Research map / 3D vision / Human pose estimation
Multi-HMR: Multi-Person Whole-Body Human Mesh Recovery in a Single Shot
Fabien Baradel, Thomas LUCAS, Matthieu Armando, Salma Galaaoui, Romain Brégier, Philippe Weinzaepfel, Gregory Rogez
Read the paperCheck your idea against 137,217 papers
AI-extracted. The measured effects below were read by AI from the paper's full text (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 |
|---|---|---|---|
| replace HRNet with a ViT backbone (ablation) | Multi-HMR with HPH · EHF PVE-All | PVE: 115.2 -> 80.1 | Improved |
| replace iterative regressor with the cross-attention Human Perception Head (ablation) | Multi-HMR ViT-B · EHF PVE-All | PVE: 65.0 -> 55.3 | Improved |
| add self-attention across person queries in HPH (ablation) | HPH · EHF PVE-All | PVE: 63.2 -> 55.3 | Improved |
| increase HPH layers from 2 to 8 (ablation) | HPH · MuPoTs PCK / EHF PVE | PCK / PVE: 76.3 -> 78.9 PCK; 55.3 -> 51.0 PVE | Improved |
| train on synthetic data (BEDLAM+AGORA) instead of real data (ablation) | Multi-HMR ViT-B · EHF PVE-All | PVE: 70.2 -> 55.3 | Improved |
| add CUFFS close-up synthetic dataset (ablation) | Multi-HMR ViT-B · EHF-H PVE-Hands | PVE: 47.4 -> 40.5 | Improved |
| mix real-world data with synthetic data (ablation) | Multi-HMR ViT-B · EHF PVE-All | PVE: 49.8 -> 61.1 | Hurt |
| increase input resolution (ablation) | Multi-HMR · MuPoTs, CMU, EHF | accuracy and speed | Mixed |
| embed flattened intrinsics matrix through a linear layer (ablation) | Multi-HMR ViT-B · MuPoTs PCK | PCK: 76.3 -> 74.8 | Hurt |
| add ray-direction camera embedding with focal length normalization (ablation) | Multi-HMR ViT-B · MuPoTs PCK | PCK: 76.3 -> 78.8 | Improved |
| give ground-truth intrinsics at train and test time (ablation) | Multi-HMR · MuPoTs MRPE | MRPE: 1345 -> 693 | Improved |
| use camera intrinsics for reconstruction metrics (ablation) | Multi-HMR · MuPoTs/3DPW/CMU reconstruction | reconstruction error: 76.3 -> 76.5 PCK | No clear effect |
Earlier methods this paper compares with
Named baselines from earlier accepted papers in this paper's results tables: better than 1 of 1 on most shared results.
- PSVT — PSVT: End-to-End Multi-Person 3D Pose and Shape Estimation With Progressive Video Transformers (CVPR 2023) · better on 3 of 3 shared results
Matched by method name in results tables read by AI. Authors' own tables tend to favour the newer method, and a comparison can be missed when a method is renamed.
Closest papers by meaning
- End-to-End Recovery of Human Shape and Pose (CVPR 2018 · similarity 0.93)
- SAM 3D Body: Robust Full-Body Human Mesh Recovery (CVPR 2026 · similarity 0.91)
- AiOS: All-in-One-Stage Expressive Human Pose and Shape Estimation (CVPR 2024 · similarity 0.91)
- MUC: Mixture of Uncalibrated Cameras for Robust 3D Human Body Reconstruction (AAAI 2025 · similarity 0.90)
- SAT-HMR: Real-Time Multi-Person 3D Mesh Estimation via Scale-Adaptive Tokens (CVPR 2025 · similarity 0.89)
- GenHMR: Generative Human Mesh Recovery (AAAI 2025 · similarity 0.89)
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.
