Traffic Sign Recognition Using Convolutional Neural Networks

  • Dr. V. Padmavathi, Burra Kamal Koushik, Pendyala Harshith, Sri Teja Vedantham

Abstract

Autonomous Driving is the foreseen future of road transportation. It involves various technologies for perception, classification, intelligent decision making so on and so forth to adhere to laws of the road. Traffic Signs Recognition is a critical part of perception technology. This paper majorly involves two actions Traffic Sign Detection and Classification from a video input. Exploiting the Shape and Color features of the Traffic Signs in each frame, Detection is done by normalizing the input image and extracting specific color channel (Blue and Red) images and detect particular shapes (circles and triangles). For classification a Convolutional Neural Network model is introduced with 7 hidden layers and optimized by ADAM algorithm which achieved an accuracy of 97% on German Traffic Sign Recognition Benchmark (GTSRB) dataset which contain over 50,000 images and 43 classes

 

Keywords: Autonomous vehicles, Image Classification,Convolutional Neural Networks, German Traffic Sign and Recognition Benchmark, Image processing

Published
2020-06-06
How to Cite
Dr. V. Padmavathi, Burra Kamal Koushik, Pendyala Harshith, Sri Teja Vedantham. (2020). Traffic Sign Recognition Using Convolutional Neural Networks. International Journal of Advanced Science and Technology, 29(05), 11523-11532. Retrieved from http://sersc.org/journals/index.php/IJAST/article/view/25263