A model of Vietnamese traffic signs classification

Abstract

With the development of modern technology today, its application to various fields is extremely important. Especially in the field of traffic safety, with a society where the occurrence of traffic accidents is very high, in which traffic sign recognition is one of the essential topics for everyone. Therefore, in this thesis, the goal is to provide a model that can identify and classify types of traffic signs in Vietnam in the most accurate way. And to complete this project, it is necessary to rely on Deep Convolutional Neural Network (CNN) and Resnet-50 model to identify and provide parameters for classification of traffic signs. Along with using data augmentation techniques to create image complexity and with convolution layers combining modern methods such as batch normalization, the initialization to improve the accuracy of the problem.