Plot brain decode (#345)

* Fix bug for MotorImagery All events, setting n_classes * Fix bug for MotorImagery All events, setting n_classes * Fix is Valid Change default value n_classes=None * Update docs/source/whats_new.rst * plot_BrainDecode first draft * [pre-commit.ci] auto fixes from pre-commit.com hooks * plot_BrainDecode * [pre-commit.ci] auto fixes from pre-commit.com hooks * plot_BrainDecode * plot_BrainDecode * Update examples/plot_BrainDecode.py Co-authored-by: Sylvain Chevallier <sylvain.chevallier@universite-paris-saclay.fr> * Update examples/plot_BrainDecode.py Co-authored-by: Sylvain Chevallier <sylvain.chevallier@universite-paris-saclay.fr> * plot_BrainDecode * plot_BrainDecode, definition of setup_seed in moabb.utils * plot_BrainDecode, definition of setup_seed in moabb.utils * brain decode try benchmark * [pre-commit.ci] auto fixes from pre-commit.com hooks * Plot_function * [pre-commit.ci] auto fixes from pre-commit.com hooks * Delete PyTorch_EEGNetv4.yml * Delete plot_benchmark_BrainDecode.py * Renaming the file * Changing init file * Changing the tutorial * Renaming the file to lower * Updating the InputSetterEEG * Adding new function to get the models from braindecode, and forcing the reinitialize model if the shape change. * Partially init the shallow model * Changing to ShallowNet * =) Solving the issue with hardcode variable and changing the way we compare the variable. * draft test * removing test * Plot_function * [pre-commit.ci] auto fixes from pre-commit.com hooks * update plot_braindecode * [pre-commit.ci] auto fixes from pre-commit.com hooks * Plot BranDecode with dynamically input setter * [pre-commit.ci] auto fixes from pre-commit.com hooks * Plot BranDecode with dynamically input setter * [pre-commit.ci] auto fixes from pre-commit.com hooks * remove plot_BrainDecode from GitIgnore * Plot_Benchmark: - EEGNetv4 with BrainDecode (in_chans) - EEGInception (in_channels) - Dataset BNCI2014001 (22channel) and BNCI2014004 (3channel) - Change in benchmark function: if a pipeline have BrainDecode in the name it is gonna set the return_epochs to True * [pre-commit.ci] auto fixes from pre-commit.com hooks * Plot_Benchmark: - EEGNetv4 with BrainDecode (in_chans) - EEGInception (in_channels) - ShallowFBCSPNet - Deep4Net - Dataset BNCI2014001 (22channel) and BNCI2014004 (3channel) - Change in benchmark function: if a pipeline have BrainDecode in the name it is gonna set the return_epochs to True * Plot_Benchmark: - EEGNetv4 with BrainDecode (in_chans) - EEGInception (in_channels) - ShallowFBCSPNet - Deep4Net - Dataset BNCI2014001 (22channel) and BNCI2014004 (3channel) - Change in benchmark function: if a pipeline have BrainDecode in the name it is gonna set the return_epochs to True * [pre-commit.ci] auto fixes from pre-commit.com hooks * precommit stuffs * cleaning line * Changing the set seed * Changing the import * Renaming BrainDecode to braindecode * Update moabb/benchmark.py * renaming to lowercase * lower case for pipeline folder of braindecode * Update examples/plot_braindecode.py Co-authored-by: Sylvain Chevallier <sylvain.chevallier@universite-paris-saclay.fr> * Update examples/plot_braindecode.py Co-authored-by: Sylvain Chevallier <sylvain.chevallier@universite-paris-saclay.fr> * Update examples/pipelines_braindecode/braindecode_Deep4Net.py Co-authored-by: Sylvain Chevallier <sylvain.chevallier@universite-paris-saclay.fr> * Update examples/pipelines_braindecode/braindecode_Deep4Net.py Co-authored-by: Sylvain Chevallier <sylvain.chevallier@universite-paris-saclay.fr> * Update