Pathlookup: A deep learning-based framework to assist visually impaired in outdoor wayfinding
MetadataShow full item record
Das, U., Namboodiri, V., & He, H. (2021). PathLookup: A deep learning-based framework to assist visually impaired in outdoor wayfinding. Paper presented at the 2021 IEEE International Conference on Pervasive Computing and Communications Workshops and Other Affiliated Events, PerCom Workshops 2021, 111-116. doi:10.1109/PerComWorkshops51409.2021.9431007
Reading and following visual signs remains the predominant mechanism for navigation and receiving wayfinding information in areas without accurate GPS coverage. This puts people who are blind or visually impaired (BVI) at a great disadvantage. There still remains a great need to provide a low-cost, easy to use, and reliable auxiliary wayfinding system within indoor and outdoor spaces that complements existing satellite-based systems. Through both a user study and a quantitative study of GPS accuracies in outdoor environments, this paper highlights the need for auxiliary outdoor wayfinding tools for people with visual impairments. A deep learning-based image localization framework called PathLookup is proposed in this work for accurately providing path advancement information for outdoor wayfinding. Evaluation results show PathLookup to be highly accurate and fast potentially proving to be a valuable tool for future integration into outdoor wayfinding systems.
Click on the DOI link to access this conference paper at the publishers website (may not be free).