print(_pilotX.shape)
if (_compX.shape[2] > _pilotX.shape[2]):
_compX = _compX[:, :, :_pilotX.shape[2]]
elif (_compX.shape[2] < _pilotX.shape[2]):
_pilotX = _pilotX[:, :, :_compX.shape[2]]
chans, samples = _pilotX.shape[1], _pilotX.shape[2]
Path(WEIGHT_PATH).mkdir(parents=True, exist_ok=True)
# stratify and split pilot data for transfer learning
skf = StratifiedKFold(TRANSFER_FOLDS, shuffle=True)
comp_avg = {'acc': 0, 'bal': 0, 'kap': 0}
pilot_avg = {'acc': 0, 'bal': 0, 'kap': 0}
for i in range(0, FOLDS):
comp_trainX, comp_valX, comp_testX = add_kernel_dim(get_fold(
_compX, FOLDS, i, test_val_rest_split),
kernels=KERNELS)
comp_trainY, comp_valY, comp_testY = onehot(
get_fold(_compY, FOLDS, i, test_val_rest_split))
# weight file path
weight_file = f"{WEIGHT_PATH}/{i+1}.h5"
# initialise model
model = EEGNet(nb_classes=CLASSES,
Chans=chans,
Samples=samples,
dropoutRate=0.5,
kernLength=64,
F1=8,
D=2,
### script start
_compX, _compY = epoch_comp(prep_comp(comp_channel_map2, GOODS), CLASSES)
_pilotX, _pilotY = epoch_pilot(prep_pilot('data/rivet/VIPA_BCIpilot_imaginedmovement.vhdr', GOODS), CLASSES)
chans, samples = _pilotX.shape[1], _pilotX.shape[2]
Path(WEIGHT_PATH).mkdir(parents=True, exist_ok=True)
pilotX = add_kernel_dim((_pilotX,), kernels=KERNELS)
pilotY = onehot((_pilotY,))
comp_avg = {'acc': 0, 'bal': 0, 'kap': 0}
pilot_avg = {'acc': 0, 'bal': 0, 'kap': 0}
for i in range(0, FOLDS):
trainX, valX, compX = add_kernel_dim(get_fold(_compX, FOLDS, i, test_val_rest_split), kernels=KERNELS)
trainY, valY, compY = onehot(get_fold(_compY, FOLDS, i, test_val_rest_split))
# weight file path
weight_file = f"{WEIGHT_PATH}/{i+1}.h5"
# initialise model
model = EEGNet(
nb_classes=CLASSES, Chans=chans, Samples=samples,
dropoutRate=0.5, kernLength=125, F1=16, D=4, F2=64,
dropoutType='Dropout'
)
# compile model
model.compile(loss='categorical_crossentropy', optimizer='adam', metrics=['accuracy'])
batch_size=16,
epochs=50,
verbose=0,
validation_data=(validation['x'], validation['y']),
callbacks=[checkpointer])
### script start
compX, compY = epoch_comp(prep_comp({}), CLASSES)
chans, samples = compX.shape[1], compX.shape[2]
Path(WEIGHT_PATH).mkdir(parents=True, exist_ok=True)
comp_avg = {'acc': 0, 'bal': 0, 'kap': 0}
for i in range(0, FOLDS):
trainX, valX, testX = add_kernel_dim(get_fold(compX, FOLDS, i,
test_val_rest_split),
kernels=KERNELS)
trainY, valY, testY = onehot(get_fold(compY, FOLDS, i,
test_val_rest_split))
# weight file path
weight_file = f"{WEIGHT_PATH}/{i+1}.h5"
# initialise model
model = EEGNet(nb_classes=CLASSES,
Chans=chans,
Samples=samples,
dropoutRate=0.5,
kernLength=125,
F1=16,
D=4,
def get_model(transfer_paths=None):
K.clear_session()
model = EEGNet(nb_classes=3,
Chans=9,
Samples=250,
dropoutRate=0.5,
kernLength=64,
F1=8,
D=2,
F2=16,
dropoutType='Dropout')
# compile model loss function and optimizer for transfer learning
model.compile(loss='categorical_crossentropy',
optimizer='adam',
metrics=['accuracy'])
# load base model
model.load_weights(BASE_WEIGHTS)
# K.clear_session()
print(transfer_paths)
if not transfer_paths or len(transfer_paths) is 0:
return (model, True)
print(len(transfer_paths))
try:
transfer_raw = read_raw_brainvision(transfer_paths[0], preload=True)
if len(transfer_paths) > 1:
for i in range(1, len(transfer_paths)):
print(i)
transfer_raw = concatenate_raws([
transfer_raw,
read_raw_brainvision(transfer_paths[i], preload=True)
])
except Exception as e:
print('failed', e)
return (model, None)
transX, transY = epoch_pilot(transfer_raw,
n_classes=3,
good_channels=GOODS,
resample=RESAMPLE,
trange=T_RANGE,
l_freq=LO_FREQ,
h_freq=HI_FREQ)
# separate 4:1 train:validation
transX, transY = stratify(transX, transY, 5)
trans_trainX, trans_valX = add_kernel_dim(get_fold(transX, 5, 0,
test_rest_split),
kernels=1)
trans_trainY, trans_valY = onehot(get_fold(transY, 5, 0, test_rest_split))
trans_valY, _ = onehot((trans_valY, []))
# perform transfer learning on the base model and selected transfer file
train(model, {
"x": trans_trainX,
"y": trans_trainY
}, {
"x": trans_valX,
"y": trans_valY
},
epochs=EPOCHS)
return (model, False)
_pilotX, _pilotY = epoch_pilot(prepall_pilot(GOODS, h_freq=LOW_PASS),
CLASSES,
resample=RESAMPLE,
trange=T_RANGE)
# _pilotX, _pilotY = epoch_pilot(prep_pilot('data/rivet/VIPA_BCIpilot_realmovement.vhdr', GOODS, l_freq=0.5, h_freq=30.), CLASSES)
chans, samples = _pilotX.shape[1], _pilotX.shape[2]
Path(WEIGHT_PATH).mkdir(parents=True, exist_ok=True)
# stratify and split pilot data for transfer learning
skf = StratifiedKFold(TRANSFER_FOLDS, shuffle=True)
comp_avg = {'acc': 0, 'bal': 0, 'kap': 0}
pilot_avg = {'acc': 0, 'bal': 0, 'kap': 0}
for i in range(0, FOLDS):
comp_trainX, comp_testX = get_fold(_compX, FOLDS, i, test_rest_split)
comp_trainY, comp_testY = get_fold(_compY, FOLDS, i, test_rest_split)
comp_trainX, comp_trainY = stratify(comp_trainX, comp_trainY, FOLDS - 1)
comp_trainX, comp_valX = get_fold(_compX, FOLDS - 1, 0, test_rest_split)
comp_trainY, comp_valY = get_fold(_compY, FOLDS - 1, 0, test_rest_split)
comp_trainX, comp_valX, comp_testX = add_kernel_dim(
(comp_trainX, comp_valX, comp_testX), kernels=KERNELS)
comp_trainY, comp_valY, comp_testY = onehot(
(comp_trainY, comp_valY, comp_testY))
# weight file path
weight_file = f"{WEIGHT_PATH}/{i+1}.h5"
# initialise model
model = EEGNet(nb_classes=CLASSES,
Chans=chans,