A shared MLP predicts Gaussian visibility from viewpoint to discard occluded primitives during 3DGS rendering, combined with an instanced Tensor-Core rasterizer for lower VRAM and better quality on composed scenes.
This includes the results on the full layout files compared with our baseline instanced rasterizer implementation
1 Pith paper cite this work. Polarity classification is still indexing.
1
Pith paper citing it
fields
cs.CV 1years
2025 1verdicts
CONDITIONAL 1representative citing papers
citing papers explorer
-
NVGS: Neural Visibility for Occlusion Culling in 3D Gaussian Splatting
A shared MLP predicts Gaussian visibility from viewpoint to discard occluded primitives during 3DGS rendering, combined with an instanced Tensor-Core rasterizer for lower VRAM and better quality on composed scenes.