Reproduce “Rethinking Sparse-View 3D Gaussian Splatting: A Few-Shot Learning Perspective on Overfitting Prevention” 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

Lejia Ye, Bin Wu, Jingjing Gao, Yiyang Wu, Xiaohu Zhang

The authors present sparse-view 3D Gaussian Splatting for dynamic-human view synthesis. It combines pseudo-view supervision, camera-pose correction, static-Gaussian dropout, and validation-based early stopping. With eight training and eight test views, it leads all three metrics on validation sequences; on Seq. 012, PSNR is 26.55 versus 24.68 for Binocular3DGS. No ablation identifies component effects; rapid motion may violate the pseudo-view assumption.

Summary by alphaXiv.

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