Recognition of Traffic Sign Using CNN and Deep Learning

Abstract

: An application of traffic sign recognition is proposed on the basis of the convolution neural network (CNN).A CNN is an artificial neural network that is used to process and recognize the image that focuses on processing pixel data. A dataset is trained, tested, and saved in order for the application to be able to detect and classify the image considered from the dataset. A Graphical User Interface (GUI) is designed for the user to try and use the application which will load the image from the dataset and classify the image as per its requirement. In the German traffic sign recognition criterion, an accuracy of 98% is obtained from the model used. Traffic Sign Recognition plays an integral part in the intelligent transportation system and has driverless vehicles and assisted driving systems are some of the applications of it [1]. The self-driving cars needed to identify each and every detail that are present on the road that includes vehicles on the road as well as pedestrians walking on the sidewalk with extreme accuracy and precision. There were no challenging and publicly available datasets in the domain for a period of time but the situations had changed in the year 2011 when Stallkamp et al [2] and Larsson and Felsberg [3] introduced datasets that includes demonstrations for traffic sign detection and classification of it.