Interactive YOLO-based Object Detection using a Polygonal Region of Interest for Airborne Surveillance Applications

Object detection in airborne surveillance is constrained by low size, weight and power. Under these constraints, detection speeds can be improved using a Region of Interest (ROI) based approach. This paper outlines an ROI-based detection utilizing the You Only Look Once (YOLO) model, augmented with mouse interaction and polygon area functions. Mouse interaction enables user-driven selection of ROIs for concentrating on specific regions of interest. Additionally, polygon area function facilitates the creation of irregularly shaped ROIs, for object tracking. Our experimental findings illustrate the efficacy of the suggested approach in achieving accurate ROI detection while maintaining the real-time performance.