isl-org/PhotorealismEnhancement

Reversing the input and output

alpkabac opened this issue · 4 comments

alpkabac commented

I could not find how to contact you direclty, and issues are not the place for this, but I was really curious if we could reverse the process. What I mean by that is if we used photorealistic images as the input, could convert them into much cartoony or fictional(?) ones?

IceHacks commented

I would assume it to be much the same process. Just use GTA pictures at the "Real Image" and the data used in the "G-Buffer Encoder."

imfunniee commented

ye similar

darthnithin commented

I would assume it to be much the same process. Just use GTA pictures at the "Real Image" and the data used in the "G-Buffer Encoder."

I assume for the dataset you would just use the game engine (GTA) pictures, and the real images as input, though you might have to annotate the images as i dont see how you could get the G-Buffer (which would come from an engine) from real pictures.

  • 👍2
srrichter commented

That's a great idea. I guess a good way to start would be even without the G-buffers here. Our experiments indicated that you can get strong results even without the G-buffers.

  • ❤️2