------------------------------------------ ================================================ FILE: util/benchmark.py ================================================ from collections import defaultdict import time from typing import Any, Counter, DefaultDict, Tuple, Dict, Optional import warnings import numpy as np import torch from torch import nn import tqdm from util.misc import nested_tensor_from_tensor_list from fvcore.nn import FlopCountAnalysis from fvcore.nn.jit_handles import Handle @torch.no_grad() def measure_average_inference_time(model, inputs, num_iters=100, warm_iters=5): ts = [] # note that warm-up iters. are excluded from the total iters. for iter_ in tqdm.tqdm(range(warm_iters + num_iters)): torch.cuda.synchronize() t_ = time.perf_counter() model(inputs) torch.cuda.synchronize() t = time.perf_counter() - t_ if iter_ >= warm_iters: ts.append(t) return sum(ts) / len(ts) def python_ops_mode_for_deform_attn(model, ops_mode): def change_ops_mode(module): if hasattr(module, "python_ops_for_test"): module.python_ops_for_test = ops_mode model.apply(change_ops_mode) @torch.no_grad() def compute_fps(model, dataset, num_iters=300, warm_iters=5, batch_size=4): print(f"computing fps.. (num_iters={num_iters}, batch_size={batch_size}) " f"warm_iters={warm_iters}, batch_size={batch_size}]") assert num_iters > 0 and warm_iters >= 0 and batch_size > 0 model.cuda() model.eval() inputs = nested_tensor_from_tensor_list( [dataset.__getitem__(0)[0].cuda() for _ in range(batch_size)]) t = measure_average_inference_time(model, inputs, num_iters, warm_iters) model.train() print(f"FPS: {1.0 / t * batch_size}") return 1.0 / t * batch_size @torch.no_grad() def compute_gflops(model, dataset, approximated=True): print(f"computing flops.. (approximated={approximated})") model.eval() python_ops_mode_for_deform_attn(model, True) if approximated: # use just a single image to approximate the full compuation # the size of the image was found heuristically images = [torch.randn((3, 850, 1040))] else: # full computation: get the first 100 images of COCO val2017 images = [] for idx in range(100): img, _ = dataset[idx] images.append(img) gflops_list = [] imsize_list = [] for img in tqdm.tqdm(images): inputs = [img.cuda()] with warnings.catch_warnings(): warnings.filterwarnings("ignore", category=RuntimeWarning) res = flop_count_without_warnings(model, (inputs,), )[0] gflops = sum(res.values()) gflops_list.append(gflops) imsize_list.append(list(img.shape)) if approximated: print(f"The image size used for approximation: [3, 850, 1040]") else: print("Average image size of first 100 image of COCO val2017 : " f"{np.array(imsize_list).mean(0)}") print(f"GFLOPs : {np.array(gflops_list).mean()}") model.train() python_ops_mode_for_deform_attn(model, False) return gflops def flop_count_without_warnings( model: nn.Module, inputs: Tuple[Any, ...], supported_ops: Optional[Dict[str, Handle]] = None, ) -> Tuple[DefaultDict[str, float], Counter[str]]: """copied and modified from fvcore.nn.flop_count.py Given a model and an input to the model, compute the per-operator Gflops of the given model. Args: model (nn.Module): The model to compute flop counts. inputs (tuple): Inputs that are passed to `model` to count flops. Inputs need to be in a tuple. supported_ops (dict(str,Callable) or None) : provide additional handlers for extra ops, or overwrite the existing handlers for convolution and matmul and einsum. The key is operator name and the value is a function that takes (inputs, outputs) of the op. We count one Multiply-Add as one FLOP. Returns: tuple[defaultdict, Counter]: A dictionary that records the number of gflops for each operation and a Counter that records the number of unsupported operations. """ if supported_ops is None: supported_ops = {} flop_counter = FlopCountAnalysis(model, inputs).set_op_handle(**supported_ops) flop_counter.unsupported_ops_warnings(False) flop_counter.uncalled_modules_warnings(False) flop_counter.tracer_warnings("no_tracer_warning") giga_flops = defaultdict(float) for op, flop in flop_counter.by_operator().items(): giga_flops[op] = flop / 1e9 return giga_flops, flop_counter.unsupported_ops() ================================================ FILE: util/box_ops.py ================================================ # ------------------------------------------------------------------------ # Deformable DETR # Copyright (c) 2020 SenseTime. All Rights Reserved. # Licensed under the Apache License, Version 2.0 [see LICENSE for details] # ------------------------------------------------------------------------ # Modified from DETR (https://github.com/facebookresearch/detr) # Copyright (c) Facebook, Inc. and its affiliates. All Rights Reserved # ------------------------------------------------------------------------ """ Utilities for bounding box manipulation and GIoU. """ import torch from torchvision.ops.boxes import box_area def box_cxcywh_to_xyxy(x): x_c, y_c, w, h = x.unbind(-1) b = [(x_c - 0.5 * w), (y_c - 0.5 * h), (x_c + 0.5 * w), (y_c + 0.5 * h)] return torch.stack(b, dim=-1) def box_xyxy_to_cxcywh(x): x0, y0, x1, y1 = x.unbind(-1) b = [(x0 + x1) / 2, (y0 + y1) / 2, (x1 - x0), (y1 - y0)] return torch.stack(b, dim=-1) # modified from torchvision to also return the union def box_iou(boxes1, boxes2): area1 = box_area(boxes1) area2 = box_area(boxes2) lt = torch.max(boxes1[:, None, :2], boxes2[:, :2]) # [N,M,2] rb = torch.min(boxes1[:, None, 2:], boxes2[:, 2:]) # [N,M,2] wh = (rb - lt).clamp(min=0) # [N,M,2] inter = wh[:, :, 0] * wh[:, :, 1] # [N,M] union = area1[:, None] + area2 - inter iou = inter / union return iou, union def generalized_box_iou(boxes1, boxes2): """ Generalized IoU from https://giou.stanford.edu/ The boxes should be in [x0, y0, x1, y1] format Returns a [N, M] pairwise matrix, where N = len(boxes1) and M = len(boxes2) """ # degenerate boxes gives inf / nan results # so do an early check assert (boxes1[:, 2:] >= boxes1[:, :2]).all() assert (boxes2[:, 2:] >= boxes2[:, :2]).all() iou, union = box_iou(boxes1, boxes2) lt = torch.min(boxes1[:, None, :2], boxes2[:, :2]) rb = torch.max(boxes1[:, None, 2:], boxes2[:, 2:]) wh = (rb - lt).clamp(min=0) # [N,M,2] area = wh[:, :, 0] * wh[:, :, 1] return iou - (area - union) / area def masks_to_boxes(masks): """Compute the bounding boxes around the provided masks The masks should be in format [N, H, W] where N is the number of masks, (H, W) are the spatial dimensions. Returns a [N, 4] tensors, with the boxes in xyxy format """ if masks.numel() == 0: return torch.zeros((0, 4), device=masks.device) h, w = masks.shape[-2:] y = torch.arange(0, h, dtype=torch.float) x = torch.arange(0, w, dtype=torch.float) y, x = torch.meshgrid(y, x) x_mask = (masks * x.unsqueeze(0)) x_max = x_mask.flatten(1).max(-1)[0] x_min = x_mask.masked_fill(~(masks.bool()), 1e8).flatten(1).min(-1)[0] y_mask = (masks * y.unsqueeze(0)) y_max = y_mask.flatten(1).max(-1)[0] y_min = y_mask.masked_fill(~(masks.bool()), 1e8).flatten(1).min(-1)[0] return torch.stack([x_min, y_min, x_max, y_max], 1) ================================================ FILE: util/dam.py ================================================ # ------------------------------------------------------------------------------------ # Sparse DETR # Copyright (c) 2021 KakaoBrain. All Rights Reserved. # Licensed under the Apache License, Version 2.0 [see LICENSE for details] # ------------------------------------------------------------------------------------ from pathlib import Path import numpy as np import torch import torch.nn.functional as F import matplotlib.pyplot as plt import matplotlib.patches as patches from util.box_ops import box_cxcywh_to_xyxy from util.misc import unwrap def idx_to_flat_grid(spatial_shapes, idx): flat_grid_shape = (idx.shape[0], int(torch.sum(spatial_shapes[..., 0] * spatial_shapes[..., 1]))) flat_grid = torch.zeros(flat_grid_shape, device=idx.device, dtype=torch.float32) flat_grid.scatter_(1, idx.to(torch.int64), 1) return