Research map / 3D vision / Human pose estimation
Self-Supervised Category-Level Articulated Object Pose Estimation with Part-Level SE(3) Equivariance
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 a self-supervised, part-by-part reconstruction approach with part-level equivariant features | Category-level articulated object pose estimation · Complete point clouds, average over seven categories | Mean per-part rotation error (°): NPCS-EPN: 5.38; Ours: 7.90 | Hurt |
| Use a self-supervised, part-by-part reconstruction approach with part-level equivariant features | Category-level articulated object pose estimation · Complete point clouds, average over seven categories | Mean per-part rotation error (°): Oracle ICP: 46.11; Ours: 7.90 | Helped |
| Use the proposed method for articulated-object part segmentation | Part segmentation · Complete point clouds, average over seven categories | Segmentation MIoU: BSP-Net: 66.22; Ours: 73.26 | Helped |
| Use part-by-part rather than whole-shape reconstruction | Shape reconstruction · Complete point clouds, average over seven categories | Chamfer L1: EPN: 0.036; Ours: 0.031 | Helped |
| Use pose-aware, part-level equivariant features instead of global equivariant features (ablation) | Proposed method · All categories, per-category per-part average | Mean per-part rotation error (°): 10.40 -> 9.21 | Helped |
| Use pose-aware, part-level equivariant features instead of global equivariant features (ablation) | Proposed method · All categories, per-category per-part average | Mean per-part translation error: 0.072 -> 0.052 | Helped |
| Use pose-aware, part-level equivariant features instead of global equivariant features (ablation) | Proposed method · All categories, per-category per-part average | Chamfer L1: 0.032 -> 0.025 | Helped |
| Accumulate part-level features rather than using whole-shape features (ablation) | Proposed method · All categories, per-category per-part average | Mean per-part rotation error (°): 12.85 -> 9.21 | Helped |
| Accumulate part-level features rather than using whole-shape features (ablation) | Proposed method · All categories, per-category per-part average | Segmentation IoU: 71.68 -> 76.90 | Helped |
| Add joint regularization to the reconstruction loss (ablation) | Proposed method with pose-aware features · All categories, per-category per-part average | Mean per-part rotation error (°): 11.74 -> 9.21 | Helped |
| Add joint regularization to the reconstruction loss (ablation) | Proposed method with pose-aware features · All categories, per-category per-part average | Mean per-part translation error: 0.070 -> 0.052 | Helped |
| Add joint regularization to the reconstruction loss (ablation) | Proposed method with pose-aware features · All categories, per-category per-part average | Chamfer L1: 0.038 -> 0.025 | Helped |
Closest papers by meaning
- EfficientCAPER: An End-to-End Framework for Fast and Robust Category-Level Articulated Object Pose Estimation (NeurIPS 2024 · similarity 0.88)
- Unsupervised Learning of Category-Level 3D Pose from Object-Centric Videos (CVPR 2024 · similarity 0.88)
- Emergence of a Shared Canonical Object Frame from In-the-Wild Videos (NeurIPS 2026 · similarity 0.88)
- SecondPose: SE(3)-Consistent Dual-Stream Feature Fusion for Category-Level Pose Estimation (CVPR 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)
- Rethinking Correspondence-based Category-Level Object Pose Estimation (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.