有時候在用caffe訓練模型的時候,會因為數據集的某些圖片損壞而出現一些莫名其妙的錯誤,因此寫了這樣一個判斷圖片是否可用的代碼,用來清洗數據
def get_image_files():
path = 'images_path'
files = os.listdir(path)
s=[]
for file in files:
s.append(file)
return s
def remove_bad_images():
s = get_image_files()
for image in s:
try:
image_file = 'images_path' +image
img = Image.open(image_file)
img.verify()
except IOError:
print(image_file)
shutil.move(image_file,'/home/data/errimg')