examples/pipelines_braindecode/braindecode_EEGInception.py Co-authored-by: Sylvain Chevallier <sylvain.chevallier@universite-paris-saclay.fr> * Update examples/pipelines_braindecode/braindecode_EEGNetv4.py Co-authored-by: Sylvain Chevallier <sylvain.chevallier@universite-paris-saclay.fr> * Modification requested * Modification requested * typos, and another small general corrections. * Fixing small things for the review * Update examples/plot_benchmark_braindecode.py Co-authored-by: Sylvain Chevallier <sylvain.chevallier@universite-paris-saclay.fr> * Update examples/plot_braindecode.py Co-authored-by: Sylvain Chevallier <sylvain.chevallier@universite-paris-saclay.fr> * Update examples/pipelines_braindecode/braindecode_ShallowFBCSPNet.py Co-authored-by: Sylvain Chevallier <sylvain.chevallier@universite-paris-saclay.fr> * fixing tests * Adding the Pipeline of braincode in the pipeline folder * removing deep4net * fixing test * adding text * changing to parametrize * Updating function * Updating test function * changing the init file * adding braindecode to the folder * testing hidden util braindecode * splitting the util tests to not download * updating the test --------- Co-authored-by: Sylvain Chevallier <sylvain.chevallier@universite-paris-saclay.fr> Co-authored-by: pre-commit-ci[bot] <66853113+pre-commit-ci[bot]@users.noreply.github.com> Co-authored-by: bruAristimunha <a.bruno@aluno.ufabc.edu.br>
NeuroTechX · Mar 19, 2023 · 14b55a1 · 14b55a1
1 parent 01a8133
commit 14b55a1
Show file tree

Hide file tree

Showing 18 changed files with 1,158 additions and 24 deletions.
diff --git a/docs/source/pipelines.rst b/docs/source/pipelines.rst
@@ -16,11 +16,11 @@ Pipelines
 
     features.LogVariance
     features.FM
-	 features.ExtendedSSVEPSignal
+    features.ExtendedSSVEPSignal
     features.AugmentedDataset
     features.StandardScaler_Epoch
-	 csp.TRCSP
-	 classification.SSVEP_CCA
+    csp.TRCSP
+	classification.SSVEP_CCA
     deep_learning.KerasDeepConvNet
     deep_learning.KerasEEGITNet
     deep_learning.KerasEEGNet_8_2
@@ -42,3 +42,5 @@ Base & Utils
     utils_deep_model.EEGNet
     utils_deep_model.EEGNet_TC
     utils_deep_model.TCN_block
+    utils_pytorch.BraindecodeDatasetLoader
+    utils_pytorch.InputShapeSetterEEG
diff --git a/docs/source/whats_new.rst b/docs/source/whats_new.rst
@@ -25,6 +25,7 @@ Enhancements
 - Adding new tutorial to benchmark with GridSearchCV (:gh:`323` by `Igor Carrara`_)
 - Add six deep learning models (Tensorflow), and build a tutorial to show to use the deep learning model (:gh:`326` by `Igor Carrara`_, `Bruno Aristimunha`_ and `Sylvain Chevallier`_)
 - Add a augmentation model to the pipeline (:gh:`326` by `Igor Carrara`_)
+- Add BrainDecode example(:gh:`340` by `Igor Carrara`_ and `Bruno Aristimunha`_)
 - Add Google Analytics to the documentation (:gh:`335` by `Bruno Aristimunha`_)
 
 Bugs

diff --git a/examples/pipelines_braindecode/braindecode_EEGInception.py b/examples/pipelines_braindecode/braindecode_EEGInception.py
@@ -0,0 +1,64 @@
+import torch
+from braindecode import EEGClassifier
+from braindecode.models import EEGInception
+from sklearn.pipeline import Pipeline
+from skorch.callbacks import EarlyStopping, EpochScoring
+from skorch.dataset import ValidSplit
+
+from moabb.pipelines.utils_pytorch import BraindecodeDatasetLoader, InputShapeSetterEEG
+
+
+# Set up GPU if it is there
+cuda = torch.cuda.is_available()
+device = "cuda" if cuda else "cpu"
+
+# Hyperparameter
+LEARNING_RATE = 0.0001
+WEIGHT_DECAY = 0
+BATCH_SIZE = 64
+SEED = 42
+VERBOSE = 1
+EPOCH = 10
+PATIENCE = 3
