ChunML

def one_shot_input_fn(filenames, labels):
    dataset = tf.data.Dataset.from_tensor_slices((filenames, labels))
    dataset = dataset.map(_parse_data).batch(1)

    iterator = dataset.make_one_shot_iterator()
    img, label = iterator.get_next()
    return img, label

ChunML

means = [123.68, 116.779, 103.939]

def _parse_data(filename, label, new_size=224):
    img_string = tf.read_file(filename)
    img = tf.image.decode_jpeg(img_string)
    img = tf.image.resize_images(img, (new_size, new_size))
    img.set_shape([new_size, new_size, 3])
    img = tf.to_float(img)
    channels = tf.split(axis=2, num_or_size_splits=3, value=img)
    for i in range(3):

ChunML

def visualize_dataset(imgs, labels):
    fig, axes = plt.subplots(ncols=4, nrows=4)
    fig.set_size_inches(10, 10)

    for i, img in enumerate(imgs):
        img += means
        np.clip(img, 0, 255, img)
        img = img[0].astype(np.uint8)

        img = Image.fromarray(img)

ChunML

img, label = one_shot_input_fn(filenames, labels)

with tf.Session() as sess:
    sess.run(tf.global_variables_initializer())
    res_imgs = []
    res_labels = []
    for _ in range(16):
        res_img, res_label = sess.run([img, label])
        res_imgs.append(res_img)
        res_labels.append(res_label)

ChunML

ChunML

def reinitializable_input_fn(filenames, labels, train_val_ratio=0.8):
    num_files = len(filenames)
    num_train_files = int(num_files * train_val_ratio)
    train_filenames = filenames[:num_train_files]
    train_labels = labels[:num_train_files]
    val_filenames = filenames[num_train_files:]
    val_labels = labels[num_train_files:]

ChunML

    train_data = tf.data.Dataset.from_tensor_slices(
        (train_filenames, train_labels))
    train_data = train_data.map(_parse_data).shuffle(1000).repeat().batch(4)

ChunML

    val_data = tf.data.Dataset.from_tensor_slices(
        (val_filenames, val_labels))
    val_data = val_data.map(_parse_data).batch(1)

ChunML

    iterator = tf.data.Iterator.from_structure(train_data.output_types,
                                               train_data.output_shapes)

ChunML

    next_element = iterator.get_next()