


The model automatically uses a few faces as the initial training material (the model use) because it's using ffhqXXXX.Ffhq stands for Flickr-Faces-HQ Dataset, it is collected by the same guys who developed StyleGAN2. So they trained styleGan with this data and shared their model. When we type ffqhXXXX instead .pkl file, we tell to styleGan to download this pre-trained model. This is why we get faces on the early step of training.
The generated results are put into point clouds and processed with grasshopper into 3D models.
The generated results are put into point clouds and processed with grasshopper into 3D models.




