size = np.int(512 / roi)
padding_width = 4
masks_all = np.zeros((len(masks) * size * size, roi + 2 * padding_width,
roi + 2 * padding_width))
inputs_all = np.zeros((len(inputs) * size * size, roi + 2 * padding_width,
roi + 2 * padding_width, 3))
previous_all = np.zeros(
(len(inputs) * size * size, roi + 2 * padding_width,
roi + 2 * padding_width, 3))
nexts_all = np.zeros((len(inputs) * size * size, roi + 2 * padding_width,
roi + 2 * padding_width, 3))
masks_all = read_masks(masks, size, roi, padding_width)
inputs_all = read_inputs(inputs, size, roi, padding_width)
previous_all = read_inputs(previouss, size, roi, padding_width)
nexts_all = read_inputs(nexts, size, roi, padding_width)
full_set_all = np.zeros(
(inputs_all.shape[0], inputs_all.shape[1], inputs_all.shape[2], 9))
full_set_all[:, :, :, 0:3] = previous_all
full_set_all[:, :, :, 3:6] = inputs_all
full_set_all[:, :, :, 6:9] = nexts_all
num_items = full_set_all.shape[0]
ind_train = np.int64(num_items * (1 - (1 / 4.8)))
full_masks_test = masks_all[ind_train:]
full_test_set = full_set_all[ind_train:]
(len(inputs_train_list) * size * size, roi + 2 * padding_width,
roi + 2 * padding_width, 3))
masks_test = np.zeros((len(masks_test_list) * size * size,
roi + 2 * padding_width, roi + 2 * padding_width))
inputs_test = np.zeros(
(len(inputs_test_list) * size * size, roi + 2 * padding_width,
roi + 2 * padding_width, 3))
previous_test = np.zeros(
(len(inputs_test_list) * size * size, roi + 2 * padding_width,
roi + 2 * padding_width, 3))
next_test = np.zeros((len(inputs_test_list) * size * size,
roi + 2 * padding_width, roi + 2 * padding_width, 3))
masks_train = read_masks(masks_train_list, size, roi, padding_width)
inputs_train = read_inputs(inputs_train_list, size, roi, padding_width)
previous_train = read_inputs(previouss_train_list, size, roi,
padding_width)
next_train = read_inputs(nexts_train_list, size, roi, padding_width)
masks_test = read_masks(masks_test_list, size, roi, padding_width)
inputs_test = read_inputs(inputs_test_list, size, roi, padding_width)
previous_test = read_inputs(previouss_test_list, size, roi, padding_width)
next_test = read_inputs(nexts_test_list, size, roi, padding_width)
# sort out black masks for training
full_masks_train = masks_train
full_masks_train = mirror_combine_data(full_masks_train)
full_masks_train = np.vstack((full_masks_train, full_masks_train))
ind = threshold_mean(full_masks_train, 0.01, 0.8)
