Discussions about research, Faceswapping and things that don't fit in the other categories here.
- Posts: 5
- Joined: Sat Mar 27, 2021 7:00 pm
- Has thanked: 4 times
- Been thanked: 1 time
Or will that destroy landmark/mask metadata?
Context: I'm working on a project with lots of low-res source footage. I want to enhance it using various techniques (denoise, super-resolution etc), but it seems highly wasteful to process the entirety of every frame when all I need is a dataset of faces.
I found some tools but they're all written for DFL and I don't know to what extent they're compatible.
- Posts: 1433
- Joined: Fri Jul 12, 2019 12:53 am
- Has thanked: 47 times
- Been thanked: 278 times
Super-ressing images and feeding into Faceswap is generally bad juju (basically, feeding an AI images generated by another AI tends to not work out so great).
However, to answer your initial question.... It depends on the software package, and whether it maintains the information that we store in the image header. I would suspect most won't.
There may be applications around that can copy the header data from one image to another, but I'm not sure tbh.