Research map / 3D vision / Human pose estimation
CAPTRA: CAtegory-Level Pose Tracking for Rigid and Articulated Objects From Point Clouds
Yijia Weng, He Wang, Qiang Zhou, Yuzhe Qin, Yueqi Duan, Qingnan Fan, Baoquan Chen, Hao Su, Leonidas J. Guibas
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 pose canonicalization (previous-frame pose) to CoordinateNet (ablation) | CoordinateNet · NOCS-REAL275 | 5deg5cm accuracy: 14.93 -> 46.74 | Improved |
| add RotationNet to canonicalized CoordinateNet (ablation) | CAPTRA · NOCS-REAL275 | 5deg5cm accuracy: 46.74 -> 62.16 | Improved |
| use RotationNet residual rotation regression instead of DSAC++ pose fitting (ablation) | canonicalized CoordinateNet · NOCS-REAL275 | 5deg5cm accuracy: 54.77 -> 62.16 | Improved |
| add CoordinateNet to canonicalized pose regression (instead of regressing scale and translation) (ablation) | CAPTRA · NOCS-REAL275 | 5deg5cm accuracy: 25.99 -> 62.16 | Improved |
| add scale, translation and corner pose losses (ablation) | CAPTRA · NOCS-REAL275 | 5deg5cm accuracy: 60.48 -> 62.16 | Improved |
| add RotationNet to canonicalized CoordinateNet (ablation) | CAPTRA · SAPIEN articulated | 5deg5cm accuracy: 95.06 -> 98.35 | Improved |
| use RotationNet instead of DSAC++ pose fitting (ablation) | canonicalized CoordinateNet · SAPIEN articulated | 5deg5cm accuracy: 95.68 -> 98.35 | Improved |
| add CoordinateNet to canonicalized pose regression (ablation) | CAPTRA · SAPIEN articulated | 5deg5cm accuracy: 21.69 -> 98.35 | Improved |
| add scale, translation and corner pose losses (ablation) | CAPTRA · SAPIEN articulated | 5deg5cm accuracy: 97.63 -> 98.35 | Improved |
| track scale instead of fixing the noisy initial scale (ablation) | CAPTRA · SAPIEN articulated | mIoU: 71.70% -> 74.00% (scale error 1.09% -> 0.30%) | Improved |
| track scale instead of fixing the noisy initial scale (ablation) | CAPTRA · NOCS-REAL275 (excluding camera) | mIoU: 73.43% -> 76.42% | Improved |
| add ground-truth joint axis rotation projection at test time (ablation) | CAPTRA · SAPIEN articulated | 5deg5cm accuracy: 98.35 -> 98.74 | No clear effect |
Later papers that compared with this method
5 later accepted papers report CAPTRA as a baseline in their results tables; 5 of them report a better number on most of the results they share with it.
- Exploring Category-level Articulated Object Pose Tracking on SE(3) Manifolds (AAAI 2026 · better on 1 of 1 shared result)
- Towards Real-World Aerial Vision Guidance With Categorical 6D Pose Tracker (TPAMI 2025 · better on 3 of 3 shared results)
- KPA-Tracker: Towards Robust and Real-Time Category-Level Articulated Object 6D Pose Tracking (AAAI 2024 · better on 5 of 5 shared results)
- Tracking and Reconstructing Hand Object Interactions from Point Cloud Sequences in the Wild (AAAI 2023 · better on 2 of 2 shared results)
- CATRE: Iterative Point Clouds Alignment for Category-Level Object Pose Refinement (ECCV 2022 · better on 2 of 3 shared results)
Matched by method name in results tables read by AI. Authors' own tables tend to favour the newer method, and a comparison can be missed when a method is renamed.
Closest papers by meaning
- KPA-Tracker: Towards Robust and Real-Time Category-Level Articulated Object 6D Pose Tracking (AAAI 2024 · similarity 0.90)
- Category-Level Articulated Object Pose Estimation (CVPR 2020 · similarity 0.90)
- Category-Level Articulated Object Pose Estimation via Pose–Shape Hypothesis Generation and Verification (ECCV 2026 · similarity 0.90)
- CATRE: Iterative Point Clouds Alignment for Category-Level Object Pose Refinement (ECCV 2022 · similarity 0.89)
- EfficientCAPER: An End-to-End Framework for Fast and Robust Category-Level Articulated Object Pose Estimation (NeurIPS 2024 · similarity 0.88)
- CAP-Net: A Unified Network for 6D Pose and Size Estimation of Categorical Articulated Parts from a Single RGB-D Image (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.
