Joint Discovery of Object States and Manipulation Actions
This work addresses the challenge of understanding object manipulations in videos for applications in robotics and computer vision, representing an incremental advance in unsupervised learning for visual tasks.
The paper tackled the problem of automatically discovering object states and manipulation actions from videos, proposing a joint model that learns to identify states and localize actions without additional supervision, and demonstrated successful discovery of seven actions and states on a new real-life dataset with improved performance in both tasks.
Many human 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 manipulation 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 manipulation actions, and introduce new optimization techniques to learn model parameters without additional supervision. We demonstrate successful discovery of seven manipulation actions and corresponding object states on a new dataset of videos depicting real-life object manipulations. We show that our joint formulation results in an improvement of object state discovery by action recognition and vice versa.