Fill and Transfer: A Simple Physics-based Approach for Containability Reasoning

Lap-Fai Yu

Noah Duncan

Sai-Kit Yeung


(a) Which direction shall we tilt the pitcher to pour the water out?
(b) Shall we use this container to hold water?
(c) Which object is a good liquid container?
(d) Shall we use a plate or a cup to carry water?

Abstract The visual perception of object affordances has emerged as a useful ingredient for building powerful computer vision and robotic applications [30]. In this paper we introduce a novel approach to reason about liquid containability - the affordance of containing liquid. Our approach analyzes container objects based on two simple physical processes: the Fill and Transfer of liquid. First, it reasons about whether a given 3D object is a liquid container and its best filling direction. Second, it proposes directions to transfer its contained liquid to the outside while avoiding spillage. We compare our simplified model with a common fluid dynamics simulation and demonstrate that our algorithm makes human-like choices about the best directions to fill containers and transfer liquid from them. We apply our approach to reason about the containability of several real-world objects acquired using a consumer-grade depth camera.

Keywords: functionality, physics-based reasoning, affordance, vision for graphics, robotics

Publications:

Acknowledgements:

Results:

Containability analysis of some objects in our dataset:



Real-world examples. (a) An input scene showing a sealed can, a bowl and a cereal box on a table. Our algorithm suggests that the bowl is a container and all its transfer directions are equally-likely, and that the table after being inverted is also a possible container. (b) An input scene showing a cup, a plate, two speakers and a piece of cloth. The cup and plate are identified as possible containers. The cup is found to be more robust to slight perturbation, as it has a smaller rate of decrease of containee volume (V ) when slightly tilted. (c)-(g) Results on other real-world objects captured by a Structure Sensor. The hole-ridden tray in (g) is identified as a non-container.