Full Code of CVxTz/EEG_classification for AI

Repository: CVxTz/EEG_classification
Branch: master
Commit: f65060d6adb8
Files: 17
Total size: 73.1 KB

Directory structure:
gitextract_jfjg8v0s/

├── .gitignore
├── LICENSE
├── README.md
├── code/
│   ├── baseline.py
│   ├── cnn_crf_model.py
│   ├── cnn_crf_model_20_folds.py
│   ├── cnn_model.py
│   ├── eda.py
│   ├── lstm_model.py
│   ├── models.py
│   ├── run.sh
│   └── utils.py
├── deepsleepnet_data/
│   ├── dhedfreader.py
│   ├── download_physionet.sh
│   ├── prepare_physionet.py
│   └── readme.md
└── requirements.txt

================================================
FILE CONTENTS
================================================

================================================
FILE: .gitignore
================================================
# Byte-compiled / optimized / DLL files
__pycache__/
*.py[cod]
*$py.class

# C extensions
*.so

# Distribution / packaging
.Python
build/
develop-eggs/
dist/
downloads/
eggs/
.eggs/
lib/
lib64/
parts/
sdist/
var/
wheels/
*.egg-info/
.installed.cfg
*.egg
MANIFEST

# PyInstaller
#  Usually these files are written by a python script from a template
#  before PyInstaller builds the exe, so as to inject date/other infos into it.
*.manifest
*.spec

# Installer logs
pip-log.txt
pip-delete-this-directory.txt

# Unit test / coverage reports
htmlcov/
.tox/
.coverage
.coverage.*
.cache
nosetests.xml
coverage.xml
*.cover
.hypothesis/
.pytest_cache/

# Translations
*.mo
*.pot

# Django stuff:
*.log
local_settings.py
db.sqlite3

# Flask stuff:
instance/
.webassets-cache

# Scrapy stuff:
.scrapy

# Sphinx documentation
docs/_build/

# PyBuilder
target/

# Jupyter Notebook
.ipynb_checkpoints

# pyenv
.python-version

# celery beat schedule file
celerybeat-schedule

# SageMath parsed files
*.sage.py

# Environments
.env
.venv
env/
venv/
ENV/
env.bak/
venv.bak/

# Spyder project settings
.spyderproject
.spyproject

# Rope project settings
.ropeproject

# mkdocs documentation
/site

# mypy
.mypy_cache/


================================================
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 [yyyy] [name of copyright owner]

   Licensed under the Apache License, Version 2.0 (the "License");
   you may not use this file except in compliance with the License.
   You may obtain a copy of the License at

       http://www.apache.org/licenses/LICENSE-2.0

   Unless required by applicable law or agreed to in writing, software
   distributed under the License is distributed on an "AS IS" BASIS,
   WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
   See the License for the specific language governing permissions and
   limitations under the License.


================================================
FILE: README.md
================================================


[![DOI](https://zenodo.org/badge/DOI/10.5281/zenodo.4060151.svg)](https://doi.org/10.5281/zenodo.4060151)


# EEG_classification
Description of the approach : https://towardsdatascience.com/sleep-stage-classification-from-single-channel-eeg-using-convolutional-neural-networks-5c710d92d38e


Sleep Stage Classification from Single Channel EEG using Convolutional Neural
Networks

*****

<span class="figcaption_hack">Photo by [Paul
M](https://unsplash.com/photos/7i9yLoUgoP8?utm_source=unsplash&utm_medium=referral&utm_content=creditCopyText)
on
[Unsplash](https://unsplash.com/search/photos/owl?utm_source=unsplash&utm_medium=referral&utm_content=creditCopyText)</span>

Quality Sleep is an important part of a healthy lifestyle as lack of it can
cause a list of
[issues](https://www.webmd.com/sleep-disorders/features/10-results-sleep-loss#1)
like a higher risk of cancer and chronic fatigue. This means that having the
tools to automatically and easily monitor sleep can be powerful to help people
sleep better.<br> Doctors use a recording of a signal called EEG which measures
the electrical activity of the brain using an electrode to understand sleep
stages of a patient and make a diagnosis about the quality if their sleep.

In this post we will train a neural network to do the sleep stage classification
automatically from EEGs.

### **Data**

In our input we have a sequence of 30s epochs of EEG where each epoch has a
label [{“W”, “N1”, “N2”, “N3”,
“REM”}](https://en.wikipedia.org/wiki/Sleep_cycle).

<span class="figcaption_hack">Fig 1 : EEG Epoch</span>

<span class="figcaption_hack">Fig 2 : Sleep stages through the night</span>

This post is based on a publicly available EEG Sleep data (
[Sleep-EDF](https://www.physionet.org/physiobank/database/sleep-edfx/) ) that
was done on 20 subject, 19 of which have 2 full nights of sleep. We use the
pre-processing scripts available in this
[repo](https://github.com/akaraspt/deepsleepnet) and split the train/test so
that no study subject is in both at the same time.

The general objective is to go from a 1D sequence like in fig 1 and predict the
output hypnogram like in fig 2.

### Model Description

Recent approaches [[1]](https://arxiv.org/pdf/1703.04046.pdf) use a sub-model
that encodes each epoch into a 1D vector of fixed size and then a second
sequential sub-model that maps each epoch’s vector into a class from [{“W”,
“N1”, “N2”, “N3”, “REM”}](https://en.wikipedia.org/wiki/Sleep_cycle).

Here we use a 1D CNN to encode each Epoch and then another 1D CNN or LSTM that
labels the sequence of epochs to create the final
[hypnogram](https://en.wikipedia.org/wiki/Hypnogram). This allows the prediction
for an epoch to take into account the context.

<span class="figcaption_hack">Sub-model 1 : Epoch encoder</span>

<span class="figcaption_hack">Sub-model 2 : Sequential model for epoch classification</span>

The full model takes as input the sequence of EEG epochs ( 30 seconds each)
where the sub-model 1 is applied to each epoch using the TimeDistributed Layer
of [Keras](https://keras.io/) which produces a sequence of vectors. The sequence
of vectors is then fed into a another sub-model like an LSTM or a CNN that
produces the sequence of output labels.<br> We also use a linear Chain
[CRF](https://en.wikipedia.org/wiki/Conditional_random_field) for one of the
models and show that it can improve the performance.

### Training Procedure

The full model is trained end-to-end from scratch using Adam optimizer with an
initial learning rate of 1e⁻³ that is reduced each time the validation accuracy
plateaus using the ReduceLROnPlateau Keras Callbacks.

<span class="figcaption_hack">Accuracy Training curves</span>

### Results

We compare 3 different models :

* CNN-CNN : This ones used a 1D CNN for the epoch encoding and then another 1D CNN
for the sequence labeling.
* CNN-CNN-CRF : This model used a 1D CNN for the epoch encoding and then a 1D
CNN-CRF for the sequence labeling.
* CNN-LSTM : This ones used a 1D CNN for the epoch encoding and then an LSTM for
the sequence labeling.

We evaluate each model on an independent test set and get the following results
:

* CNN-CNN : F1 = 0.81, ACCURACY = 0.87
* CNN-CNN-CRF : F1 = 0.82, ACCURACY =0.89
* CNN-LSTM : F1 = 0.71, ACCURACY = 0.76

The CNN-CNN-CRF outperforms the two other models because the CRF helps learn the
transition probabilities between classes. The LSTM based model does not work as
well because it is most sensitive to hyper-parameters like the optimizer and the
batch size and requires extensive tuning to perform well.

<span class="figcaption_hack">Ground Truth Hypnogram</span>

<span class="figcaption_hack">Predicted Hypnogram using CNN-CNN-CRF</span>

Source code available here :
[https://github.com/CVxTz/EEG_classification](https://github.com/CVxTz/EEG_classification)

I look forward to your suggestions and feedback.

