Neural Network Based Traffic Sign Recognition for Autonomous Driving

Autonomous and self-driving cars represent a certitude for the near future and it is expected to be safer, time saving and to bring in environmental benefits. The road infrastructure still embeds traffic signs that have to be recognized with high accuracy by the sensorial system mounted on an autonomous car. This paper deals with a new Traffic Sign Recognitions system based on a pretrained neural network for general objects and retrained for a large number of traffic signs collected in real environment under different metrorological conditions. The proposed Traffic Sign Recognition system is able to classify signs in new scenarios with more than 96% accuracy.