From Pixels to Predicates: Recent Advances in Computational and Robotic Vision