Reproduce “NVGS: Neural Visibility for Occlusion Culling in 3D Gaussian Splatting” 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
Brent Zoomers, Florian Hahlbohm, Joni Vanherck, Lode Jorissen, Marcus Magnor, Nick Michiels
NVGS learns per-asset visibility to cull Gaussian splats that will not contribute to a view before rendering, using a small neural network. It reuses predictions across asset instances, reducing frame-time work. In the CROWD comparison, NVGS reports 48.6 PSNR and 4.5 GB peak VRAM, versus V3DG’s 43.2 PSNR and 20.4 GB; the combined-system result does not isolate the neural component.
Summary by alphaXiv.
Tip: change arxiv.org to autoarxiv.org in any arXiv abstract link to open the paper here.