Understanding human-object interaction is a fundamental challenge in computer vision and robotics. Crucial to it is the ability to infer 'object affordances' from visual data, namely the types of interaction supported by ...