Reproduce “SplatAnything: Gaussian Spacing Normalization for Feed-forward 3DGS across Arbitrary Scenes” 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
Hao Yu, Zhenwei Wang, Yongzhen Hu, Yuning Gong, Junta Wu, Xiaowei Zhou, Hujun Bao, Tengfei Wang and 2 more
SplatAnything uses feed-forward 3D Gaussian Splatting to reconstruct scenes from images without per-scene optimization. It introduces Gaussian Spacing Normalization (GSNorm), which scales geometry using median neighboring-point spacing rather than mean camera distance. In a four-view ablation, GSNorm scores 0.198/0.245 DISTS/LPIPS, below RDNorm and scale calibration. On pose-free Mip-NeRF 360, LPIPS is 0.463 versus Mirror2-GS’s 0.528, within benchmark settings.
Summary by alphaXiv.
Tip: change arxiv.org to autoarxiv.org in any arXiv abstract link to open the paper here.