Reproduce “SOAR: Self-Occluded Avatar Recovery from a Single Video In the Wild” 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
Zhuoyang Pan, Angjoo Kanazawa, Hang Gao
SOAR builds a single poseable human avatar from casual video by combining a globally consistent surfel representation with structural normal supervision and diffusion-based refinement. The method uses observed frames to constrain visible regions and a learned prior to propose unseen appearance. On seven DNA-Rendering sequences, it reports PSNR 23.16 and LPIPS 0.055, outperforming listed baselines, though unseen details remain unverified.
Summary by alphaXiv.
Tip: change arxiv.org to autoarxiv.org in any arXiv abstract link to open the paper here.