Advancements in autonomous vehicle technology depend on the development of object detection systems that efficiently balance speed and accuracy. This study evaluates real-time object detection models, with an emphasis on all YOLO kinds from v1 to v10. It draws attention to significant enhancements in accuracy and speed, remarkably with YOLOv9 and YOLOv10, which triumph a high Mean Average Precision (mAP) of 0.98, surpassing earlier YOLO versions and opposing algorithms. We reveal that YOLOv10 stands out for its finest trade-off between accuracy and computational efficiency, making it a powerful nominee for autonomous vehicle applications. The paper highlights the worth of picking the suitable model based on the specific requirements of the vehicle system. By utilizing extensive datasets such as Berkeley DeepDrive 100K and VisDrone, we prove that YOLOv10 can detect crucial road objects, including vehicles, traffic signs, and pedestrians, training the model for real-world deployment.
Evaluating Real-Time Object Detection Models for Autonomous Vehicular Vision applications
12.05.2025
482508 byte
Aufsatz (Konferenz)
Elektronische Ressource
Englisch