Research map / 3D vision / Human pose estimation
SMOC-Net: Leveraging Camera Pose for Self-Supervised Monocular Object Pose Estimation
Read the paperCheck 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 |
|---|---|---|---|
| Use SMOC-Net with camera-pose-derived relative-pose constraints and the camera-pose-guided refiner | GDR-Net-based self-supervised pose estimation · LineMOD | Average Recall of ADD(-S) (%): 91.3 vs 89.9 | Helped |
| Use SMOC-Net instead of Self6D++ trained without depth images | Self-supervised pose estimation · Occluded-LineMOD | Average Recall of ADD(-S) (%): 59.8 vs 63.3 | Helped |
| Add depth images during training | Self6D++ · Occluded-LineMOD | Average Recall of ADD(-S) (%): 59.8 -> 64.7 | Helped |
| Add the rotation loss Lrot (ablation) | SMOC-Net · LineMOD | Average Recall of ADD(-S) (%): 90.0 -> 91.3 | Helped |
| Add the translation loss Ltrans (ablation) | SMOC-Net · LineMOD | Average Recall of ADD(-S) (%): 90.8 -> 91.3 | Helped |
| Use the camera-pose-guided refiner instead of DeepIM (ablation) | SMOC-Net · LineMOD | Average Recall of ADD(-S) (%): 90.4 -> 91.3 | Helped |
| Add unannotated real-image training to synthetic-data training (ablation) | SMOC-Net · LineMOD | Average Recall of ADD(-S) (%): 77.5 -> 91.3 | Helped |
| Use SMOC-Net's training procedure instead of Self6D++ | Self-supervised pose estimation · LineMOD | training time per epoch (s): 37 -> 8 | Helped |
Closest papers by meaning
- NOPE: Novel Object Pose Estimation from a Single Image (CVPR 2024 · similarity 0.88)
- Co-op: Correspondence-based Novel Object Pose Estimation (CVPR 2025 · similarity 0.87)
- MRC-Net: 6-DoF Pose Estimation with MultiScale Residual Correlation (CVPR 2024 · similarity 0.86)
- Reloc3r: Large-Scale Training of Relative Camera Pose Regression for Generalizable, Fast, and Accurate Visual Localization (CVPR 2025 · similarity 0.85)
- One2Any: One-Reference 6D Pose Estimation for Any Object (CVPR 2025 · similarity 0.85)
- Emergence of a Shared Canonical Object Frame from In-the-Wild Videos (NeurIPS 2026 · similarity 0.85)
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.