flat_grid def attn_map_to_flat_grid(spatial_shapes, level_start_index, sampling_locations, attention_weights): # sampling_locations: [N, n_layers, Len_q, n_heads, n_levels, n_points, 2] # attention_weights: [N, n_layers, Len_q, n_heads, n_levels, n_points] N, n_layers, _, n_heads, *_ = sampling_locations.shape sampling_locations = sampling_locations.permute(0, 1, 3, 2, 5, 4, 6).flatten(0, 2).flatten(1, 2) # [N * n_layers * n_heads, Len_q * n_points, n_levels, 2] attention_weights = attention_weights.permute(0, 1, 3, 2, 5, 4).flatten(0, 2).flatten(1, 2) # [N * n_layers * n_heads, Len_q * n_points, n_levels] rev_spatial_shapes = torch.stack([spatial_shapes[..., 1], spatial_shapes[..., 0]], dim=-1) # hw -> wh (xy) col_row_float = sampling_locations * rev_spatial_shapes col_row_ll = col_row_float.floor().to(torch.int64) zero = torch.zeros(*col_row_ll.shape[:-1], dtype=torch.int64, device=col_row_ll.device) one = torch.ones(*col_row_ll.shape[:-1], dtype=torch.int64, device=col_row_ll.device) col_row_lh = col_row_ll + torch.stack([zero, one], dim=-1) col_row_hl = col_row_ll + torch.stack([one, zero], dim=-1) col_row_hh = col_row_ll + 1 margin_ll = (col_row_float - col_row_ll).prod(dim=-1) margin_lh = -(col_row_float - col_row_lh).prod(dim=-1) margin_hl = -(col_row_float - col_row_hl).prod(dim=-1) margin_hh = (col_row_float - col_row_hh).prod(dim=-1) flat_grid_shape = (attention_weights.shape[0], int(torch.sum(spatial_shapes[..., 0] * spatial_shapes[..., 1]))) flat_grid = torch.zeros(flat_grid_shape, dtype=torch.float32, device=attention_weights.device) zipped = [(col_row_ll, margin_hh), (col_row_lh, margin_hl), (col_row_hl, margin_lh), (col_row_hh, margin_ll)] for col_row, margin in zipped: valid_mask = torch.logical_and( torch.logical_and(col_row[..., 0] >= 0, col_row[..., 0] < rev_spatial_shapes[..., 0]), torch.logical_and(col_row[..., 1] >= 0, col_row[..., 1] < rev_spatial_shapes[..., 1]), ) idx = col_row[..., 1] * spatial_shapes[..., 1] + col_row[..., 0] + level_start_index idx = (idx * valid_mask).flatten(1, 2) weights = (attention_weights * valid_mask * margin).flatten(1) flat_grid.scatter_add_(1, idx, weights) return flat_grid.reshape(N, n_layers, n_heads, -1) def compute_corr(flat_grid_topk, flat_grid_attn_map, spatial_shapes): if len(flat_grid_topk.shape) == 1: flat_grid_topk = flat_grid_topk.unsqueeze(0) flat_grid_attn_map = flat_grid_attn_map.unsqueeze(0) tot = flat_grid_attn_map.sum(-1) hit = (flat_grid_topk * flat_grid_attn_map).sum(-1) corr = [hit / tot] flat_grid_idx = 0 for shape in spatial_shapes: level_range = np.arange(int(flat_grid_idx), int(flat_grid_idx + shape[0] * shape[1])) tot = (flat_grid_attn_map[:, level_range]).sum(-1) hit = (flat_grid_topk[:, level_range] * flat_grid_attn_map[:, level_range]).sum(-1) flat_grid_idx += shape[0] * shape[1] corr.append(hit / tot) return corr ================================================ FILE: util/misc.py ================================================ # ------------------------------------------------------------------------------------ # Sparse DETR # Copyright (c) 2021 KakaoBrain. All Rights Reserved. # Licensed under the Apache License, Version 2.0 [see LICENSE for details] # ------------------------------------------------------------------------------------ # Modified from Deformable DETR (https://github.com/fundamentalvision/Deformable-DETR) # Copyright (c) 2020 SenseTime. All Rights Reserved. # ------------------------------------------------------------------------ # Modified from DETR (https://github.com/facebookresearch/detr) # Copyright (c) Facebook, Inc. and its affiliates. All Rights Reserved # ------------------------------------------------------------------------ """ Misc functions, including distributed helpers. Mostly copy-paste from torchvision references. """ import os import subprocess import time from collections import defaultdict, deque import datetime import pickle import socket from typing import Optional, List import torch import torch.nn as nn import torch.distributed as dist from torch import Tensor from torch.nn.parallel import DistributedDataParallel # needed due to empty tensor bug in pytorch and torchvision 0.5 import torchvision if float(torchvision.