paddleseg数据集txt文件生成

import glob
import os
import cv2

#数据清洗
def generate_list(img_path,lab_path,file_list):
    write_list = []
    img_list = glob.glob(os.path.join(img_path , "*jpg"))
    #lab_list = glob.glob(os.path.join(lab_path, "*png"))
    print(img_list)
    for i in img_list:
#数据检查，一一对应
        #img = cv2.imread(i)

        #i.replace(img_path,lab_path)
        lab_p = i.replace(img_path,lab_path)
        #img = cv2.imread(img_p,0)
        #print(img)
        #print(img_p)

        #形成列表
        write_list.append(i + " " + lab_p)
        print(write_list)

    with open(file_list, "w") as f:
        for i in write_list:
            changetype = i.rsplit('.',1)[0]
            changetype += '.png'
            f.write(changetype.replace('E:\\NEt\\PaddleSeg-release-2.6\\data\\buildings\\',"").replace('\\','/') + '\n')


if __name__ =="__main__":

    img_path = r'E:\NEt\PaddleSeg-release-2.6\data\buildings\JPEGImages'         #图像目录
    lab_path = r'E:\NEt\PaddleSeg-release-2.6\data\buildings\Annotations'        #标签目录
    file_list = r'E:\NEt\PaddleSeg-release-2.6\data\buildings\train.txt'         #txt文件
    generate_list(img_path,lab_path,file_list)

两种标签文件均可用 ，换路径时需要按照split中改写

 绝对路径亲测可用，代码还待完善，需手动划分测试集和验证集

paddeseg数据格式