+
+# Create the dataset
+create_dataset = BraindecodeDatasetLoader()
+
+# Set EEG Inception model
+model = EEGInception(in_channels=1, n_classes=2, input_window_samples=100)
+
+# Define a Skorch classifier
+clf = EEGClassifier(
+    module=model,
+    criterion=torch.nn.CrossEntropyLoss,
+    optimizer=torch.optim.Adam,
+    optimizer__lr=LEARNING_RATE,
+    batch_size=BATCH_SIZE,
+    max_epochs=EPOCH,
+    train_split=ValidSplit(0.2, random_state=SEED),
+    device=device,
+    callbacks=[
+        EarlyStopping(monitor="valid_loss", patience=PATIENCE),
+        EpochScoring(
+            scoring="accuracy", on_train=True, name="train_acc", lower_is_better=False
+        ),
+        EpochScoring(
+            scoring="accuracy", on_train=False, name="valid_acc", lower_is_better=False
+        ),
+        InputShapeSetterEEG(
+            params_list=["in_channels", "input_window_samples", "n_classes"],
+        ),
+    ],
+    verbose=VERBOSE,  # Not printing the results for each epoch
+)
+
+# Create the pipelines
+pipes = Pipeline([("Braindecode_dataset", create_dataset), ("EEGInception", clf)])
+
+# this is what will be loaded
+PIPELINE = {
+    "name": "braindecode_EEGInception",
+    "paradigms": ["LeftRightImagery", "MotorImagery"],
+    "pipeline": pipes,
+    "citations": "https://doi.org/10.1109/TNSRE.2020.3048106",
+}
diff --git a/examples/pipelines_braindecode/braindecode_EEGNetv4.py b/examples/pipelines_braindecode/braindecode_EEGNetv4.py
@@ -0,0 +1,64 @@
+import torch
+from braindecode import EEGClassifier
+from braindecode.models import EEGNetv4
+from sklearn.pipeline import Pipeline
+from skorch.callbacks import EarlyStopping, EpochScoring
+from skorch.dataset import ValidSplit
+
+from moabb.pipelines.utils_pytorch import BraindecodeDatasetLoader, InputShapeSetterEEG
+
+
+# Set up GPU if it is there
+cuda = torch.cuda.is_available()
+device = "cuda" if cuda else "cpu"
+
+# Hyperparameter
+LEARNING_RATE = 0.0001
+WEIGHT_DECAY = 0
+BATCH_SIZE = 64
+SEED = 42
+VERBOSE = 1
+EPOCH = 10
+PATIENCE = 3
+
+# Create the dataset
+create_dataset = BraindecodeDatasetLoader()
+
+# Set EEGNetv4 model
+model = EEGNetv4(in_chans=1, n_classes=2, input_window_samples=100)
+
+# Define a Skorch classifier
+clf = EEGClassifier(
+    module=model,
+    criterion=torch.nn.CrossEntropyLoss,
+    optimizer=torch.optim.Adam,
+    optimizer__lr=LEARNING_RATE,
+    batch_size=BATCH_SIZE,
+    max_epochs=EPOCH,
+    train_split=ValidSplit(0.2, random_state=SEED),
+    device=device,
+    callbacks=[
+        EarlyStopping(monitor="valid_loss", patience=PATIENCE),
+        EpochScoring(
+            scoring="accuracy", on_train=True, name="train_acc", lower_is_better=False
+        ),
+        EpochScoring(
+            scoring="accuracy", on_train=False, name="valid_acc", lower_is_better=False
+        ),
+        InputShapeSetterEEG(
+            params_list=["in_chans", "input_window_samples", "n_classes"],
+        ),
+    ],
+    verbose=VERBOSE,  # Not printing the results for each epoch
+)
+
+# Create the pipelines
+pipes = Pipeline([("braindecode_dataset", create_dataset), ("EEGNetv4", clf)])
+
+# this is what will be loaded
+PIPELINE = {
+    "name": "braindecode_EEGNetv4",
+    "paradigms": ["LeftRightImagery", "MotorImagery"],
+    "pipeline": pipes,
+    "citations": "https://doi.org/10.1088/1741-2552/aace8c",
+}
diff --git a/examples/pipelines_braindecode/braindecode_ShallowFBCSPNet.py b/examples/pipelines_braindecode/braindecode_ShallowFBCSPNet.py
@@ -0,0 +1,66 @@
+import torch
+from braindecode import EEGClassifier
