Understanding Geographic Attentions of Crowd from Photographing Information

We propose a framework for sharing photographs that reduces a load on communication network. When interesting events occur, many people take photographs of the events. Collecting photographs of such events and quickly building image database would enable us to share information in real-time more concretely than when using text based methods. However, a simple approach sending all photographs to a server immediately may cause congestion of a network infrastructure. In this paper, we propose a digital photographs classification method based on a main object of a photograph without using pixel information. Our method uses photographing information, that includes information about location and azimuth of a camera. Photographing information is automatically embedded into the photographs when it was taken. Our method determines main objects of photographs (socalled subject) and classifies the photographs based on these subjects. We can suppress congestion because only the best photographs transmitted to a server. In addition, classification accuracy is high because of the effect of collective intelligence. In this paper, we confirmed the effectiveness of our method through a series of experiments using a commercially available digital camera.