Urban traffic congestion continues to challenge transportation systems. Traditional traffic signals operate on fixed cycles, ignoring real-time road dynamics. This paper presents a real-time smart traffic management system that uses YOLOv5 [4] for vehicle detection and counting within user-defined ROIs, Firebase [5] for cloud storage, and a trained Random Forest model [6] for adaptive signal timing prediction. The solution integrates Python, OpenCV, Firebase Admin SDK, and Streamlit [7] to provide responsive traffic control and a dynamic dashboard. This system achieves high detection reliability and adaptive decision-making while being deployable on low-resource systems.