As we mentioned in the previous section, the generator network is heavily inspired by the architecture of U-Net. The architecture of U-Net is almost the same as that of an auto-encoder network. One major difference between them is that the U-Net network has skip connections between the layers in the encoder, and the decoder parts of the generator network and auto-encoder doesn't have skip connections. The U-Net network consists of two networks: the encoder network and the decoder network. The following diagram illustrates the architecture of U-Net at a basic level:

The previous diagram should give you an understanding of the architecture of a U-Net. As you can see, the output of the first layer is directly merged with the last layer. The output of the second layer is merged with the second last layer and so on. If n is the total number of layers, there are skip connections between the i^th layer in the encoder network and the (n-i)th layer in the decoder network. The i^th layer can be any layer among these layers. Let's take a closer look at both of the networks one by one.