Deep Learning-based Traffic Violation Detection using YOLOv8 and PaddleOCR

The wearing of a helmet on motorcycle riding minimizes the risk of head injuries sustained during traffic accidents. Advanced algorithms combined with traffic surveillance systems help detect helmet and license plate information as well as triple riding. Although conventional modes of manual enforcement fail, modern deep learning techniques are unable to handle small-object detection, such as helmets and license plates. This study suggests an improved YOLOv8n-based model, which uses lightweight convolutional modules and weighted connections with a rather complex network to increase detection precision while reducing computational load. The sort of object that they were recognizing, namely the helmet, triple riders, and license plates, was shown to be really competent, high mAP and FPS with fewer model parameters, making it particularly important for real-life traffic environments and the limitations of processing capacity of devices. The study uncovered an efficient automated means of enforcement of traffic violations, which supports enforcement of road safety while weaning to reducing manual monitoring.