Research map / 3D vision / Human pose estimation
DeepIM: Deep Iterative Matching for 6D Pose Estimation
Yi Li, Gu Wang, Xiangyang Ji, Yu Xiang, Dieter Fox
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 |
|---|---|---|---|
| refine PoseCNN poses with DeepIM | PoseCNN · LINEMOD | 6D Pose accuracy (%): 62.7 -> 88.6 | Improved |
| refine PoseCNN poses with DeepIM | PoseCNN · Occlusion LINEMOD | 6D Pose accuracy (%): 26.9 -> 55.5 | Improved |
| refine PoseCNN poses with DeepIM (RGB) | PoseCNN · YCB-Video | AUC of ADD(-S): 61.31 -> 81.9 | Improved |
| use tracking from previous frame instead of PoseCNN init | DeepIM · YCB-Video | AUC of ADD(-S): 81.9 -> 85.9 | Improved |
| add iterative training (train 2 iterations vs 1), tested with 2 iterations (ablation) | DeepIM · LINEMOD | 6D Pose accuracy (%): 79.0 -> 88.7 | Improved |
| test with 4 iterations when trained without iteration (ablation) | DeepIM train_iter=1 · LINEMOD | 6D Pose accuracy (%): 77.9 -> 76.1 | Hurt |
| train/test with 4 iterations instead of 2 (ablation) | DeepIM · LINEMOD | 6D Pose accuracy (%): 88.7 -> 88.6 | No clear effect |
| add zoom-in on bounding boxes (ablation) | DeepIM · LINEMOD | 6D Pose accuracy (%): 65.2 -> 88.6 | Improved |
| use disentangled transformation instead of camera frame (ablation) | DeepIM (no zoom) · LINEMOD | 6D Pose accuracy (%): 44.3 -> 65.2 | Improved |
| use disentangled representation instead of model frame (ablation) | DeepIM · LINEMOD | 5cm 5deg accuracy (%): 79.2 -> 86.6 | Improved |
| use point matching loss instead of distance loss (ablation) | DeepIM · LINEMOD | 6D Pose accuracy (%): 79.2 -> 88.6 | Improved |
| train one shared network instead of separate per-object networks (ablation) | DeepIM · LINEMOD | 6D Pose accuracy (%): 87.6 -> 88.6 | Improved |
Closest papers by meaning
- Single-Stage 6D Object Pose Estimation (CVPR 2020 · similarity 0.92)
- DCL-Net: Deep Correspondence Learning Network for 6D Pose Estimation (ECCV 2022 · similarity 0.91)
- Coupled Iterative Refinement for 6D Multi-Object Pose Estimation (CVPR 2022 · similarity 0.90)
- Segmentation-Driven 6D Object Pose Estimation (CVPR 2019 · similarity 0.89)
- FoundationPose: Unified 6D Pose Estimation and Tracking of Novel Objects (CVPR 2024 · similarity 0.89)
- SingRef6D: Monocular Novel Object Pose Estimation with a Single RGB Reference (NeurIPS 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.
