Research map / 3D vision / Human pose estimation
Self-Supervised Learning of 3D Human Pose Using Multi-View Geometry
Check 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 multi-view epipolar geometry to self-supervise a monocular 3D pose estimator | EpipolarPose · Human3.6M | MPJPE: 76.60 vs 118.41 | Helped |
| Add a refinement unit to the self-supervised model (ablation) | EpipolarPose SS · Human3.6M | MPJPE: 76.60 -> 60.56 | Helped |
| Use ground-truth 2D poses instead of estimated 2D poses for triangulation (ablation) | EpipolarPose SS · Human3.6M | MPJPE: 76.60 -> 55.08 | Helped |
| Use an H36M-trained 2D detector instead of an MPII-trained detector (ablation) | Triangulation supervision · Human3.6M triangulation | MPJPE: 45.86 -> 28.37 | Helped |
| Add camera rotation R to the triangulation-based supervision (ablation) | EpipolarPose SS · Human3.6M | NMPJPE: 77.75 -> 75.25 | Helped |
| Add camera rotation R to the triangulation-based supervision (ablation) | EpipolarPose S1 · MPI-INF-3DHP | NMPJPE: 119.86 -> 115.37 | Helped |
| Remove camera rotation R (ablation) | Ground-truth 2D triangulation · Human3.6M triangulation | PMPJPE: 2.13 -> 15.06 | Hurt |
| Use an MPII-trained detector rather than ground-truth 2D keypoints for triangulation (ablation) | Triangulation supervision · Human3.6M triangulation | MPJPE: 4.38 -> 45.86 | Hurt |
| Train with only Subject 1 rather than the complete self-supervised training set (ablation) | EpipolarPose · Human3.6M | MPJPE: 76.60 -> 65.35 | Helped |
| Use self-supervised training instead of full 3D supervision (ablation) | EpipolarPose · Human3.6M | MPJPE: 51.83 -> 76.60 | Hurt |
| Use self-supervised training instead of full 3D supervision (ablation) | EpipolarPose · MPI-INF-3DHP | PCK: 77.5 -> 64.7 | Hurt |
| Train using only Subject 1 rather than all available unlabeled multi-view footage (ablation) | EpipolarPose · Human3.6M | MPJPE: 76.60 -> 65.35 | Helped |
Closest papers by meaning
- SelfPose3d: Self-Supervised Multi-Person Multi-View 3d Pose Estimation (CVPR 2024 · similarity 0.89)
- FreeMan: Towards Benchmarking 3D Human Pose Estimation under Real-World Conditions (CVPR 2024 · similarity 0.88)
- ManiPose: Manifold-Constrained Multi-Hypothesis 3D Human Pose Estimation (NeurIPS 2024 · similarity 0.87)
- COMPOSE: Hypergraph Cover Optimization for Multi-view 3D Human Pose Estimation (NeurIPS 2026 · similarity 0.87)
- MV-SSM: Multi-View State Space Modeling for 3D Human Pose Estimation (CVPR 2025 · similarity 0.87)
- Self-Supervised 3D Human Mesh Recovery from a Single Image with Uncertainty-Aware Learning (AAAI 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.