Research map / 3D vision / Human pose estimation
Single-Stage 6D Object Pose Estimation
Check your idea against 108,967 papers
AI-extracted summary. Written by AI from the paper's title only (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 |
|---|---|---|---|
| Replace PVNet's RANSAC-based fusion with the single-stage pose network | PVNet · Occluded-LINEMOD | ADD-0.1d: 40.8 -> 43.3 | Helped |
| Replace PVNet's RANSAC-based fusion with the single-stage pose network | PVNet · Occluded-LINEMOD | REP-5px: 61.1 -> 62.3 | Helped |
| Replace PVNet's RANSAC-based fusion with the single-stage pose network | PVNet · Occluded-LINEMOD | total time (ms): 40 -> 22 | Helped |
| Replace SegDriven's RANSAC-based post-processing with the pose network | SegDriven · Occluded-LINEMOD | ADD-0.1d: 27.0 -> 32.3 | Helped |
| Replace correspondence fusion with the single-stage pose network | SegDriven · YCB-Video | ADD-0.1d: 39.0 -> 53.9 | Helped |
| Replace correspondence fusion with the single-stage pose network | PVNet · YCB-Video | REP-5px: 47.4 -> 48.7 | Helped |
| Ignore fixed keypoint-cluster order by using one max-pooling operation over all correspondences (ablation) | Single-stage pose network · Synthetic data | pose accuracy | Hurt |
| Use the single-stage pose network instead of RANSAC-based PnP | Correspondence-based pose estimation · Synthetic data with increasing correspondence noise | pose accuracy | Mixed |
| Use the single-stage network rather than voting-based PnP | PVNet-style 3D-point-to-2D-vector correspondences · Synthetic data with increasing correspondence noise | pose accuracy | Mixed |
| Add instance normalization or batch normalization (ablation) | Grouped feature aggregation network · Pose estimation from correspondence clusters | performance | No clear effect |
| Use shared 3D keypoints across object types instead of object-specific keypoints (ablation) | Correspondence-extraction network · Real-data object pose estimation | accuracy | No clear effect |
| Represent correspondence coordinates as grid-cell centers plus offsets instead of direct image coordinates (ablation) | Correspondence-extraction network · Real-data object pose estimation | performance | Helped |
Closest papers by meaning
- One2Any: One-Reference 6D Pose Estimation for Any Object (CVPR 2025 · similarity 0.91)
- FoundationPose: Unified 6D Pose Estimation and Tracking of Novel Objects (CVPR 2024 · similarity 0.90)
- SingRef6D: Monocular Novel Object Pose Estimation with a Single RGB Reference (NeurIPS 2025 · similarity 0.89)
- Pos3R: 6D Pose Estimation for Unseen Objects Made Easy (CVPR 2025 · similarity 0.89)
- FAR: Flexible Accurate and Robust 6DoF Relative Camera Pose Estimation (CVPR 2024 · similarity 0.88)
- Any6D: Model-free 6D Pose Estimation of Novel Objects (CVPR 2025 · 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.