Reproduce “Fill2SR: Repurposing Inpainting Diffusion Transformers for Real-World Super-Resolution” 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
Xingfu Yi, Xiaoxue Yu
Fill2SR restores enlarged images from degraded real-world inputs by placing aligned low-quality image latents in a pretrained inpainting model’s masked-image input slot and generating restorations from noise. With RCDT, it raises RealLQ250’s 4× no-reference MUSIQ score from 62.1914 to 71.5498, while synthetic LPIPS worsens slightly; these quality estimates do not verify recovered details.
Summary by alphaXiv.
Tip: change arxiv.org to autoarxiv.org in any arXiv abstract link to open the paper here.