Add bi2012 dependency

doc/api.rst

@@ -57,4 +57,5 @@ Datasets
     get_mne_sample
     get_linearly_separable_dataset
     get_qiskit_dataset
+    get_bi2012_dataset
 

doc/requirements.txt

@@ -2,7 +2,7 @@ sphinx-gallery
 sphinx-bootstrap_theme
 numpydoc
 cython
-mne[data]>=0.24
+mne==0.20.1
 seaborn
 scikit-learn
 joblib
@@ -11,3 +11,4 @@ cvxpy==1.1.12
 qiskit==0.20.0
 scipy==1.7.3
 git+https://github.com/pyRiemann/pyRiemann#egg=pyriemann
+braininvaders2012>=0.1

pyriemann_qiskit/datasets/__init__.py

@@ -1,8 +1,10 @@
 from .utils import (get_mne_sample,
                     get_linearly_separable_dataset,
-                    get_qiskit_dataset)
+                    get_qiskit_dataset,
+                    get_bi2012_dataset)
 
 
 __all__ = ["get_mne_sample",
            "get_linearly_separable_dataset",
-           "get_qiskit_dataset"]
+           "get_qiskit_dataset",
+           "get_bi2012_dataset"]

pyriemann_qiskit/datasets/utils.py

@@ -1,8 +1,10 @@
 import numpy as np
-from mne import io, read_events, pick_types, Epochs
+from mne import io, read_events, pick_types, Epochs, find_events
 from mne.datasets import sample
 from qiskit.ml.datasets import ad_hoc_data
 from sklearn.datasets import make_classification
+from sklearn.preprocessing import LabelEncoder
+from braininvaders2012.dataset import BrainInvaders2012
 
 
 def get_mne_sample(n_trials=10):
@@ -39,6 +41,7 @@ def get_mne_sample(n_trials=10):
         https://mne.tools/stable/overview/datasets_index.html
 
     """
+
     data_path = sample.data_path()
 
     # Set parameters and read data
@@ -123,3 +126,94 @@ def get_linearly_separable_dataset():
     X += 2 * rng.uniform(size=X.shape)
 
     return(X, y)
+
+
+def get_bi2012_dataset(training=True,
+                       fmin=1, fmax=24, sampling=128, tmin=0.1, tmax=0.7,
+                       baseline=None, verbose=False):
+    """Return sample data from the brain invaders 2012 dataset.
+
+    Return an iterator over the brain invaders 2012 dataset [1]_.
+    Each call to the _next_ method return the epochs `X` and the predicted
+    target vector `y` relative to `X` for a subject in the dataset.
+
+    The dataset contains the EEG recording of 26 participants playing a BCI
+    version of the vintage game Brain Invaders, based on the
+    oddball paradigm (P300).
+    Each subject participated in a Training and Online session of the game.
+
+    A detailed description of the dataset can be found in [2]_.
+
+    Parameters
+    ----------
+    training : bool (default:True)
+        If true will only download and analyze data related
+        to the training session.
+    fmin : int (default:1)
+        Minimum frequence (Hz) for passband filtering (data preprocessing).
+    fmax : int (default:24)
+        Maximum frequence (Hz) for passband filtering (data preprocessing).
+    sampling : int (default:128)
+        Data were initially sampled at 128Hz. You can resample the data using
+        this parameter.
+    tmin : int (default:0.1)
+        Start time before event.
+    tmax : int (default:0.7)
+        Stop time before event.
+    baseline : None | tuple[2] (default:None)
+        The time interval to apply baseline correction [3]_.
+    verbose : bool (default:False)
+        Print all traces if true.
+
+    Returns
+    -------
+    dataset: Iterator
+       An iterator over the brain invaders 2012 dataset.
+
+    References
+    ----------
+    .. [1] Available from: \
+        https://github.com/plcrodrigues/py.BI.EEG.2012-GIPSA
+    .. [2] G. F. P. Van Veen et al. \
+        ‘Building Brain Invaders: EEG data of an experimental validation’
+        (mai 2019). Available at:  \
+        https://hal.archives-ouvertes.fr/hal-02126068
+    .. [3] https://mne.tools/0.20/generated/mne.Epochs.html
+
+    """
+
+    dataset = BrainInvaders2012(Training=training, Online=not training)
+    subjects = dataset.subject_list
+
+    class Iterator():
+        def _iter_(self):
+            self.index = 0
+
+        def _next_(self):
+            if self.index == len(subjects - 1):
+                raise StopIteration
+
+            subject = subjects[self.index]
+            self.index += 1
+            data = dataset._get_single_subject_data(subject)
+            run = "run_" + ("training" if training else "online")
+            raw = data['session_1'][run]
+
+            # filter data and resample
+            raw.filter(fmin, fmax, verbose=verbose)
+            raw.resample(sampling)
+
+            # detect the events and cut the signal into epochs
+            events = find_events(raw=raw, shortest_event=1, verbose=verbose)
+            event_id = {'NonTarget': 1, 'Target': 2}
+            epochs = Epochs(raw, events, event_id, tmin, tmax,
+                            baseline=baseline, verbose=verbose, preload=True)
+            epochs.pick_types(eeg=True)
+
+            # get trials and labels
+            X = epochs.get_data()
+            y = events[:, -1]
+            y = LabelEncoder().fit_transform(y)
+            return (X, y)
+
+    return Iterator()

setup.py

@@ -38,9 +38,11 @@
                         'scipy==1.5.4; python_version < "3.7.0"',
                         'scipy==1.7.3; python_version >= "3.7.0"',
                         'docplex',
+                        'braininvaders2012>=0.1',
+                        'mne==0.20.1',
                         'tqdm'
                         ],
-      extras_require={'docs': ['sphinx-gallery', 'sphinx-bootstrap_theme', 'numpydoc', 'mne', 'seaborn'],
-                      'tests': ['pytest', 'seaborn', 'flake8', 'mne', 'pooch', 'tqdm']},
+      extras_require={'docs': ['sphinx-gallery', 'sphinx-bootstrap_theme', 'numpydoc', 'seaborn'],
+                      'tests': ['pytest', 'seaborn', 'flake8', 'pooch', 'tqdm']},
       zip_safe=False,
 )