The AG-VPReID 2025 competition builds on the success of AG-ReID 2023, offering an advanced challenge in video-based person re-identification across aerial and ground environments. This competition introduces the new AG-VPReID dataset, which significantly expands the scope of aerial-ground person ReID research with high-altitude video data.

What sets this competition apart is the unprecedented scale and complexity of the dataset, featuring 3,027 identities (including 1,693 distractors), 13,507 tracklets, and approximately 3.7 million frames. The data was collected using two UAVs flying at extreme altitudes of 80-120 meters, two ground-based CCTV cameras, and two wearable cameras on a university campus.

The video-based nature of this dataset introduces temporal dynamics that can be leveraged for better identification but also presents unique challenges such as motion inconsistencies and temporal misalignment. Combined with the extreme viewpoint variations, significant scale differences, low resolution at high altitudes, occlusions, motion blur, and variable lighting conditions, this competition provides a realistic test bed for developing robust person re-identification algorithms for real-world surveillance applications.

Participants will develop algorithms to match individuals across these challenging aerial and ground video sequences, with performance evaluated using mean Average Precision (mAP) and Cumulative Matching Characteristics (CMC) metrics. This competition represents a significant advancement in aerial-ground person ReID research, pushing the boundaries from image-based to video-based analysis and from moderate (45m) to extreme (120m) altitudes.