torch ==1.2.0
our dataset have upload to Baidu NetDisk
The link: https://pan.baidu.com/s/1ZNxA81lRXqSQYSmUVUKLtQ
extraction code: vmhe .
in this dataset, there are two type of images, one is the template image, which named as xxxx_temp.jpg, and the other is the defect images, which named as xxxx_test.jpg. The structure of directory is based on the VOC structure.
run check.py and load two images into this script, one is the template image and the other is the defect image.
run train.py and start to train model