Traffic sign recognition using convolutional network

Abstract

Automatic traffic sign recognition is not only practical support and safety for the driver, but is also essential for automatic driving systems. The article proposes a method to identify traffic signs in Vietnam based on a CNN convolutional network model. The model has been tested with the GTSRB (German Traffic Sign Recognition Benchmark) standard dataset with more than 50,000 different traffic signs with an accuracy of more than 95.96%. In addition, the model was also tested with more than 100 actual photos taken on some streets in Hanoi. The results show that the proposed model is not only effective with the standard data set, but also works well in the real environment.