__version__[:3]) < 0.5: import math from torchvision.ops.misc import _NewEmptyTensorOp def _check_size_scale_factor(dim, size, scale_factor): # type: (int, Optional[List[int]], Optional[float]) -> None if size is None and scale_factor is None: raise ValueError("either size or scale_factor should be defined") if size is not None and scale_factor is not None: raise ValueError("only one of size or scale_factor should be defined") if not (scale_factor is not None and len(scale_factor) != dim): raise ValueError( "scale_factor shape must match input shape. " "Input is {}D, scale_factor size is {}".format(dim, len(scale_factor)) ) def _output_size(dim, input, size, scale_factor): # type: (int, Tensor, Optional[List[int]], Optional[float]) -> List[int] assert dim == 2 _check_size_scale_factor(dim, size, scale_factor) if size is not None: return size # if dim is not 2 or scale_factor is iterable use _ntuple instead of concat assert scale_factor is not None and isinstance(scale_factor, (int, float)) scale_factors = [scale_factor, scale_factor] # math.floor might return float in py2.7 return [ int(math.floor(input.size(i + 2) * scale_factors[i])) for i in range(dim) ] elif float(torchvision.__version__[:3]) < 0.7: from torchvision.ops import _new_empty_tensor from torchvision.ops.misc import _output_size class SmoothedValue(object): """Track a series of values and provide access to smoothed values over a window or the global series average. """ def __init__(self, window_size=20, fmt=None): if fmt is None: fmt = "{median:.4f} ({global_avg:.4f})" self.deque = deque(maxlen=window_size) self.total = 0.0 self.count = 0 self.fmt = fmt def update(self, value, n=1): self.deque.append(value) self.count += n self.total += value * n def synchronize_between_processes(self): """ Warning: does not synchronize the deque! """ if not is_dist_avail_and_initialized(): return t = torch.tensor([self.count, self.total], dtype=torch.float64, device='cuda') dist.barrier() dist.all_reduce(t) t = t.tolist() self.count = int(t[0]) self.total = t[1] @property def median(self): d = torch.tensor(list(self.deque)) return d.median().item() @property def avg(self): d = torch.tensor(list(self.deque), dtype=torch.float32) return d.mean().item() @property def global_avg(self): return self.total / self.count @property def max(self): return max(self.deque) @property def value(self): return self.deque[-1] def __str__(self): return self.fmt.format( median=self.median, avg=self.avg, global_avg=self.global_avg, max=self.max, value=self.value) def unwrap(wrapped_module): if isinstance(wrapped_module, DistributedDataParallel): module = wrapped_module.module else: module = wrapped_module return module def check_unused_parameters(model, loss_dict, weight_dict): print("=== Check unused parameters ===") # print unused parameters print(f"set(loss_dict) - set(weight_dict) = {set(loss_dict.keys()) - set(weight_dict.keys())}") print(f"set(weight_dict) - set(loss_dict) = {set(weight_dict.keys()) - set(loss_dict.keys())}") unused_params = [name for name, param in unwrap(model).named_parameters() if param.grad is None and not name.startswith('backbone')] if unused_params: raise RuntimeError(f"Unused parameters: {unused_params}") else: print("All the parameters are used.") def all_gather(data): """ Run all_gather on arbitrary picklable data (not necessarily tensors) Args: data: any picklable object Returns: list[data]: list of data gathered from each rank """ world_size = get_world_size() if world_size == 1: return [data] # serialized to a Tensor buffer = pickle.dumps(data) storage = torch.ByteStorage.from_buffer(buffer) tensor = torch.ByteTensor(storage).to("cuda") # obtain Tensor size of each rank local_size = torch.tensor([tensor.numel()], device="cuda") size_list = [torch.tensor([0], device="cuda") for _ in range(world_size)] dist.all_gather(size_list, local_size) size_list = [int(size.item()) for size in size_list] max_size = max(size_list) # receiving Tensor from all ranks # we pad the tensor because torch all_gather does not support # gathering tensors of different shapes tensor_list = [] for _ in size_list: tensor_list.append(torch.empty((max_size,), dtype=torch.uint8, device="cuda")) if local_size != max_size: padding = torch.empty(size=(max_size - local_size,), dtype=torch.uint8, device="cuda") tensor = torch.cat((tensor, padding), dim=0) dist.all_gather(tensor_list, tensor) data_list = [] for size, tensor in zip(size_list, tensor_list): buffer = tensor.cpu().numpy().tobytes()[:size] data_list.append(pickle.loads(buffer)) return data_list def reduce_dict(input_dict, average=True): """ Args: input_dict (dict): all the values will be reduced average (bool): whether to do average or sum Reduce the values in the dictionary from all processes so that all processes have the averaged results. Returns a dict with the same fields as input_dict, after reduction. """ world_size = get_world_size() if world_size < 2: return input_dict with torch.no_grad(): names = [] values = [] # sort the keys so that they are consistent across processes for k in sorted(input_dict.keys()): names.append(k) values.append(input_dict[k]) values = torch.stack(values, dim=0) dist.all_reduce(values) if average: values /= world_size reduced_dict = {k: v for k, v in zip(names, values)} return reduced_dict class MetricLogger(object): def __init__(self, delimiter="\t"): self.meters = defaultdict(SmoothedValue) self.delimiter = delimiter def update(self, **kwargs): for k, v in kwargs.items(): if isinstance(v, torch.Tensor): v = v.item() assert isinstance(v, (float, int)) self.meters[k].update(v) def __getattr__(self, attr): if attr in self.meters: return self.meters[attr] if attr in self.__dict__: return self.__dict__[attr] raise AttributeError("'{}' object has no attribute '{}'".format( type(self).__name__, attr)) def __str__(self): loss_str = [] for name, meter in self.meters.items(): loss_str.append( "{}: {}".format(name, str(meter)) ) return self.delimiter.join(loss_str) def synchronize_between_processes(self): for meter in self.meters.values(): meter.synchronize_between_processes() def add_meter(self, name, meter): self.meters[name] = meter def log_every(self, iterable, print_freq, header=None): i = 0 if not header: header = '' start_time = time.time() end = time.time() iter_time = SmoothedValue(fmt='{avg:.4f}') data_time = SmoothedValue(fmt='{avg:.4f}') space_fmt = ':' + str(len(str(len(iterable)))) + 'd' if torch.cuda.is_available(): log_msg = self.delimiter.join([ header, '[{0' + space_fmt + '}/{1}]', 'eta: {eta}', '{meters}', 'time: {time}', 'data: {data}', 'max mem: {memory:.0f}' ]) else: log_msg = self.delimiter.join([ header, '[{0' + space_fmt + '}/{1}]', 'eta: {eta}', '{meters}', 'time: {time}', 'data: {data}' ]) MB = 1024.0 * 1024.0 for obj in iterable: data_time.update(time.time() - end) yield obj iter_time.update(time.time() - end) if i % print_freq == 0 or i == len(iterable) - 1: eta_seconds = iter_time.global_avg * (len(iterable) - i) eta_string = str(datetime.timedelta(seconds=int(eta_seconds))) if torch.cuda.is_available(): print(log_msg.format( i, len(iterable), eta=eta_string, meters=str(self), time=str(iter_time), data=str(data_time), memory=torch.cuda.max_memory_allocated() / MB)) else: print(log_msg.format( i, len(iterable), eta=eta_string, meters=str(self), time=str(iter_time), data=str(data_time))) i += 1 end = time.time() total_time = time.time() - start_time total_time_str = str(datetime.timedelta(seconds=int(total_time))) print('{} Total time: {} ({:.4f} s / it)'.format( header, total_time_str, total_time / len(iterable))) def get_sha(): cwd = os.path.dirname(os.path.abspath(__file__)) def _run(command): return subprocess.check_output(command, cwd=cwd).decode('ascii').strip() sha = 'N/A' diff = "clean" branch = 'N/A' try: sha = _run(['git', 'rev-parse', 'HEAD']) subprocess.check_output(['git', 'diff'], cwd=cwd) diff = _run(['git', 'diff-index', 'HEAD']) diff = "has uncommited changes" if diff else "clean" branch = _run(['git', 'rev-parse', '--abbrev-ref', 'HEAD']) except Exception: pass message = f"sha: {sha}, status: {diff}, branch: {branch}" return message def collate_fn(batch): batch = list(zip(*batch)) batch[0] = nested_tensor_from_tensor_list(batch[0]) return tuple(batch) def _max_by_axis(the_list): # type: (List[List[int]]) -> List[int] maxes = the_list[0] for sublist in the_list[1:]: for index, item in enumerate(sublist): maxes[index] = max(maxes[index], item) return maxes def nested_tensor_from_tensor_list(tensor_list: List[Tensor]): # TODO make this more general if tensor_list[0].ndim == 3: # TODO make it support different-sized images max_size = _max_by_axis([list(img.shape) for img in tensor_list]) # min_size = tuple(min(s) for s in zip(*[img.shape for img in tensor_list])) batch_shape = [len(tensor_list)] + max_size b, c, h, w = batch_shape dtype = tensor_list[0].dtype device = tensor_list[0].device tensor = torch.zeros(batch_shape, dtype=dtype, device=device) mask = torch.ones((b, h, w), dtype=torch.bool, device=device) for img, pad_img, m in zip(tensor_list, tensor, mask): pad_img[: img.shape[0], : img.shape[1], : img.shape[2]].copy_(img) m[: img.shape[1], :img.shape[2]] = False else: raise ValueError('not supported') return NestedTensor(tensor, mask) class NestedTensor(object): def __init__(self, tensors, mask: Optional[Tensor]): self.tensors = tensors self.mask = mask def to(self, device, non_blocking=False): # type: (Device) -> NestedTensor # noqa cast_tensor = self.tensors.to(device, non_blocking=non_blocking) mask = self.mask if mask is not None: assert mask is not None cast_mask = mask.to(device, non_blocking=non_blocking) else: cast_mask = None return NestedTensor(cast_tensor, cast_mask) def record_stream(self, *args, **kwargs): self.tensors.record_stream(*args, **kwargs) if self.mask is not None: self.mask.record_stream(*args, **kwargs) def decompose(self): return self.tensors, self.mask def __repr__(self): return str(self.tensors) def setup_for_distributed(is_master): """ This function disables printing when not in master process """ import builtins as __builtin__ builtin_print = __builtin__.print def print(*args, **kwargs): force = kwargs.pop('force', False) if is_master or force: builtin_print(*args, **kwargs) __builtin__.print = print def is_dist_avail_and_initialized(): if not dist.is_available(): return False if not dist.is_initialized(): return False return True def get_world_size(): if not is_dist_avail_and_initialized(): return 1 return dist.get_world_size() def get_rank(): if not is_dist_avail_and_initialized(): return 0 return dist.get_rank() def get_local_size(): if not is_dist_avail_and_initialized(): return 1 return int(os.environ['LOCAL_SIZE']) def get_local_rank(): if not is_dist_avail_and_initialized(): return 0 return int(os.environ['LOCAL_RANK']) def is_main_process(): return get_rank() == 0 def save_on_master(*args, **kwargs): if is_main_process(): torch.save(*args, **kwargs) def _check_if_valid_ip(ip): try: socket.inet_aton(ip) # legal except socket.error: # Not legal return False return True def _maybe_gethostbyname(addr): """to be compatible with Braincloud on which one can access the nodes by their task names. Each node has to wait until all the tasks in the group are up on the cloud.""" if _check_if_valid_ip(addr): # If IP address is given, do nothing return addr # Otherwise, find the IP address by hostname done = False retry = 0 print(f"Get URL by the given hostname '{addr}' in Braincloud..") while not done: try: addr = socket.gethostbyname(addr) done = True except: retry += 1 print(f"Retrying count: {retry}") time.sleep(3) print(f"Found the host by IP address: {addr}") return addr def init_distributed_mode(args): if 