def test_conv_channel(self):
fake_image_list = split_eeg.split_eeg_signal_axes(FAKE_EEG_SIGNAL,
split_dim=1)
kernel_tensor, input_shape = \
concat_eeg.conv_eeg_signal_channel(fake_image_list, 256, 1)
# check the size of the shape is equal to the concat size
self.assertEqual(kernel_tensor.get_shape().as_list(), input_shape)
def test_conv_channel(self):
fake_image_list = split_eeg.split_eeg_signal_axes(FAKE_EEG_SIGNAL,
split_dim=1)
kernel_tensor, input_shape = \
concat_eeg.conv_eeg_signal_channel(fake_image_list, 256, 1)
# check the size of the shape is equal to the concat size
self.assertEqual(kernel_tensor.get_shape().as_list(), input_shape)
def test_conv_channel_time(self):
# first split the eeg signal of (10, 256, 256, 1) to 256 * (10, 1, 256, 1)
fake_image_list = split_eeg.split_eeg_signal_axes(FAKE_EEG_SIGNAL,
split_dim=1)
# then concat the eeg signal use 256 channel kernel index to (10, 256*5, 256, 1)
kernel_tensor, input_shape = \
concat_eeg.conv_eeg_signal_channel(fake_image_list, 256, 1)
self.assertEqual(kernel_tensor.get_shape().as_list(), [10, 256*5, 256, 4])
self.assertEqual(input_shape, [10, 256*5, 256, 4])
# then split the new eeg signal of (10, 1280, 256, 1) to 256 * (10, 1280, 1, 1)
fake_image_list = split_eeg.split_eeg_signal_axes(kernel_tensor,
split_dim=2)
# then concat the eeg signal use 256 channel kernel index to (10, 256*5, 256*5, 1)
kernel_tensor, input_shape = \
concat_eeg.conv_eeg_signal_time(fake_image_list, np.arange(0, 256))
# check the size of the shape is equal to the concat size
self.assertEqual(kernel_tensor.get_shape().as_list(), [10, 256*5, 256*5, 4])
self.assertEqual(input_shape, [10, 256*5, 256*5, 4])
def test_conv_channel_time(self):
# first split the eeg signal of (10, 256, 256, 1) to 256 * (10, 1, 256, 1)
fake_image_list = split_eeg.split_eeg_signal_axes(FAKE_EEG_SIGNAL,
split_dim=1)
# then concat the eeg signal use 256 channel kernel index to (10, 256*5, 256, 1)
kernel_tensor, input_shape = \
concat_eeg.conv_eeg_signal_channel(fake_image_list, 256, 1)
self.assertEqual(kernel_tensor.get_shape().as_list(),
[10, 256 * 5, 256, 4])
self.assertEqual(input_shape, [10, 256 * 5, 256, 4])
# then split the new eeg signal of (10, 1280, 256, 1) to 256 * (10, 1280, 1, 1)
fake_image_list = split_eeg.split_eeg_signal_axes(kernel_tensor,
split_dim=2)
# then concat the eeg signal use 256 channel kernel index to (10, 256*5, 256*5, 1)
kernel_tensor, input_shape = \
concat_eeg.conv_eeg_signal_time(fake_image_list, np.arange(0, 256))
# check the size of the shape is equal to the concat size
self.assertEqual(kernel_tensor.get_shape().as_list(),
[10, 256 * 5, 256 * 5, 4])
self.assertEqual(input_shape, [10, 256 * 5, 256 * 5, 4])
