Developed a CNN-based deep learning model using TensorFlow and Keras to accurately recognize and classify traffic symbols, applying state-of-the-art machine learning techniques in the domain of computer vision. Built a complete traffic recognition system in Python, including data collection, manual annotation, and image preprocessing techniques (resizing, normalization, augmentation) to enhance training quality and improve model detection accuracy. Achieved over 95% model accuracy in recognizing more than 40 types of traffic signs from the GTSRB dataset. Integrated OpenCV for real-time video stream processing, enabling the system to detect traffic signs dynamically and reliably under varying conditions. Engineered the solution to support driver-assist applications, especially in adverse weather or low-visibility scenarios, improving situational awareness and road safety. Explored potential for deployment on embedded systems for scalability in smart vehicle infrastructure.