Video Autoencoder: self-supervised disentanglement of static 3D structure and motion | IEEE Conference Publication | IEEE Xplore