Research map / 3D vision / Human pose estimation
Exploiting View-Specific Appearance Similarities Across Classes for Zero-Shot Pose Prediction: A Metric Learning Approach
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 |
|---|---|---|---|
| learn a metric (J-VC) instead of plain kNN in feature space (ablation) | kNN pose/class prediction · 3DObjects | total pose accuracy: 71.68 -> 80.84 | Helped |
| add shared global metric plus class-specific metrics (MMJ-VC) over single joint metric (J-VC) (ablation) | metric learning · 3DObjects | total pose accuracy: 80.84 -> 82.00 | Helped |
| MMJ-VC vs J-VC (ablation) | metric learning · PASCAL3D+ supervised | Acc pi/6: 51.31 -> 51.84 | Helped |
| learn separate pose metrics per class without joint multi-task training (MM-VC) vs MMJ-VC (ablation) | MMJ-VC · PASCAL3D+ supervised | Acc pi/6: 48.55 -> 51.84 | Helped |
| joint multi-task learning (MMJ-VC) instead of separate metrics (MM-VC) (ablation) | zero-shot pose · 3DObjects | Acc(phi,theta) mean: 20.7 -> 22.6 | Helped |
| MMJ-VC vs plain kNN for zero-shot pose (ablation) | zero-shot pose · PASCAL3D+ | Acc pi/6 mean: 38.34 -> 42.33 | Helped |
| separate per-class metrics without joint training (MM-VC) vs plain kNN (ablation) | zero-shot pose · 3DObjects | Acc(phi,theta) mean: 20.7 -> 20.7 | Hurt |
| rescore R-CNN proposals with metric-model confidence (ablation) | MMJ-VC + R-CNN · 3DObjects bicycle | AP/AVP: 98.57/72.09 -> 99.06/83.61 | Helped |
| rescore R-CNN proposals with metric-model confidence (ablation) | MMJ-VC + R-CNN · 3DObjects car | AP/AVP: 99.07/86.61 -> 99.06/88.88 | Mixed |
| MMJ-VC + R-CNN vs prior detectors with viewpoint | MMJ-VC · PASCAL3D+ 4 views mean | AP/AVP: 26.8/19.5 -> 44.7/33.4 | Helped |
Closest papers by meaning
- Unsupervised Learning of Category-Level 3D Pose from Object-Centric Videos (CVPR 2024 · similarity 0.87)
- NOPE: Novel Object Pose Estimation from a Single Image (CVPR 2024 · similarity 0.87)
- Universal Features Guided Zero-Shot Category-Level Object Pose Estimation (AAAI 2025 · similarity 0.85)
- UNOPose: Unseen Object Pose Estimation with an Unposed RGB-D Reference Image (CVPR 2025 · similarity 0.85)
- Orient Anything: Learning Robust Object Orientation Estimation from Rendering 3D Models (ICML 2025 · similarity 0.85)
- GIVEPose: Gradual Intra-class Variation Elimination for RGB-based Category-Level Object Pose Estimation (CVPR 2025 · 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.