Reproduce “Do Underwater Image Quality Metrics Agree with Human Perception? A Benchmark with Human Choices, Simple Edits and Generative Drift” 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
Rishii Kumar Singh, Abhinav Rajput, Dinesh Kumar Vishwakarma
The authors test whether 73 no-reference underwater image-quality metrics agree with human judgments across five references and choices between raw photos and edits. They find no metric agrees consistently: the best worst-case Cohen’s κ is 0.23, below the study raters’ median of 0.41. For 32 photos, UCIQE scored a 20-step edit above raw in every case, although people preferred it in only 1% of votes.
Summary by alphaXiv.
Tip: change arxiv.org to autoarxiv.org in any arXiv abstract link to open the paper here.