Reproduce “IntrinSync: Joint Intrinsic Decomposition and Reciprocal Rendering” 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
Zheng Gu, Rui Huang, Xilu Zhang, Jingbo Zhang, Min Lu, Zhida Sun, Dani Lischinski, Daniel Cohen-Or and 1 more
IntrinSync estimates five aligned maps—albedo, shading, surface normals, roughness, and metallicity—from an RGB image, predicting them jointly and training inverse and forward processes to check one another. Its dual cycle aligns predictions across both directions. On the synthetic outdoor MatrixCity test set, albedo reaches 27.867 dB PSNR versus Ouroboros’s 22.034 dB, measuring agreement with ground truth, not physical correctness.
Summary by alphaXiv.
Tip: change arxiv.org to autoarxiv.org in any arXiv abstract link to open the paper here.