Traffic Sign Recognition Using a Customized Convolutional Neural Network

Abstract

Recognition of traffic signs is fundamentally a multiclass classification problem that is an essential part of autonomous driving system that aid vehicles recognize and obey road regulations. Due to their power to learn and generalize from data, neural networks have become a useful approach for solving complex problems presented by image classification tasks. Such systems are considered as a major component of an intelligent transportation system which enhances road safety and prevents from potential hazards. In the underlying research study, a customized Convolutional Neural Network (CNN) has been exploited for the categorization of traffic signs based on the German Traffic Sign Recognition Benchmark (GTSRB) dataset. The proposed model integrated with data augmentation and a robust architecture of CNN excels with an overwhelming accuracy of approximately 97% and can easily be deployed in real products like intelligent and automated traffic management systems, road safety solutions and self-driven vehicles etc.