Research map / Label-efficient and robust learning / Recognition and long-tail
Adversarial Open-World Person Re-Identification
Xiang Li, Ancong Wu, Wei-Shi Zheng
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 APN-generated imposters to training (ablation) | APN · Market-1501 Set Verification, FTR 0.1% | TTR (%): 34.62 -> 43.85 | Improved |
| add APN-generated imposters to training (ablation) | APN · CUHK01 Set Verification, FTR 0.1% | TTR (%): 44.44 -> 55.56 | Improved |
| add APN-generated imposters to training (ablation) | APN · CUHK03 Set Verification, FTR 0.1% | TTR (%): 61.90 -> 66.67 | Improved |
| add weight sharing between target discriminator and feature extractor (ablation) | APN · Market-1501 Set Verification, FTR 1% | TTR (%): 65.38 -> 82.31 | Improved |
| add target discriminator Dt (ablation) | APN · Market-1501 Set Verification, FTR 0.1% | TTR (%): 36.92 -> 43.85 | Improved |
| add person discriminator Dp (ablation) | APN · Market-1501 Set Verification, FTR 0.1% | TTR (%): 35.38 -> 43.85 | Improved |
| generate imposters without person discriminator instead of no imposters (ablation) | APN w/o Dp · Market-1501 Set Verification, FTR 0.1% | TTR (%): 34.62 -> 35.38 | No clear effect |
| use LSRI instead of LSRO label smoothing for imposters (ablation) | APN · Market-1501 Set Verification, FTR 0.1% | TTR (%): 40.00 -> 43.85 | Improved |
| use multi-shot gallery (two images per target) instead of single-shot (ablation) | APN · Market-1501 Set Verification, FTR 0.1% | TTR (%): 23.85 -> 43.85 | Improved |
| increase target proportion from 0.5% to 5% (ablation) | APN · Market-1501 Set Verification, FTR 0.1% | TTR (%): 95.08 -> 38.82 | Hurt |
| use APN instead of ResNet-50 baseline | ResNet-50 · Market-1501 Set Verification, FTR 0.1% | TTR (%): 34.62 -> 43.85 | Improved |
| use APN imposters instead of DeepFool adversarial samples | ResNet-50 · Market-1501 Set Verification, FTR 0.1% | TTR (%): 34.62 -> 43.85 | Improved |
Closest papers by meaning
- advPattern: Physical-World Attacks on Deep Person Re-Identification via Adversarially Transformable Patterns (ICCV 2019 · similarity 0.90)
- Deep Learning for Person Re-Identification: A Survey and Outlook (TPAMI 2021 · similarity 0.89)
- Towards Open-World Person Re-Identification by One-Shot Group-Based Verification (TPAMI 2016 · similarity 0.89)
- Towards Robust Person Re-Identification by Defending Against Universal Attackers (TPAMI 2022 · similarity 0.88)
- Recover and Identify: A Generative Dual Model for Cross-Resolution Person Re-Identification (ICCV 2019 · similarity 0.87)
- Feature-Level Adversarial Attacks and Ranking Disruption for Visible-Infrared Person Re-identification (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.