+from braindecode.models import ShallowFBCSPNet
+from sklearn.pipeline import Pipeline
+from skorch.callbacks import EarlyStopping, EpochScoring
+from skorch.dataset import ValidSplit
+
+from moabb.pipelines.utils_pytorch import BraindecodeDatasetLoader, InputShapeSetterEEG
+
+
+# Set up GPU if it is there
+cuda = torch.cuda.is_available()
+device = "cuda" if cuda else "cpu"
+
+# Hyperparameter
+LEARNING_RATE = 0.0001
+WEIGHT_DECAY = 0
+BATCH_SIZE = 64
+SEED = 42
+VERBOSE = 1
+EPOCH = 10
+PATIENCE = 3
+
+# Create the dataset
+create_dataset = BraindecodeDatasetLoader()
+
+# Set Shallow Filter Bank CSP Net model
+model = ShallowFBCSPNet(
+    in_chans=1, n_classes=2, input_window_samples=100, final_conv_length="auto"
+)
+
+# Define a Skorch classifier
+clf = EEGClassifier(
+    module=model,
+    criterion=torch.nn.CrossEntropyLoss,
+    optimizer=torch.optim.Adam,
+    optimizer__lr=LEARNING_RATE,
+    batch_size=BATCH_SIZE,
+    max_epochs=EPOCH,
+    train_split=ValidSplit(0.2, random_state=SEED),
+    device=device,
+    callbacks=[
+        EarlyStopping(monitor="valid_loss", patience=PATIENCE),
+        EpochScoring(
+            scoring="accuracy", on_train=True, name="train_acc", lower_is_better=False
+        ),
+        EpochScoring(
+            scoring="accuracy", on_train=False, name="valid_acc", lower_is_better=False
+        ),
+        InputShapeSetterEEG(
+            params_list=["in_chans", "input_window_samples", "n_classes"],
+        ),
+    ],
+    verbose=VERBOSE,  # Not printing the results for each epoch
+)
+
+# Create the pipelines
+pipes = Pipeline([("braindecode_dataset", create_dataset), ("ShallowFBCSPNet", clf)])
+
+# this is what will be loaded
+PIPELINE = {
+    "name": "braindecode_ShallowFBCSPNet",
+    "paradigms": ["LeftRightImagery", "MotorImagery"],
+    "pipeline": pipes,
+    "citations": "https://doi.org/10.1002/hbm.23730",
+}
diff --git a/examples/plot_benchmark_braindecode.py b/examples/plot_benchmark_braindecode.py
@@ -0,0 +1,121 @@
+"""
+=======================================================================
+Benchmarking on MOABB with Braindecode (PyTorch) deep net architectures
+=======================================================================
+This example shows how to use MOABB to benchmark a set of Braindecode pipelines (deep learning
+architectures) on all available datasets.
+For this example, we will use only 2 datasets to keep the computation time low, but this benchmark is designed
+to easily scale to many datasets.
+"""
+# Authors: Igor Carrara <igor.carrara@inria.fr>
+#          Bruno Aristimunha <b.aristimunha@gmail.com>
+#          Sylvain Chevallier <sylvain.chevallier@universite-paris-saclay.fr>
+#
+# License: BSD (3-clause)
+
+import os
+
+import matplotlib.pyplot as plt
+import torch
+from absl.logging import ERROR, set_verbosity
+
+from moabb import benchmark, set_log_level
+from moabb.analysis.plotting import score_plot
+from moabb.datasets import BNCI2014001, BNCI2014004
+from moabb.utils import setup_seed
+
+
+set_log_level("info")
+# Avoid output Warning
+set_verbosity(ERROR)
+os.environ["TF_CPP_MIN_LOG_LEVEL"] = "3"
+
+# Print Information PyTorch
+print(f"Torch Version: {torch.__version__}")
+
+# Set up GPU if it is there
+cuda = torch.cuda.is_available()
+device = "cuda" if cuda else "cpu"
+print("GPU is", "AVAILABLE" if cuda else "NOT AVAILABLE")
+
+###############################################################################
+# In this example, we will use only 2 subjects from the dataset ``BNCI2014001`` and ``BNCI2014004``.
