How to decide upon the normalisation that need to be used for image classification or cv problems , Is there any standard on what to choose when in particularly with Image data ?


marked as duplicate by Reinstate Monica neural-networks Aug 15 at 14:04

This question has been asked before and already has an answer. If those answers do not fully address your question, please ask a new question.

  • $\begingroup$ The most comprehensive answer regarding images is written in the context of VGG-16 but the principles in my answer are generic to any inputs. It's typically the case that there is not a large difference between alternative methods unless you're working in a very specific context with some distinguishing characteristics (MNIST auto-encoder using BCE loss, for example). If your question is more specific than these duplicates, please edit your question so that it may be reopened. $\endgroup$ – Reinstate Monica Aug 15 at 14:06