Research map / 3D vision / Human pose estimation
End-to-end Dynamic Matching Network for Multi-view Multi-person 3d Pose Estimation
Congzhentao Huang, Shuai Jiang, Yang Li, Ziyue Zhang, Jason Traish, Chen Deng, Sam Ferguson, Richard Yi Da Xu
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 |
|---|---|---|---|
| train 2d detection, matching and 3d estimation end-to-end instead of separate steps (ablation) | multi-view 3d pose model · Shelf (PCP@0.5) | PCP@0.5 average: 96.67 -> 97.39 | Improved |
| use our 3d-space dynamic matching instead of person re-id (ablation) | multi-view 3d pose model · Shelf (PCP@0.5) | PCP@0.5 average: 95.68 -> 97.39 | Improved |
| use our 3d-space dynamic matching instead of epipolar geometry (ablation) | multi-view 3d pose model · Shelf (PCP@0.5) | PCP@0.5 average: 93.44 -> 97.39 | Improved |
| use our matching instead of person re-id (time cost) (ablation) | multi-view 3d pose model · Shelf (PCP@0.5) | time (s): 6.73 -> 0.96 | Improved |
| use our matching instead of epipolar geometry (time cost) (ablation) | multi-view 3d pose model · Shelf (PCP@0.5) | time (s): 0.64 -> 0.96 | Hurt |
| add learnable weight matrix to point triangulation (ablation) | 3d pose estimator · Shelf (PCP@0.5) | PCP@0.5 average: 94.00 -> 97.39 | Improved |
| replace learnable triangulation network with weighted point triangulation (ablation) | 3d pose estimator · Shelf (PCP@0.5) | PCP@0.5 average: 97.13 -> 97.39 | Improved |
| end-to-end vs multi-step on Actor 2 (ablation) | multi-view 3d pose model · Shelf (PCP@0.5), Actor 2 | PCP@0.5: 95.16 -> 96.22 | Improved |
| our model vs Dong et al. on Shelf | best prior method · Shelf (PCP@0.5) | PCP@0.5 average: 96.76 -> 97.39 | Improved |
| our model vs Dong et al. on Campus | best prior method · Campus (PCP@0.5) | PCP@0.5 average: 95.85 -> 96.71 | Improved |
| our model vs prior best on occluded lower arms of Actor 2 | Dong et al. · Shelf (PCP@0.5), Actor 2 lower arms | PCP@0.5: 71.62 -> 81.08 | Improved |
Closest papers by meaning
- Fast and Robust Multi-Person 3D Pose Estimation From Multiple Views (CVPR 2019 · similarity 0.85)
- Dual Networks Based 3D Multi-Person Pose Estimation From Monocular Video (TPAMI 2022 · similarity 0.85)
- Graph and Temporal Convolutional Networks for 3D Multi-person Pose Estimation in Monocular Videos (AAAI 2021 · similarity 0.84)
- 3D Pictorial Structures Revisited: Multiple Human Pose Estimation (TPAMI 2016 · similarity 0.84)
- Fast and Robust Multi-Person 3D Pose Estimation and Tracking From Multiple Views (TPAMI 2021 · similarity 0.84)
- COMPOSE: Hypergraph Cover Optimization for Multi-view 3D Human Pose Estimation (NeurIPS 2026 · similarity 0.82)
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.
