I’ve been looking at some other GAN-generators like Stylegan2 Ada and it’s Pytorch/Tensorflow implementations. The artworks they generate are extremely good and they’re trained on small sets of images (1-2K). E.g. the model based on Metfaces photo set generates portraits that look like real portraits and don’t have any ‘scars’ or disformed functions like most the DAE portraits have.
Are there any plans to update DAE with something similar to improve the quality of the generated artworks?