Research map / 3D vision / Human pose estimation
Deep Model-Based 6D Pose Refinement in RGB
Fabian Manhardt, Wadim Kehl, Nassir Navab, Federico Tombari
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 the learned refinement to SSD-6D poses | SSD-6D · Hinterstoisser | VSS: 0.66 -> 0.83 | Improved |
| use learned refinement instead of 2D edge-based ICP | refinement · Hinterstoisser | VSS: 0.71 -> 0.83 | Improved |
| use RGB learned refinement instead of 3D cloud-based ICP | refinement · Hinterstoisser | VSS: 0.84 -> 0.83 | No clear effect |
| add refinement on SSD-6D poses | SSD-6D · Hinterstoisser | rotation error (deg): 27.96 -> 16.17 | Improved |
| add refinement on SSD-6D poses | SSD-6D · Hinterstoisser | ADD (%): 7.4 -> 34.1 | Improved |
| use RGB refinement instead of 3D ICP with depth | refinement · Hinterstoisser | ADD (%): 90.9 -> 34.1 | Hurt |
| add refinement on SSD-6D poses | SSD-6D · Occlusion | ADD (%): 6.2 -> 27.5 | Improved |
| add refinement on SSD-6D poses | SSD-6D · Occlusion | rotation error (deg): 34.42 -> 24.36 | Improved |
| replace MSE regression loss with the visual (contour overlap) loss (ablation) | refinement network · Tejani | VSS: 0.676 -> 0.817 | Improved |
| use RGB-only method versus RGB tracker of Tjaden | refinement · Tejani | VSS: 0.304 -> 0.817 | Improved |
| use depth-free method versus RGB-D tracker of Kehl | refinement · Tejani | VSS: 0.627 -> 0.817 | Improved |
| increase initial rotation perturbation to 45 degrees on a textured object (bvise) | refinement · Hinterstoisser perturbation | rotation error: 60% below 5 degrees, 80% below 10 degrees | Mixed |
Closest papers by meaning
- PR-GCN: A Deep Graph Convolutional Network With Point Refinement for 6D Pose Estimation (ICCV 2021 · similarity 0.88)
- CDPN: Coordinates-Based Disentangled Pose Network for Real-Time RGB-Based 6-DoF Object Pose Estimation (ICCV 2019 · similarity 0.88)
- RefPose: Leveraging Reference Geometric Correspondences for Accurate 6D Pose Estimation of Unseen Objects (CVPR 2025 · similarity 0.88)
- SingRef6D: Monocular Novel Object Pose Estimation with a Single RGB Reference (NeurIPS 2025 · similarity 0.88)
- Boosting 6D Object Pose Estimation via Monocular Depth Cues (ECCV 2026 · similarity 0.88)
- SCFlow2: Plug-and-Play Object Pose Refiner with Shape-Constraint Scene Flow (CVPR 2025 · similarity 0.86)
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.
