Closed this issue 7 years ago · 1 comments
Why not do image normalisation for training
it is actually used, u can check the preprocess_input function for more details. But this implementation neither includes strong image augmentation nor hard example selection. These sections would be implemented later