Research map / Detection and segmentation / Open-set and few-shot detection
Learning Semantic Part-Based Models from Google Images
Davide Modolo, Vittorio Ferrari
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 polishing of web part instances (A0 to A1) (ablation) | part appearance model · PASCAL-Part val | mAP: 12.9 -> 15.7 | Improved |
| add viewpoint-specific part location model to A1 (ablation) | A1 · PASCAL-Part val | mAP: 15.7 -> 18.3 | Improved |
| mine more part instances using object-part association and update appearance (A2) (ablation) | A1 · PASCAL-Part val | mAP: 18.3 -> 22.0 | Improved |
| combine A2 with location model L2 (ablation) | A2 · PASCAL-Part val | mAP: 22.0 -> 23.5 | Improved |
| adapt appearance to PASCAL VOC domain (A3) (ablation) | A2 + L2 · PASCAL-Part val | mAP: 23.5 -> 25.4 | Improved |
| adapt location model to VOC domain (L3 instead of L2) (ablation) | A3 · PASCAL-Part val | mAP: 26.8 -> 27.2 | Improved |
| learn from web part images with all stages instead of naive part-name queries (ablation) | part detector · PASCAL-Part val | mAP: 12.9 to 27.2 AP | Improved |
| train viewpoint classifier on Google Images instead of VOC manual annotations | viewpoint classifier · PASCAL VOC val | mean AP: 49.3 -> 53.7 | Improved |
| add web part models to R-CNN (VOC 2010) | R-CNN · PASCAL VOC 2010 val | mean AP: 57.4 -> 59.6 | Improved |
| add web part models to R-CNN (VOC 2012) | R-CNN · PASCAL VOC 2012 val | mean AP: 56.5 -> 58.5 | Improved |
| add web object images to R-CNN training data | R-CNN · PASCAL VOC 2010 val | mean AP: 57.4 -> 56.6 | Hurt |
| add web part instances to fully supervised part detector | A_FS1 · PASCAL-Part val | mAP: 43.9 -> 41.5 | Hurt |
Closest papers by meaning
- Beyond Holistic Object Recognition: Enriching Image Understanding With Part States (CVPR 2018 · similarity 0.91)
- Objects as Context for Detecting Their Semantic Parts (CVPR 2018 · similarity 0.90)
- SPDA-CNN: Unifying Semantic Part Detection and Abstraction for Fine-Grained Recognition (CVPR 2016 · similarity 0.90)
- "PartImageNet: A Large, High-Quality Dataset of Parts" (ECCV 2022 · similarity 0.89)
- Semantic Part Detection via Matching: Learning to Generalize to Novel Viewpoints From Limited Training Data (ICCV 2019 · similarity 0.88)
- Multi-Part Object Representations via Graph Structures and Co-Part Discovery (NeurIPS 2026 · 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.
