Traffic Sign Recognition using Transfer Learning and Deep Convolutional Neural Networks

There is a rapid increase in popularity of the self-driving automobiles. Road safety while keeping the navigation systems efficient necessitates an efficient traffic sign recognition (TSR) system. Exiting methods are computationally complex. This work proposes a novel approach leveraging the adaptability of the transfer learning (TL) models. A pre-trained convolutional neural network (CNN) is used with transfer learning to improve the system performance. The model is trained on the German traffic sign recognition benchmark (GTSRB) dataset. The proposed work aims to recognize a wider range of traffic signs while keeping computational costs low. The comparison analysis with existing and standard models validates the effectiveness of the proposed model. The suggested approach may significantly increase the recognition capabilities prompting a safer transportation system.