Closed alex-moon closed 3 years ago
Closing this issue for now as we have ISR. I would like to try to get SinGAN working as I believe it has better outcomes for this use case, but I've found the respective codebases unworkable. Leave it alone for now.
The results of the work against VC-1 are promising, but the most promising outputs are very blurry due to the compounding effect of inpainting and low vc iterations. Can we add another library that'll do sharpening in between steps?
Possible candidates:
https://github.com/alexjc/neural-enhance https://github.com/tamarott/SinGAN https://github.com/idealo/image-super-resolution