Traffic Management in India Using YOLOv9 for Emergency and Regular Vehicle Detection

India's urban centers like Bangalore and Hyderabad are losing billions of dollars to traffic congestion, causing crippling delays for ambulances carrying heart attacks or accident victims. This model is trained with specific hyperparameters to get the best performance, and then it is validated with accuracy, precision, and recall metrics as well as IoU (Intersection over Union). The paper introduces a system to detect emergency vehicles like ambulances, police vehicles, etc in traffic using the India Driving Dataset (IDD), Indian Vehicle Dataset and YOLOv9t model experiencing congestion in Indian urbanizing cities which is designed for both single frame detection of bounding boxes at test-time and hard example mining within each iteration. The model was trained and validated on an 80–20 split with mAP value of 0.765 and precision of 0.852 making it suitable for assisting dynamic traffic signal timing adjustments based on the presence, location, or direction of travel by emergency vehicles. This enhances efficiency, productivity, and a greater system to reduce emergency response times reduces traffic-congestion. In ongoing work, we aim to further tune the system for a wider set of traffic conditions and integrate it with existing citywide infrastructure. These results illustrate the model's high accuracy and practical use, showing how emergency response time may be improved greatly with a much lower source of traffic congestion on roads. The economic implications of the system are also further reinforced, as lower congestion can help raise productivity and efficiency.