Robust Traffic Signs Classification using Deep Convolutional Neural Network

Smart traffic management systems have recently piqued the interest of scientists and researchers due to the enormous growth in the number of vehicles. In fact, Intelligent Transportation Systems (ITS) can handle numerous problems by using computer vision such as; traffic sign detection, recognition, and classification. Lately, Deep Convolutional Neural Network (DCNN) has been exceedingly used in traffic signs classification thanks to the powerful feature extraction and robust prediction. However, the majority of related work focuses on one aspect, the accuracy, or the parameters requirement, which makes the task unsuitable for real-time or practical uses. To address this issue, we propose a novel efficient, and lightweight neural network for traffic signs classification in road scenes. Our proposed network is able to save parameter resources while maintaining high accuracy. We mention that we have used the Belgium Traffic Sign dataset (BelgiumTS) to prove the efficiency of our proposed model in terms of accuracy and parameters requirements.