What is the general approach to defect detection in deep learning?
Would the approach be better if we try to learn the positive images (defects in images) as much as possible or we try to learn the negative images (images without blemishes) and try to single out the defects as some anomalies
Can someone point me to some architecture?
REgards
