Method For Traffic Violation Detection Using Deep Learning

Urban areas commonly encounter traffic congestion. One of its primary causes is motorists violating traffic regulations, such as travelling at excessive speeds. Besides risking the safety of other road users, speeding exacerbates traffic capacity. Artificial Intelligence will detect instances where vehicles exceed the maximum speed limit on a thoroughfare to address this issue. The YOLOv8 algorithm, a commonly utilised Convolutional Neural Network Architecture in Computer Vision for object detection and classification, is employed in this study to detect instances of speeding vehicles in live video footage. The research model can accurately identify and categorise vehicles in real-time CCTV footage, showcasing superior bounding box prediction performance. However, the model's low Recall rate and Mean Average Precision scores at both IoU thresholds warrant further exploration. Issues such as misidentification of objects remain prevalent in the model, which may stem from inaccurate classification of object categories, computational restrictions, and partiality in the data. Issues such as misidentification of objects remain prevalent in the model, which may stem from inaccurate classification of object categories, computational restrictions, and partiality in the data. Nonetheless, this study reveals multiple possibilities for enhancing the model, which includes employing new data sets and incorporating supplementary detections.