Gravar-mail: Deep Networks Can Resemble Human Feed-forward Vision in Invariant Object Recognition