Research map / 3D vision / Human pose estimation
Generalizable and Actionable Parts Pose Estimation with Symmetry Annotation-Free Learning Strategy
wenxiao chen, Xueyu Yuan, Liu Liu, Di Wu, Dan Guo
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 symmetry-aware design (rotational parts) (ablation) | SAFAG · GAPartNet seen | rotation error (deg): 25.01 -> 3.35 | Improved |
| add symmetry-aware design (mirror parts) (ablation) | SAFAG · GAPartNet unseen | rotation error (deg): 56.27 -> 14.32 | Improved |
| replace HS-Layer with HyperS3 backbone (ablation) | SAFAG · GAPartNet seen | avg rotation error (deg): 4.50 -> 3.23 | Improved |
| replace HS-Layer with HyperS3 backbone (ablation) | SAFAG · GAPartNet unseen | avg rotation error (deg): 14.15 -> 10.83 | Improved |
| use explicit equivalent solution loss instead of symmetry-axis error loss (ablation) | SAFAG · GAPartNet seen Slider Lid | rotation error (deg): 22.23 -> 0.46 | Improved |
| use explicit equivalent solution loss instead of symmetry-axis error loss (ablation) | SAFAG · GAPartNet unseen Hinge Knob | rotation error (deg): 48.41 -> 28.98 | Improved |
| correctly label rotational symmetry rather than mirror (ablation) | SAFAG · GAPartNet seen rotational parts | rotation error (deg): 10.06 -> 3.35 | Improved |
| correctly label rotational symmetry rather than asymmetric (ablation) | SAFAG · GAPartNet seen rotational parts | rotation error (deg): 25.01 -> 3.35 | Improved |
| increase candidates K from 32 to 128 (ablation) | SAFAG · GAPartNet | rotation error (deg): 3.70 -> 3.23 seen; 4.68 -> 3.99 unseen | No clear effect |
| increase Neq from 18 to 72 (ablation) | SAFAG · GAPartNet | rotation error (deg): 5.97 -> 3.27 seen; 11.77 -> 3.55 unseen | Improved |
| increase Neq from 36 to 72 (ablation) | SAFAG · GAPartNet | training FPS: 4.33 -> 2.55 | Hurt |
| use SAFAG instead of RFMPose / GenPose++ generative models | category-level pose methods · GAPartNet | avg rotation error seen (deg): 17.03 (RFMPose) -> 3.23 | Improved |
Closest papers by meaning
- Towards Symmetry-sensitive Pose Estimation: A Rotation Representation for Symmetric Object Classes (IJCV 2026 · similarity 0.86)
- Two by Two: Learning Multi-Task Pairwise Objects Assembly for Generalizable Robot Manipulation (CVPR 2025 · similarity 0.85)
- Vision Foundation Model Enables Generalizable Object Pose Estimation (NeurIPS 2024 · similarity 0.84)
- Articulate your NeRF: Unsupervised articulated object modeling via conditional view synthesis (NeurIPS 2024 · similarity 0.84)
- Unsupervised Learning of Category-Level 3D Pose from Object-Centric Videos (CVPR 2024 · similarity 0.84)
- CAP-Net: A Unified Network for 6D Pose and Size Estimation of Categorical Articulated Parts from a Single RGB-D Image (CVPR 2025 · similarity 0.84)
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.
