Reproduce “HARMONI: Aligning Human and Scene Priors for Multi-View 4D Reconstruction” 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
Sangmin Kim, Minhyuk Hwang, Geonho Cha, Dongyoon Wee, Jaesik Park
HARMONI reconstructs people and scenes in aligned 3D coordinates by using scene depth to place the body and visible body proportions to calibrate scene scale. It combines estimates across views through confidence-weighted fusion, without test-time optimization. On RICH, it reports W-MPJPE of 132.4 mm versus 184.9 mm for Human3R; on EgoHumans, 0.37 m versus HSfM’s 1.04 m, with ground-truth identities.
Summary by alphaXiv.
Tip: change arxiv.org to autoarxiv.org in any arXiv abstract link to open the paper here.