Monocular Tracking with a Mixture of View-Dependent Learned Models