Sums of Gaussians for Markerless Motion Capture

Abstract: Capturing the motion of humans in realtime without interfering with the person or the environment has many uses, e.g. classical entertainment (movies, computer games), sports analytics, healthcare, assisted refereeing, and various virtual reality applications to name just a few.  In this talk I will detail an approach and several extensions for the underlying pose optimization problem. By representing both the body and the video images as sums of gaussians the problem becomes efficient to optimize and analytically differentiable.