Joint Discovery of Object States and Manipulating Actions

Abstract: Many activities involve object manipulations aiming to modify the object state. Examples of common state changes include full/empty bottle, open/closed door, and attached/detached car wheel. In this work, we seek to automatically discover the states of objects and the associated manipulating actions. Given a set of videos for a particular task, we propose a joint model that learns to identify object states and to localize state-modifying actions. Our model is formulated as a discriminative clustering cost with constraints. We assume a consistent temporal order for the changes in object states and manipulating actions, and learn the model without additional supervision. Our method is validated on a new dataset of videos depicting real-life object manipulations. We demonstrate successful discovery of seven manipulating actions and corresponding object states. Moreover, we show the improvement of object state discovery by action recognition and vice versa.

Joint work with Jean-Baptiste Alayrac, Ivan Laptev and Simon Lacoste-Julien.

 Josef Sivic:  www.di.ens.fr/~josef/