Traffic sign recognition for controlling intelligent vehicle

Abstract

Deep learning is a subfield of artificial intelligence that has emerged this recent years with several applications. We are particularly interested in using convolutional neural networks for traffic sign images classification. Our traning framework is based on the AlexNet pre-trained model. To improve its prediction accuracy we use fine tuning techniques. The generated prediction model is then interfaced with a mobile robot equipped with a Raspberry Pi hardware board and a camera. Test environment shows how well robot system takes naviguation decisions based on the classification pre-trained model.