A Culturally-Aware Benchmark for Person Re-Identification in Modest Attire

Person Re-Identification (ReID) is a fundamental task in computer vision withcritical applications in surveillance and security. Despite progress in recentyears, most existing ReID models often struggle to generalize across diversecultural contexts, particularly in Islamic regions like Iran, where modestclothing styles are prevalent. Existing datasets predominantly feature Westernand East Asian fashion, limiting their applicability in these settings. Toaddress this gap, we introduce Iran University of Science and Technology PersonRe-Identification (IUST_PersonReId), a dataset designed to reflect the uniquechallenges of ReID in new cultural environments, emphasizing modest attire anddiverse scenarios from Iran, including markets, campuses, and mosques.Experiments on IUST_PersonReId with state-of-the-art models, such as SemanticControllable Self-supervised Learning (SOLIDER) and Contrastive Language-ImagePretraining Re-Identification (CLIP-ReID), reveal significant performance dropscompared to benchmarks like Market1501 and Multi-Scene MultiTime (MSMT17),specifically, SOLIDER shows a drop of 50.75% and 23.01% Mean Average Precision(mAP) compared to Market1501 and MSMT17 respectively, while CLIP-ReID exhibitsa drop of 38.09% and 21.74% mAP, highlighting the challenges posed by occlusionand limited distinctive features. Sequence-based evaluations show improvementsby leveraging temporal context, emphasizing the dataset's potential foradvancing culturally sensitive and robust ReID systems. IUST_PersonReId offersa critical resource for addressing fairness and bias in ReID research globally.