Implementation of A Style-Based Generator Architecture for Generative Adversarial Networks (https://arxiv.org/abs/1812.04948) in PyTorch
Usage:
for celebA
python train.py --mixing -d {folder} PATH
for FFHQ
python train.py --mixing --loss r1 --sched -d {folder}
I have mixed styles at 4^2 - 8^2 scale. I can't get samples as dramatic as samles in the original paper. I think my model too dependent on 4^2 scale features - it seems like that much of details determined in that scale, so little variations can be acquired after it.
Trained high resolution model on FFHQ. I think result seems more interesting.