For Supervised ML we have three classes (good, cracked, corroded). Image size is 250x250, we create different resolution in run time.
The attached zip file has two folders (train and test) and each of them have three sub folders (good, cracked and corroded).