Research map / 3D vision / Human pose estimation
Cross View Fusion for 3D Human Pose Estimation
Haibo Qiu, Chunyu Wang, Jingdong Wang, Naiyan Wang, Wenjun Zeng
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 |
|---|---|---|---|
| add cross-view feature fusion (vs Single), trained on H36M+MPII, triangulation (ablation) | Single baseline · Human3.6M test | MPJPE (mm) average: 36.28 -> 27.90 | Improved |
| add cross-view feature fusion, trained on H36M only, triangulation (ablation) | Single baseline · Human3.6M test | MPJPE (mm) average: 94.54 -> 38.29 | Improved |
| add cross-view feature fusion, trained on H36M only, RPSM (ablation) | Single baseline · Human3.6M test | MPJPE (mm) average: 47.82 -> 31.17 | Improved |
| add the fusion network instead of sum or max over epipolar lines (ablation) | Sum/Max baselines · Human3.6M 2D wrist | joint detection rate wrist (%): 91.57 -> 95.01 | Improved |
| add recursive refinement (RPSM) instead of triangulation, H36M only (ablation) | Single · Human3.6M test | MPJPE (mm) average: 94.54 -> 47.82 | Improved |
| add recursive refinement (RPSM) instead of triangulation, H36M+MPII (ablation) | Fusion · Human3.6M test | MPJPE (mm) average: 27.90 -> 26.21 | Improved |
| add recursive refinement (RPSM) instead of triangulation on the Sitting action (ablation) | Fusion +MPII · Human3.6M test | MPJPE (mm) Sitting: 40.47 -> 32.12 | Improved |
| run RPSM refinement iterations t=10 instead of PSM (t=0) (ablation) | Single-H36M · Human3.6M test (RPSM iterations) | MPJPE (mm): 95.23 -> 47.82 | Improved |
| run RPSM refinement iterations t=10 instead of PSM (t=0) (ablation) | Fusion-+MPII · Human3.6M test (RPSM iterations) | MPJPE (mm): 77.28 -> 26.21 | Improved |
| increase RPSM iterations from t=5 to t=10 (ablation) | Fusion-+MPII · Human3.6M test (RPSM iterations) | MPJPE (mm): 26.26 -> 26.21 | No clear effect |
| add cross-view fusion on Total Capture (ablation) | Single-RPSM · Total Capture | MPJPE (mm) mean: 41 -> 29 | Improved |
| add training data MPII to H36M for 2D estimation, no fusion, triangulation (ablation) | Single · Human3.6M test | MPJPE (mm) average: 94.54 -> 36.28 | Improved |
Closest papers by meaning
- Multi-View Multi-Person 3D Pose Estimation With Plane Sweep Stereo (CVPR 2021 · similarity 0.91)
- Cross-View Tracking for Multi-Human 3D Pose Estimation at Over 100 FPS (CVPR 2020 · similarity 0.90)
- Fast and Robust Multi-Person 3D Pose Estimation From Multiple Views (CVPR 2019 · similarity 0.89)
- Unsupervised Cross-Modal Alignment for Multi-Person 3D Pose Estimation (ECCV 2020 · similarity 0.89)
- CoDePose: Multi-View 3D Human Pose Estimation via Coupled 2D-3D Denoising Diffusion (ECCV 2026 · similarity 0.89)
- PoseIRM: Enhance 3D Human Pose Estimation on Unseen Camera Settings via Invariant Risk Minimization (CVPR 2024 · 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.
