I have been training a CNN Autoencoder on binary images (pixels are either 0 or 1) of size 64x64. The m