Question

似乎我在弄乱准备数据集的步骤。无法找到正确的答案或在文档中查找正确的解决方案。我用下面的###指出了问题所在。

def parse_file(data_path):


    imagepaths = list()
    labels = list()
    # a working parser for os is here

    imagepaths = tf.constant(imagepaths, dtype=tf.string)
    labels = tf.constant(labels, dtype=tf.float32)

    return imagepaths, labels


def parse_image(imagepath, label):

    image_string = tf.read_file(imagepath)
    image_decoded = tf.image.decode_png(image_string, channels=3)
    # The image size is 425x425.
    image_resized = tf.image.resize_images(image_decoded, [img_size, img_size])
    image_normalized = image_resized * 1.0/255
    print(image_normalized)
    print(label)
    return image_normalized, label

dataset = tf.data.Dataset.from_tensor_slices((parsed_files))
dataset = dataset.map(parse_image)
dataset = dataset.batch(batch_size)

iterator = dataset.make_initializable_iterator()
iterator = iterator.get_next()

x = tf.placeholder(tf.float32, [None, img_size, img_size, channels])
y = tf.placeholder(tf.float32, [None, 1])

（模型在这里无关紧要。）

with tf.Session() as sess:

    ### AttributeError: 'tuple' object has no attribute 'initializer'
    sess.run(iterator.initializer)
    batch_x, batch_y = iterator.get_next()
    test1, test2 = sess.run([batch_x, batch_y])
    total_batch = int(total_input[0] / batch_size)
    # define the iterator for the network
    for epoch in range(epochs):
        avg_cost = 0
        for i in range(total_batch):
            batch_x, batch_y = sess.run(iterator)
            _, c = sess.run([optimiser, cross_entropy], feed_dict={x: batch_x, y: batch_y})
            avg_cost += c / total_batch

        test_acc = sess.run(accuracy,feed_dict={x: test_x, y: np.expand_dims(test_y, axis=-1)})
        print("Epoch:", (epoch + 1), "cost =", "{:.3f}".format(avg_cost), " test accuracy: {:.3f}".format(test_acc))
        summary = sess.run(merged, feed_dict={x: test_x, y: np.expand_dims(test_y, axis=-1)})

    print("\nTraining complete!")
    print(sess.run(accuracy, feed_dict={x: test_x, y: np.expand_dims(test_y, axis=-1)}))

Answer 1

我没有使用tf.Datasets的经验，但这可能会出错：

iterator = dataset.make_initializable_iterator()
iterator = iterator.get_next()

首先，创建一个迭代器，然后使用.get_next方法从其中查询数据来覆盖它。这显然给你一个元组。然后，您可以这样做：

sess.run(iterator.initializer)

您会收到错误消息，因为迭代器不再是make_initializable_iterator（）中的迭代器。您是否尝试过：

iterator = dataset.make_initializable_iterator()
with tf.Session() as sess:
    sess.run(iterator.initializer)

在那之后您可能会得到更多错误，但是也许是错误的，因为我不习惯使用tf.Datasets。

看看这个示例，我发现了here：

max_value = tf.placeholder(tf.int64, shape=[])
dataset = tf.data.Dataset.range(max_value)
iterator = dataset.make_initializable_iterator()
next_element = iterator.get_next()

# Initialize an iterator over a dataset with 10 elements.
sess.run(iterator.initializer, feed_dict={max_value: 10})
for i in range(10):
    value = sess.run(next_element)
    assert i == value

# Initialize the same iterator over a dataset with 100 elements.
sess.run(iterator.initializer, feed_dict={max_value: 100})
for i in range(100):
    value = sess.run(next_element)
    assert i == value

Tensorflow CNN'tuple'对象没有属性'initializer'

1 个答案: