Reproduce “Unsupervised Real-world Image Super Resolution via Domain-distance Aware Training” 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
Yunxuan Wei, Shuhang Gu, Yawei Li, Longcun Jin
DASR addresses unsupervised real-world image super-resolution by training with synthetic low-resolution and high-resolution pairs alongside unlabeled real low-resolution images. It weights synthetic-pair supervision according to patch-level resemblance to real inputs. On RealSR’s 100-pair validation set, DASR’s LPIPS is 0.228 versus FSSR’s 0.265, and MOS is 2.05 versus 2.45; paired-data ESRGAN scores better on both, at 0.199 LPIPS and 1.35 MOS.
Summary by alphaXiv.
Tip: change arxiv.org to autoarxiv.org in any arXiv abstract link to open the paper here.