[CVPR 2022] StyleSwin: Transformer-based GAN for High-resolution Image Generation

StyleSwin: Transformer-based GAN for High-resolution Image Generation

[Open Sourced] StyleSwin: CVPR 2022 Transformer Based GAN for High Resolution Image Generation Paper Implementation by Microsoft Research Asia

The original StyleSwin implementation utilized Swin Transformer blocks in the generator and included a style injection mechanism, achieving state-of-the-art results in high-resolution image generation. By incorporating ADA, we have generated high-resolution medical images despite data constraints.

No description available

Implementation of Microsoft Research's "StyleSwin: Transformer-based GAN for High-resolution Image Generation" https://arxiv.org/abs/2112.10762

StyleSwin based Image-to-image translation

No description available