StyleGAN 2

A reimplmentation of StyleGAN 2 in Pytorch with friendly UI allowing to observe real-time training footstep

Requirements

• CUDA 11.3 and CUDNN v8.2.0
• Python 3.7.16
• Pytorch 1.9.0
• Anaconda3 2022.10

Showcase

Below are the generated samples trained on Danbooru2019 Portraits

References

T. Karras, S. Laine, M. Aittala, J. Hellsten, J. Lehtinen, and T. Aila, ‘Analyzing and Improving the Image Quality of StyleGAN’, in Proc. CVPR, 2020.