Stakeholders of tourism and science of tourism request tools to understand and analyze the tourist practices and tourist mobility. While these ones are still only partially known, the rise of image hosting websites opens a new trend to follow. This new source of information allows to apprehend frequentation of a tourist site according to seasons, years, photographer country of origin. Furthermore whatever the site size : city, county, country, continent or any delimited area. Focusing on a specific destination, the set of user photos characterizes his path and the items that illustrate, from his point of view, the destination. In this paper we propose a new methodology to automatically build path visual description from image hosting websites. To describe visually a path, a clustering process provides the most-representative images for each point of interest along the path. To illustrate how this work permits inference of imaginary, we present our results on New-York city.