Abstract
Indoor signage plays an important role in finding specific destinations and way-finding especially for blind and visually impaired people (VIP). In this paper, we developed a new indoor signage classifier using deep convolutional neural Network (DCNN). Computer vision-based systems using cameras-based present a potential intermediate to assist blind and VIP persons on accessing unfamiliar buildings. Experiments were performed on a new dataset taken in an indoor building in France. The proposed dataset present 800 natural images divided into 4 indoor signs. Results achieved show that our proposed approach presents very encouraging results coming to 99.8% as recognition precision rate.