Traffic sign recognition plays a pivotal role in modern intelligent transportation systems, contributing significantly to traffic management and road safety. This thesis presents a comprehensive investigation into the utilization of deep learning techniques for enhanced traffic sign recognition. The research delves into edge detection methodologies, deep learning architectures, and classification techniques specifically tailored for the identification of traffic and road signs. A diverse range of deep learning models, including Convolutional Neural Networks (CNN), VGG19, ResNet50, ResNet101, and ResNet152, are scrutinized for their effectiveness in traffic sign classification. The evaluation is conducted on a comprehensive dataset encompassing various road signs captured under diverse environmental conditions. The classification pipeline integrates edge detection algorithms such as Canny, Sobel, and Prewitt in conjunction with the selected neural network models. Experimental results exhibit notable performance disparities among the evaluated architectures. CNN achieves the highest accuracy of 96% when combined with the Prewitt edge detection method, while VGG19 attains 95% accuracy under the same conditions. ResNet50 achieves a peak accuracy of 96% with the Prewitt edge detection technique, while ResNet101 demonstrates the capability to achieve 97% accuracy when utilizing the Canny edge detection method. Remarkably, ResNet152 emerges as the top-performing model, achieving an impressive accuracy rate of 98% when employing the Sobel edge detection method, along with an exceptional F1-Score.