Research map / 3D vision / Human pose estimation
Diff9D: Diffusion-Based Domain-Generalized Category-Level 9-DoF Object Pose Estimation
Jian Liu, Wei Sun, Hui Yang, Pengchao Deng, Chongpei Liu, Nicu Sebe, Hossein Rahmani, Ajmal Saeed Mian
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 RGB image condition (ablation) | Diff9D · REAL275 | 3D75 mAP: 32.9 -> 44.1 | Improved |
| add point cloud condition (ablation) | Diff9D · REAL275 | 3D75 mAP: 2.9 -> 44.1 | Improved |
| add time step condition (ablation) | Diff9D · REAL275 | 3D75 mAP: 0.0 -> 44.1 | Improved |
| add shape prior condition on top of shape estimator and encoder (ablation) | Diff9D · REAL275 | 3D75 mAP: 44.1 -> 43.8 | No clear effect |
| add shape estimator and encoder (ablation) | Diff9D · REAL275 | 3D75 mAP: 39.3 -> 44.1 | Improved |
| add NOCS shape regression and encoding (ablation) | Diff9D · REAL275 | 3D75 mAP: 42.7 -> 44.1 | Improved |
| use self-attention instead of cross-attention in denoiser (ablation) | Diff9D · REAL275 | 3D75 mAP: 16.7 -> 44.1 | Improved |
| add skip connections in denoiser (ablation) | Diff9D · REAL275 | 3D75 mAP: 41.2 -> 44.1 | Improved |
| use 7 transformer blocks instead of 3 (ablation) | Diff9D · REAL275 | 5deg2cm mAP: 32.2 -> 36.5 | Improved |
| increase transformer blocks beyond 7 (9, 11) (ablation) | Diff9D · REAL275 | 5deg2cm mAP: 36.5 -> 36.5 | No clear effect |
| use 3 DDIM sampling steps instead of 1 (ablation) | Diff9D · REAL275 | 3D75 mAP: 0.0 -> 44.1 | Improved |
| increase sampling steps from 3 to 1000 (ablation) | Diff9D · REAL275 | 3D75 mAP: 44.1 -> 44.3 | No clear effect |
Later papers that compared with this method
2 later accepted papers report Diff9D as a baseline in their results tables; 2 of them report a better number on most of the results they share with it.
- GenCOPE: Syn2Real Generalized Category-Level Object Pose Estimation for Robotic Picking (NeurIPS 2026 · better on 13 of 14 shared results)
- G2FM: A Geodesic Flow Matching Framework with Geometric Prior for Category-Level 9-DoF Pose Estimation (ECCV 2026 · better on 9 of 12 shared results)
Earlier methods this paper compares with
Named baselines from earlier accepted papers in this paper's results tables: better than 1 of 1 on most shared results.
- SAR-Net — SAR-Net: Shape Alignment and Recovery Network for Category-Level 6D Object Pose and Size Estimation (CVPR 2022) · better on 1 of 1 shared result
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
- Object Pose Estimation via the Aggregation of Diffusion Features (CVPR 2024 · similarity 0.90)
- GenCOPE: Syn2Real Generalized Category-Level Object Pose Estimation for Robotic Picking (NeurIPS 2026 · similarity 0.90)
- Vision Foundation Model Enables Generalizable Object Pose Estimation (NeurIPS 2024 · similarity 0.88)
- 6D-Diff: A Keypoint Diffusion Framework for 6D Object Pose Estimation (CVPR 2024 · similarity 0.88)
- $\text{Di}^2\text{Pose}$: Discrete Diffusion Model for Occluded 3D Human Pose Estimation (NeurIPS 2024 · similarity 0.87)
- 3D-Aware Hypothesis & Verification for Generalizable Relative Object Pose Estimation (ICLR 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.
