Hi, There are sorts of pre-trained CNN data base for deep learning in webs, my problem is how can I do for my own images (images are not in any database)!! The code explained here: https://github.com/vedaldi/practical-cnn/blob/master/exercise1.m is about CNN for one image,and explain for different steps.however, any body can help me is it fine for training new images? if yes, how can I train my images (100 RGB images)? Do I need to for each image separately and save them as a .mat file?