I tried training data using Stylegan3, locally. It will take 73.8 days!
1771.2 hours of letting my computer run day and night!
I am using PyTorch and have a i9-13900KF, Nvidia RTX 4090, 64GB RAM & 4 TB SSD.
It really give me an idea of how much computing horsepower this all takes!
Question, is there a lighter weight model from Hugging Face I can use that people have tried that will give similar like results?
Goal: Train an image generator to use my face. So I could say ‘draw a magician in a photo realistic style’ it would draw the magician with my face, and features.
Is this possible?
Data. I tool some videos of myself - just shoulder up. And auto cut the videos into frames. It is me in different lighting, shaved, unshaved, different hats, lighting, me talking, looking around. I tried to keep the proportions the same. ie same camera set up, image size (all transformed to 512x512)
I am not a programmer, I am an IT systems guy. The closest I do to programming I do is some scripting.
So… Am I crazy!?
Thoughts on how to do this, if possible.
I am also looking forward to hearing from others on their journeys in this space!