python删除文件夹里损坏的图片

Posted 2022-04-01 _刘文凯_

今天学习pytoch时发现有好多图片损坏了

报错如下：
PIL.UnidentifiedImageError: cannot identify image file <_io.BufferedReader name=’./train\\cat\\052cat.jpg’>

原因

图片没有被正确的下载

解决

写个小程序把所有损坏的图片都删除

import cv2
import os
import numpy as np
fn_all = './test/cat/'
for fn in os.listdir(fn_all):
    img = cv2.imread(fn_all + fn)
    if type(img) != type(np.array([1])):
        os.remove(fn_all + fn)

小程序改进：

import cv2
import os
import numpy as np
fn_all = ['./val/','./train/','./test/']
for fn_a in fn_all:
  for fn in os.walk(fn_n):
    if fn[-1] != []:
        for i in fn[-1]:
            img_f = fn[0] +'/'+ i
            if type(cv2.imread(img_f)) != type(np.array([1])):
                os.remove(img_f)