+#
+# Running the benchmark
+# ---------------------
+#
+# The benchmark is run using the ``benchmark`` function. You need to specify the
+# folder containing the pipelines, the kind of evaluation, and the paradigm
+# to use. By default, the benchmark will use all available datasets for all
+# paradigms listed in the pipelines. You could restrict to specific evaluation and
+# paradigm using the ``evaluations`` and ``paradigms`` arguments.
+#
+# To save computation time, the results are cached. If you want to re-run the
+# benchmark, you can set the ``overwrite`` argument to ``True``.
+#
+# It is possible to indicate the folder to cache the results and the one to save
+# the analysis & figures. By default, the results are saved in the ``results``
+# folder, and the analysis & figures are saved in the ``benchmark`` folder.
+#
+# This code is implemented to run on CPU. If you're using a GPU, do not use multithreading
+# (i.e. set n_jobs=1)
+#
+# In order to allow the benchmark function to work with return_epoch=True (Required to use Braindecode(
+# we need to call each pipeline as "braindecode_xxx...", with xxx the name of the model to be
+# handled correctly by the benchmark function.
+
+# Set up reproducibility of Tensorflow
+setup_seed(42)
+
+# Restrict this example only to the first two subjects of BNCI2014001
+dataset = BNCI2014001()
+dataset2 = BNCI2014004()
+dataset.subject_list = dataset.subject_list[:2]
+dataset2.subject_list = dataset2.subject_list[:2]
+datasets = [dataset, dataset2]
+
+results = benchmark(
+    pipelines="./pipelines_braindecode",
+    evaluations=["CrossSession"],
+    paradigms=["LeftRightImagery"],
+    include_datasets=datasets,
+    results="./results/",
+    overwrite=False,
+    plot=False,
+    output="./benchmark/",
+    n_jobs=-1,
+)
+
+###############################################################################
+# The deep learning architectures implemented in MOABB using Braindecode are:
+#
+# - Shallow Convolutional Network [1]_
+# - Deep Convolutional Network [1]_
+# - EEGNetv4 [2]_
+# - EEGInception [3]_
+#
+# Benchmark prints a summary of the results. Detailed results are saved in a
+# pandas dataframe, and can be used to generate figures. The analysis & figures
+# are saved in the ``benchmark`` folder.
+
+score_plot(results)
+plt.show()
+
+##############################################################################
+# References
+# ----------
+# .. [1] Schirrmeister, R. T., Springenberg, J. T., Fiederer, L. D. J.,
+#    Glasstetter, M., Eggensperger, K., Tangermann, M., ... & Ball, T. (2017).
+#    `Deep learning with convolutional neural networks for EEG decoding and
+#    visualization <https://doi.org/10.1002/hbm.23730>`_.
+#    Human brain mapping, 38(11), 5391-5420.
+# .. [2] Lawhern, V. J., Solon, A. J., Waytowich, N. R., Gordon, S. M.,
+#    Hung, C. P., & Lance, B. J. (2018). `EEGNet: a compact convolutional neural
+#    network for EEG-based brain-computer interfaces.
+#    <https://doi.org/10.1088/1741-2552/aace8c>`_
+#    Journal of neural engineering, 15(5), 056013.
+# .. [3] Santamaria-Vazquez, E., Martinez-Cagigal, V., Vaquerizo-Villar,
+#    F., & Hornero, R. (2020). `EEG-inception: A novel deep convolutional neural network
+#    for assistive ERP-based brain-computer interfaces.
+#    <https://doi.org/10.1109/TNSRE.2020.3048106>`_
+#    IEEE Transactions on Neural Systems and Rehabilitation Engineering