Research map / 3D vision / Human pose estimation
AlignPose: Generalizable 6D Pose Estimation via Multi-view Feature-metric Alignment
Anna Šárová Mikeštíková, Médéric Fourmy, Martin Cifka, Josef Sivic, Vladimir Petrik
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 |
|---|---|---|---|
| aggregate single-view candidates across 4 views without NMS (ablation) | Co-op candidates · YCB-V | AP: 69.5 -> 33.9 | Hurt |
| add 3D NMS after aggregation (ablation) | AlignPose · YCB-V | AP: 33.9 -> 62.3 | Improved |
| add 3D NMS after aggregation (ablation) | AlignPose · T-LESS | AP: 48.6 -> 80.2 | Improved |
| add multi-view feature-metric refinement after NMS (ablation) | AlignPose · YCB-V | AP: 62.3 -> 83.3 | Improved |
| add multi-view feature-metric refinement after NMS (ablation) | AlignPose · T-LESS | AP: 80.2 -> 92.4 | Improved |
| render templates online rather than retrieve offline (ablation) | AlignPose · YCB-V | AP: 81.7 -> 83.3 | Improved |
| render templates online rather than retrieve offline (ablation) | AlignPose · T-LESS | AP: 89.2 -> 92.4 | Improved |
| score by average feature-metric loss across views instead of max (ablation) | AlignPose · YCB-V | AP: 81.6 -> 83.3 | Improved |
| use DINOv2-L foundation features instead of dense SIFT (ablation) | AlignPose · YCB-V | AP: 70.7 -> 83.3 | Improved |
| swap DINOv2-L layer 18 for DINOv3-L layer 18 (ablation) | AlignPose · YCB-V | AP: 83.3 -> 84.1 | No clear effect |
| replace CosyPose multi-view with AlignPose refinement | FoundPose candidates · HouseCat6D (metallic) | AP: 27.1 -> 46.5 | Improved |
| use 3PT detections instead of FOSOD detections (ablation) | AlignPose · IPD | AP: 61.2 -> 79.8 | Improved |
Closest papers by meaning
- UNOPose: Unseen Object Pose Estimation with an Unposed RGB-D Reference Image (CVPR 2025 · similarity 0.92)
- Any6D: Model-free 6D Pose Estimation of Novel Objects (CVPR 2025 · similarity 0.91)
- Co-op: Correspondence-based Novel Object Pose Estimation (CVPR 2025 · similarity 0.90)
- RefPose: Leveraging Reference Geometric Correspondences for Accurate 6D Pose Estimation of Unseen Objects (CVPR 2025 · similarity 0.89)
- Pos3R: 6D Pose Estimation for Unseen Objects Made Easy (CVPR 2025 · similarity 0.89)
- SingRef6D: Monocular Novel Object Pose Estimation with a Single RGB Reference (NeurIPS 2025 · 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.
