On-Board Real-Time Pedestrian Detection for Micro Unmanned Aerial Vehicles Based on YOLO-v8

Nowadays, allowing unmanned aerial vehicles (UAVs) to accompany humans in daily life has become a hot topic. Pedestrian detection plays an important role on this application with its accuracy and real-time detection. In this paper, we design an on-board real-time pedestrian detection method for micro UAVs based on YOLO-v8 network. More specifically, several custom network models of small scales are fine-tuned based on YOLO-v8 to detect pedestrians in real-time. Besides, our detection methods are implemented and deployed on a custom micro UAV. Through simulation and real-world experiments, the results show that the custom detection network models based on YOLO-v8 can accurately detect pedestrians at up to 34 FPS on Jetson Xavier NX.