Reproduce “MoSoLoco: Monocular 3D Human Localization through Visual Latent Features and Social Context” 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
Yasamin Borhani, Javad Khoramdel, Yang Gao, Po-Chien Luan, Alexandre Alahi
MoSoLoco reuses a frozen pose detector’s joint-aligned visual features alongside 2D skeletons to estimate people’s ground-plane positions from monocular images. Its optional same-frame attention module incorporates nearby detections, without temporal tracking. On JRDB, MoSoLoco-Trans with DETRPose-S achieves 0.248 m ADE versus 0.322 m for MonoTransmotion, a 23.0% reduction. This comparison covers accepted matches under evaluation policy, not missed detections.
Summary by alphaXiv.
Tip: change arxiv.org to autoarxiv.org in any arXiv abstract link to open the paper here.