DCGAN is initialized with random weights, so a random code plugged into the network would produce a totally random image. Having said that, as you might imagine, the network has countless parameters that we can tweak, plus the objective is to locate a environment of those parameters which makes samples produced from random codes look like the c