[[1] DeepSleepNet: a Model for Automatic Sleep Stage Scoring based on Raw
Single-Channel EEG](https://arxiv.org/pdf/1703.04046.pdf)

How to cite:
```
@software{mansar_youness_2020_4060151,
  author       = {Mansar Youness},
  title        = {CVxTz/EEG\_classification: v1.0},
  month        = sep,
  year         = 2020,
  publisher    = {Zenodo},
  version      = {v1.0},
  doi          = {10.5281/zenodo.4060151},
  url          = {https://doi.org/10.5281/zenodo.4060151}
}
```


================================================
FILE: code/baseline.py
================================================
import numpy as np
from glob import glob
import os
from sklearn.model_selection import train_test_split

base_path = "/media/ml/data_ml/EEG/deepsleepnet/data_npy"

files = glob(os.path.join(base_path, "*.npz"))
train_val, test = train_test_split(files, test_size=0.15, random_state=1337)

train, val = train_test_split(train_val, test_size=0.1, random_state=1337)

train_dict = {k: np.load(k) for k in train}
test_dict = {k: np.load(k) for k in test}
val_dict = {k: np.load(k) for k in val}





================================================
FILE: code/cnn_crf_model.py
================================================
from models import get_model_cnn_crf
import numpy as np
from utils import gen, chunker, WINDOW_SIZE, rescale_array
from keras.callbacks import ModelCheckpoint, EarlyStopping, ReduceLROnPlateau
from sklearn.metrics import f1_score, accuracy_score, classification_report
from glob import glob
import os
from sklearn.model_selection import train_test_split
from tqdm import tqdm
import matplotlib.pyplot as plt


base_path = "/media/ml/data_ml/EEG/deepsleepnet/data_npy"

files = sorted(glob(os.path.join(base_path, "*.npz")))

ids = sorted(list(set([x.split("/")[-1][:5] for x in files])))
#split by test subject
train_ids, test_ids = train_test_split(ids, test_size=0.15, random_state=1338)

train_val, test = [x for x in files if x.split("/")[-1][:5] in train_ids],\
                  [x for x in files if x.split("/")[-1][:5] in test_ids]

train, val = train_test_split(train_val, test_size=0.1, random_state=1337)

train_dict = {k: np.load(k) for k in train}
test_dict = {k: np.load(k) for k in test}
val_dict = {k: np.load(k) for k in val}

model = get_model_cnn_crf()

file_path = "cnn_crf_model.h5"
# model.load_weights(file_path)

checkpoint = ModelCheckpoint(file_path, monitor='val_acc', verbose=1, save_best_only=True, mode='max')
early = EarlyStopping(monitor="val_acc", mode="max", patience=20, verbose=1)
redonplat = ReduceLROnPlateau(monitor="val_acc", mode="max", patience=5, verbose=2)
callbacks_list = [checkpoint, early, redonplat]  # early

model.fit_generator(gen(train_dict, aug=False), validation_data=gen(val_dict), epochs=100, verbose=2,
                    steps_per_epoch=1000, validation_steps=300, callbacks=callbacks_list)
model.load_weights(file_path)


preds = []
gt = []

for record in tqdm(test_dict):
    all_rows = test_dict[record]['x']
    record_y_gt = []
    record_y_pred = []
    for batch_hyp in chunker(range(all_rows.shape[0])):


        X = all_rows[min(batch_hyp):max(batch_hyp)+1, ...]
        Y = test_dict[record]['y'][min(batch_hyp):max(batch_hyp)+1]

        X = np.expand_dims(X, 0)

        X = rescale_array(X)

        Y_pred = model.predict(X)
        Y_pred = Y_pred.argmax(axis=-1).ravel().tolist()

        gt += Y.ravel().tolist()
        preds += Y_pred

        record_y_gt += Y.ravel().tolist()
        record_y_pred += Y_pred

    # fig_1 = plt.figure(figsize=(12, 6))
    # plt.plot(record_y_gt)
    # plt.title("Sleep Stages")
    # plt.ylabel("Classes")
    # plt.xlabel("Time")
    # plt.show()
    #
    # fig_2 = plt.figure(figsize=(12, 6))
    # plt.plot(record_y_pred)
    # plt.title("Predicted Sleep Stages")
    # plt.ylabel("Classes")
    # plt.xlabel("Time")
    # plt.show()



f1 = f1_score(gt, preds, average="macro")

print("Seq Test f1 score : %s "% f1)

acc = accuracy_score(gt, preds)

print("Seq Test accuracy score : %s "% acc)

print(classification_report(gt, preds))

================================================
FILE: code/cnn_crf_model_20_folds.py
================================================
from models import get_model_cnn_crf
import numpy as np
from utils import gen, chunker, WINDOW_SIZE, rescale_array
from keras.callbacks import ModelCheckpoint, EarlyStopping, ReduceLROnPlateau
from sklearn.metrics import f1_score, accuracy_score, classification_report
from glob import glob
import os
from sklearn.model_selection import train_test_split
from tqdm import tqdm
import matplotlib.pyplot as plt


base_path = "/media/ml/data_ml/EEG/deepsleepnet/data_npy"

files = sorted(glob(os.path.join(base_path, "*.npz")))

ids = list(set([x.split("/")[-1][:5] for x in files]))
list_f1 = []
list_acc = []
preds = []
gt = []
for id in ids:
    test_ids = {id}
    train_ids = set([x.split("/")[-1][:5] for x in files]) - test_ids

    train_val, test = [x for x in files if x.split("/")[-1][:5] in train_ids],\
                      [x for x in files if x.split("/")[-1][:5] in test_ids]

    train, val = train_test_split(train_val, test_size=0.1, random_state=1337)

    train_dict = {k: np.load(k) for k in train}
    test_dict = {k: np.load(k) for k in test}
    val_dict = {k: np.load(k) for k in val}

    model = get_model_cnn_crf(lr=0.0001)

    file_path = "cnn_crf_model_20_folds.h5"
    # model.load_weights(file_path)

    checkpoint = ModelCheckpoint(file_path, monitor='val_acc', verbose=1, save_best_only=True, mode='max')
    early = EarlyStopping(monitor="val_acc", mode="max", patience=20, verbose=1)
    redonplat = ReduceLROnPlateau(monitor="val_acc", mode="max", patience=5, verbose=2)
    callbacks_list = [checkpoint, redonplat]  # early

    model.fit_generator(gen(train_dict, aug=False), validation_data=gen(val_dict), epochs=40, verbose=2,
                        steps_per_epoch=1000, validation_steps=300, callbacks=callbacks_list)
    model.load_weights(file_path)




    for record in tqdm(test_dict):
        all_rows = test_dict[record]['x']
        record_y_gt = []
        record_y_pred = []
        for batch_hyp in chunker(range(all_rows.shape[0])):


            X = all_rows[min(batch_hyp):max(batch_hyp)+1, ...]
            Y = test_dict[record]['y'][min(batch_hyp):max(batch_hyp)+1]

            X = np.expand_dims(X, 0)

            X = rescale_array(X)

            Y_pred = model.predict(X)
            Y_pred = Y_pred.argmax(axis=-1).ravel().tolist()

            gt += Y.ravel().tolist()
            preds += Y_pred

            record_y_gt += Y.ravel().tolist()
            record_y_pred += Y_pred


f1 = f1_score(gt, preds, average="macro")

acc = accuracy_score(gt, preds)

print("acc %s, f1 %s"%(acc, f1))



================================================
FILE: code/cnn_model.py
================================================
from models import get_model_cnn
import numpy as np
from utils import gen, chunker, WINDOW_SIZE, rescale_array
from keras.callbacks import ModelCheckpoint, EarlyStopping, ReduceLROnPlateau
from sklearn.metrics import f1_score, accuracy_score, classification_report
from glob import glob
import os
from sklearn.model_selection import train_test_split
from tqdm import tqdm


base_path = "/media/ml/data_ml/EEG/deepsleepnet/data_npy"

files = sorted(glob(os.path.join(base_path, "*.npz")))

ids = sorted(list(set([x.split("/")[-1][:5] for x in files])))
#split by test subject
train_ids, test_ids = train_test_split(ids, test_size=0.15, random_state=1338)

train_val, test = [x for x in files if x.split("/")[-1][:5] in train_ids],\
                  [x for x in files if x.split("/")[-1][:5] in test_ids]

train, val = train_test_split(train_val, test_size=0.1, random_state=1337)

train_dict = {k: np.load(k) for k in train}
test_dict = {k: np.load(k) for k in test}
val_dict = {k: np.load(k) for k in val}

model = get_model_cnn()

file_path = "cnn_model.h5"
# model.load_weights(file_path)

checkpoint = ModelCheckpoint(file_path, monitor='val_acc', verbose=1, save_best_only=True, mode='max')
early = EarlyStopping(monitor="val_acc", mode="max", patience=20, verbose=1)
redonplat = ReduceLROnPlateau(monitor="val_acc", mode="max", patience=5, verbose=2)
callbacks_list = [checkpoint, early, redonplat]  # early

model.fit_generator(gen(train_dict, aug=False), validation_data=gen(val_dict), epochs=100, verbose=2,
                    steps_per_epoch=1000, validation_steps=300, callbacks=callbacks_list)
model.load_weights(file_path)


preds = []
gt = []

for record in tqdm(test_dict):
    all_rows = test_dict[record]['x']
    for batch_hyp in chunker(range(all_rows.shape[0])):


        X = all_rows[min(batch_hyp):max(batch_hyp)+1, ...]
        Y = test_dict[record]['y'][min(batch_hyp):max(batch_hyp)+1]

