批量读取数据next_batch函数的实现

输入的是训练数据路径data_path，label_path，还有batch_size；

输出的是：

 train_data_batch，shape 是( [batch_size,img_row,  img_col, 3]) # 可相应变化
 train_label_batch，shape是([batch_size, img_row, img_col]) # 可相应变化

import cv2
import os
import numpy as np

def next_batch(data_path, lable_path, batch_size):
    train_temp = np.random.randint(low=0, high=Train_nums + 1, size=batch_size) # 生成元素的值在[low,high)区间，随机选取
    train_data_batch = np.zeros([batch_size,img_row,  img_col, 3]) # 其中[img_row,  img_col, 3]是原数据的shape，相应变化
    train_label_batch = np.zeros([batch_size, img_row, img_col]) #
    count = 0 # 后面就是读入图像，并打包成四维的batch
    img_list = os.listdir(data_path)
    for i in train_temp:
        img_path = os.path.join(data_path, img_list[i])  # 图片文件
        label_path = os.path.join(lable_path, img_list[i])  # label的图片名要和img对应上
        train_data_batch[count, :, :, :] = cv2.imread(img_path)
        train_label_batch[count, :, :] = cv2.imread(label_path)
        count+=1
    return train_data_batch, train_label_batch