DCGAN is initialized with random weights, so a random code plugged in the network would deliver a very random image. Nevertheless, when you might imagine, the network has an incredible number of parameters that we can tweak, as well as intention is to find a location of such parameters that makes samples generated from random codes appear like the