        X = np.expand_dims(X, 0)

        X = rescale_array(X)

        Y_pred = model.predict(X)
        Y_pred = Y_pred.argmax(axis=-1).ravel().tolist()

        gt += Y.ravel().tolist()
        preds += Y_pred



f1 = f1_score(gt, preds, average="macro")

print("Seq Test f1 score : %s "% f1)

acc = accuracy_score(gt, preds)

print("Seq Test accuracy score : %s "% acc)

print(classification_report(gt, preds))

================================================
FILE: code/eda.py
================================================
import os
import h5py
import numpy as np
import matplotlib.pyplot as plt
import datetime as dt
import collections
import librosa

path = "/media/ml/data_ml/EEG/deepsleepnet/data_npy/SC4061E0.npz"

data = np.load(path)

x = data['x']
y = data['y']

fig_1 = plt.figure(figsize=(12, 6))
plt.plot(x[100, ...].ravel())
plt.title("EEG Epoch")
plt.ylabel("Amplitude")
plt.xlabel("Time")
plt.show()

fig_2 = plt.figure(figsize=(12, 6))
plt.plot(y.ravel())
plt.title("Sleep Stages")
plt.ylabel("Classes")
plt.xlabel("Time")
plt.show()

================================================
FILE: code/lstm_model.py
================================================
from models import get_model_lstm
import numpy as np
from utils import gen, chunker, WINDOW_SIZE, rescale_array
from keras.callbacks import ModelCheckpoint, EarlyStopping, ReduceLROnPlateau
from sklearn.metrics import f1_score, accuracy_score, classification_report
from glob import glob
import os
from sklearn.model_selection import train_test_split
from tqdm import tqdm


base_path = "/media/ml/data_ml/EEG/deepsleepnet/data_npy"

files = sorted(glob(os.path.join(base_path, "*.npz")))

ids = sorted(list(set([x.split("/")[-1][:5] for x in files])))
#split by test subject
train_ids, test_ids = train_test_split(ids, test_size=0.15, random_state=1338)

train_val, test = [x for x in files if x.split("/")[-1][:5] in train_ids],\
                  [x for x in files if x.split("/")[-1][:5] in test_ids]

train, val = train_test_split(train_val, test_size=0.1, random_state=1337)

train_dict = {k: np.load(k) for k in train}
test_dict = {k: np.load(k) for k in test}
val_dict = {k: np.load(k) for k in val}

model = get_model_lstm()

file_path = "lstm_model.h5"
# model.load_weights(file_path)

checkpoint = ModelCheckpoint(file_path, monitor='val_acc', verbose=1, save_best_only=True, mode='max')
early = EarlyStopping(monitor="val_acc", mode="max", patience=20, verbose=1)
redonplat = ReduceLROnPlateau(monitor="val_acc", mode="max", patience=5, verbose=2)
callbacks_list = [checkpoint, early, redonplat]  # early

model.fit_generator(gen(train_dict, aug=False), validation_data=gen(val_dict), epochs=100, verbose=2,
                    steps_per_epoch=1000, validation_steps=300, callbacks=callbacks_list)
model.load_weights(file_path)


preds = []
gt = []

for record in tqdm(test_dict):
    all_rows = test_dict[record]['x']
    for batch_hyp in chunker(range(all_rows.shape[0])):


        X = all_rows[min(batch_hyp):max(batch_hyp)+1, ...]
        Y = test_dict[record]['y'][min(batch_hyp):max(batch_hyp)+1]

        X = np.expand_dims(X, 0)

        X = rescale_array(X)

        Y_pred = model.predict(X)
        Y_pred = Y_pred.argmax(axis=-1).ravel().tolist()

        gt += Y.ravel().tolist()
        preds += Y_pred



f1 = f1_score(gt, preds, average="macro")

print("Seq Test f1 score : %s "% f1)

acc = accuracy_score(gt, preds)

print("Seq Test accuracy score : %s "% acc)

print(classification_report(gt, preds))

================================================
FILE: code/models.py
================================================
from keras import optimizers, losses, activations, models
from keras.layers import Dense, Input, Dropout, Convolution1D, MaxPool1D, GlobalMaxPool1D, GlobalAveragePooling1D, \
    concatenate, SpatialDropout1D, TimeDistributed, Bidirectional, LSTM
from keras_contrib.layers import CRF

from utils import WINDOW_SIZE

def get_model():
    nclass = 5
    inp = Input(shape=(3000, 1))
    img_1 = Convolution1D(16, kernel_size=5, activation=activations.relu, padding="valid")(inp)
    img_1 = Convolution1D(16, kernel_size=5, activation=activations.relu, padding="valid")(img_1)
    img_1 = MaxPool1D(pool_size=2)(img_1)
    img_1 = SpatialDropout1D(rate=0.01)(img_1)
    img_1 = Convolution1D(32, kernel_size=3, activation=activations.relu, padding="valid")(img_1)
    img_1 = Convolution1D(32, kernel_size=3, activation=activations.relu, padding="valid")(img_1)
    img_1 = MaxPool1D(pool_size=2)(img_1)
    img_1 = SpatialDropout1D(rate=0.01)(img_1)
    img_1 = Convolution1D(32, kernel_size=3, activation=activations.relu, padding="valid")(img_1)
    img_1 = Convolution1D(32, kernel_size=3, activation=activations.relu, padding="valid")(img_1)
    img_1 = MaxPool1D(pool_size=2)(img_1)
    img_1 = SpatialDropout1D(rate=0.01)(img_1)
    img_1 = Convolution1D(256, kernel_size=3, activation=activations.relu, padding="valid")(img_1)
    img_1 = Convolution1D(256, kernel_size=3, activation=activations.relu, padding="valid")(img_1)
    img_1 = GlobalMaxPool1D()(img_1)
    img_1 = Dropout(rate=0.01)(img_1)

    dense_1 = Dropout(rate=0.01)(Dense(64, activation=activations.relu, name="dense_1")(img_1))
    dense_1 = Dropout(rate=0.05)(Dense(64, activation=activations.relu, name="dense_2")(dense_1))
    dense_1 = Dense(nclass, activation=activations.softmax, name="dense_3")(dense_1)

    model = models.Model(inputs=inp, outputs=dense_1)
    opt = optimizers.Adam(0.001)

    model.compile(optimizer=opt, loss=losses.sparse_categorical_crossentropy, metrics=['acc'])
    model.summary()
    return model

def get_base_model():
    inp = Input(shape=(3000, 1))
    img_1 = Convolution1D(16, kernel_size=5, activation=activations.relu, padding="valid")(inp)
    img_1 = Convolution1D(16, kernel_size=5, activation=activations.relu, padding="valid")(img_1)
    img_1 = MaxPool1D(pool_size=2)(img_1)
    img_1 = SpatialDropout1D(rate=0.01)(img_1)
    img_1 = Convolution1D(32, kernel_size=3, activation=activations.relu, padding="valid")(img_1)
    img_1 = Convolution1D(32, kernel_size=3, activation=activations.relu, padding="valid")(img_1)
    img_1 = MaxPool1D(pool_size=2)(img_1)
    img_1 = SpatialDropout1D(rate=0.01)(img_1)
    img_1 = Convolution1D(32, kernel_size=3, activation=activations.relu, padding="valid")(img_1)
    img_1 = Convolution1D(32, kernel_size=3, activation=activations.relu, padding="valid")(img_1)
    img_1 = MaxPool1D(pool_size=2)(img_1)
    img_1 = SpatialDropout1D(rate=0.01)(img_1)
    img_1 = Convolution1D(256, kernel_size=3, activation=activations.relu, padding="valid")(img_1)
    img_1 = Convolution1D(256, kernel_size=3, activation=activations.relu, padding="valid")(img_1)
    img_1 = GlobalMaxPool1D()(img_1)
    img_1 = Dropout(rate=0.01)(img_1)

    dense_1 = Dropout(0.01)(Dense(64, activation=activations.relu, name="dense_1")(img_1))

    base_model = models.Model(inputs=inp, outputs=dense_1)
    opt = optimizers.Adam(0.001)

    base_model.compile(optimizer=opt, loss=losses.sparse_categorical_crossentropy, metrics=['acc'])
    #model.summary()
    return base_model


def get_model_cnn():
    nclass = 5

    seq_input = Input(shape=(None, 3000, 1))
    base_model = get_base_model()
    # for layer in base_model.layers:
    #     layer.trainable = False
    encoded_sequence = TimeDistributed(base_model)(seq_input)
    encoded_sequence = SpatialDropout1D(rate=0.01)(Convolution1D(128,
                                                               kernel_size=3,
                                                               activation="relu",
                                                               padding="same")(encoded_sequence))
    encoded_sequence = Dropout(rate=0.05)(Convolution1D(128,
                                                               kernel_size=3,
                                                               activation="relu",
                                                               padding="same")(encoded_sequence))

