Studying the world by mining photo-sharing websites
Indiana University, United States
The popularity of photo-sharing websites has created immense collections of images online, with Flickr and Facebook alone hosting over 100 billion images. Each of these photos is an observation of what a small part of the world looked like at a particular point in time and space, as well as a record of where its photographer was and what he or she was paying attention to. When aggregated together and combined with the non-visual metadata available on photo sharing sites (including timestamps, geo-tags, and captions), these billions of photos are a rich source of information about the world and about human activity. In this talk I'll discuss some of our recent work in data mining and computer vision that aims to unlock this latent information from photo-sharing sites. I'll focus on recent lines of work including reconstructing maps and models of cities and landmarks, mining photos for evidence of environmental phenomena of interest to biologists and ecologists, and studying human preferences and interactions.