'RANK' in os.environ and 'WORLD_SIZE' in os.environ: os.environ["MASTER_ADDR"] = _maybe_gethostbyname(os.environ["MASTER_ADDR"]) args.rank = int(os.environ["RANK"]) args.world_size = int(os.environ['WORLD_SIZE']) args.gpu = int(os.environ['LOCAL_RANK']) args.dist_url = 'env://' os.environ['LOCAL_SIZE'] = str(torch.cuda.device_count()) elif 'SLURM_PROCID' in os.environ: proc_id = int(os.environ['SLURM_PROCID']) ntasks = int(os.environ['SLURM_NTASKS']) node_list = os.environ['SLURM_NODELIST'] num_gpus = torch.cuda.device_count() addr = subprocess.getoutput( 'scontrol show hostname {} | head -n1'.format(node_list)) os.environ['MASTER_PORT'] = os.environ.get('MASTER_PORT', '29500') os.environ['MASTER_ADDR'] = addr os.environ['WORLD_SIZE'] = str(ntasks) os.environ['RANK'] = str(proc_id) os.environ['LOCAL_RANK'] = str(proc_id % num_gpus) os.environ['LOCAL_SIZE'] = str(num_gpus) args.dist_url = 'env://' args.world_size = ntasks args.rank = proc_id args.gpu = proc_id % num_gpus else: print('Not using distributed mode') args.distributed = False return args.distributed = True torch.cuda.set_device(args.gpu) args.dist_backend = 'nccl' print('| distributed init (rank {}): {}'.format( args.rank, args.dist_url), flush=True) torch.distributed.init_process_group(backend=args.dist_backend, init_method=args.dist_url, world_size=args.world_size, rank=args.rank) torch.distributed.barrier() setup_for_distributed(args.rank == 0) @torch.no_grad() def accuracy(output, target, topk=(1,)): """Computes the precision@k for the specified values of k""" if target.numel() == 0: return [torch.zeros([], device=output.device)] maxk = max(topk) batch_size = target.size(0) _, pred = output.topk(maxk, 1, True, True) pred = pred.t() correct = pred.eq(target.view(1, -1).expand_as(pred)) res = [] for k in topk: correct_k = correct[:k].view(-1).float().sum(0) res.append(correct_k.mul_(100.0 / batch_size)) return res def interpolate(input, size=None, scale_factor=None, mode="nearest", align_corners=None): # type: (Tensor, Optional[List[int]], Optional[float], str, Optional[bool]) -> Tensor """ Equivalent to nn.functional.interpolate, but with support for empty batch sizes. This will eventually be supported natively by PyTorch, and this class can go away. """ if float(torchvision.__version__[:3]) < 0.7: if input.numel() > 0: return torch.nn.functional.interpolate( input, size, scale_factor, mode, align_corners ) output_shape = _output_size(2, input, size, scale_factor) output_shape = list(input.shape[:-2]) + list(output_shape) if float(torchvision.__version__[:3]) < 0.5: return _NewEmptyTensorOp.apply(input, output_shape) return _new_empty_tensor(input, output_shape) else: return torchvision.ops.misc.interpolate(input, size, scale_factor, mode, align_corners) def get_total_grad_norm(parameters, norm_type=2): parameters = list(filter(lambda p: p.grad is not None, parameters)) norm_type = float(norm_type) device = parameters[0].grad.device total_norm = torch.norm(torch.stack([torch.norm(p.grad.detach(), norm_type).to(device) for p in parameters]), norm_type) return total_norm def inverse_sigmoid(x, eps=1e-5): x = x.clamp(min=0, max=1) x1 = x.clamp(min=eps) x2 = (1 - x).clamp(min=eps) return torch.log(x1/x2) def scale_learning_rate(args): print("==============") if 'WORLD_SIZE' in os.environ: world_size = int(os.environ['WORLD_SIZE']) else: world_size = 1 batch_size = args.batch_size * world_size scale = (batch_size / 16) ** 0.5 print(f'Global_batch({batch_size}) = local_batch({args.batch_size}) x world_size({world_size})') print(f'Scaling factor(x{scale:.3f}) = sqrt( global_batch({batch_size}) / 16 )') for name in ['lr', 'lr_backbone']: lr_origin = getattr(args, name) lr_new = lr_origin * scale setattr(args, name, lr_new) print(f'LR scaled ({name}) : {lr_origin:.4e} -> {lr_new:.4e}') print("==============") return args ================================================ FILE: util/plot_utils.py ================================================ # ------------------------------------------------------------------------ # Deformable DETR # Copyright (c) 2020 SenseTime. All Rights Reserved. # Licensed under the Apache License, Version 2.0 [see LICENSE for details] # ------------------------------------------------------------------------ # Modified from DETR (https://github.com/facebookresearch/detr) # Copyright (c) Facebook, Inc. and its affiliates. All Rights Reserved # ------------------------------------------------------------------------ """ Plotting utilities to visualize training logs. """ import torch import pandas as pd import seaborn as sns import matplotlib.pyplot as plt from pathlib import Path, PurePath def plot_logs(logs, fields=('class_error', 'loss_bbox_unscaled', 'mAP'), ewm_col=0, log_name='log.txt'): ''' Function to plot specific fields from training log(s). Plots both training and test results. :: Inputs - logs = list containing Path objects, each pointing to individual dir with a log file - fields = which results to plot from each log file - plots both training and test for each field. - ewm_col = optional, which column to use as the exponential weighted smoothing of the plots - log_name = optional, name of log file if different than default 'log.txt'. :: Outputs - matplotlib plots of results in fields, color coded for each log file. - solid lines are training results, dashed lines are test results. ''' func_name = "plot_utils.py::plot_logs" # verify logs is a list of Paths (list[Paths]) or single Pathlib object Path, # convert single Path to list to avoid 'not iterable' error if not isinstance(logs, list): if isinstance(logs, PurePath): logs = [logs] print(f"{func_name} info: logs param expects a list argument, converted to list[Path].") else: raise ValueError(f"{func_name} - invalid argument for logs parameter.\n \ Expect list[Path] or single Path obj, received {type(logs)}") # verify valid dir(s) and that every item in list is Path object for i, dir in enumerate(logs): if not isinstance(dir, PurePath): raise ValueError(f"{func_name} - non-Path object in logs argument of {type(dir)}: \n{dir}") if dir.exists(): continue raise ValueError(f"{func_name} - invalid directory in logs argument:\n{dir}") # load log file(s) and plot dfs = [pd.read_json(Path(p) / log_name, lines=True) for p in logs] fig, axs = plt.subplots(ncols=len(fields), figsize=(16, 5)) for df, color in zip(dfs, sns.color_palette(n_colors=len(logs))): for j, field in enumerate(fields): if field == 'mAP': coco_eval = pd.DataFrame(pd.np.stack(df.test_coco_eval.dropna().values)[:, 1]).ewm(com=ewm_col).mean() axs[j].plot(coco_eval, c=color) else: df.interpolate().ewm(com=ewm_col).mean().plot( y=[f'train_{field}', f'test_{field}'], ax=axs[j], color=[color] * 2, style=['-', '--'] ) for ax, field in zip(axs, fields): ax.legend([Path(p).name for p in logs]) ax.set_title(field) def plot_precision_recall(files, naming_scheme='iter'): if naming_scheme == 'exp_id': # name becomes exp_id names = [f.parts[-3] for f in files] elif naming_scheme == 'iter': names = [f.stem for f in files] else: raise ValueError(f'not supported {naming_scheme}') fig, axs = plt.subplots(ncols=2, figsize=(16, 5)) for f, color, name in zip(files, sns.color_palette("Blues", n_colors=len(files)), names): data = torch.load(f) # precision is n_iou, n_points, n_cat, n_area, max_det precision = data['precision'] recall = data['params'].recThrs scores = data['scores'] # take precision for all classes, all areas and 100 detections precision = precision[0, :, :, 0, -1].mean(1) scores = scores[0, :, :, 0, -1].mean(1) prec = precision.mean() rec = data['recall'][0, :, 0, -1].mean() print(f'{naming_scheme} {name}: mAP@50={prec * 100: 05.1f}, ' + f'score={scores.mean():0.3f}, ' + f'f1={2 * prec * rec / (prec + rec + 1e-8):0.3f}' ) axs[0].plot(recall, precision, c=color) axs[1].plot(recall, scores, c=color) axs[0].set_title('Precision / Recall') axs[0].legend(names) axs[1].set_title('Scores / Recall') axs[1].legend(names) return fig, axs