Research map / 3D vision / Human pose estimation
Towards Human-Level 3D Relative Pose Estimation: Generalizable, Training-Free, With Single Reference
Yuan Gao, Yajing Luo, Junhong Wang, Kui Jia, Gui-Song Xia
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 back-surface culling (ablation) | render-and-compare pose refinement · LineMOD | Mean Err: 38.09 -> 29.93 | Improved |
| add semantic maps to RGB (ablation) | render-and-compare pose refinement · LineMOD | Acc@5: 19.62 -> 24.32 | Improved |
| add RGB maps to semantic maps (ablation) | render-and-compare pose refinement · LineMOD | Acc@5: 19.22 -> 24.32 | Improved |
| use DINOv2 instead of LoFTR features (ablation) | render-and-compare pose refinement · LineMOD | Acc@5: 5.24 -> 19.22 | Improved |
| use DINOv2 instead of Stable Diffusion features (ablation) | render-and-compare pose refinement · LineMOD | Acc@5: 12.12 -> 19.22 | Improved |
| add LoFTR features to RGB (ablation) | render-and-compare pose refinement · LineMOD | Acc@5: 19.62 -> 14.80 | Hurt |
| add SD features to RGB (ablation) | render-and-compare pose refinement · LineMOD | Mean Err: 36.26 -> 33.78 | Improved |
| use uniform instead of random initialization (ablation) | render-and-compare pose refinement · LineMOD | Acc@5: 23.08 -> 24.32 | Improved |
| add differentiable-renderer refinement after initialization (ablation) | Ours · LineMOD | Acc@5: 6.66 -> 24.32 | Improved |
| increase uniform initialization samples beyond 16000 (ablation) | Ours · LineMOD | Acc@10: 4000 to 16000 marginal; saturates after 16000 | No clear effect |
| increase refinement iterations beyond 120 (ablation) | Ours · LineMOD | Acc@10: 30 to 120 marginal; saturates after 120 | No clear effect |
| training-free method versus supervised DVMNet (in-dataset) | Ours · LineMOD | Acc@5: 2.72 -> 24.32 | Improved |
Closest papers by meaning
- 3D-Aware Hypothesis & Verification for Generalizable Relative Object Pose Estimation (ICLR 2024 · similarity 0.91)
- UNOPose: Unseen Object Pose Estimation with an Unposed RGB-D Reference Image (CVPR 2025 · similarity 0.90)
- DVMNet: Computing Relative Pose for Unseen Objects Beyond Hypotheses (CVPR 2024 · similarity 0.88)
- Structure-Aware Correspondence Learning for Relative Pose Estimation (CVPR 2025 · similarity 0.88)
- SchemaPose: RGB-Based Category-Level Pose Estimation with Parametric Category Schema (NeurIPS 2026 · similarity 0.88)
- ImageNet3D: Towards General-Purpose Object-Level 3D Understanding (NeurIPS 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.
