Research map / Detection and segmentation / Weakly supervised detection and segmentation
We Don't Need No Bounding-Boxes: Training Object Class Detectors Using Only Human Verification
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 Yes/No human verification to MIL-based training | MIL WSOL object detector · PASCAL VOC 2007 | mAP: 27% -> 45% | Helped |
| Add Yes/No human verification to MIL-based localization | MIL WSOL · PASCAL VOC 2007 trainval | CorLoc: 43% -> 83% | Helped |
| Train only on positively verified detections (ablation) | MIL-style detector training · PASCAL VOC 2007 trainval | CorLoc: 43% -> 82% | Helped |
| Remove each negatively verified proposal from the search space (ablation) | Retraining with verified detections · PASCAL VOC 2007 trainval | CorLoc: 82% -> 92% | Helped |
| Remove proposals overlapping negatively verified detections (ablation) | Retrain + remove Neg · PASCAL VOC 2007 trainval | CorLoc: 92% -> 95% | Helped |
| Use Part/Container/Mixed/Missed labels for targeted proposal removal (ablation) | Retrain + remove ExtNeg · PASCAL VOC 2007 trainval | CorLoc: 95% -> 96% | Helped |
| Replace Yes/No human verification with five-way YPCMM verification (ablation) | Human-verified detector training · PASCAL VOC 2007 human annotation | annotation time: 5.8 hours -> 7.7 hours | Hurt |
| Replace Yes/No human verification with five-way YPCMM verification (ablation) | Human-verified detector training · PASCAL VOC 2007 | CorLoc: 83% -> 81% | Mixed |
| Replace AlexNet with VGG16 (ablation) | Yes/No human verification · PASCAL VOC 2007 test, complete training set | mAP: 50% -> 58% | Helped |
| Use the complete training set instead of the reduced training set (ablation) | Yes/No human verification with AlexNet · PASCAL VOC 2007 test | mAP: 45% -> 50% | Helped |
| Replace verification training with full supervision | AlexNet detector, reduced PASCAL VOC 2007 training set · PASCAL VOC 2007 test | mAP: 45% -> 51% | Helped |
| Use verification instead of full supervision with the same annotation time | PASCAL VOC 2007 detector training · PASCAL VOC 2007 test | mAP: 33% -> 45%; 30% -> 45% | Helped |
Closest papers by meaning
- Training Object Class Detectors With Click Supervision (CVPR 2017 · similarity 0.93)
- Open-Vocabulary Object Detection Using Captions (CVPR 2021 · similarity 0.90)
- Track and Transfer: Watching Videos to Simulate Strong Human Supervision for Weakly-Supervised Object Detection (CVPR 2016 · similarity 0.89)
- Bounding Box Regression With Uncertainty for Accurate Object Detection (CVPR 2019 · similarity 0.88)
- Object Detection Benchmarks are Incomplete: The Role of Label Errors and Annotation Uncertainty (NeurIPS 2026 · similarity 0.86)
- OpenBox: Annotate Any Bounding Boxes in 3D (NeurIPS 2025 · similarity 0.86)
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.