AI Paper Atlas

Research map / Adversarial robustness and security / Attacks on text and graphs

Adversarial Metric Attack and Defense for Person Re-Identification

Song Bai, Yingwei Li, Yuyin Zhou, Qizhu Li, Philip H. S. Torr

TPAMI 2020 · Adversarial robustness and security › Attacks on text and graphs · approach: Adversarial perturbations beyond images · 91 citations (OpenAlex)

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 changedWhereMeasuredEffect
attack the metric distance instead of the classification lossI-FGSM on B1 · Market-1501 white-boxmAP: 52.93 -> 0.367Improved
use iterative I-FGSM instead of single-step FGSM for white-box metric attackmetric attack on B1 · Market-1501 white-boxmAP: 7.054 -> 0.367Improved
add momentum (MI-FGSM) for black-box transfermetric attack from B1 to B3 · Market-1501 black-boxmAP: 29.42 -> 25.53Improved
attack an ensemble of three models instead of a single modelMI-FGSM metric attack, tested on B1 · Market-1501 black-boxmAP: 22.29 -> 14.94Improved
attack an ensemble of three models instead of a single modelMI-FGSM metric attack, tested on HACNN · Market-1501 black-boxmAP: 37.98 -> 30.45Improved
use attention-based re-ID model (HACNN) instead of plain baseline B1-B3 as the victimMI-FGSM black-box attack from B1 · Market-1501 black-boxmAP: 25.53 -> 37.98Improved
use XQDA Mahalanobis metric on deep features instead of EuclideanB1 clean evaluation · Market-1501mAP: 77.52 -> 75.86Hurt
evaluate with a different metric than the one attacked (Euclidean attack, XQDA test) (ablation)metric attack · Market-1501mAPImproved
train metric-preserving model on clean plus adversarial training imagesB1 tested on B3 adversarial gallery · Market-1501mAP: 24.72 -> 70.46Improved
train metric-preserving model on clean plus adversarial training imagesB1 on clean gallery · Market-1501mAP: 77.52 -> 76.69Hurt
train metric-preserving model on clean plus adversarial training imagesB4 triplet model on clean gallery · Market-1501mAP: 67.86 -> 59.87Hurt
train metric-preserving model on clean plus adversarial training imagesB4 white-box adversarial gallery · Market-1501mAP: 0.211 -> 57.45Improved

Closest papers by meaning

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.

Open scime draft

Get the next scime Dispatch and a monthly field update

Free. We send a confirmation link first; every email has a one-click unsubscribe. Privacy