Deep Learning Based Indian Traffic Sign Detection And Recognition System

Abstract

Traffic sign detection and recognition is a critical component of intelligent transportation systems and autonomous driving, particularly in complex and unstructured environments such as Indian road networks. This study proposes a deep learning-based framework for robust detection and classification of Indian traffic signs using a combination of object detection and Convolution Neural Network (CNN) models. The methodology integrates a real-time detection model to localize traffic signs in input images, followed by a CNN-based classifier to accurately categorize the detected signs into predefined classes such as mandatory, cautionary, and informatory signs. A comprehensive dataset comprising diverse Indian traffic scenarios is utilized, incorporating variations in illumination, occlusion, weather conditions, and sign degradation. Data preprocessing techniques, including normalization and augmentation, are employed to enhance model generalization. The system is implemented on a cloud-based platform using GPU acceleration, enabling efficient training and scalability. Transfer learning strategies are further incorporated to improve performance and reduce training time. Experimental results demonstrate that the proposed model achieves high detection accuracy and classification performance, outperforming conventional methods under challenging real-world conditions. The framework also exhibits real-time processing capability, making it suitable for deployment in advanced driver assistance systems and smart mobility solutions. The proposed approach contributes to improving road safety and supports the development of intelligent, automated traffic monitoring systems in developing countries.