Evaluation of YOLOv8n as a Suitable Tool for UAV Navigation

Abstract

As the world continues to embrace the intelligent automation of robots, it is also faced with significant technological challenges. One of these problems includes unarmed aerial vehicles (UAVs) and the avoidance of obstacles. In order to solve this issue, drones should be able to collect information from their environment accurately. Therefore, in this research, an investigation of different lightweight object detectors is exhibited, in order to select the most optimal algorithm to solve this challenge. The research demonstrated that You-Only-Look-Once (YOLO) was the most optimal model for this use case (in terms of speed) and thus one of its best-performing lightweight versions, YOLOv8n, was selected and thoroughly evaluated. Moreover, the challenge of developing an algorithm for a UAV, that uses a standard 2D camera for people avoidance, is as well presented.