Research map / 3D vision / Human pose estimation
Shape-Constraint Recurrent Flow for 6D Object Pose Estimation
Yang Hai, Rui Song, Jiaojiao Li, Yinlin Hu
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 |
|---|---|---|---|
| constrain correlation lookup using pose-induced flow from the target's 3D shape (ablation) | recurrent pose refinement with RANSAC-PnP · YCB-V | ADD-0.1d: 61.9 -> 67.6 | Improved |
| replace RANSAC-PnP with a learned pose regressor (ablation) | recurrent pose refinement without shape-constraint lookup · YCB-V | ADD-0.1d: 61.9 -> 63.6 | Improved |
| add a learned pose regressor while retaining shape-constraint lookup (ablation) | recurrent pose refinement · YCB-V | ADD-0.1d: 67.6 -> 70.5 | Improved |
| use more recurrent inference iterations (ablation) | SCFlow · YCB-V | ADD-0.1d: over 11.5% | Improved |
| train with mixed real and PBR data rather than real images alone (ablation) | SCFlow · LINEMOD | ADD-0.1d: 99.3 -> 99.3 | No clear effect |
| train with mixed real and PBR data rather than real images alone (ablation) | SCFlow · LINEMOD-Occluded | ADD-0.1d: 66.4 -> 67.0 | Improved |
| train with mixed real and PBR data rather than real images alone (ablation) | SCFlow · YCB-V | ADD-0.1d: 70.5 -> 73.2 | Improved |
| use shape-constraint recurrent refinement instead of PFA | PoseCNN-initialized pose refinement · YCB-V | ADD-0.1d: 61.9 -> 70.5 | Improved |
| remove reliance on RANSAC-PnP through learned pose regression | pose refinement · YCB-V | refinement time (ms): 37 -> 17 | 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.
- RNNPose — RNNPose: Recurrent 6-DoF Object Pose Refinement With Robust Correspondence Field Estimation and Pose Optimization (CVPR 2022) · better on 2 of 2 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
- SCFlow2: Plug-and-Play Object Pose Refiner with Shape-Constraint Scene Flow (CVPR 2025 · similarity 0.91)
- GenFlow: Generalizable Recurrent Flow for 6D Pose Refinement of Novel Objects (CVPR 2024 · similarity 0.90)
- Pseudo Flow Consistency for Self-Supervised 6D Object Pose Estimation (ICCV 2023 · similarity 0.89)
- Rectified Point Flow: Generic Point Cloud Pose Estimation (NeurIPS 2025 · similarity 0.87)
- RefPose: Leveraging Reference Geometric Correspondences for Accurate 6D Pose Estimation of Unseen Objects (CVPR 2025 · similarity 0.86)
- SingRef6D: Monocular Novel Object Pose Estimation with a Single RGB Reference (NeurIPS 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.
