Reproduce “ClearGS: Reliability-Aware Gaussian Splatting from Handheld Videos” 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
Xuanzhi Liu, Xinyi Wu, Hang Pan, Wensi Huang, Zhenyao Wu, Ruize Han, Song Wang
ClearGS reconstructs scenes from degraded handheld video by separating each frame’s appearance reliability from its geometric usefulness, then supplementing observations with selected restorations from the same video. On GS2E severe blur, across 10 scenes, it reports LPIPS 0.350 versus DiFiX3D’s 0.388, CLIP-IQA 0.293 versus 0.253, and MUSIQ 43.18 versus 36.93 at observed poses; results show perceptual improvements under tested conditions, not universal recovery.
Summary by alphaXiv.
Tip: change arxiv.org to autoarxiv.org in any arXiv abstract link to open the paper here.