Reproduce “RefGaussian: Disentangling Reflections from 3D Gaussian Splatting for Realistic 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
Rui Zhang, Tianyue Luo, Weidong Yang, Ben Fei, Jingyi Xu, Qingyuan Zhou, Keyi Liu, Ying He
RefGaussian models reflective scenes by giving each Gaussian separate transmitted and reflected appearances, then blending them with a learned reflection map. It targets novel-view rendering from photographs. On six-scene RFFR, it averages 34.856 PSNR and 0.957 SSIM, versus 33.978 and 0.956 for 3D-GS. Its average LPIPS is worse than 3D-GS; results are strongest on conspicuously reflective scenes.
Summary by alphaXiv.
Tip: change arxiv.org to autoarxiv.org in any arXiv abstract link to open the paper here.