CycleGAN
CycleGAN copied to clipboard
network applicability
I have a question, can this network do the task of image restoration? such as, unsupervised SR, unsupervised inverse halftone. Or does it only handle the task of style transformation class, going from one style X to another style Y, where there is no more content in Y than there is in X
It can be applied to SR and inverse halftone. But it might not work as well as supervised methods (such as supervised SR). For SR, it's quite easy to get supervised data.