Research map / 3D vision / Human pose estimation
Generative Category-level Object Pose Estimation via Diffusion Models
Jiyao Zhang, Mingdong Wu, Hao Dong
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 |
|---|---|---|---|
| rank candidates by energy model and mean-pool instead of random pick (ablation) | GenPose · REAL275 | 5deg2cm mAP: 13.5 -> 52.1 | Improved |
| add mean pooling over candidates (random ranking) (ablation) | GenPose · REAL275 | 5deg2cm mAP: 13.5 -> 49.4 | Improved |
| use energy-based ranking instead of random ranking before pooling (ablation) | GenPose · REAL275 | 5deg2cm mAP: 49.4 -> 52.1 | Improved |
| filter low-energy candidates (delta=60%) instead of keeping all (ablation) | GenPose K=50 · REAL275 | 10deg2cm mAP: 69.7 -> 72.4 | Improved |
| filter too aggressively (delta=20%) (ablation) | GenPose K=50 · REAL275 | 10deg2cm mAP: 72.4 -> 70.2 | Hurt |
| increase candidates K from 10 to 50 (ablation) | GenPose · REAL275 | 10deg2cm mAP: 67.9 -> 72.4 | Improved |
| increase candidates K from 50 to 100 (ablation) | GenPose · REAL275 | 10deg2cm mAP: 72.4 -> 72.6 | No clear effect |
| generative diffusion formulation instead of regression | category-level pose estimation · REAL275 | 5deg5cm mAP: 48.1 -> 60.9 vs RBP-Pose | Improved |
| generative model on unseen symmetric categories vs RBP-Pose | GenPose · REAL275 bowl unseen | 5deg2cm mAP: 0.0 -> 64.5 | Improved |
| adapt estimator to tracking by warm-starting from previous frame | GenPose · REAL275 tracking | 5deg5cm / FPS: 71.5 vs 57.2 (5deg5cm); 17.18 vs 89.21 FPS | Mixed |
Earlier methods this paper compares with
Named baselines from earlier accepted papers in this paper's results tables: better than 0 of 1 on most shared results.
- CATRE — CATRE: Iterative Point Clouds Alignment for Category-Level Object Pose Refinement (ECCV 2022) · better on 1 of 2 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
- A Visual Navigation Perspective for Category-Level Object Pose Estimation (ECCV 2022 · similarity 0.92)
- Diff9D: Diffusion-Based Domain-Generalized Category-Level 9-DoF Object Pose Estimation (TPAMI 2025 · similarity 0.91)
- Zero-Shot Category-Level Object Pose Estimation (ECCV 2022 · similarity 0.90)
- DICArt: Advancing Category-level Articulated Object Pose Estimation in Discrete State-Spaces (CVPR 2026 · similarity 0.90)
- Object Pose Estimation via the Aggregation of Diffusion Features (CVPR 2024 · similarity 0.89)
- Unsupervised Learning of Category-Level 3D Pose from Object-Centric Videos (CVPR 2024 · similarity 0.89)
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.
