TFRecords文件的存储与读取

将cats和dogs两个文件夹各1000张图片存储为：train.tfrecords

#将图片文件生成train record
import os
import tensorflow as tf
from PIL import Image
#生成cats和dogs的record文件
path='./data/train'
filenames=os.listdir(path)
writer=tf.python_io.TFRecordWriter('./data/train/train.tfrecords')
classes=['cats','dogs']#两类
for index,name in enumerate(classes):print(index,name)
#for name in os.listdir(path):
    class_path=path+os.sep+namefor img_name in os.listdir(class_path):img_path=class_path+os.sep+img_nameimg=Image.open(img_path)img=img.resize((500,500))img_raw=img.tobytes()#图片转化成二进制
        example=tf.train.Example(features=tf.train.Features(feature={'label':tf.train.Feature(int64_list=tf.train.Int64List(value=[index])),
            'image':tf.train.Feature(bytes_list=tf.train.BytesList(value=[img_raw]))}))writer.write(example.SerializeToString())

变为

读取过程：

# 读取cats与dogs的train.tfrecords文件
import tensorflow as tf
import cv2
import os
os.environ['TF_CPP_MIN_LOG_LEVEL'] = '2'

filename='./data/train/train.tfrecords'
#读取并解析.tfrecords文件
def read_and_decode(filename):filename_queue=tf.train.string_input_producer([filename])# 按队列的形式读取
    reader=tf.TFRecordReader()_,serialized_example=reader.read(filename_queue)#返回文件名和文件
    features=tf.parse_single_example(serialized_example,
                            features={'label':tf.FixedLenFeature([],tf.int64),#与存储的类型一致
                                'image':tf.FixedLenFeature([],tf.string)})img=tf.decode_raw(features['image'],tf.uint8)img=tf.reshape(img,shape=[500,500,3])img = tf.cast(img, dtype=tf.float32) * (1.0 / 128) - 0.5
    label = tf.cast(features['label'], dtype=tf.int32)return img,labelimg,label=read_and_decode(filename)img_batch,label_batch=tf.train.shuffle_batch([img,label],batch_size=1,
                                             capacity=10,min_after_dequeue=1)
init=tf.global_variables_initializer()
with tf.Session() as sess:sess.run(init)threads=tf.train.start_queue_runners(sess=sess)for i in range(10):label=sess.run(label_batch)imgcv2=sess.run(img_batch)imgcv2.resize((500,500,3))print(label)cv2.imshow('img',imgcv2)cv2.waitKey()