3DNN: Viewpoint Invariant 3D Geometry Matching for Scene Understanding

   Abstract