Research map / 3D vision / Human pose estimation
Lightweight Multi-View 3D Pose Estimation Through Camera-Disentangled Representation
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 |
|---|---|---|---|
| Use camera-conditioned canonical feature fusion instead of independent per-view prediction (ablation) | 3D pose estimation · TotalCapture test (seen cameras) | MPJPE (mm), mean: 39.3 -> 27.5 | Helped |
| Use camera-conditioned canonical feature fusion instead of independent per-view prediction (ablation) | 3D pose estimation · Human3.6M test (no additional training data) | MPJPE (mm), mean: 43.2 -> 30.2 | Helped |
| Add feature fusion across views (ablation) | 3D pose estimation · TotalCapture test (seen cameras) | MPJPE (mm), mean: 39.3 -> 31.8 | Helped |
| Add camera-conditioned canonical transforms to feature fusion (ablation) | 3D pose estimation · TotalCapture test (seen cameras) | MPJPE (mm), mean: 31.8 -> 28.6 | Helped |
| Back-propagate through the triangulation layer using the 3D target metric (ablation) | Canonical Fusion · TotalCapture test (seen cameras) | MPJPE (mm), mean: 28.6 -> 27.5 | Helped |
| Add feature fusion across views (ablation) | 3D pose estimation · Human3.6M test (no additional training data) | MPJPE (mm), mean: 43.2 -> 35.4 | Helped |
| Add camera-conditioned canonical transforms to feature fusion (ablation) | 3D pose estimation · Human3.6M test (no additional training data) | MPJPE (mm), mean: 35.4 -> 32.5 | Helped |
| Back-propagate through the triangulation layer using the 3D target metric (ablation) | Canonical Fusion · Human3.6M test (no additional training data) | MPJPE (mm), mean: 32.5 -> 30.2 | Helped |
| Use camera-conditioned canonical fusion rather than camera-unaware Fusion (ablation) | 3D pose estimation · TotalCapture test (unseen cameras 2,4,6,8) | MPJPE (mm), mean: 78.9 -> 38.2 | Helped |
| Use camera-unaware Fusion instead of independent per-view prediction (ablation) | 3D pose estimation · TotalCapture test (unseen cameras 2,4,6,8) | MPJPE (mm), mean: 48.2 -> 78.9 | Hurt |
| Use Canonical Fusion rather than the volumetric Iskakov method | 3D pose estimation · Human3.6M test (additional training data) | MPJPE: 21.0 vs 20.8 | No clear effect |
| Use Canonical Fusion rather than the volumetric Iskakov method | 3D pose estimation · Human3.6M test (additional training data) | inference time: 0.04s vs 2.30s | Helped |
Closest papers by meaning
- Weakly-Supervised Discovery of Geometry-Aware Representation for 3D Human Pose Estimation (CVPR 2019 · similarity 0.88)
- Coarse-To-Fine Volumetric Prediction for Single-Image 3D Human Pose (CVPR 2017 · similarity 0.88)
- A Single 2D Pose with Context is Worth Hundreds for 3D Human Pose Estimation (NeurIPS 2023 · similarity 0.88)
- 3D Bounding Box Estimation Using Deep Learning and Geometry (CVPR 2017 · similarity 0.88)
- SVTformer: Spatial-View-Temporal Transformer for Multi-View 3D Human Pose Estimation (AAAI 2025 · similarity 0.87)
- Multiple View Geometry Transformers for 3D Human Pose Estimation (CVPR 2024 · 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.