Reproduce “Unsupervised Learning for Color Constancy” 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
Nikola Banić, Karlo Koščević, Sven Lončarić
Color Tiger learns illumination structure from uncalibrated images by pooling rough estimates, clustering them into two centers, and using estimates to select a center. Across eight NUS camera datasets, its median angular error is 1.70°, versus 1.77° for Color Dog. Its geometric-mean aggregate is 2.09, versus 1.44 for FFCC, so the results support label-free estimation, not overall superiority.
Summary by alphaXiv.
Tip: change arxiv.org to autoarxiv.org in any arXiv abstract link to open the paper here.