TensorFlow Read TFRecords

Example Code:

import tensorflow as tf

record_path = '/tmp/tfrecord_write_test/part-r-00000'

filename_queue = tf.train.string_input_producer([record_path])
reader = tf.TFRecordReader()
_, serialized_example = reader.read(filename_queue)
features = tf.parse_single_example(serialized_example,
                                   features={
                                       'label': tf.FixedLenFeature([], tf.float32),
                                       'features': tf.FixedLenFeature([2], tf.float32),
                                   })

with tf.Session() as sess:
    init_op = tf.initialize_all_variables()
    sess.run(init_op)
    coord = tf.train.Coordinator()
    threads = tf.train.start_queue_runners(coord=coord)
    for i in range(2):
        example = sess.run([features])#在会话中取出image和label
        print(example)

你可能感兴趣的:(TensorFlow)