Reproduce “WorldSR: Harnessing World Knowledge Search for Grounded Image 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
Junyuan Deng, Xinyi Wu, Zhenyao Wu, Song Wang, Yu Wang, Congchao Zhu
WorldSR addresses image super-resolution hallucinations by identifying scene entities, retrieving and filtering visual references for them, then using those references to guide reconstruction. In the Face benchmark, it reaches identity similarity (IDS) of 0.8199, versus 0.7705 for InstantRestore, a reported 6.4% improvement. IDS measures identity consistency, not general image fidelity; WorldSR does not lead on every pixel-based measure.
Summary by alphaXiv.
Tip: change arxiv.org to autoarxiv.org in any arXiv abstract link to open the paper here.