Deep learning in postprocessing Thread

Hello!

Does anyone have experience on using generative neural networks at postprocessing? For example, for improving the realism or adding details.

I was wondering if that would heavily affect the framerate or the glitches would be to obvious.