Research map / 3D vision / Human pose estimation
Domes to Drones: Self-Supervised Active Triangulation for 3D Human Pose Reconstruction
Read the paperCheck 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 |
|---|---|---|---|
| use learned RL viewpoint policy instead of Max-Azim heuristic | ACTOR · Panoptic multi-people, 3 cameras (hip-aligned) | 3d error (mm/joint): 103 mm/joint improvement at 3 cameras, 78 at 6 | Helped |
| use learned policy instead of random cameras | ACTOR · Panoptic multi-people, 3 cameras | 3d error total (mm/joint): 469.9 -> 281.5 | Helped |
| add camera history and auxiliary state to the deep feature blob (ablation) | ACTOR · Panoptic multi-people, 3 cameras | 3d error total (mm/joint): 372.9 -> 281.5 | Helped |
| add camera history and auxiliary state to the deep feature blob (ablation) | ACTOR · Panoptic single-person, auto | 3d error total (mm/joint): 80.2 -> 74.6 | Helped |
| add temporal fusion of 3d reconstructions (ablation) | ACTOR · Panoptic multi-people, 3 cameras | 3d error total (mm/joint): 565.2 -> 281.5 | Helped |
| add temporal fusion of 3d reconstructions (ablation) | ACTOR · Panoptic multi-people, 10 cameras | 3d error total (mm/joint): 136.6 -> 122.1 | Helped |
| add temporal fusion of 3d reconstructions (ablation) | ACTOR · Panoptic single-person, auto | 3d error total (mm/joint): 75.9 -> 74.6 | Helped |
| use ground-truth oracle camera selection instead of ACTOR | ACTOR · Panoptic multi-people, 3 cameras | 3d error total (mm/joint): 281.5 -> 147.6 | Helped |
| stop automatically once all joints triangulated (auto mode) vs random | ACTOR · Panoptic multi-people, auto | 3d error total (mm/joint): 142.7 -> 125.6 | Helped |
| fine-tune policy for drone and triangulate 5 selected views instead of all 33 | ACTOR · Drone scene | 2d reprojection error (pixels): 35.4 -> 11.5 | Helped |
Closest papers by meaning
- Deep Reinforcement Learning for Active Human Pose Estimation (AAAI 2020 · similarity 0.91)
- 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)
- Sparse-view Pose Estimation and Reconstruction via Analysis by Generative Synthesis (NeurIPS 2024 · similarity 0.88)
- LEAP: Liberate Sparse-View 3D Modeling from Camera Poses (ICLR 2024 · similarity 0.88)
- ODHSR: Online Dense 3D Reconstruction of Humans and Scenes from Monocular Videos (CVPR 2025 · similarity 0.87)
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.