Unsupervised Learning of Depth and Ego-Motion from Video

   Abstract