Reproduce “PanoImager: Geometry-Guided Novel View Synthesis and Reconstruction from Sparse Panoramic Views” 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

Zhisong Xu, Takeshi Oishi

PanoImager addresses sparse panoramic robotic mapping by generating geometry-guided views and fitting them with captured images into a shared 3D Gaussian scene. It splits panoramas into perspective views, estimates poses and depth, then uses those cues to complete additional viewpoints. On OmniScenes, visible-view PSNR/SSIM/LPIPS scores are 30.86/0.893/0.148 versus baseline bests of 25.67/0.890/0.217; these measure image fidelity, not metric geometry.

Summary by alphaXiv.

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