Abstract
In this paper, we teach a machine to discover the laws of physics from videostreams. We assume no prior knowledge of physics, beyond a temporal stream ofbounding boxes. The problem is very difficult because a machine must learn notonly a governing equation (e.g. projectile motion) but also the existence ofgoverning parameters (e.g. velocities). We evaluate our ability to discoverphysical laws on videos of elementary physical phenomena, such as projectilemotion or circular motion. These elementary tasks have textbook governingequations and enable ground truth verification of our approach.
Quick Read (beta)
loading the full paper ...