<![CDATA[Traffic congestion at urban intersections can significantly reduce transportation efficiency and obstruct emergency vehicles that require rapid access to critical locations. This study proposes DeepTraffic, an integrated deep-learning-based traffic-control system designed to detect vehicles, classify traffic density, identify emergency vehicles, and optimize traffic-light timing in real time. An empirical experimental system-development approach was employed by integrating Camera/CCTV, an emergency sensor, YOLO-based vehicle detection, vehicle classification and counting, traffic-density estimation, an emergency-priority decision algorithm, and a Raspberry Pi controller connected to traffic-light hardware. Traffic conditions were classified into five categories based on detected vehicle density, namely empty, normal, busy, congested, and highly congested, with adaptive green-light durations assigned according to the corresponding traffic state. Emergency vehicles, including ambulances, fire engines, and police vehicles, were assigned priority conditions that activate the corresponding green phase to facilitate rapid intersection passage. System evaluation was designed to assess vehicle-detection performance using precision, recall, F1-score, mAP@0.5, and frames per second, while traffic-density classification and signal-control performance were evaluated through classification accuracy and signal-duration execution. Emergency-priority performance was assessed through activation rate, response time, waiting-time reduction, and queue-length reduction. The proposed framework integrates real-time computer vision with adaptive signal control and emergency prioritization, providing a reproducible foundation for intelligent traffic management and future smart-city transportation applications. ]]>
