Reproduce “GS-CPE: Unified 6-Degree-of-Freedom Camera Pose Estimation via 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

Huaiyuan Weng, Chul Min Yeum, Su-Min Kang

GS-CPE localizes a camera in a mapped scene by matching query-image features to a rendered 3D Gaussian Splatting map, then refining pose through visibility-masked, multi-scale image alignment and adaptive re-rendering. On 7Scenes, its average per-scene median errors are 0.91 cm and 0.29°, versus 1.4 cm and 0.37° for GSplatLoc; it is not best in every scene and requires prebuilt maps.

Summary by alphaXiv.

Tip: change arxiv.org to autoarxiv.org in any arXiv abstract link to open the paper here.