Second version, trained on pics generated with the first version.
First two pics depict images generated with this version.
Third pic is generated from the old version.
Last 3 are the training results of this version.
