Research map / Adversarial robustness and security / Attacks on text and graphs
advPattern: Physical-World Attacks on Deep Person Re-Identification via Adversarially Transformable Patterns
Zhibo Wang, Siyan Zheng, Mengkai Song, Qian Wang, Alireza Rahimpour, Hairong Qi
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 adversarial pattern on clothes (digital Evading Attack, testing set) | re-ID Model A · PRCS | rank-1: 87.9% -> 4.2% | Hurt |
| add adversarial pattern on clothes (digital Evading Attack, testing set) | re-ID Model B · PRCS | rank-1: 84.7% -> 10.4% | Hurt |
| add physical adversarial pattern worn by adversary (Evading Attack) | re-ID Model A · PRCS, 14 physical positions | rank-1: 87.9% -> 27.1% | Hurt |
| add physical adversarial pattern (Impersonation Attack) | re-ID Model B · PRCS, 14 physical positions | rank-1 as target: 47.1% rank-1 and 67.9% mAP | Improved |
| evaluate pattern on unseen testing-set images instead of generating set | advPattern Evading, Model A · PRCS | rank-1 of adversary: 0.0% -> 4.2% | No clear effect |
| target an identity from Market1501 instead of PRCS (digital Impersonation) | advPattern, Model A · digital, generating set | rank-1 as target: 86.7% -> 68.0% | Hurt |
| target an identity from Market1501 instead of PRCS (digital Impersonation, testing set) | advPattern, Model A · digital, testing set | rank-1 as target: 74.4% -> 55.8% | Hurt |
| move to larger angle and farther distance (P13 vs P1) | advPattern Evading, Model A · physical | rank-1 of adversary: 0.00% -> 80.0% | Hurt |
| move from digital to physical world (Evading, Model A average) | advPattern · PRCS | rank-1 of adversary: 4.2% -> 27.1% | Hurt |
| use mask, TV smoothness, degradation function and printable color range for physical robustness | advPattern · PRCS physical | rank-1 | No clear effect |
Closest papers by meaning
- Towards Robust Person Re-Identification by Defending Against Universal Attackers (TPAMI 2022 · similarity 0.90)
- Adversarial Open-World Person Re-Identification (ECCV 2018 · similarity 0.90)
- Learning to Attack Real-World Models for Person Re-identification via Virtual-Guided Meta-Learning (AAAI 2021 · similarity 0.89)
- Feature-Level Adversarial Attacks and Ranking Disruption for Visible-Infrared Person Re-identification (NeurIPS 2024 · similarity 0.88)
- DAP: A Dynamic Adversarial Patch for Evading Person Detectors (CVPR 2024 · similarity 0.86)
- Revisiting Adversarial Patches for Designing Camera-Agnostic Attacks against Person Detection (NeurIPS 2024 · 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.
