x_train.shape[0] x_train.shape[1] x_train.shape[2] layer = [] inp = Input(shape=(x_train.shape[1],)) print(inp) inputs=[] for i in range(x_train.shape[1]): x = layers.Dense(units = 64, activation = "sigmoid")(inp) x = layers.Dense(units = 64, activation = "sigmoid")(x) x = layers.Dense(units = 64, activation = "sigmoid")(x) layer.append(x) inputs.append(inp) print(inputs) len(inputs) x.shape # x와 같은 크기의 텐서를 만들어준다 x = tf.zeros_like(x) x.shape # layer 리스트에 Dense들이 하나씩 들어가있다 print(layer) layer for k in layer:
for i, char in enumerate(ff): char = char.strip() char_list.append(char) char2id = {j: i for i, j in enumerate(char_list)} id2char = {i: j for i, j in enumerate(char_list)} with open(label_txt_path, 'r', encoding='UTF-8') as f: ff = f.readlines() for i, line in enumerate(ff): line = line.strip() img_name = line.split(' ')[0] label = line.split()[1:] label = list(map(int, label)) img_names.append(img_name) labels.append(label) # img_names = random.shuffle(img_names) # labels = random.shuffle(labels) assert len(img_names) == len(labels), "len(img_names) !=len(labels)" length_data = len(img_names) train_img_names = img_names[:int(0.9 * length_data)] train_labels = labels[:int(0.9 * length_data)] test_img_names = img_names[int(0.9 * length_data):] test_labels = labels[int(0.9 * length_data):] print('train_img_nums:', len(train_img_names)) print('train_label_nums:', len(train_labels))