Finding the path to a destination is an indispensable skill in our daily life. Thanks to the global navigation satellite system (GNSS), navigation systems such as Google Map facilitates this task very well for the uses of vehicle and pedestrian in outdoor environments. However, due to the lack of indoor localization solution, the navigation system for the indoor environment still remains in the realm of research. On the other hand, indoor wayfinding tasks become not trivial anymore, because of the intricate structure of modern buildings with vast information, especially for seniors and the visually impaired. In this dissertation, we introduce a navigation framework to provide an affordable solution to both the travelers and the venue providers. The framework provides a solution in two phases of an upcoming trip. The first phase is a Virtual Reality (VR) pre-travel training system allowing the travelers to learn the structure, functionalities, and necessary cues for wayfinding in the target environment prior to the physical visit. A novel approach to represent a virtual environment of the physical space was proposed and implemented in this system. It leverages a computer vision-based solution to Simultaneous Localization and Mapping (SLAM) to enable panoramic videos for immersive annotation in VR. Based on the performance tests in 3 different indoor environments, we show that the total preparation time to deploy this system in a new environment is proportional to the length of the recording path of panoramic videos at a rate of less than a minute per meter. This is a significant reduction compared to traditional 3D modeling approaches. While the cost savings are substantial, the framework enables the users to explore the virtual environment only along the recorded paths. The author envisions that this is sufficient for the purpose of learning an unfamiliar environment. The second phase is a wayfinding assistance system providing on-site guidance through a mobile application when the user arrives at the target building in person. This system can provide the user's current location and orientation, navigation instructions to a chosen destination, and other contextual information in the environment. The deployment process utilizes a Structure-from-Motion (SfM) pipeline to generate a spatial map for localization. The video recordings collected for the pre-travel training system can be reused as parts of the input in this process. With the performance tests in the three testing environments, it shows that in contrast to the pre-travel training system, the total processing time is approximately proportional to the coverage area. Every 3 square meters will take about 1 minute. A novel multi-layer graph-based data structure is also proposed and implemented to organize the building information and expedite the navigation instruction generation. A human trial with 9 participants using this system was also conducted in the Campus Center of the University of Massachusetts, Amherst.