    #out = TimeDistributed(Dense(nclass, activation="softmax"))(encoded_sequence)
    out = Convolution1D(nclass, kernel_size=3, activation="softmax", padding="same")(encoded_sequence)

    model = models.Model(seq_input, out)

    model.compile(optimizers.Adam(0.001), losses.sparse_categorical_crossentropy, metrics=['acc'])
    model.summary()

    return model

def get_model_lstm():
    nclass = 5

    seq_input = Input(shape=(None, 3000, 1))
    base_model = get_base_model()
    for layer in base_model.layers:
        layer.trainable = False
    encoded_sequence = TimeDistributed(base_model)(seq_input)
    encoded_sequence = Bidirectional(LSTM(100, return_sequences=True))(encoded_sequence)
    encoded_sequence = Dropout(rate=0.5)(encoded_sequence)
    encoded_sequence = Bidirectional(LSTM(100, return_sequences=True))(encoded_sequence)
    #out = TimeDistributed(Dense(nclass, activation="softmax"))(encoded_sequence)
    out = Convolution1D(nclass, kernel_size=1, activation="softmax", padding="same")(encoded_sequence)

    model = models.Model(seq_input, out)

    model.compile(optimizers.Adam(0.001), losses.sparse_categorical_crossentropy, metrics=['acc'])
    model.summary()

    return model

def get_model_cnn_crf(lr=0.001):
    nclass = 5

    seq_input = Input(shape=(None, 3000, 1))
    base_model = get_base_model()
    # for layer in base_model.layers:
    #     layer.trainable = False
    encoded_sequence = TimeDistributed(base_model)(seq_input)
    encoded_sequence = SpatialDropout1D(rate=0.01)(Convolution1D(128,
                                                               kernel_size=3,
                                                               activation="relu",
                                                               padding="same")(encoded_sequence))
    encoded_sequence = Dropout(rate=0.05)(Convolution1D(128,
                                                               kernel_size=3,
                                                               activation="linear",
                                                               padding="same")(encoded_sequence))

    #out = TimeDistributed(Dense(nclass, activation="softmax"))(encoded_sequence)
    # out = Convolution1D(nclass, kernel_size=3, activation="linear", padding="same")(encoded_sequence)

    crf = CRF(nclass, sparse_target=True)

    out = crf(encoded_sequence)


    model = models.Model(seq_input, out)

    model.compile(optimizers.Adam(lr), crf.loss_function, metrics=[crf.accuracy])
    model.summary()

    return model


================================================
FILE: code/run.sh
================================================
#python cnn_model.py > cnn_logs.txt
#python cnn_crf_model.py > cnn_crf_logs.txt
#python lstm_model.py > lstm_logs.txt
python cnn_crf_model_20_folds.py > cnn_crf_folds_logs.txt

================================================
FILE: code/utils.py
================================================
import h5py
import numpy as np
import random

WINDOW_SIZE = 100

def rescale_array(X):
    X = X / 20
    X = np.clip(X, -5, 5)
    return X


def aug_X(X):
    scale = 1 + np.random.uniform(-0.1, 0.1)
    offset = np.random.uniform(-0.1, 0.1)
    noise = np.random.normal(scale=0.05, size=X.shape)
    X = scale * X + offset + noise
    return X

def gen(dict_files, aug=False):
    while True:
        record_name = random.choice(list(dict_files.keys()))
        batch_data = dict_files[record_name]
        all_rows = batch_data['x']

        for i in range(10):
            start_index = random.choice(range(all_rows.shape[0]-WINDOW_SIZE))

            X = all_rows[start_index:start_index+WINDOW_SIZE, ...]
            Y = batch_data['y'][start_index:start_index+WINDOW_SIZE]

            X = np.expand_dims(X, 0)
            Y = np.expand_dims(Y, -1)
            Y = np.expand_dims(Y, 0)

            if aug:
                X = aug_X(X)
            X = rescale_array(X)

            yield X, Y


def chunker(seq, size=WINDOW_SIZE):
    return (seq[pos:pos + size] for pos in range(0, len(seq), size))

================================================
FILE: deepsleepnet_data/dhedfreader.py
================================================
#Source : https://github.com/akaraspt/deepsleepnet

'''
Reader for EDF+ files.
TODO:
 - add support for log-transformed channels:
   http://www.edfplus.info/specs/edffloat.html and test with
   data generated with
   http://www.edfplus.info/downloads/software/NeuroLoopGain.zip.
 - check annotations with Schalk's Physiobank data.
Copyright (c) 2012 Boris Reuderink.
'''

import re, datetime, operator, logging
import numpy as np
from collections import namedtuple

EVENT_CHANNEL = 'EDF Annotations'
log = logging.getLogger(__name__)

class EDFEndOfData: pass


def tal(tal_str):
  '''Return a list with (onset, duration, annotation) tuples for an EDF+ TAL
  stream.
  '''
  exp = '(?P<onset>[+\-]\d+(?:\.\d*)?)' + \
    '(?:\x15(?P<duration>\d+(?:\.\d*)?))?' + \
    '(\x14(?P<annotation>[^\x00]*))?' + \
    '(?:\x14\x00)'

  def annotation_to_list(annotation):
    return unicode(annotation, 'utf-8').split('\x14') if annotation else []

  def parse(dic):
    return (
      float(dic['onset']),
      float(dic['duration']) if dic['duration'] else 0.,
      annotation_to_list(dic['annotation']))

  return [parse(m.groupdict()) for m in re.finditer(exp, tal_str)]


def edf_header(f):
  h = {}
  assert f.tell() == 0  # check file position
  assert f.read(8) == '0       '

  # recording info)
  h['local_subject_id'] = f.read(80).strip()
  h['local_recording_id'] = f.read(80).strip()

  # parse timestamp
  (day, month, year) = [int(x) for x in re.findall('(\d+)', f.read(8))]
  (hour, minute, sec)= [int(x) for x in re.findall('(\d+)', f.read(8))]
  h['date_time'] = str(datetime.datetime(year + 2000, month, day,
    hour, minute, sec))

  # misc
  header_nbytes = int(f.read(8))
  subtype = f.read(44)[:5]
  h['EDF+'] = subtype in ['EDF+C', 'EDF+D']
  h['contiguous'] = subtype != 'EDF+D'
  h['n_records'] = int(f.read(8))
  h['record_length'] = float(f.read(8))  # in seconds
  nchannels = h['n_channels'] = int(f.read(4))

  # read channel info
  channels = range(h['n_channels'])
  h['label'] = [f.read(16).strip() for n in channels]
  h['transducer_type'] = [f.read(80).strip() for n in channels]
  h['units'] = [f.read(8).strip() for n in channels]
  h['physical_min'] = np.asarray([float(f.read(8)) for n in channels])
  h['physical_max'] = np.asarray([float(f.read(8)) for n in channels])
  h['digital_min'] = np.asarray([float(f.read(8)) for n in channels])
  h['digital_max'] = np.asarray([float(f.read(8)) for n in channels])
  h['prefiltering'] = [f.read(80).strip() for n in channels]
  h['n_samples_per_record'] = [int(f.read(8)) for n in channels]
  f.read(32 * nchannels)  # reserved

  assert f.tell() == header_nbytes
  return h


class BaseEDFReader:
  def __init__(self, file):
    self.file = file


  def read_header(self):
    self.header = h = edf_header(self.file)

