Mobile Application Based Indoor Routing System Using Transfer Learning

Nowadays, indoor routing in places with complex multi-storey architecture such as hospitals, shopping malls, parking garages and public buildings is traditionally carried out using signage or devices in a fixed position. When we examine the literature, it is generally seen that indoor orientation studies for certain needs are seen. The fact that the routing systems are fixed, and the signage is not an effective tool constitutes the motivation of this study. In this study, an image-based mobile application that is hardware-independent and adaptable to other interior spaces has been implemented using a mobile device. The application basically consists of two parts. In the first part, transfer learning based MobileNetV2 architecture is used to determine the initial store location. The proposed model detects the store signage image taken from the camera with 96% success. In the second part, the user is successfully guided to the target using the Dijkstra algorithm. With the developed mobile application, the user can reach the targets on the same or different floors in the fastest way without wasting time and without asking anyone. The application was tried in real time in a shopping center and successful results are obtained.