Reproduce “LUMEN: Geometry-Guided Diffusion Model for Spatially-Varying HDR Illumination Estimation” 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
Hongze Chen, Fangyu Lin, Runze Zhang, Lingting Zhu, Weikai Chen, Zeyu Hu, Yingda Yin, Keyang Luo and 3 more
LUMEN estimates location-specific illumination by rendering reconstructed scene geometry as a partial panorama, then using a diffusion model to complete it into a full HDR environment map. It combines this geometric cue with input-view appearance. On synthetic Infinigen scenes, mirror-sphere SSIM is 0.849 versus 0.511 for adapted 4DLighting; the comparison does not isolate the partial-panorama mechanism.
Summary by alphaXiv.
Tip: change arxiv.org to autoarxiv.org in any arXiv abstract link to open the paper here.