    # calculate ranges for rescaling
    self.dig_min = h['digital_min']
    self.phys_min = h['physical_min']
    phys_range = h['physical_max'] - h['physical_min']
    dig_range = h['digital_max'] - h['digital_min']
    assert np.all(phys_range > 0)
    assert np.all(dig_range > 0)
    self.gain = phys_range / dig_range


  def read_raw_record(self):
    '''Read a record with data and return a list containing arrays with raw
    bytes.
    '''
    result = []
    for nsamp in self.header['n_samples_per_record']:
      samples = self.file.read(nsamp * 2)
      if len(samples) != nsamp * 2:
        raise EDFEndOfData
      result.append(samples)
    return result


  def convert_record(self, raw_record):
    '''Convert a raw record to a (time, signals, events) tuple based on
    information in the header.
    '''
    h = self.header
    dig_min, phys_min, gain = self.dig_min, self.phys_min, self.gain
    time = float('nan')
    signals = []
    events = []
    for (i, samples) in enumerate(raw_record):
      if h['label'][i] == EVENT_CHANNEL:
        ann = tal(samples)
        time = ann[0][0]
        events.extend(ann[1:])
        # print(i, samples)
        # exit()
      else:
        # 2-byte little-endian integers
        dig = np.fromstring(samples, '<i2').astype(np.float32)
        phys = (dig - dig_min[i]) * gain[i] + phys_min[i]
        signals.append(phys)

    return time, signals, events


  def read_record(self):
    return self.convert_record(self.read_raw_record())


  def records(self):
    '''
    Record generator.
    '''
    try:
      while True:
        yield self.read_record()
    except EDFEndOfData:
      pass


def load_edf(edffile):
  '''Load an EDF+ file.
  Very basic reader for EDF and EDF+ files. While BaseEDFReader does support
  exotic features like non-homogeneous sample rates and loading only parts of
  the stream, load_edf expects a single fixed sample rate for all channels and
  tries to load the whole file.
  Parameters
  ----------
  edffile : file-like object or string
  Returns
  -------
  Named tuple with the fields:
    X : NumPy array with shape p by n.
      Raw recording of n samples in p dimensions.
    sample_rate : float
      The sample rate of the recording. Note that mixed sample-rates are not
      supported.
    sens_lab : list of length p with strings
      The labels of the sensors used to record X.
    time : NumPy array with length n
      The time offset in the recording for each sample.
    annotations : a list with tuples
      EDF+ annotations are stored in (start, duration, description) tuples.
      start : float
        Indicates the start of the event in seconds.
      duration : float
        Indicates the duration of the event in seconds.
      description : list with strings
        Contains (multiple?) descriptions of the annotation event.
  '''
  if isinstance(edffile, basestring):
    with open(edffile, 'rb') as f:
      return load_edf(f)  # convert filename to file

  reader = BaseEDFReader(edffile)
  reader.read_header()

  h = reader.header
  log.debug('EDF header: %s' % h)

  # get sample rate info
  nsamp = np.unique(
    [n for (l, n) in zip(h['label'], h['n_samples_per_record'])
    if l != EVENT_CHANNEL])
  assert nsamp.size == 1, 'Multiple sample rates not supported!'
  sample_rate = float(nsamp[0]) / h['record_length']

  rectime, X, annotations = zip(*reader.records())
  X = np.hstack(X)
  annotations = reduce(operator.add, annotations)
  chan_lab = [lab for lab in reader.header['label'] if lab != EVENT_CHANNEL]

  # create timestamps
  if reader.header['contiguous']:
    time = np.arange(X.shape[1]) / sample_rate
  else:
    reclen = reader.header['record_length']
    within_rec_time = np.linspace(0, reclen, nsamp, endpoint=False)
    time = np.hstack([t + within_rec_time for t in rectime])

  tup = namedtuple('EDF', 'X sample_rate chan_lab time annotations')
  return tup(X, sample_rate, chan_lab, time, annotations)


