Comparative study of Traffic Sign Classification Models

Abstract

Autonomous vehicle technology has experienced a revolution due to the development of artificial intelligence, which incorporates complex models and algorithms. One of the most important developments in computer science and robotic intelligence is self-driving cars. Although several models have already been developed and are in use, everyone in this subject is certain that there is still a tonne of untapped potential. In order to perform more accurately and effectively than conventional models, convolutional neural networks are utilised in this project’s feature categorization. The image is first converted to grayscale by the algorithm, then it is divided into three layers. A sophisticated convolutional neural network that incorporates crucial data from traffic signs and images classifies them simultaneously into the right categories. The system’s effectiveness has been shown by the results. We categorise traffic signs using convolutional networks (ConvNets) as part of the German Traffic Sign Recognition Benchmark dataset (GTSRB). With biological inspiration, ConvNets are multi-stage topologies that automatically pick up the invariant characteristic hierarchy.. The capsule network’s most current accuracy on the German Traffic Sign Recognition Benchmark dataset was 98.44%. Index Terms: Neural Network, CNN, LeNet, Traffic Sign Classifier, GTSRB, ALVINN, and ReLU.