Efficient Vehicle Detection System Using YOLOv8 on Jetson Nano Board

The progression of intelligent transportation systems heavily hinges on the efficacy of vehicle identification and detection techniques. Within this realm, numerous methodologies have been proposed, among which the YOLOv8 model has emerged as a notable contender, lauded for its remarkable accuracy and widespread adoption across diverse domains. The following one is it operates in real-time object detection, which makes this a huge development in the YOLO series. In terms of operation, this suggests that it finds the best balance between speed and optimum, which makes it ideal for applications that need speedy and accurateobject recognition. In our case, we propose employing the YOLOv8 model in the recognition of nine types of vehicles and free categorization from the processing of a revised image dataset. The reprocessing of the image dataset by adding labels and including 204 validation photos and 2,674 images for training operation is performed. Afterward, the training in the case of the parameters holds with epoch = 30, batch = 8, and image size = 640, which includes the results of training showing accuracy for the training performed at 99.1%.