<ul><li>A new method called Modality Unified Attack (MUA) is proposed for person re-identification (re-id) models.</li><li>The method focuses on generating modality unified adversarial examples (AEs) to attack omni-modality re-id models.</li><li>It introduces modality-specific adversarial generators and cross-modality disruption techniques.</li><li>Experiment results show high mean mAP Drop Rates for the attacked re-id models.</li></ul>

Modality Unified Attack for Omni-Modality Person Re-Identification

Discover more