Reproduce “DeLiRa: Self-Supervised Depth, Light, and Radiance Fields” 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
Vitor Guizilini, Igor Vasiljevic, Jiading Fang, Rares Ambrus, Sergey Zakharov, Vincent Sitzmann, Adrien Gaidon
DeLiRa combines volumetric rendering with a multi-view photometric loss, using predicted depth to check agreement between overlapping images, and decodes depth, color, and radiance from a shared scene representation. On ScanNet-Frontal, its depth-field decoder reports AbsRel 0.054 versus NeRF’s 0.393, with posed images and no depth supervision. This evidence is limited to that benchmark and protocol, not arbitrarily sparse views.
Summary by alphaXiv.
Tip: change arxiv.org to autoarxiv.org in any arXiv abstract link to open the paper here.