GeoGSplat: geometry priors top DTU and NRGBD with just 65K Gaussians
zhenjun_zhao · x · 2026-10-29
GeoGSplat is a geometry-grounded feed-forward Gaussian Splatting model that explicitly injects geometric priors into the non-pixel-aligned framework via a dual-branch encoder (Anchors for coarse 3D structure, Triplane for fine features) and a hierarchical decoder.
- Beats pixel-aligned and non-pixel-aligned SOTA on DTU and NRGBD for both novel view synthesis (PSNR 20.48/24.78) and 3D reconstruction (CD 0.018/0.017)
- Uses only 65K Gaussians vs 150K-643K for competitors
- Advantage grows as target viewpoints move farther from input cameras
More from Research
- Information-theoretic scaling laws predict memorization and FID in diffusion models — LucaAmb · 2026-10-29
- ICLR reviewer hands out scores of 1,1,1,1 and 2 across five papers — chhaviyadav_ · 2026-10-29
- kalomaze calls out hnet baseline: effective depth unmatched, no wall clock reported — kalomaze · 2026-10-29
- AVERT Study Launched: Can Autonomous Vehicles Reduce Trauma? — zakkohane · 2026-10-29
- Bad binder design drove only ~3% of Phase 2/3 failures — the rest elude AlphaFold — clarejtbirch · 2026-10-29
- Demo shows emergent 'in-context tokenization' beating fixed tokenizers — willdepue · 2026-10-29