Artifacts after projection to SWAGAN latent space

rosinality / stylegan2-pytorch

Implementation of Analyzing and Improving the Image Quality of StyleGAN (StyleGAN 2) in PyTorch

MIT License

2.74k stars 623 forks source link

Artifacts after projection to SWAGAN latent space #284

Open kostagiolasn opened 2 years ago

kostagiolasn commented 2 years ago

Hi! When trying to perform a projection to a trained SWAGAN's latent space, the generations I receive tend to include some artifacts, thus rendering the overall image generation unrealistic. Do you have any clues about why is this happening?

Thank you in advance!

rosinality commented 2 years ago

Generally projection process is not very stable and consistent. Does projection with stylegan2 works well with your samples?

seungjun-moon commented 2 years ago

Where did you get the pretrained-SWAGAN weights? Thanx in advance.