Performance Analysis of YOLOv5, YOLOv7, YOLOv8, and YOLOv9 on Road Environment Object Detection: Comparative Study
Ghita Ikmel, El Amrani El Idrissi Najiba · 2024
This article presents a comparative analysis of different versions of the You Only Look Once (YOLO) object detection algorithms, with a focus on their performance in various environments. Using a dataset captured under various conditions, we evaluated the effectiveness of YOLOv5, YOLOv7, YOLOv8, and the newest YOLOv9 in object detection. Our study aimed to evaluate whether advancements in the newer versions of YOLO translated into improved object detection performance and to what extent, especially concerning object detection in road environments. Through extensive evaluation, we observed a gradual improvement in object detection performance from YOLOv5 to YOLOv9. YOLOv9 achieved the shortest inference time, outperforming previous versions, while maintaining competitive precision and recall values. YOLOv5 specially YOLOv5x achieved the highest mean Average Precision (mAP). Furthermore, our comparison with prior studies using earlier versions of YOLO underscores the continuous evolution and improvement of YOLO detectors over time.