[sorenjan]

Is this basically teaching the network how to do pattern processing, like line and corner detection, and then using that trained network as a starting point when training on real images?