Aerial-Ground Person Re-ID

Summary: The paper introduces research on person re-identification in aerial imagery, provides a large-scale aerial person ReID dataset named PRAI-1581, and experimental results show that re-identifying persons in aerial imagery is a challenging problem.

Summary: Unmanned aerial vehicles (UAVs) are increasingly used for monitoring public spaces, providing a cost-effective way of data collection, particularly in detecting, tracking, and identifying subjects of interest in crowds. The P-DESTRE dataset is the first to offer UAV-based data for pedestrian long-term re-identification research, with consistent ID annotations across different days. The paper also compares the effectiveness of state-of-the-art pedestrian detection, tracking, short/long term re-identification techniques in well-known surveillance datasets as baselines for the P-DESTRE data.