Lecture: Sightseeing in the eyes of neural networks

Often we remember a city, that we visited as a tourist, by its landmarks e.g., Tokyo by its tower or the ancient Buddhist temple. However, a local would probably describe Tokyo by other attributes such as its narrow pedestrian sidewalks. What is the typical urban scene of a city that discriminates it from other cities? Or the most distinguishing visual urban features of Tokyo once compared to an American city e.g., Pittsburgh? Perhaps it is not easy to find your experts, who have seen (or perhaps lived) in both places, to answer the former question. What about an AI assistant who has explored all the streets of both cities and now is ready to share with you, its observation and perception? This work showcases how intelligent visual models, the so-called artificial neural networks, can be used to create knowledge from publicly available street-view imagery.