DCGAN is initialized with random weights, so a random code plugged into your network would deliver a totally random impression. Even so, as you might imagine, the network has millions of parameters that we can tweak, as well as target is to find a setting of such parameters which makes samples generated from random codes look like the schooling kno