As a future challenge, considering the practicality of the system, it is necessary to reduce the number of query images taken while keeping the accuracy of localization, because requiring users to take four query images every time would reduce the convenience of the system.

At the same time, it is also necessary to evaluate

• performance in places where people often get lost, such as train stations
• accuracy when query images include moving objects such as pedestrians
• accuracy when the capture orientation of a query is different from that of the reference image
• and to verify the accuracy when using single-shot panoramic images to reduce the cost of creating the reference images.

Last but not least, the feature extractor should be trained on image data collected in a particular facility, so that it can learn the characteristics specific to the facility and to estimate the current position with higher accuracy.