Improving Content-Invariance in Gated Autoencoders for 2D and 3D Object Rotation | Read Paper on Bytez