Drone detection with YOLOv5

Abstract

We propose a fast and accurate method for visual drone detection based on YOLOv5 architecture providing state-of-the-art performance. The proposed method aims to drone detection in combat and real-world environments for military use based on visual detection in the visible and infrared spectrum. The method provides precision/recall of 99.1/98.5% and 99.0/95.3% for RGB and infrared videos from the AntiUAV dataset.