Research map / Adversarial robustness and security / Attacks on text and graphs
Learning to Attack Real-World Models for Person Re-identification via Virtual-Guided Meta-Learning
Read the paperCheck 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 virtual-guided meta-learning for a universal perturbation | IDE and PCB re-ID models · unseen target domains in Table 1 | mAP and rank-1: IDE mAP: 8.2 -> 4.9 (Duke → Market); PCB mAP: 21.6 -> 19.5 (Duke → Market) | Helped |
| add real extra training data without meta-learning (ablation) | baseline universal perturbation · Duke → MSMT and Market → MSMT | mAP: Duke → MSMT: 5.6 -> 5.1; Market → MSMT: 5.8 -> 5.7 | Mixed |
| add PersonX without meta-learning (ablation) | baseline universal perturbation · Duke → MSMT and Market → MSMT | mAP: Duke → MSMT: 5.6 -> 4.8; Market → MSMT: 5.8 -> 5.0 | Helped |
| add meta-learning with real data as meta-test (ablation) | universal perturbation · Duke → MSMT and Market → MSMT | mAP: Duke → MSMT: 5.1 -> 4.6; Market → MSMT: 5.7 -> 5.5 | Mixed |
| add meta-learning with PersonX as meta-test rather than real extra data (ablation) | universal perturbation · Duke → MSMT and Market → MSMT | mAP: Duke → MSMT: 4.6 -> 3.5; Market → MSMT: 5.5 -> 3.4 | Helped |
| train MisRank with PersonX in addition to source data | MisRank · unseen target domains | mAP: Duke → Market: 28.2 -> 38.5; Duke → MSMT: 11.7 -> 20.9 | Hurt |
Closest papers by meaning
- Similarity Distribution Based Membership Inference Attack on Person Re-identification (AAAI 2023 · similarity 0.91)
- Meta Batch-Instance Normalization for Generalizable Person Re-Identification (CVPR 2021 · similarity 0.90)
- Smoothing Adversarial Domain Attack and P-Memory Reconsolidation for Cross-Domain Person Re-Identification (CVPR 2020 · similarity 0.90)
- Learning to Generalize Unseen Domains via Memory-based Multi-Source Meta-Learning for Person Re-Identification (CVPR 2021 · similarity 0.89)
- Alice Benchmarks: Connecting Real World Re-Identification with the Synthetic (ICLR 2024 · similarity 0.87)
- Feature-Level Adversarial Attacks and Ranking Disruption for Visible-Infrared Person Re-identification (NeurIPS 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.