The Visual Centrifuge: Model-Free Layered Video Representations | IEEE Conference Publication | IEEE Xplore