Research map / 3D vision / Human pose estimation
Accurate and Efficient Object Pose Estimation via the Aggregation of Diffusion Features
Tianfu Wang, Guosheng Hu, H. G. Wang
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 |
|---|---|---|---|
| replace ResNet template-pose features with aggregated Stable Diffusion features (CWA) | Template-pose · Occluded-LINEMOD (unseen) | accuracy (%): 77.9 -> 85.9 | Improved |
| use diffusion features (CWA) instead of Template-pose | Template-pose · LINEMOD (unseen) | accuracy (%): 94.4 -> 97.9 | Improved |
| replace vanilla aggregation with nonlinear aggregation (ablation) | Ours (VA) · Occluded-LINEMOD (unseen) | accuracy (%): 84.6 -> 85.5 | Improved |
| replace nonlinear aggregation with context-aware weight aggregation (ablation) | Ours (NA) · Occluded-LINEMOD (unseen) | accuracy (%): 85.5 -> 85.9 | Improved |
| replace CWA with confidence adaptive aggregation (early exit) (ablation) | Ours (CWA) · LINEMOD | MACs / FPS: 929B -> 678B MACs; 9.2 -> 15.5 FPS | Improved |
| replace CWA with confidence adaptive aggregation | Ours (CWA) · LINEMOD (unseen) | accuracy (%): 97.9 -> 97.7 | No clear effect |
| lower input resolution from 512 to 128 with CAA | Ours (CAA) · T-LESS | Recall VSD: 70.82 -> 69.52 | No clear effect |
| replace Stable Diffusion v1-5 with DINOv2 features (ablation) | Ours (CWA) · Occluded-LINEMOD (unseen) | accuracy (%): 85.9 -> 78.1 | Hurt |
| replace Stable Diffusion v1-5 with OpenCLIP features (ablation) | Ours (CWA) · Occluded-LINEMOD (unseen) | accuracy (%): 85.9 -> 81.1 | Hurt |
| replace SD-V1-5 with SD-V2-0 (ablation) | Ours (CWA) · LINEMOD (unseen) | accuracy (%): 97.9 -> 98.4 | No clear effect |
| raise confidence threshold from 0.65 to 0.7 (ablation) | Ours (CAA) · Occluded-LINEMOD (unseen) | accuracy (%): 83.0 -> 85.5 | Improved |
| raise confidence threshold from 0.7 to 0.75 (ablation) | Ours (CAA) · Occluded-LINEMOD (unseen) | accuracy / MACs: 85.5 -> 85.6; 719B -> 863B MACs | No clear effect |
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.
- GigaPose — GigaPose: Fast and Robust Novel Object Pose Estimation via One Correspondence (CVPR 2024) · better on 2 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
- Object Pose Estimation via the Aggregation of Diffusion Features (CVPR 2024 · similarity 0.97)
- Pose without Guesses: Generalizable Object Pose Estimation From a Single Reference (IJCV 2026 · similarity 0.90)
- 3D-Aware Hypothesis & Verification for Generalizable Relative Object Pose Estimation (ICLR 2024 · similarity 0.90)
- Confronting Ambiguity in 6D Object Pose Estimation via Score-Based Diffusion on SE(3) (CVPR 2024 · similarity 0.88)
- DiffusionRegPose: Enhancing Multi-Person Pose Estimation using a Diffusion-Based End-to-End Regression Approach (CVPR 2024 · similarity 0.87)
- Cameras as Rays: Pose Estimation via Ray Diffusion (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.
