Reproduce “Efficient and Faithful High-Resolution Image Editing with Sparse Cross-Resolution Denoising” 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

Maosen Zhao, Xianfang Zeng, Pengtao Chen, Peng Ye, Gang Yu, Jiayuan Fan, Tao Chen

FLASH enables high-resolution instruction-guided editing by anchoring sparse local refinement to an evolving 1K edit within the same system. On VINS-4KEval, at 4K, direct FLASH (R=3) records HaarPSI 0.4067 and keep-region mSSIM 0.7621, versus 0.2426 and 0.4598 for VINS. These source-fidelity measures do not establish instruction completion.

Summary by alphaXiv.

Tip: change arxiv.org to autoarxiv.org in any arXiv abstract link to open the paper here.