DCGAN is initialized with random weights, so a random code plugged in the network would create a very random impression. Nonetheless, as you may think, the network has numerous parameters that we are able to tweak, and the target is to find a setting of those parameters that makes samples crea