================================================
FILE: deepsleepnet_data/download_physionet.sh
================================================
#Source : https://github.com/akaraspt/deepsleepnet
wget https://www.physionet.org/physiobank/database/sleep-edfx/sleep-cassette/SC4001E0-PSG.edf
wget https://www.physionet.org/physiobank/database/sleep-edfx/sleep-cassette/SC4001E0-PSG.edf.hyp
wget https://www.physionet.org/physiobank/database/sleep-edfx/sleep-cassette/SC4001EC-Hypnogram.edf
wget https://www.physionet.org/physiobank/database/sleep-edfx/sleep-cassette/SC4002E0-PSG.edf
wget https://www.physionet.org/physiobank/database/sleep-edfx/sleep-cassette/SC4002E0-PSG.edf.hyp
wget https://www.physionet.org/physiobank/database/sleep-edfx/sleep-cassette/SC4002EC-Hypnogram.edf
wget https://www.physionet.org/physiobank/database/sleep-edfx/sleep-cassette/SC4011E0-PSG.edf
wget https://www.physionet.org/physiobank/database/sleep-edfx/sleep-cassette/SC4011E0-PSG.edf.hyp
wget https://www.physionet.org/physiobank/database/sleep-edfx/sleep-cassette/SC4011EH-Hypnogram.edf
wget https://www.physionet.org/physiobank/database/sleep-edfx/sleep-cassette/SC4012E0-PSG.edf
wget https://www.physionet.org/physiobank/database/sleep-edfx/sleep-cassette/SC4012E0-PSG.edf.hyp
wget https://www.physionet.org/physiobank/database/sleep-edfx/sleep-cassette/SC4012EC-Hypnogram.edf
wget https://www.physionet.org/physiobank/database/sleep-edfx/sleep-cassette/SC4021E0-PSG.edf
wget https://www.physionet.org/physiobank/database/sleep-edfx/sleep-cassette/SC4021E0-PSG.edf.hyp
wget https://www.physionet.org/physiobank/database/sleep-edfx/sleep-cassette/SC4021EH-Hypnogram.edf
wget https://www.physionet.org/physiobank/database/sleep-edfx/sleep-cassette/SC4022E0-PSG.edf
wget https://www.physionet.org/physiobank/database/sleep-edfx/sleep-cassette/SC4022E0-PSG.edf.hyp
wget https://www.physionet.org/physiobank/database/sleep-edfx/sleep-cassette/SC4022EJ-Hypnogram.edf
wget https://www.physionet.org/physiobank/database/sleep-edfx/sleep-cassette/SC4031E0-PSG.edf
wget https://www.physionet.org/physiobank/database/sleep-edfx/sleep-cassette/SC4031E0-PSG.edf.hyp
wget https://www.physionet.org/physiobank/database/sleep-edfx/sleep-cassette/SC4031EC-Hypnogram.edf
wget https://www.physionet.org/physiobank/database/sleep-edfx/sleep-cassette/SC4032E0-PSG.edf
wget https://www.physionet.org/physiobank/database/sleep-edfx/sleep-cassette/SC4032E0-PSG.edf.hyp
wget https://www.physionet.org/physiobank/database/sleep-edfx/sleep-cassette/SC4032EP-Hypnogram.edf
wget https://www.physionet.org/physiobank/database/sleep-edfx/sleep-cassette/SC4041E0-PSG.edf
wget https://www.physionet.org/physiobank/database/sleep-edfx/sleep-cassette/SC4041E0-PSG.edf.hyp
wget https://www.physionet.org/physiobank/database/sleep-edfx/sleep-cassette/SC4041EC-Hypnogram.edf
wget https://www.physionet.org/physiobank/database/sleep-edfx/sleep-cassette/SC4042E0-PSG.edf
wget https://www.physionet.org/physiobank/database/sleep-edfx/sleep-cassette/SC4042E0-PSG.edf.hyp
wget https://www.physionet.org/physiobank/database/sleep-edfx/sleep-cassette/SC4042EC-Hypnogram.edf
wget https://www.physionet.org/physiobank/database/sleep-edfx/sleep-cassette/SC4051E0-PSG.edf
wget https://www.physionet.org/physiobank/database/sleep-edfx/sleep-cassette/SC4051E0-PSG.edf.hyp
wget https://www.physionet.org/physiobank/database/sleep-edfx/sleep-cassette/SC4051EC-Hypnogram.edf
wget https://www.physionet.org/physiobank/database/sleep-edfx/sleep-cassette/SC4052E0-PSG.edf
wget https://www.physionet.org/physiobank/database/sleep-edfx/sleep-cassette/SC4052E0-PSG.edf.hyp
wget https://www.physionet.org/physiobank/database/sleep-edfx/sleep-cassette/SC4052EC-Hypnogram.edf
wget https://www.physionet.org/physiobank/database/sleep-edfx/sleep-cassette/SC4061E0-PSG.edf
wget https://www.physionet.org/physiobank/database/sleep-edfx/sleep-cassette/SC4061E0-PSG.edf.hyp
wget https://www.physionet.org/physiobank/database/sleep-edfx/sleep-cassette/SC4061EC-Hypnogram.edf
wget https://www.physionet.org/physiobank/database/sleep-edfx/sleep-cassette/SC4062E0-PSG.edf
wget https://www.physionet.org/physiobank/database/sleep-edfx/sleep-cassette/SC4062E0-PSG.edf.hyp
wget https://www.physionet.org/physiobank/database/sleep-edfx/sleep-cassette/SC4062EC-Hypnogram.edf
wget https://www.physionet.org/physiobank/database/sleep-edfx/sleep-cassette/SC4071E0-PSG.edf
wget https://www.physionet.org/physiobank/database/sleep-edfx/sleep-cassette/SC4071E0-PSG.edf.hyp
wget https://www.physionet.org/physiobank/database/sleep-edfx/sleep-cassette/SC4071EC-Hypnogram.edf
wget https://www.physionet.org/physiobank/database/sleep-edfx/sleep-cassette/SC4072E0-PSG.edf
wget https://www.physionet.org/physiobank/database/sleep-edfx/sleep-cassette/SC4072E0-PSG.edf.hyp
wget https://www.physionet.org/physiobank/database/sleep-edfx/sleep-cassette/SC4072EH-Hypnogram.edf
wget https://www.physionet.org/physiobank/database/sleep-edfx/sleep-cassette/SC4081E0-PSG.edf
wget https://www.physionet.org/physiobank/database/sleep-edfx/sleep-cassette/SC4081E0-PSG.edf.hyp
wget https://www.physionet.org/physiobank/database/sleep-edfx/sleep-cassette/SC4081EC-Hypnogram.edf
wget https://www.physionet.org/physiobank/database/sleep-edfx/sleep-cassette/SC4082E0-PSG.edf
wget https://www.physionet.org/physiobank/database/sleep-edfx/sleep-cassette/SC4082E0-PSG.edf.hyp
wget https://www.physionet.org/physiobank/database/sleep-edfx/sleep-cassette/SC4082EP-Hypnogram.edf
wget https://www.physionet.org/physiobank/database/sleep-edfx/sleep-cassette/SC4091E0-PSG.edf
wget https://www.physionet.org/physiobank/database/sleep-edfx/sleep-cassette/SC4091E0-PSG.edf.hyp
wget https://www.physionet.org/physiobank/database/sleep-edfx/sleep-cassette/SC4091EC-Hypnogram.edf
wget https://www.physionet.org/physiobank/database/sleep-edfx/sleep-cassette/SC4092E0-PSG.edf
wget https://www.physionet.org/physiobank/database/sleep-edfx/sleep-cassette/SC4092E0-PSG.edf.hyp
wget https://www.physionet.org/physiobank/database/sleep-edfx/sleep-cassette/SC4092EC-Hypnogram.edf
wget https://www.physionet.org/physiobank/database/sleep-edfx/sleep-cassette/SC4101E0-PSG.edf
wget https://www.physionet.org/physiobank/database/sleep-edfx/sleep-cassette/SC4101E0-PSG.edf.hyp
wget https://www.physionet.org/physiobank/database/sleep-edfx/sleep-cassette/SC4101EC-Hypnogram.edf
wget https://www.physionet.org/physiobank/database/sleep-edfx/sleep-cassette/SC4102E0-PSG.edf
wget https://www.physionet.org/physiobank/database/sleep-edfx/sleep-cassette/SC4102E0-PSG.edf.hyp
wget https://www.physionet.org/physiobank/database/sleep-edfx/sleep-cassette/SC4102EC-Hypnogram.edf
wget https://www.physionet.org/physiobank/database/sleep-edfx/sleep-cassette/SC4111E0-PSG.edf
wget https://www.physionet.org/physiobank/database/sleep-edfx/sleep-cassette/SC4111E0-PSG.edf.hyp
wget https://www.physionet.org/physiobank/database/sleep-edfx/sleep-cassette/SC4111EC-Hypnogram.edf
wget https://www.physionet.org/physiobank/database/sleep-edfx/sleep-cassette/SC4112E0-PSG.edf
wget https://www.physionet.org/physiobank/database/sleep-edfx/sleep-cassette/SC4112E0-PSG.edf.hyp
wget https://www.physionet.org/physiobank/database/sleep-edfx/sleep-cassette/SC4112EC-Hypnogram.edf
wget https://www.physionet.org/physiobank/database/sleep-edfx/sleep-cassette/SC4121E0-PSG.edf
wget https://www.physionet.org/physiobank/database/sleep-edfx/sleep-cassette/SC4121E0-PSG.edf.hyp
wget https://www.physionet.org/physiobank/database/sleep-edfx/sleep-cassette/SC4121EC-Hypnogram.edf
wget https://www.physionet.org/physiobank/database/sleep-edfx/sleep-cassette/SC4122E0-PSG.edf
wget https://www.physionet.org/physiobank/database/sleep-edfx/sleep-cassette/SC4122E0-PSG.edf.hyp
wget https://www.physionet.org/physiobank/database/sleep-edfx/sleep-cassette/SC4122EV-Hypnogram.edf
wget https://www.physionet.org/physiobank/database/sleep-edfx/sleep-cassette/SC4131E0-PSG.edf
wget https://www.physionet.org/physiobank/database/sleep-edfx/sleep-cassette/SC4131E0-PSG.edf.hyp
wget https://www.physionet.org/physiobank/database/sleep-edfx/sleep-cassette/SC4131EC-Hypnogram.edf
wget https://www.physionet.org/physiobank/database/sleep-edfx/sleep-cassette/SC4141E0-PSG.edf
wget https://www.physionet.org/physiobank/database/sleep-edfx/sleep-cassette/SC4141E0-PSG.edf.hyp
wget https://www.physionet.org/physiobank/database/sleep-edfx/sleep-cassette/SC4141EU-Hypnogram.edf
wget https://www.physionet.org/physiobank/database/sleep-edfx/sleep-cassette/SC4142E0-PSG.edf
wget https://www.physionet.org/physiobank/database/sleep-edfx/sleep-cassette/SC4142E0-PSG.edf.hyp
wget https://www.physionet.org/physiobank/database/sleep-edfx/sleep-cassette/SC4142EU-Hypnogram.edf
wget https://www.physionet.org/physiobank/database/sleep-edfx/sleep-cassette/SC4151E0-PSG.edf
wget https://www.physionet.org/physiobank/database/sleep-edfx/sleep-cassette/SC4151E0-PSG.edf.hyp
wget https://www.physionet.org/physiobank/database/sleep-edfx/sleep-cassette/SC4151EC-Hypnogram.edf
wget https://www.physionet.org/physiobank/database/sleep-edfx/sleep-cassette/SC4152E0-PSG.edf
wget https://www.physionet.org/physiobank/database/sleep-edfx/sleep-cassette/SC4152E0-PSG.edf.hyp
wget https://www.physionet.org/physiobank/database/sleep-edfx/sleep-cassette/SC4152EC-Hypnogram.edf
wget https://www.physionet.org/physiobank/database/sleep-edfx/sleep-cassette/SC4161E0-PSG.edf
wget https://www.physionet.org/physiobank/database/sleep-edfx/sleep-cassette/SC4161E0-PSG.edf.hyp
wget https://www.physionet.org/physiobank/database/sleep-edfx/sleep-cassette/SC4161EC-Hypnogram.edf
wget https://www.physionet.org/physiobank/database/sleep-edfx/sleep-cassette/SC4162E0-PSG.edf
wget https://www.physionet.org/physiobank/database/sleep-edfx/sleep-cassette/SC4162E0-PSG.edf.hyp
wget https://www.physionet.org/physiobank/database/sleep-edfx/sleep-cassette/SC4162EC-Hypnogram.edf
wget https://www.physionet.org/physiobank/database/sleep-edfx/sleep-cassette/SC4171E0-PSG.edf
wget https://www.physionet.org/physiobank/database/sleep-edfx/sleep-cassette/SC4171E0-PSG.edf.hyp
wget https://www.physionet.org/physiobank/database/sleep-edfx/sleep-cassette/SC4171EU-Hypnogram.edf
wget https://www.physionet.org/physiobank/database/sleep-edfx/sleep-cassette/SC4172E0-PSG.edf
wget https://www.physionet.org/physiobank/database/sleep-edfx/sleep-cassette/SC4172E0-PSG.edf.hyp
wget https://www.physionet.org/physiobank/database/sleep-edfx/sleep-cassette/SC4172EC-Hypnogram.edf
wget https://www.physionet.org/physiobank/database/sleep-edfx/sleep-cassette/SC4181E0-PSG.edf
wget https://www.physionet.org/physiobank/database/sleep-edfx/sleep-cassette/SC4181E0-PSG.edf.hyp
wget https://www.physionet.org/physiobank/database/sleep-edfx/sleep-cassette/SC4181EC-Hypnogram.edf
wget https://www.physionet.org/physiobank/database/sleep-edfx/sleep-cassette/SC4182E0-PSG.edf
wget https://www.physionet.org/physiobank/database/sleep-edfx/sleep-cassette/SC4182E0-PSG.edf.hyp
wget https://www.physionet.org/physiobank/database/sleep-edfx/sleep-cassette/SC4182EC-Hypnogram.edf
wget https://www.physionet.org/physiobank/database/sleep-edfx/sleep-cassette/SC4191E0-PSG.edf
wget https://www.physionet.org/physiobank/database/sleep-edfx/sleep-cassette/SC4191E0-PSG.edf.hyp
wget https://www.physionet.org/physiobank/database/sleep-edfx/sleep-cassette/SC4191EP-Hypnogram.edf
wget https://www.physionet.org/physiobank/database/sleep-edfx/sleep-cassette/SC4192E0-PSG.edf
wget https://www.physionet.org/physiobank/database/sleep-edfx/sleep-cassette/SC4192E0-PSG.edf.hyp
wget https://www.physionet.org/physiobank/database/sleep-edfx/sleep-cassette/SC4192EV-Hypnogram.edf


