This is a small model which could work as a good example of a project involving a generative adversarial network (GAN). The model aims to generate new portraits based on 128 input features. The provided network has been trained on 9376 different portraits from the lfw dataset. Below an example of the output could be seen displaying generated portrait combined with the given input. The total training time took 20 epochs.
Probably bacause of the dataset, the network seems to favor old dudes in suits