Reconstructing Humans and Objects in Interaction using Large Reconstruction Models
MILO reconstructs human-object interactions from a single image by using Large Reconstruction Models (LRMs) as a geometric scaffold. It simplifies reconstruction by interpreting the LRM's mesh, which preserves relative positions and proximity cues. The method estimates 3D keypoints, fits a parametric body model (SMPL-H), and refines the fit to produce anatomically plausible human meshes tightly aligned with the scene.