Research map / Detection and segmentation / Weakly supervised detection and segmentation
Training Object Class Detectors With Click Supervision
Check 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 |
|---|---|---|---|
| Add one center click per object to MIL | Reference MIL on PASCAL VOC 2007 · PASCAL VOC 2007 | CorLoc and mAP: +28.8% CorLoc and +16.3% mAP | Helped |
| Use two-click center estimates rather than one click (ablation) | One-click supervision on PASCAL VOC 2007 · PASCAL VOC 2007 | CorLoc: +0.8% | Helped |
| Use two-click center estimates rather than one click (ablation) | One-click supervision on PASCAL VOC 2007 · PASCAL VOC 2007 | mAP: +0.9% | Helped |
| Add a box-area estimate from the distance between two clicks (ablation) | Two-click center supervision on PASCAL VOC 2007 · PASCAL VOC 2007 | CorLoc: +5.2% | Helped |
| Add a box-area estimate from the distance between two clicks (ablation) | Two-click center supervision on PASCAL VOC 2007 · PASCAL VOC 2007 | mAP: +3.2% | Helped |
| Add simulated center-click supervision to MIL | Reference MIL on MS COCO · MS COCO | CorLoc and mAP: 24.2% -> 51.8% CorLoc; 8.9% -> 18.3% mAP | Helped |
| Add a qualification test (ablation) | Click annotation protocol · PASCAL VOC 2007 trainval annotation study | Error distance: 43.8 -> 29.4 | Helped |
| Use synthetic polygons instead of real images in the qualification test (ablation) | Click annotation protocol · PASCAL VOC 2007 trainval annotation study | Error distance: 29.4 -> 29.3 | No clear effect |
| Add quality control with hidden evaluation images (ablation) | Polygon qualification click annotation protocol · PASCAL VOC 2007 trainval annotation study | Error distance: 29.3 -> 21.2 | Helped |
| Use center-clicks rather than click-anywhere annotations | One-click MIL localization · PASCAL VOC 2007 trainval | CorLoc: 55.5% -> 67.2% | Helped |
| Use random clicks inside objects instead of center-clicks | One-click MIL localization · PASCAL VOC 2007 trainval | CorLoc: 43.4% | No clear effect |
| Use two simulated clicks and estimate box area | Reference MIL on MS COCO · MS COCO | CorLoc and mAP: 58.6% CorLoc; 19.3% mAP | Helped |
Closest papers by meaning
- We Don't Need No Bounding-Boxes: Training Object Class Detectors Using Only Human Verification (CVPR 2016 · similarity 0.93)
- Rethinking the Route Towards Weakly Supervised Object Localization (CVPR 2020 · similarity 0.90)
- Open-Vocabulary Object Detection Using Captions (CVPR 2021 · similarity 0.89)
- Bridging the Gap Between Classification and Localization for Weakly Supervised Object Localization (CVPR 2022 · similarity 0.89)
- Weakly Supervised Object Localization With Progressive Domain Adaptation (CVPR 2016 · similarity 0.88)
- Weakly Supervised Object Boundaries (CVPR 2016 · 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.