DCGAN is initialized with random weights, so a random code plugged to the network would deliver a very random picture. However, while you may think, the network has a lot of parameters that we are able to tweak, plus the objective is to locate a placing of these parameters that makes samples created from random codes look like the teaching info.Pro