Research map / Detection and segmentation / CNN features
Robust Object Detection with Inaccurate Bounding Boxes
Chengxin Liu, Kewei Wang, Hao Lu, Zhiguo Cao, Ziming Zhang
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 |
|---|---|---|---|
| train with the object bag instance selection loss (ablation) | FasterRCNN · VOC 2007 (40% box noise) | mAP@0.5: 42.5 -> 56.0 | Improved |
| train with the object bag instance selection loss (ablation) | FasterRCNN · VOC 2007 (30% box noise) | mAP@0.5: 60.1 -> 66.9 | Improved |
| train with the object bag instance selection loss (ablation) | FasterRCNN · COCO (40% box noise) | mAP@0.5: 28.9 -> 32.6 | Improved |
| add object-aware instance selection (OA-IS) (ablation) | OA-MIL FasterRCNN · VOC 2007 (40% box noise) | mAP@0.5: 56.0 -> 63.3 | Improved |
| add object-aware instance selection (OA-IS) (ablation) | OA-MIL FasterRCNN · COCO (40% box noise) | mAP@0.5: 32.6 -> 39.8 | Improved |
| add object-aware instance extension (OA-IE) (ablation) | OA-MIL FasterRCNN · COCO (40% box noise) | mAP@0.5: 39.8 -> 42.6 | Improved |
| add object-aware instance extension (OA-IE) (ablation) | OA-MIL FasterRCNN · VOC 2007 (10% box noise) | mAP@0.5: 77.2 -> 77.4 | No clear effect |
| apply OA-MIL to FasterRCNN | FasterRCNN · VOC 2007 (40% box noise) | mAP@0.5: 42.5 -> 63.8 | Improved |
| apply OA-MIL to RetinaNet | RetinaNet · VOC 2007 (40% box noise) | mAP@0.5: 45.0 -> 53.4 | Improved |
| apply OA-MIL to FasterRCNN trained on noisy GWHD | FasterRCNN · GWHD | Val ADA: 0.608 -> 0.639 | Improved |
| apply OA-MIL to FasterRCNN trained on clean GWHD | FasterRCNN · GWHD | Test ADA: 0.511 -> 0.530 | Improved |
| use Co-teaching or SD-LocNet instead of vanilla training | FasterRCNN · VOC 2007 (40% box noise) | mAP@0.5: 42.5 -> 43.7 | No clear effect |
Earlier methods this paper compares with
Named baselines from earlier accepted papers in this paper's results tables: better than 1 of 1 on most shared results.
- FreeAnchor — FreeAnchor: Learning to Match Anchors for Visual Object Detection (NeurIPS 2019) · better on 1 of 1 shared result
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
- Bounding Box Regression With Uncertainty for Accurate Object Detection (CVPR 2019 · similarity 0.90)
- We Don't Need No Bounding-Boxes: Training Object Class Detectors Using Only Human Verification (CVPR 2016 · similarity 0.89)
- Training Object Class Detectors With Click Supervision (CVPR 2017 · similarity 0.89)
- Acquisition of Localization Confidence for Accurate Object Detection (ECCV 2018 · similarity 0.89)
- ObjectBox: From Centers to Boxes for Anchor-Free Object Detection (ECCV 2022 · similarity 0.88)
- Bounding Box Stability against Feature Dropout Reflects Detector Generalization across Environments (ICLR 2024 · 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.
