Introducing AOD 4: A dataset for air borne object detection.

This paper introduces an airborne object dataset comprising 22,516 images categorizing four classes of airborne objects: airplanes, helicopters, drones, and birds. The dataset was compiled from YouTube-8 M, Anti-UAV, and Ahmed Mohsen's dataset hosted on Roboflow. Videos were sourced from the first two platforms and converted into individual frames, whereas the latter dataset already consisted of images. Following collection, the dataset underwent labelling and annotation processes utilizing Roboflow's annotation tool, resulting in 7,900 annotations per class. Researchers can leverage this dataset to develop and refine algorithms for airborne object detection and tracking, with potential applications spanning military surveillance, border security, and public safety.