TFRecord: TensorFlow dataset storage format

tf.train.Example TFRecord can be understood as a list file composed of  a series of serialized  elements, each of which is composed tf.train.Example of several  tf.train.Feature dictionaries.

write record

• Read the data element into memory;

• Convert the element into  tf.train.Example an object (each  consists tf.train.Example of several  tf.train.Feature dictionaries, so you need to create a Feature dictionary first);

• Serialize  tf.train.Example the object to a string and write it to a TFRecord file via a predefined one  tf.io.TFRecordWriter .

import tensorflow as tf
import os

data_dir = './catsdogs'
train_cats_dir = data_dir + '/train/Cat/'
train_dogs_dir = data_dir + '/train/Dog/'

train_cat_filenames = [train_cats_dir + filename for filename in os.listdir(train_cats_dir)]
train_dog_filenames = [train_dogs_dir + filename for filename in os.listdir(train_dogs_dir)]
train_filenam