Yufei’s research interests generally lie at the intersection of computer vision, robotics, and machine learning. Her long-term goal is to build agents that can perceive the physical interactions among objects for generic categories, understand the consequences of their own interactions with the physical world, and even hallucinate the potential effect had specific interactions occurred. Towards this grand goal, Yufei has investigated several key aspects of this problem including 3D spatial inference for generic objects, understanding dynamics in complex scenes, and reducing human supervision for scalable solutions. This could potentially be applied to understanding large-scale videos, robot learning from demonstration, and AR/VR systems.