Performance Analysis by the Number of Learning Images on Anti-Drone Object Detection System with YOLO

Abstract

Recently, machine learning based real-time anti-drone object detection systems have attracted great attention to protect multi-use facilities or national important facilities from drones. This paper studies the performance and relationship analysis on the anti-drone object detection system by the number of learning images with YOLO network based on transfer learning, in order to provide guidelines that can be applied in real environments where learning data is difficult to obtain, such as terrorist and wartime situations, and sudden drone/UAV infiltration situations, and so on. © 2024, Korean Institute of Communications and Information Sciences. All rights reserved.