DCGAN is initialized with random weights, so a random code plugged into the network would crank out a very random picture. However, while you might imagine, the network has a lot of parameters that we will tweak, plus the aim is to locate a environment of these parameters which makes samples generated from random codes appear like the instruction k