Traffic Sign Recognition Using CNN

Manuscript submitted November 20, 2024; revised November 29, 2024; accepted December 25, 2024; published January 17, 2025

Abstract—Traffic Sign Recognition (TSR) is a crucial component of intelligent transportation systems, aiming to enhance road safety and support autonomous vehicle navigation. This paper focuses on developing a traffic sign recognition system using Convolutional Neural Networks (CNNs). The system employs two distinct models: a custom sequential CNN and a pre-trained VGG19 model. Both models were trained on the German Traffic Sign Recognition Benchmark (GTSRB) dataset, which comprises 43 classes of traffic signs, over 30 epochs. The paper investigated two training scenarios. In the first case, both models were trained with a batch size of 8 and a learning rate of 0.001. In the second case, the batch size was increased to 16, and the learning rate was decreased to 0.0001. The models were evaluated based on various metrics, including accuracy, validation curves, test accuracy, confusion matrix, F1-Score, precision, and recall. Results showed that the second case (batch size of 16 and learning rate of 0.0001) yielded better overall performance, particularly in test accuracy. In this scenario, the sequential model achieved a training accuracy of 99.77% and a validation accuracy of 97.48%. The VGG19 model outperformed the sequential model, achieving a training accuracy of 99.94% and a validation accuracy of 98.46%. The results of this paper contribute to the advancement of traffic sign recognition systems, supporting their implementation in real-world intelligent transportation and autonomous driving applications..

keywords—Convolutional Neural Network (CNN), German Traffic Sign Recognition Benchmark (GTSRB), German Traffic Sign Recognition Benchmark (GTSRB), VGG

Cite: Wajeh E. Elside, Ahmed J. Abougarair"Traffic Sign Recognition Using CNN," Journal of Advances in Artificial Intelligence vol. 3, no. 1, pp. 19-39, 2025. doi: 10.18178/JAAI.2025.3.1.19-39

Copyright © 2025 by the authors. This is an open access article distributed under the Creative Commons Attribution License which permits unrestricted use, distribution, and reproduction in any medium, provided the original work is properly cited (CC BY 4.0).