Reproduce “Light Field Networks: Neural Scene Representations with Single-Evaluation 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
Vincent Sitzmann, Semon Rezchikov, William T. Freeman, Joshua B. Tenenbaum, Fredo Durand
LFNs render novel views by mapping each oriented ray directly to color in one network evaluation, avoiding repeated ray samples. A hypernetwork learns a prior over light fields from scenes, while test-time latent-code optimization fits an image. On 13 ShapeNet categories, LFN averages 24.95 dB PSNR, above SRN and DVR, but pixelNeRF scores 26.80 dB; the claim applies to simple scenes.
Summary by alphaXiv.
Tip: change arxiv.org to autoarxiv.org in any arXiv abstract link to open the paper here.