Research map / Detection and segmentation / Open-set and few-shot detection
Robust and Accurate Object Detection via Adversarial Learning
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 |
|---|---|---|---|
| Apply Det-AdvProp during fine-tuning | EfficientDet detectors · COCO val | mAP: 0.3--1.1 mAP | Helped |
| Apply Det-AdvProp during fine-tuning | EfficientDet detectors · COCO-C | mAP: 0.8--3.8 mAP | Helped |
| Apply Det-AdvProp during fine-tuning | EfficientDet detectors · PASCAL VOC 2012 | mAP: 0.2--1.3 mAP | Helped |
| Replace targeted with non-targeted attacks (ablation) | Det-AdvProp on EfficientDet-D0-D2 · COCO val | mAP: 34.3 -> 34.0; 40.2 -> 40.1; 43.5 -> 43.4 | Hurt |
| Replace targeted with non-targeted attacks (ablation) | Det-AdvProp on EfficientDet-D3-D5 · COCO-C | mAP: 30.1 -> 30.8; 31.8 -> 32.8; 32.4 -> 33.7 | Helped |
| Increase targeted attack strength from ε=1 to ε=2 (ablation) | Det-AdvProp on EfficientDet-D0-D2 · COCO val | mAP: 34.7 -> 34.2; 40.5 -> 40.0; 43.8 -> 43.5 | Hurt |
| Increase targeted attack strength from ε=1 to ε=2 (ablation) | Det-AdvProp on EfficientDet-D5 · COCO val | mAP: 51.5 -> 51.7 | Helped |
| Generate adversarial examples from classification and localization separately, then retain the stronger example using one auxiliary batchnorm (ablation) | Det-AdvProp · COCO val | mAP: 51.3 -> 51.8 | Helped |
| Use three batchnorms and retain separate classification- and localization-sourced examples (ablation) | Det-AdvProp on EfficientDet-D5 · COCO val | mAP: 51.8 -> 51.3 | Hurt |
| Use three batchnorms and retain separate classification- and localization-sourced examples (ablation) | Det-AdvProp on EfficientDet-D5 · COCO-C | relative rPC (%): 65.1 (+3.9) -> 65.3 (+4.1) | Helped |
| Use adversarial training instead of Det-AdvProp | EfficientDet-D1 and D3 · COCO val | mAP: 40.5 -> 38.0; 47.6 -> 45.3 | Hurt |
| Combine Det-AdvProp with AutoAugment | EfficientDet detectors · COCO-C | mAP: +1.3; +2.3; +2.2; +3.2; +3.8; +3.6 | Helped |
Closest papers by meaning
- Class-Aware Robust Adversarial Training for Object Detection (CVPR 2021 · similarity 0.89)
- Transferable Adversarial Attacks for Object Detection Using Object-Aware Significant Feature Distortion (AAAI 2024 · similarity 0.87)
- Adapting Object Detectors via Selective Cross-Domain Alignment (CVPR 2019 · similarity 0.87)
- Cross-Domain Adaptive Teacher for Object Detection (CVPR 2022 · similarity 0.87)
- Scale-Aware Automatic Augmentation for Object Detection (CVPR 2021 · similarity 0.86)
- KDAT: Inherent Adversarial Robustness via Knowledge Distillation with Adversarial Tuning for Object Detection Models (AAAI 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.