SplatAnything：面向任意场景前馈3DGS的高斯间距归一化

AI 摘要

研究者提出 SplatAnything，通过高斯间距归一化（GSNorm）对场景级高斯间距统计量进行规范化，解决现有前馈 3DGS 方法中高斯尺度监督不一致的问题。该方法属于 InfiniSplat 系列 v3，并建立了覆盖多数据集、不同输入稀疏度和外推视角的系统性基准。实验显示其在场景尺度与视角大幅变化下达到 SOTA，可应用于自动驾驶、机器人导航、AR/VR 和游戏开发。

正文节选

SplatAnything: Gaussian Spacing Normalization for Feed-forward 3DGS across Arbitrary Scenes Abstract SplatAnything applies Gaussian Spacing Normalization to canonicalize scene-level Gaussian spacing statistics, improving feed-forward 3DGS reconstruction across varying scene scales and viewpoints. We identify a key limitation of existing feed-forward 3DGS methods: Gaussian attributes are typically predicted in a radial-distance-normalized space that canonicalizes a global radial statistic of the

发布时间：—

抓取时间：2026-11-09 15:48

来源机构：Hugging Face