Research map / 3D vision / Human pose estimation
Fusing Local Similarities for Retrieval-Based 3D Orientation Estimation of Unseen Objects
Chen Zhao, Yinlin Hu, Mathieu Salzmann
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 adaptive fusion instead of simple averaging of local similarities (ablation) | Ours · LineMOD unseen Split #1 | Rota. Acc. (%): 8.7% increase in Rota. Acc. over Avg | Improved |
| combine sigmoid and softmax in the fusion weights instead of sigmoid only (ablation) | Ours · LineMOD unseen Split #1 | Rota. Acc. (%): 88.57 -> 89.55 | Improved |
| combine sigmoid and softmax in the fusion weights instead of softmax only (ablation) | Ours · LineMOD unseen Split #1 | Rota. Acc. (%): 89.03 -> 89.55 | Improved |
| add local normalization of input images (ablation) | Ours · LineMOD unseen Split #1 | Rota. Acc. (%): 85.91 -> 89.25 | Improved |
| add weighted infoNCE loss instead of standard infoNCE (ablation) | Ours · LineMOD unseen Split #1 | Rota. Acc. (%): 85.91 -> 87.14 | Improved |
| add weighted infoNCE on top of local normalization (ablation) | Ours · LineMOD unseen Split #1 | Rota. Acc. (%): 89.25 -> 89.55 | Improved |
| use multi-scale local feature maps instead of a global descriptor from the smallest map (ablation) | Ours · LineMOD unseen Split #1 | Rota. Acc. (%) | Improved |
| use fast retrieval instead of exhaustive greedy search (ablation) | Ours · LineMOD unseen Split #1 | Rota. Acc. (%): 95.93 -> 89.55 Rota. Acc.; 30.74s -> 0.42s | Mixed |
| increase the number of reference images (ablation) | Ours · LineMOD unseen Split #1 | Rota. Acc. (%) | Improved |
| patch-level retrieval vs global-descriptor retrieval (LD, NetVLAD, MPE) on unseen objects | Ours · LineMOD unseen objects (mean) | Rota. Acc. (%): 47.35 -> 81.52 (best baseline MPE vs Ours) | Improved |
| patch-level retrieval vs HOG on T-LESS | Ours · T-LESS | Rota. Acc. (%): 74.22 -> 78.73 | Improved |
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.
- GDR-Net — GDR-Net: Geometry-Guided Direct Regression Network for Monocular 6D Object Pose Estimation (CVPR 2021) · better on 3 of 3 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
- Orient Anything: Learning Robust Object Orientation Estimation from Rendering 3D Models (ICML 2025 · similarity 0.87)
- 3D-Aware Hypothesis & Verification for Generalizable Relative Object Pose Estimation (ICLR 2024 · similarity 0.87)
- UNOPose: Unseen Object Pose Estimation with an Unposed RGB-D Reference Image (CVPR 2025 · similarity 0.87)
- Tango3D: Towards Alignment for Global and Local 2D-3D Correspondence (NeurIPS 2026 · similarity 0.86)
- ImageNet3D: Towards General-Purpose Object-Level 3D Understanding (NeurIPS 2024 · similarity 0.85)
- RefPose: Leveraging Reference Geometric Correspondences for Accurate 6D Pose Estimation of Unseen Objects (CVPR 2025 · similarity 0.85)
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.
