search

rosinality / stylegan2-pytorch

Implementation of Analyzing and Improving the Image Quality of StyleGAN (StyleGAN 2) in PyTorch
MIT License
2.72k stars 620 forks source link

Project DeepFashion dataset to latent space #258

Open vinduon opened 3 years ago

vinduon commented 3 years ago

Hi,

I do not know that whether we can project deepfashion (or human full body) dataset to latent space of stylegan2?

rosinality commented 3 years ago

Generally you cannot get high quality results. (Actually it is hard to project even face photos to the latent spaces.)

vinduon commented 3 years ago

Firstly, thank you for your response.

Yeah, I also think that is hard for the model to reach high quality images. But I think we can get average or just able to see human full body (no need super details). Because I see some repo using stylegan2 can generate different scenes from LSUN dataset (which has various structural features).

rosinality commented 3 years ago

You can may try with extend latent codes. (that is, allowing latent codes distinct, so it you can have 18 latent vectors.) Like --w_plus options in the projector.py.