Reproduce “Styleformer: Transformer based Generative Adversarial Networks with Style Vector” on OpenResearch
Turn your favorite coding agents into research agents that review literature, develop hypotheses, and run experiments.
Universal app for Apple silicon and Intel. Requires macOS 11 or later.
About this paper
Jeeseung Park, Younggeun Kim
Styleformer adapts style modulation and demodulation to self-attention in a GAN generator, aiming to coordinate distant image regions while retaining layer-wise style control. For high resolutions, it uses Linformer-projected attention or combines low-resolution Styleformer blocks with high-resolution StyleGAN2 blocks. On unconditional CIFAR-10, it reports FID 2.82 versus 2.92 for StyleGAN2-ADA; this modest comparison uses results reported across papers.
Summary by alphaXiv.
Tip: change arxiv.org to autoarxiv.org in any arXiv abstract link to open the paper here.