How to find deepfakes' structure mentioned in this paper? i can't find any paper or materials,code seems too difficult,can't reflect the simple model structure at all

[qym77]