Modern metropolitan environments continue to face a major problem with traffic congestion, which increases travel time, fuel consumption, and pollution. The Smart Traffic Management System (STMS) suggested in this paper uses an improved YOLO-based deep learning model for dynamic traffic flow optimization and real-time vehicle recognition. To evaluate real-time traffic footage, precisely identify different vehicle kinds, and calculate traffic density, the system combines computer vision and artificial intelligence (AI). This analysis enables an intelligent control system to dynamically modify traffic signals and reroute automobiles to alleviate congestion. Real-world traffic datasets are used to assess the suggested system, which shows excellent detection accuracy and responsiveness in real time. Comparative findings demonstrate how effective the strategy is in enhancing traffic flow and reducing bottlenecks. The results show that combining adaptive signal control and AI-driven real-time traffic monitoring can greatly improve urban mobility and sustainability, opening the door to smarter and more effective cities. According to experimental data, the suggested system's mean Average Precision (mAP) of 92.4% indicates its high level of vehicle identification accuracy. Additionally, the model maintains an Intersection over Union (IoU) score of 0.85, guaranteeing accurate vehicle localization. The technology also efficiently adjusts to different traffic situations, minimizing congestion and improving traffic flow in real time.