================================================
FILE: deepsleepnet_data/prepare_physionet.py
================================================
#Source : https://github.com/akaraspt/deepsleepnet

import argparse
import glob
import math
import ntpath
import os
import shutil
import urllib
import urllib2

from datetime import datetime

import numpy as np
import pandas as pd

from mne import Epochs, pick_types, find_events
from mne.io import concatenate_raws, read_raw_edf

import dhedfreader


# Label values
W = 0
N1 = 1
N2 = 2
N3 = 3
REM = 4
UNKNOWN = 5

stage_dict = {
    "W": W,
    "N1": N1,
    "N2": N2,
    "N3": N3,
    "REM": REM,
    "UNKNOWN": UNKNOWN
}

class_dict = {
    0: "W",
    1: "N1",
    2: "N2",
    3: "N3",
    4: "REM",
    5: "UNKNOWN"
}

ann2label = {
    "Sleep stage W": 0,
    "Sleep stage 1": 1,
    "Sleep stage 2": 2,
    "Sleep stage 3": 3,
    "Sleep stage 4": 3,
    "Sleep stage R": 4,
    "Sleep stage ?": 5,
    "Movement time": 5
}

EPOCH_SEC_SIZE = 30


def main():
    parser = argparse.ArgumentParser()
    parser.add_argument("--data_dir", type=str, default="/data/physionet_sleep",
                        help="File path to the CSV or NPY file that contains walking data.")
    parser.add_argument("--output_dir", type=str, default="/data/physionet_sleep/eeg_fpz_cz",
                        help="Directory where to save outputs.")
    parser.add_argument("--select_ch", type=str, default="EEG Fpz-Cz",
                        help="File path to the trained model used to estimate walking speeds.")
    args = parser.parse_args()

    # Output dir
    if not os.path.exists(args.output_dir):
        os.makedirs(args.output_dir)
    else:
        shutil.rmtree(args.output_dir)
        os.makedirs(args.output_dir)

    # Select channel
    select_ch = args.select_ch

    # Read raw and annotation EDF files
    psg_fnames = glob.glob(os.path.join(args.data_dir, "*PSG.edf"))
    ann_fnames = glob.glob(os.path.join(args.data_dir, "*Hypnogram.edf"))
    psg_fnames.sort()
    ann_fnames.sort()
    psg_fnames = np.asarray(psg_fnames)
    ann_fnames = np.asarray(ann_fnames)

    for i in range(len(psg_fnames)):
        # if not "ST7171J0-PSG.edf" in psg_fnames[i]:
        #     continue

        raw = read_raw_edf(psg_fnames[i], preload=True, stim_channel=None)
        sampling_rate = raw.info['sfreq']
        raw_ch_df = raw.to_data_frame(scaling_time=100.0)[select_ch]
        raw_ch_df = raw_ch_df.to_frame()
        raw_ch_df.set_index(np.arange(len(raw_ch_df)))

        # Get raw header
        f = open(psg_fnames[i], 'r')
        reader_raw = dhedfreader.BaseEDFReader(f)
        reader_raw.read_header()
        h_raw = reader_raw.header
        f.close()
        raw_start_dt = datetime.strptime(h_raw['date_time'], "%Y-%m-%d %H:%M:%S")

        # Read annotation and its header
        f = open(ann_fnames[i], 'r')
        reader_ann = dhedfreader.BaseEDFReader(f)
        reader_ann.read_header()
        h_ann = reader_ann.header
        _, _, ann = zip(*reader_ann.records())
        f.close()
        ann_start_dt = datetime.strptime(h_ann['date_time'], "%Y-%m-%d %H:%M:%S")

        # Assert that raw and annotation files start at the same time
        assert raw_start_dt == ann_start_dt

        # Generate label and remove indices
        remove_idx = []    # indicies of the data that will be removed
        labels = []        # indicies of the data that have labels
        label_idx = []
        for a in ann[0]:
            onset_sec, duration_sec, ann_char = a
            ann_str = "".join(ann_char)
            label = ann2label[ann_str]
            if label != UNKNOWN:
                if duration_sec % EPOCH_SEC_SIZE != 0:
                    raise Exception("Something wrong")
                duration_epoch = int(duration_sec / EPOCH_SEC_SIZE)
                label_epoch = np.ones(duration_epoch, dtype=np.int) * label
                labels.append(label_epoch)
                idx = int(onset_sec * sampling_rate) + np.arange(duration_sec * sampling_rate, dtype=np.int)
                label_idx.append(idx)

                print "Include onset:{}, duration:{}, label:{} ({})".format(
                    onset_sec, duration_sec, label, ann_str
                )
            else:
                idx = int(onset_sec * sampling_rate) + np.arange(duration_sec * sampling_rate, dtype=np.int)
                remove_idx.append(idx)

                print "Remove onset:{}, duration:{}, label:{} ({})".format(
                    onset_sec, duration_sec, label, ann_str
                )
        labels = np.hstack(labels)
        
        print "before remove unwanted: {}".format(np.arange(len(raw_ch_df)).shape)
        if len(remove_idx) > 0:
            remove_idx = np.hstack(remove_idx)
            select_idx = np.setdiff1d(np.arange(len(raw_ch_df)), remove_idx)
        else:
            select_idx = np.arange(len(raw_ch_df))
        print "after remove unwanted: {}".format(select_idx.shape)

        # Select only the data with labels
        print "before intersect label: {}".format(select_idx.shape)
        label_idx = np.hstack(label_idx)
        select_idx = np.intersect1d(select_idx, label_idx)
        print "after intersect label: {}".format(select_idx.shape)

        # Remove extra index
        if len(label_idx) > len(select_idx):
            print "before remove extra labels: {}, {}".format(select_idx.shape, labels.shape)
            extra_idx = np.setdiff1d(label_idx, select_idx)
            # Trim the tail
            if np.all(extra_idx > select_idx[-1]):
                n_trims = len(select_idx) % int(EPOCH_SEC_SIZE * sampling_rate)
                n_label_trims = int(math.ceil(n_trims / (EPOCH_SEC_SIZE * sampling_rate)))
                select_idx = select_idx[:-n_trims]
                labels = labels[:-n_label_trims]
            print "after remove extra labels: {}, {}".format(select_idx.shape, labels.shape)

        # Remove movement and unknown stages if any
        raw_ch = raw_ch_df.values[select_idx]

        # Verify that we can split into 30-s epochs
        if len(raw_ch) % (EPOCH_SEC_SIZE * sampling_rate) != 0:
            raise Exception("Something wrong")
        n_epochs = len(raw_ch) / (EPOCH_SEC_SIZE * sampling_rate)

        # Get epochs and their corresponding labels
        x = np.asarray(np.split(raw_ch, n_epochs)).astype(np.float32)
        y = labels.astype(np.int32)

        assert len(x) == len(y)

        # Select on sleep periods
        w_edge_mins = 30
        nw_idx = np.where(y != stage_dict["W"])[0]
        start_idx = nw_idx[0] - (w_edge_mins * 2)
        end_idx = nw_idx[-1] + (w_edge_mins * 2)
        if start_idx < 0: start_idx = 0
        if end_idx >= len(y): end_idx = len(y) - 1
        select_idx = np.arange(start_idx, end_idx+1)
        print("Data before selection: {}, {}".format(x.shape, y.shape))
        x = x[select_idx]
        y = y[select_idx]
        print("Data after selection: {}, {}".format(x.shape, y.shape))

        # Save
        filename = ntpath.basename(psg_fnames[i]).replace("-PSG.edf", ".npz")
        save_dict = {
            "x": x, 
            "y": y, 
            "fs": sampling_rate,
            "ch_label": select_ch,
            "header_raw": h_raw,
            "header_annotation": h_ann,
        }
        np.savez(os.path.join(args.output_dir, filename), **save_dict)

        print "\n=======================================\n"


if __name__ == "__main__":
    main()


================================================
FILE: deepsleepnet_data/readme.md
================================================
The files in this folders were copied with minor modifications from #Source : https://github.com/akaraspt/deepsleepnet

#To get the dataset :

cd data
chmod +x download_physionet.sh
./download_physionet.sh

###Those scripts taken from the deepsleepnet only work with python2
python2 prepare_physionet.py --data_dir data --output_dir data/eeg_fpz_cz --select_ch 'EEG Fpz-Cz'

This subfolder is under the following license : 

Copyright 2017 Akara Supratak and Hao Dong.  All rights reserved.

