def substract_mean_pixel_value(train_data):
shape = np.shape(train_data)
if True:
train_data -= np.mean(train_data[0:shape[0]], axis=(2, 3))
else:
train_data -= np.mean(train_data[0:shape[0]], axis=(1, 2))
return train_data
def substract_mean_image(train_data):
shape = np.shape(train_data)
mean_image = np.mean(train_data[0:shape[0]], axis=0)
train_data -= mean_image
return train_data