Research map / Detection and segmentation / Open-set and few-shot detection
Semantic Part Detection via Matching: Learning to Generalize to Novel Viewpoints From Limited Training Data
Yutong Bai, Qing Liu, Lingxi Xie, Weichao Qiu, Yan Zheng, Alan L. Yuille
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 |
|---|---|---|---|
| add 3D CAD-model matching (Ours) instead of end-to-end Faster R-CNN | semantic part detection · VSP sedan, 16 training samples, L0 | mAP (%): 31.64 -> 42.25 | Improved |
| use matching approach with 64 training samples instead of Faster R-CNN | semantic part detection · VSP sedan, 64 training samples, L0 | mAP (%): 55.43 -> 45.39 | Hurt |
| use matching approach instead of Faster R-CNN under heaviest occlusion | semantic part detection · VSP sedan, 64 training samples, L3 | mAP (%): 13.75 -> 16.75 | Improved |
| use matching approach instead of DeepVoting under occlusion L2 | semantic part detection · VSP sedan, 16 training samples, L2 | mAP (%): 13.92 -> 23.87 | Improved |
| increase training images from 16 to 64 for Ours (ablation) | Ours · VSP sedan, L0 | mAP (%): 42.25 -> 45.39 | Improved |
| increase training images from 16 to 64 for Faster R-CNN (ablation) | Faster R-CNN · VSP sedan, L0 | mAP (%): 31.64 -> 55.43 | Improved |
| use matching approach instead of Faster R-CNN on unseen elevation viewpoints | semantic part detection · VSP sedan, unseen elevation 20 deg, 16 samples | mAP (%): 16.02 -> 45.32 | Improved |
| use matching approach instead of DeepVoting on unseen viewpoints | semantic part detection · VSP sedan, unseen elevation 20 deg, 64 samples | mAP (%): 33.82 -> 45.88 | Improved |
| transfer sedan-trained model to minivan (prototype shift) for Ours (ablation) | Ours · VSP, 32 training samples | mAP (%): 47.27 -> 35.58 | Hurt |
| transfer sedan-trained model to SUV for Ours (ablation) | Ours · VSP, 32 training samples | mAP (%): 47.27 -> 42.80 | Hurt |
| use matching approach instead of Faster R-CNN on bicycle | semantic part detection · VSP bicycle | accuracy (%): 44.02 -> 67.16 | Improved |
| use matching approach instead of DeepVoting on motorbike | semantic part detection · VSP motorbike | accuracy (%): 31.88 -> 42.81 | Improved |
Earlier methods this paper compares with
Named baselines from earlier accepted papers in this paper's results tables: better than 2 of 2 on most shared results.
- DeepVoting — DeepVoting: A Robust and Explainable Deep Network for Semantic Part Detection Under Partial Occlusion (CVPR 2018) · better on 11 of 12 shared results
- Faster R-CNN — Faster R-CNN: Towards Real-Time Object Detection with Region Proposal Networks (TPAMI 2016) · better on 11 of 12 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
- Semantic Correspondence: Unified Benchmarking and a Strong Baseline (TPAMI 2025 · similarity 0.89)
- Bridging Viewpoint Gaps: Geometric Reasoning Boosts Semantic Correspondence (CVPR 2025 · similarity 0.88)
- SemAlign3D: Semantic Correspondence between RGB-Images through Aligning 3D Object-Class Representations (CVPR 2025 · similarity 0.87)
- CADTalk: An Algorithm and Benchmark for Semantic Commenting of CAD Programs (CVPR 2024 · similarity 0.87)
- 3DCoMPaT200: Language Grounded Large-Scale 3D Vision Dataset for Compositional Recognition (NeurIPS 2024 · similarity 0.87)
- MESC-3D:Mining Effective Semantic Cues for 3D Reconstruction from a Single Image (CVPR 2025 · similarity 0.87)
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.
