Research map / 3D vision / Human pose estimation
Templates for 3D Object Pose Estimation Revisited: Generalization to New Objects and Robustness to Occlusions
Van Nguyen Nguyen, Yinlin Hu, Yang Xiao, Mathieu Salzmann, Vincent Lepetit
Check 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 |
|---|---|---|---|
| use local rather than global image representations | Template pose estimation · Unseen Occlusion-LM, average across splits | Acc15: 29.9 -> 76.3 | Improved |
| use local rather than global image representations | Template pose estimation · Unseen LM, average across splits | Acc15: 45.1 -> 93.5 | Improved |
| add template mask M to discard background features (ablation) | Ours · Unseen Occlusion-LM | Acc15: 24.1 -> 76.3 | Improved |
| add occlusion mask O with threshold δ=0.2 (ablation) | Ours · Unseen O-LM average | Acc15: 74.5 -> 76.3 | Improved |
| use occlusion mask O (ablation) | Ours · Unseen O-LM | Acc15: Ape: 53.3 -> 53.8; Can: 84.9 -> 89.7; Cat: 45.1 -> 45.1; Driller: 81.5 -> 84.4; Duck: 87.3 -> 87.2; Eggbox: 72.6 -> 76.9 | Mixed |
| reduce template count from 92K to 21K (ablation) | Ours on T-LESS · T-LESS average | Recall VSD: 58.87 -> 58.25 | No clear effect |
| initialize ResNet50 with MOCOv2 features rather than training from scratch (ablation) | Ours on T-LESS · T-LESS average | Recall VSD: 53.81 -> 58.25 | Improved |
| set occlusion threshold δ to 0.2 rather than omit O (ablation) | Ours · Unseen O-LM average | Acc15: 74.5 -> 76.3 | Improved |
| use higher-resolution local feature maps and masks (ablation) | Ours · Unseen objects in LM and O-LM | pose error | Improved |
| increase local feature dimension (ablation) | Ours · LM and O-LM | pose error | No clear effect |
Closest papers by meaning
- GigaPose: Fast and Robust Novel Object Pose Estimation via One Correspondence (CVPR 2024 · similarity 0.89)
- Co-op: Correspondence-based Novel Object Pose Estimation (CVPR 2025 · similarity 0.89)
- 3D-Aware Hypothesis & Verification for Generalizable Relative Object Pose Estimation (ICLR 2024 · similarity 0.88)
- Pos3R: 6D Pose Estimation for Unseen Objects Made Easy (CVPR 2025 · similarity 0.88)
- UNOPose: Unseen Object Pose Estimation with an Unposed RGB-D Reference Image (CVPR 2025 · similarity 0.88)
- Unsupervised Learning of Category-Level 3D Pose from Object-Centric Videos (CVPR 2024 · similarity 0.88)
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.
