In this work, we address the issue of geometric verification, with a focus on modeling large-scale landmark image collections gathered from the internet. In particular, we show that we can compute a