                                 Apache License
                           Version 2.0, January 2004
                        http://www.apache.org/licenses/

   TERMS AND CONDITIONS FOR USE, REPRODUCTION, AND DISTRIBUTION

   1. Definitions.

      "License" shall mean the terms and conditions for use, reproduction,
      and distribution as defined by Sections 1 through 9 of this document.

      "Licensor" shall mean the copyright owner or entity authorized by
      the copyright owner that is granting the License.

      "Legal Entity" shall mean the union of the acting entity and all
      other entities that control, are controlled by, or are under common
      control with that entity. For the purposes of this definition,
      "control" means (i) the power, direct or indirect, to cause the
      direction or management of such entity, whether by contract or
      otherwise, or (ii) ownership of fifty percent (50%) or more of the
      outstanding shares, or (iii) beneficial ownership of such entity.

      "You" (or "Your") shall mean an individual or Legal Entity
      exercising permissions granted by this License.

      "Source" form shall mean the preferred form for making modifications,
      including but not limited to software source code, documentation
      source, and configuration files.

      "Object" form shall mean any form resulting from mechanical
      transformation or translation of a Source form, including but
      not limited to compiled object code, generated documentation,
      and conversions to other media types.

      "Work" shall mean the work of authorship, whether in Source or
      Object form, made available under the License, as indicated by a
      copyright notice that is included in or attached to the work
      (an example is provided in the Appendix below).

      "Derivative Works" shall mean any work, whether in Source or Object
      form, that is based on (or derived from) the Work and for which the
      editorial revisions, annotations, elaborations, or other modifications
      represent, as a whole, an original work of authorship. For the purposes
      of this License, Derivative Works shall not include works that remain
      separable from, or merely link (or bind by name) to the interfaces of,
      the Work and Derivative Works thereof.

      "Contribution" shall mean any work of authorship, including
      the original version of the Work and any modifications or additions
      to that Work or Derivative Works thereof, that is intentionally
      submitted to Licensor for inclusion in the Work by the copyright owner
      or by an individual or Legal Entity authorized to submit on behalf of
      the copyright owner. For the purposes of this definition, "submitted"
      means any form of electronic, verbal, or written communication sent
      to the Licensor or its representatives, including but not limited to
      communication on electronic mailing lists, source code control systems,
      and issue tracking systems that are managed by, or on behalf of, the
      Licensor for the purpose of discussing and improving the Work, but
      excluding communication that is conspicuously marked or otherwise
      designated in writing by the copyright owner as "Not a Contribution."

      "Contributor" shall mean Licensor and any individual or Legal Entity
      on behalf of whom a Contribution has been received by Licensor and
      subsequently incorporated within the Work.

   2. Grant of Copyright License. Subject to the terms and conditions of
      this License, each Contributor hereby grants to You a perpetual,
      worldwide, non-exclusive, no-charge, royalty-free, irrevocable
      copyright license to reproduce, prepare Derivative Works of,
      publicly display, publicly perform, sublicense, and distribute the
      Work and such Derivative Works in Source or Object form.

   3. Grant of Patent License. Subject to the terms and conditions of
      this License, each Contributor hereby grants to You a perpetual,
      worldwide, non-exclusive, no-charge, royalty-free, irrevocable
      (except as stated in this section) patent license to make, have made,
      use, offer to sell, sell, import, and otherwise transfer the Work,
      where such license applies only to those patent claims licensable
      by such Contributor that are necessarily infringed by their
      Contribution(s) alone or by combination of their Contribution(s)
      with the Work to which such Contribution(s) was submitted. If You
      institute patent litigation against any entity (including a
      cross-claim or counterclaim in a lawsuit) alleging that the Work
      or a Contribution incorporated within the Work constitutes direct
      or contributory patent infringement, then any patent licenses
      granted to You under this License for that Work shall terminate
      as of the date such litigation is filed.

   4. Redistribution. You may reproduce and distribute copies of the
      Work or Derivative Works thereof in any medium, with or without
      modifications, and in Source or Object form, provided that You
      meet the following conditions:

      (a) You must give any other recipients of the Work or
          Derivative Works a copy of this License; and

      (b) You must cause any modified files to carry prominent notices
          stating that You changed the files; and

      (c) You must retain, in the Source form of any Derivative Works
          that You distribute, all copyright, patent, trademark, and
          attribution notices from the Source form of the Work,
          excluding those notices that do not pertain to any part of
          the Derivative Works; and

      (d) If the Work includes a "NOTICE" text file as part of its
          distribution, then any Derivative Works that You distribute must
          include a readable copy of the attribution notices contained
          within such NOTICE file, excluding those notices that do not
          pertain to any part of the Derivative Works, in at least one
          of the following places: within a NOTICE text file distributed
          as part of the Derivative Works; within the Source form or
          documentation, if provided along with the Derivative Works; or,
          within a display generated by the Derivative Works, if and
          wherever such third-party notices normally appear. The contents
          of the NOTICE file are for informational purposes only and
          do not modify the License. You may add Your own attribution
          notices within Derivative Works that You distribute, alongside
          or as an addendum to the NOTICE text from the Work, provided
          that such additional attribution notices cannot be construed
          as modifying the License.

      You may add Your own copyright statement to Your modifications and
      may provide additional or different license terms and conditions
      for use, reproduction, or distribution of Your modifications, or
      for any such Derivative Works as a whole, provided Your use,
      reproduction, and distribution of the Work otherwise complies with
      the conditions stated in this License.

   5. Submission of Contributions. Unless You explicitly state otherwise,
      any Contribution intentionally submitted for inclusion in the Work
      by You to the Licensor shall be under the terms and conditions of
      this License, without any additional terms or conditions.
      Notwithstanding the above, nothing herein shall supersede or modify
      the terms of any separate license agreement you may have executed
      with Licensor regarding such Contributions.

   6. Trademarks. This License does not grant permission to use the trade
      names, trademarks, service marks, or product names of the Licensor,
      except as required for reasonable and customary use in describing the
      origin of the Work and reproducing the content of the NOTICE file.

   7. Disclaimer of Warranty. Unless required by applicable law or
      agreed to in writing, Licensor provides the Work (and each
      Contributor provides its Contributions) on an "AS IS" BASIS,
      WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or
      implied, including, without limitation, any warranties or conditions
      of TITLE, NON-INFRINGEMENT, MERCHANTABILITY, or FITNESS FOR A
      PARTICULAR PURPOSE. You are solely responsible for determining the
      appropriateness of using or redistributing the Work and assume any
      risks associated with Your exercise of permissions under this License.

   8. Limitation of Liability. In no event and under no legal theory,
      whether in tort (including negligence), contract, or otherwise,
      unless required by applicable law (such as deliberate and grossly
      negligent acts) or agreed to in writing, shall any Contributor be
      liable to You for damages, including any direct, indirect, special,
      incidental, or consequential damages of any character arising as a
      result of this License or out of the use or inability to use the
      Work (including but not limited to damages for loss of goodwill,
      work stoppage, computer failure or malfunction, or any and all
      other commercial damages or losses), even if such Contributor
      has been advised of the possibility of such damages.

   9. Accepting Warranty or Additional Liability. While redistributing
      the Work or Derivative Works thereof, You may choose to offer,
      and charge a fee for, acceptance of support, warranty, indemnity,
      or other liability obligations and/or rights consistent with this
      License. However, in accepting such obligations, You may act only
      on Your own behalf and on Your sole responsibility, not on behalf
      of any other Contributor, and only if You agree to indemnify,
      defend, and hold each Contributor harmless for any liability
      incurred by, or claims asserted against, such Contributor by reason
      of your accepting any such warranty or additional liability.

   END OF TERMS AND CONDITIONS

   APPENDIX: How to apply the Apache License to your work.

      To apply the Apache License to your work, attach the following
      boilerplate notice, with the fields enclosed by brackets "{}"
      replaced with your own identifying information. (Don't include
      the brackets!)  The text should be enclosed in the appropriate
      comment syntax for the file format. We also recommend that a
      file or class name and description of purpose be included on the
      same "printed page" as the copyright notice for easier
      identification within third-party archives.

   Copyright 2017 Akara Supratak and Hao Dong

   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: requirements.txt
================================================
librosa==0.5.1
numpy==1.15.2
Keras==2.2.2
tqdm==4.23.2
keras_contrib==2.0.8
h5py==2.8.0
matplotlib==2.1.0
scikit_learn==0.20.0