Performance Comparison of YOLOv5 and YOLOv8 Architectures in Human Detection using Aerial Images
Indri Purwita Sary, Safrian Andromeda, Edmund Ucok Armin
Abstract
The development of UAV technology has reached the stage of implementing artificial intelligence, control, and sensing. Cameras as UAV data inputs are employed to ensure flight safety, search for missing persons, and disaster evacuation. Human detection using cameras while flying is the focus of this article. The application of human detection in pedestrian areas using aerial image data is used as the dataset in the deep learning input process. The architectures discussed in this study are YOLOv5 and YOLOv8. The precision, recall, and F1-score values are used as comparisons to evaluate the performance of these architectures. When both architecture performances are applied, YOLOv8 outperforms YOLOv5. The achieved performance of YOLOv8 is a precision of 84.62%, recall of 75.93%, and F1-score of 79.98%.