deblur-gan
deblur-gan copied to clipboard
the Normalization
In official paper and pytorch code, instance normalization is proposed after conv while batch normalization is used in this keras code. Is there any influence on the final result? Have you test the difference between them?