tensorflow批次读取图像数据

处理单张图片

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


def read_single_image(image_path):
    image_file = tf.read_file(image_path)
    print('image_file类型：',type(image_file))
    image_date = tf.image.decode_jpeg(image_file) #图像解码成unit8张量【w,h,c]
    print('image_data类型：',type(image_date))
    return image_date

def main():
    with tf.name_scope('read_single_image'):
        image = read_single_image('./images/image.jpg')
        with tf.Session() as sess:
            image_numpy = sess.run(image)
            print(image_numpy.shape)
            plt.imshow(image_numpy)
            plt.show()

if __name__ == '__main__':
    main()

批量处理：

img_dir = './images/'
image_path = glob.glob(img_dir + '*.jpg')
image_path += glob.glob(img_dir + '*jpeg')

'''
1.构造图片文件队列
2.构造文件阅读器
3.读取文件数据
4.处理图片数据
5.批处理
'''    
def read_images(filelist):
#1.构造图片文件队列
    filename_queue = tf.train.string_input_producer(filelist)
#2.图像读取器，将文件的全部内容作为值输出的读取器
#return :读取器实例
#read(filename_queue)：输出是一个文件名（key)和文件的内容（value)
    reader = tf.WholeFileReader()
    key, value = reader.read(filename_queue)
#3.读取文件数据
    images = tf.image.decode_jpeg(value)
#处理图片数据
    images_resize = tf.image.resize_images(images,[100,100])
    images_resize.set_shape([100,100,3])

#批处理
    images_resize_batch = tf.train.batch([images_resize],batch_size=4,num_threads=1,capacity=4)
    return images_resize_batch

def main():
    filelist = [os.path.join(os.getcwd(),file) for file in image_path]
    image_batch = read_images(filelist)
    print(image_batch.shape)
    init = tf.global_variables_initializer()
    with tf.Session() as sess:
        sess.run(init)
        coord = tf.train.Coordinator()
        threads = tf.train.start_queue_runners(sess,coord)
        img_data = sess.run(image_batch)
        print(img_data)
        coord.request_stop()
        coord.join(threads)
if __name__ == '__main__'
    main()

还有一种简单的方法：使用tf,data这个API,关于API的详细内容参考博客（https://blog.csdn.net/qq_34914551/article/details/96834647）

img_dir = './images/'
image_path = glob.glob(img_dir + '*.jpg')
image_path += glob.glob(img_dir + '*jpeg')



'''
1. 获取所有图片的路径
2. 写好预处理的代码
3. 定义好数据生成器
'''
def pre_image(filename):
    image = tf.read_file(filename)
    image = tf.image.decode_jpeg(image,channels=3)
    image_resize = tf.image.resize_images(image,[100,100])
    return image_resize

def data_generator(batchsize, shuffle=True):
    with tf.Session() as sess:
        dataset = tf.data.Dataset().from_tensor_slices(image_path)
        dataset = dataset.map(pre_image)
        dataset = dataset.batch(batchsize)
        if shuffle:
            dataset = dataset.shuffle(buffer_size=4)
        iterator = dataset.make_one_shot_iterator()
        #iterator = dataset.make_initializable_iterator()
        #sess.run(iterator.initializer)
        batch_image = iterator.get_next()
        while 1:
            try:
                batch_data = sess.run(batch_image)
                print(batch_data)
                print(batch_data.shape)
            except tf.errors.OutOfRangeError:
                print("iterator done")
                break

if __name__ == '__main__':
    data_generator(4)