Integrating computer vision and complex networks for urban analysis

Abstract

Nowadays, different types of data on cities are massively and decentrally produced, given improved technologies and reduced cost of image acquisition sensors. Two approaches are frequent in the use of visual data as a source of information about cities. In the first case, computer vision techniques are used and characteristic analyzes extracted from images are performed. In the second case, numerical or categorical data are used and these are modeled by complex networks. In the first approach, we have less physical understanding of the phenomena and, second, we do not take advantage of the myriad of visual data available. In this project, we propose the use of visual data from different sources and scales for the extraction of knowledge about the cities. In particular, we will use ground-level imagery such as from monitoring cameras, and from mobile phones and from remote sensing to model complex systems in an attempt to increase the understanding of urban problems such as mobility, crime and urbanization. Effective methods of computer vision, including methods based on deep learning, will be used in conjunction with complex network modeling techniques for knowledge extraction. The candidate plans to submit a BEPE plan to a center of excellence in complex network research.