I'm pretty sure there's not an autoencoder involved, it just looks like a vanilla conv net. This is the implementation: http://torontodeeplearning.github.io/co

I'm pretty sure there's not an autoencoder involved, it just looks like a vanilla conv net. This is the implementation: http://torontodeeplearning.github.io/convnet/ http://torontodeeplearning.github.io/convnet/