tensorflow 读、存取 图像 数据的 TFRecord 方法 （示例）

1.     利用TFRecord 格式   读、存 取    Mnist数据集的方法

存取   Mnist数据集的方法     （TFRecord格式）

import tensorflow as tf
from tensorflow.examples.tutorials.mnist import input_data
import numpy as np

def _float32_feature(value):
    return tf.train.Feature(float_list=tf.train.FloatList(value=[value]))

def _int64_feature(value):
    return tf.train.Feature(int64_list=tf.train.Int64List(value=[value]))

def _bytes_feature(value):
    return tf.train.Feature(bytes_list=tf.train.BytesList(value=[value]))

mnist=input_data.read_data_sets('./data', dtype=tf.uint8, one_hot=True)
"""
print(mnist.train.images)
print(mnist.train.labels)
print(mnist.test.images)
print(mnist.test.labels)
"""
train_images=mnist.train.images
train_labels=mnist.train.labels
#test_images=mnist.test.images
#test_labels=mnist.test.labels

train_num=mnist.train.num_examples
#test_num=mnist.test.num_examples 

pixels=train_images.shape[1]   # 784 = 28*28

file_out='./data/output.tfrecords'
writer=tf.python_io.TFRecordWriter(file_out)

for index in range(train_num):
    image_raw=train_images[index].tostring() #转换为bytes序列   

    example=tf.train.Example(features=tf.train.Features(feature={
               'pixels': _int64_feature(pixels),
               'label':_int64_feature(np.argmax(train_labels[index])),
               'x':_float32_feature(0.1),
               'image_raw':_bytes_feature(image_raw)}))

    writer.write(example.SerializeToString())
writer.close()

读取   Mnist数据集的方法     （TFRecord格式）

import tensorflow as tf

reader=tf.TFRecordReader()

files=tf.train.match_filenames_once('./data/output.*')

#filename_queue=tf.train.string_input_producer(['./data/output.tfrecords'])
filename_queue=tf.train.string_input_producer(files)

_, serialized_example=reader.read(filename_queue)

features=tf.parse_single_example(serialized_example,
                   features={
                           'image_raw':tf.FixedLenFeature([], tf.string),
                           'pixels':tf.FixedLenFeature([], tf.int64),
                           'label':tf.FixedLenFeature([], tf.int64),
                           'x':tf.FixedLenFeature([], tf.float32)
                            })

#print(features['image_raw'])    # tensor string （bytes tensor      string tensor）

# necessary operation
# bytes_list   to   uint8_list
image=tf.decode_raw(features['image_raw'], tf.uint8)
#print(image)    # tensor uint8

label=tf.cast(features['label'], tf.int32)
pixels=tf.cast(features['pixels'], tf.int32)
#image.set_shape([pixels**0.5, pixels**0.5])
image.set_shape([784])

batch_size=128
image_batch, label_batch, pixels_batch=tf.train.batch([image, label, pixels], batch_size=batch_size, capacity=1000+3*batch_size)

coord=tf.train.Coordinator()

with tf.Session() as sess:
    sess.run(tf.local_variables_initializer())
    threads=tf.train.start_queue_runners(sess=sess, coord=coord)

    for i in range(3):
        print(sess.run([image_batch, label_batch, pixels_batch]))

    coord.request_stop()
    coord.join(threads)

==================================================================

2.     利用TFRecord 格式    存取   文件夹内图片的方法

import matplotlib.pyplot as plt
import tensorflow as tf
import numpy as np

def _float32_feature(value):
    return tf.train.Feature(float_list=tf.train.FloatList(value=[value]))

def _int64_feature(value):
    return tf.train.Feature(int64_list=tf.train.Int64List(value=[value]))

def _bytes_feature(value):
    return tf.train.Feature(bytes_list=tf.train.BytesList(value=[value]))

file_out='./data/output1.tfrecords'
writer=tf.python_io.TFRecordWriter(file_out)

files = tf.gfile.Glob('./data/*.jpg')
sess=tf.Session()
for file in files:
    image_raw_data = tf.gfile.FastGFile(file,'rb').read()

    img_data = tf.image.decode_jpeg(image_raw_data)  # tensor
    img_data = sess.run(img_data) # np.array int
    resized = img_data.tostring() # np.array string uint8

    example=tf.train.Example(features=tf.train.Features(feature={
               'y':_int64_feature(1),
               'x':_float32_feature(0.1),
               'image_raw':_bytes_feature(resized)}))

    writer.write(example.SerializeToString())
    writer.close()

文件读取过程使用   输入队列 ：

import matplotlib.pyplot as plt
import tensorflow as tf
import numpy as np

def _float32_feature(value):
    return tf.train.Feature(float_list=tf.train.FloatList(value=[value]))

def _int64_feature(value):
    #value类型应为：int,long,float
    #return tf.train.Feature(int64_list=tf.train.Int64List(value=[value]))

    #value类型应为：[int],[long],[float], 这里为int的list类型
    return tf.train.Feature(int64_list=tf.train.Int64List(value=value))

def _bytes_feature(value):
    return tf.train.Feature(bytes_list=tf.train.BytesList(value=[value]))

file_out='./data/output2.tfrecords'
writer=tf.python_io.TFRecordWriter(file_out)

files = tf.train.match_filenames_once('./data/*.jpg')

# string_input_producer会产生一个文件名队列
filename_queue = tf.train.string_input_producer(files, shuffle=False, num_epochs=3)
# reader从文件名队列中读数据。对应的方法是reader.read
reader = tf.WholeFileReader()
key, value = reader.read(filename_queue)

img_data = tf.image.decode_jpeg(value)  # np.array 转换为 tensor
#print(sess.run([key, img_data]))
#print(img_data.get_shape())
img_data.set_shape([None, None, 3])

img_float = tf.image.convert_image_dtype(img_data, tf.float32)
img_float = tf.image.resize_images(img_float, [300, 300], method=0)

with tf.Session() as sess:
    # tf.train.string_input_producer定义了一个epoch变量，要对它进行初始化
    tf.local_variables_initializer().run()

    x=np.array([[1,1,1,1],[1,1,1,1]])

    coord = tf.train.Coordinator()

    # 使用start_queue_runners之后，才会开始填充队列
    threads = tf.train.start_queue_runners(sess=sess, coord=coord)

    for _ in range(3):
        resized=sess.run(img_data)
        resized = resized.tostring()

        example=tf.train.Example(features=tf.train.Features(feature={
               'x':_int64_feature(x.reshape(x.size).tolist()),
               'x2':_int64_feature([1,1]),
               'y':_float32_feature(0.1),
               'image_raw':_bytes_feature(resized)}))

        writer.write(example.SerializeToString())
    writer.close()

    coord.request_stop()
    coord.join(threads)