Detection of Two-Wheelers Traffic Violations and Automated Ticketing Using YOLOv8

Incorporating video surveillance technology for helmet recognition and triple riding identification with the eChallan system already in place is one possible way to improve road safety in India. Two-wheeler accidents have been steadily rising across the nation. Thus it is imperative that these problems are addressed. In order to address the difficulties of automatically identifying helmets and identifying cases of triple riding, we have investigated the application of computer vision and picture preprocessing techniques. Based on deep neural network models, we suggest using the YOLO (You Only Look Once) strategy. Specifically, we use YOLOv8s to produce feature maps of different sizes, with CSP-Darknet53 serving as the foundation. The model's neck is then utilized to combine these feature maps together. The SSP and PANet components that make up the model's neck produce a feature map once more, which is then sent to the head of the model. The model's architecture's YOLO layers are in charge of forecasting variables including confidence ratings, object classifications, and bounding box positions. With a mean average accuracy (mAP0.5:0.95) of 0.426, our technique has produced encouraging results, suggesting that it may be able to recognize helmet usage and triple riding offenses, hence improving road safety.