Research map / Adversarial robustness and security / Attacks on text and graphs
Universal Physical Camouflage Attacks on Object Detectors
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 |
|---|---|---|---|
| Use UPC_rc, combining RPN and classifier/regressor attacks with internal deformation simulation | Faster R-CNN · AttackScenes, standing, walking, sitting | p0.5: 0.85 -> 0.07; 0.48 -> 0.04; 0.87 -> 0.46 | Helped |
| Add internal deformation simulation Tc (ablation) | UPC · AttackScenes, standing, walking, sitting | p0.5: 0.66 -> 0.07; 0.33 -> 0.04; 0.76 -> 0.46 | Helped |
| Jointly attack RPN and classifier/regressor instead of attacking only the classifier (ablation) | UPC · AttackScenes, standing, walking, sitting | p0.5: 0.18 -> 0.07; 0.06 -> 0.04; 0.54 -> 0.46 | Helped |
| Increase the number of painted pattern regions from 3 to 8 (ablation) | UPC · AttackScenes, standing, walking, sitting | p0.5: 0.39 -> 0.07; 0.26 -> 0.04; 0.70 -> 0.46 | Helped |
| Use natural-image patterns rather than UPC-generated camouflage | Faster R-CNN · AttackScenes, standing, walking, sitting | p0.5: 0.96 -> 0.07; 0.93 -> 0.04; 0.95 -> 0.46 | Hurt |
| Use 8-pattern UPC camouflage in physical stationary scenes | Faster R-CNN · Physical stationary, standing | p0.5: 1.0 -> 0.17 | Helped |
| Apply UPC camouflage during motion, where blur degrades its appearance | Faster R-CNN · Physical motion | detection precision: 26% (150/583); 21% (80/377); 17% (37/219); 34% (240/713); 15% (118/804); 24% (240/997) | Mixed |
| Transfer the pattern across training datasets or detector architectures | UPC · Transferability testing, walking humans | p0.5: 0.95 -> 0.08; 1.0 -> 0.25 | Helped |
| Use UPC camouflage against car objects | Faster R-CNN · Car, physical photos | correct car detection rate: 24% (29/120) | Helped |
| Add a defense preprocessing method to UPC attacks | UPC · Defense evaluation, FR-VGG16-0712 | p0.5: 0.07 -> 0.20; 0.07 -> 0.12; 0.07 -> 0.28; 0.07 -> 0.17; 0.07 -> 0.10; 0.07 -> 0.11; 0.07 -> 0.34; 0.07 -> 0.16 | Mixed |
Closest papers by meaning
- PhyCamo: A Robust Physical Camouflage via Contrastive Learning for Multi-View Physical Adversarial Attack (AAAI 2025 · similarity 0.88)
- RAUCA: A Novel Physical Adversarial Attack on Vehicle Detectors via Robust and Accurate Camouflage Generation (ICML 2024 · similarity 0.88)
- Adversarial Attacks on Event-Based Pedestrian Detectors: A Physical Approach (AAAI 2025 · similarity 0.87)
- CNCA: Toward Customizable and Natural Generation of Adversarial Camouflage for Vehicle Detectors (NeurIPS 2024 · similarity 0.87)
- Full-Distance Evasion of Pedestrian Detectors in the Physical World (NeurIPS 2024 · similarity 0.86)
- UV-Attack: Physical-World Adversarial Attacks on Person Detection via Dynamic-NeRF-based UV Mapping (ICLR 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.