Research map / 3D vision / Human pose estimation
MAMMA: Markerless Accurate Multi-person Motion Acquisition
Hanz Cuevas Velasquez, Anastasios Yiannakidis, Soyong Shin, Giorgio Becherini, Markus Höschle, Joachim Tesch, Taylor Obersat, Tsvetelina Alexiadis, Eni Halilaj, Michael J. Black
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 |
|---|---|---|---|
| condition landmark network on SAM2 segmentation masks (ablation) | MammaNet · Harmony4D two-person (IoU>0.5) | 2D landmark error (px): 31.96 -> 18.33 | Improved |
| condition landmark network on SAM2 masks (ablation) | MammaNet · RICH single person | 2D landmark error (px): 8.55 -> 8.83 | No clear effect |
| condition on masks (ablation) | MammaNet · CHI3D two-person | 2D landmark error (px): 6.22 -> 4.36 | Improved |
| use one learnable query per landmark instead of a single query for all landmarks (ablation) | MammaNet vs CameraHMR decoder · MOYO | 2D landmark error (px): 12.53 -> 11.40 | Improved |
| add person-person contact optimization stage (ablation) | MAMMA · Harmony4D, CHI3D, MammaEval-D | mean penetration depth (mm): 10.50 -> 8.46 | Improved |
| add contact optimization stage (ablation) | MAMMA · 3D fitting MPJPE benchmarks | MPJPE: Harmony4D 45.26 -> 45.35; CHI3D 38.01 -> 37.96 | No clear effect |
| optimize floor contact (ablation) | MAMMA · 3D fitting | reconstruction accuracy | No clear effect |
| markerless MAMMA versus Vicon+MoSh++ marker-based pipeline | MAMMA · MammaEval-Extra | held-out marker error (mm): 21.619mm vs 22.481mm | No clear effect |
| match people across views via symmetric epipolar affinity of landmarks | MAMMA correspondence · Hi4D, Harmony4D, CHI3D, MammaEval-D | identity matching accuracy: 100% accuracy | Improved |
Closest papers by meaning
- Motions as Queries: One-Stage Multi-Person Holistic Human Motion Capture (CVPR 2025 · similarity 0.89)
- Capturing Humans in Motion: Temporal-Attentive 3D Human Pose and Shape Estimation From Monocular Video (CVPR 2022 · similarity 0.88)
- ArtTrack: Articulated Multi-Person Tracking in the Wild (CVPR 2017 · similarity 0.87)
- Capturing Closely Interacted Two-Person Motions with Reaction Priors (CVPR 2024 · similarity 0.87)
- CoMotion: Concurrent Multi-person 3D Motion (ICLR 2025 · similarity 0.86)
- HumanMM: Global Human Motion Recovery from Multi-shot Videos (CVPR 2025 · similarity 0.86)
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.
