Revolutionizing Urban Mobility: YOLO D-NET-Based AIoT Solutions for Sustainable Traffic Management

This study presents an innovative approach for addressing the complex issues associated with urban traffic management in Dhaka, Bangladesh, by incorporating the YOLOv8 model into an AIoT-based system. A customized dataset, comprising 200 data points captured from Dhaka, was utilized for training and testing the model. The YOLOv8 model produced impressive results, identifying CNG with a high precision of 89%, cars, and buses with 84% accuracy, and people with 84% accuracy. The model demonstrated an 80% accuracy rate in identifying Lanes, demonstrating its effectiveness in managing road infrastructure. The precision-confidence graphs further validate the model’s effectiveness, with discernible peaks aligning with accurately identified objects. Motorcycle detection exhibited a 60% accuracy rate, and an observed only 0.4 misclassification rate between Cars and Motorcycle suggests potential areas for refinement. These results emphasize how well the YOLOv8 model works for real-time object detection in Dhaka’s dynamic and clogged traffic environment. The study lays a foundation for future improvements in object detection accuracy and overall urban mobility through the study’s insightful contributions to developing an adaptive traffic management system. With the help of AIoT technologies, traffic management strategies in densely populated cities experiencing rapid urbanization could be revolutionized, as demonstrated by the YOLOv8 model.