Research map / 3D vision / Human pose estimation
REMIPS: Physically Consistent 3D Reconstruction of Multiple Interacting People under Weak Supervision
Read the paperCheck your idea against 108,967 papers
AI-extracted summary. Written by AI from the paper's title only (no summary yet). Numbers and claims can be wrong or missing context, so check the paper before you cite or build on it.
Measured effects reported in the paper
| What was changed | Where | Measured | Effect |
|---|---|---|---|
| add self-contact, interaction-contact and collision losses (Lic, Lsc, Lics, Lscs) jointly (ablation) | REMIPS · CHI3D | MPJPE (mm): Lics removed: 127.1 -> 120.8 with all losses | Helped |
| add interaction contact signature loss Lics (ablation) | REMIPS · CHI3D | MPVPE (mm): 142.1 -> 134.7 | Helped |
| add self-contact signature loss Lscs (ablation) | REMIPS · CHI3D | MPJPE (mm): 125.3 -> 120.8 | Helped |
| add interpenetration collision loss Lic (ablation) | REMIPS · CHI3D | MPJPE (mm): 122.7 -> 120.8 | Helped |
| add self-collision loss Lsc (ablation) | REMIPS · CHI3D | MPJPE (mm): 123.5 -> 120.8 | Helped |
| provide pseudo ground-truth camera intrinsics as input (ablation) | REMIPS · CHI3D | MPJPE (mm): 139.0 -> 120.8 | Helped |
| provide large image context (crop around all persons) instead of tight single-person crop (ablation) | REMIPS · Panoptic | MPJPE (mm): 141.7 -> 138.2 | Helped |
| feed all person tokens jointly rather than one bounding box at a time (ablation) | REMIPS · Panoptic | MPJPE (mm): 138.3 vs 138.2 | No clear effect |
| REMIPS instead of SPIN in single-person tight-crop setting | single-person pose estimation · Panoptic | MPJPE (mm): 164.9 -> 141.7 | Helped |
| downscale input image to simulate distance from camera (ablation) | REMIPS · Panoptic | MPJPE (mm): 138.3 -> 140.2 (2x), 151.7 (4x), 210.5 (8x) | Hurt |
| REMIPS vs HUND on single-person Human3.6M P1 | REMIPS · Human3.6M P1 | MPJPE / MPJPE-PA (mm): MPJPE-PA 66.0 -> 64.3; MPJPE 91.8 -> 96.1 | Mixed |
Closest papers by meaning
- MultiPly: Reconstruction of Multiple People from Monocular Video in the Wild (CVPR 2024 · similarity 0.90)
- UniCon3R: Unified Contact-aware 4D Human-Scene Reconstruction from Monocular Video (NeurIPS 2026 · similarity 0.88)
- ContactField: Implicit Field Representation for Multi-Person Interaction Geometry (NeurIPS 2024 · similarity 0.87)
- Reconstructing People, Places, and Cameras (CVPR 2025 · similarity 0.87)
- Closely Interactive Human Reconstruction with Proxemics and Physics-Guided Adaption (CVPR 2024 · similarity 0.87)
- ODHSR: Online Dense 3D Reconstruction of Humans and Scenes from Monocular Videos (CVPR 2025 · similarity 0.87)
Writing a paper in this area? scime draft is a free LaTeX editor with AI included, real-time co-authors and a check of your references and venue rules before you submit. It also imports OpenAI Prism projects.