Visual Grounding of Learned Physical Models

TL;DR

This paper introduces a neural model that visually grounds physical properties and predicts future states of objects, enabling rapid adaptation and accurate physical reasoning in complex environments.

Contribution

The work presents a novel neural approach combining visual and dynamics priors for physical reasoning and prediction, with effective inference of physical properties from limited observations.

Findings

Model infers physical properties within few observations

Enables quick adaptation to unseen scenarios

Accurately predicts future states in complex environments

Abstract

Humans intuitively recognize objects' physical properties and predict their motion, even when the objects are engaged in complicated interactions. The abilities to perform physical reasoning and to adapt to new environments, while intrinsic to humans, remain challenging to state-of-the-art computational models. In this work, we present a neural model that simultaneously reasons about physics and makes future predictions based on visual and dynamics priors. The visual prior predicts a particle-based representation of the system from visual observations. An inference module operates on those particles, predicting and refining estimates of particle locations, object states, and physical parameters, subject to the constraints imposed by the dynamics prior, which we refer to as visual grounding. We demonstrate the effectiveness of our method in environments involving rigid objects, deformable materials, and fluids. Experiments show that our model can infer the physical properties within a few observations, which allows the model to quickly adapt to unseen scenarios and make accurate predictions into the future.