Aerial-Ground Mixed-view Person Retrieval dataset

We introduce AGMix-PR(Aerial-Ground Mixed-view Person Retrieval dataset), a novel benchmark dataset specifically designed for text-to-image person retrieval in aerial-ground multi-view mixed scenarios. AGMix-PR integrates over 8,000 pedestrian identities captured from both UAV-mounted cameras and fixed ground surveillance systems across diverse urban and suburban environments, resulting in more than 70,000 image-text pairs. The dataset is meticulously curated to reflect real-world challenges, such as significant scale variations, drastic viewpoint changes, and complex occlusion patterns, all of which are rarely present in conventional TIPR datasets.