Cryo-Electron Microscopy (Cryo-EM) is a Nobel prize-winning technology for\ndetermining the 3D structure of particles at near-atomic resolution. A\nfundamental step in the recovering of the 3D single-particle structure is to\nalign its 2D projections; thus, the construction of a canonical representation\nwith a fixed rotation angle is required. Most approaches use discrete\nclustering which fails to capture the continuous nature of image rotation,\nothers suffer from low-quality image reconstruction. We propose a novel method\nthat leverages the recent development in the generative adversarial networks.\nWe introduce an encoder-decoder with a rotation angle classifier. In addition,\nwe utilize a discriminator on the decoder output to minimize the reconstruction\nerror. We demonstrate our approach with the Cryo-EM 5HDB and the rotated MNIST\ndatasets showing substantial improvement over recent methods.\n