| Summary: Within the scope of this study, an audio augmented reality application that Works with smartphones recording live images has been developed to facilitate visually impaired individuals lives. The application provides object detection and obstacle notification-depth extraction via instant phone camera image. It also provides navigation via online maps and presents these outputs to the visually impaired user with audible feedback. While performing these functions, it benefits from the problem-solving capabilities of deep learning techniques. Thanks to the use of deep learning models via web services, which are very costly in terms of computational load and storage, the application has become an easily
accessible, economical solution that can produce results even in the low-end or mid-end devices without causing a load on the user device. In addition, an open-source and domestic software product, open to the participation of developers around the world, was obtained.
Keywords: Augmented Reality, Object Detection, Navigation, Monocular Depth Extraction |
