Military Drone Swarm & Saturation Attack Dataset

This dataset addresses one of the most challenging scenarios in modern defense AI: detecting and tracking high-density drone swarms during saturation attacks. Each image contains 5–15 drones per frame at pixel sizes of 10–50 pixels, pushing the limits of small object detection algorithms. Environments include urban cityscapes, mountainous terrain, coastal areas, and open fields. The dataset features adverse weather (fog, rain, dust storms) and varied lighting from dawn to dusk. Annotations in YOLO format provide individual bounding boxes for each drone in the swarm. Designed to benchmark Counter-UAS tracking algorithms, multi-object trackers, and swarm behavior analysis. The open-source sample provides 117 images; the full package includes 1,000+ images with 5 evaluation videos.