Research map / Adversarial robustness and security / Attacks on text and graphs
Adversarial T-shirt! Evading Person Detectors in A Physical World
Kaidi Xu, Gaoyuan Zhang, Sijia Liu, Quanfu Fan, Mengshu Sun, Hongge Chen, Pin-Yu Chen, Yanzhi Wang, Xue Lin
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 |
|---|---|---|---|
| add TPS transformation (ablation) | adversarial T-shirt attack · Digital, Faster R-CNN single-detector | ASR (%): 34% -> 61% | Improved |
| add TPS transformation (ablation) | adversarial T-shirt attack · Digital, YOLOv2 single-detector | ASR (%): 48% -> 74% | Improved |
| add TPS transformation (ablation) | adversarial T-shirt attack · Physical, YOLOv2 average | ASR (%): 37% -> 57% | Improved |
| add TPS transformation (ablation) | adversarial T-shirt attack · Physical, Faster R-CNN average | ASR (%): 26% -> 47% | Improved |
| use min-max ensemble instead of averaging (ablation) | advT-TPS ensemble · Digital ensemble, Faster R-CNN | ASR (%): 32% -> 47% | Improved |
| use min-max ensemble instead of averaging (ablation) | advT-TPS ensemble · Digital ensemble, YOLOv2 | ASR (%): 60% -> 53% | Hurt |
| add TPS transformation (ablation) | adversarial T-shirt attack · Physical pose: crouching | ASR (%): 27% -> 53% | Improved |
| add TPS transformation (ablation) | adversarial T-shirt attack · Physical pose: sitting | ASR (%): 26% -> 32% | Improved |
| add TPS transformation (ablation) | adversarial T-shirt attack · Physical pose: running | ASR (%): 52% -> 63% | Improved |
| add TPS transformation (ablation) | adversarial T-shirt attack · Physical complex scene: parking lot | ASR (%): 53% -> 65% | Improved |
| add TPS transformation (ablation) | adversarial T-shirt attack · Physical complex scene: crossroad | ASR (%): 51% -> 54% | No clear effect |
| lower TV regularization lambda from 3 to 1 (ablation) | advT-TPS, YOLOv2 · Digital vs physical | ASR (%): 82% digital; 51% physical | Mixed |
Closest papers by meaning
- Adversarial Texture for Fooling Person Detectors in the Physical World (CVPR 2022 · similarity 0.89)
- Beyond Digital Domain: Fooling Deep Learning Based Recognition System in Physical World (AAAI 2020 · similarity 0.87)
- Adversarial Attacks on Event-Based Pedestrian Detectors: A Physical Approach (AAAI 2025 · similarity 0.87)
- DAP: A Dynamic Adversarial Patch for Evading Person Detectors (CVPR 2024 · similarity 0.86)
- UV-Attack: Physical-World Adversarial Attacks on Person Detection via Dynamic-NeRF-based UV Mapping (ICLR 2025 · similarity 0.86)
- Full-Distance Evasion of Pedestrian Detectors in the Physical World (NeurIPS 2024 · similarity 0.85)
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.
