Figure 3.

The Sketch Generation Module. The system is composed of three main components: the training dataset of real faces ($X_{real}$), the training dataset of generated faces ($X_{fake}$) and the generated sketches (Y). The Generator network $G_{XY}$ converts real face images ($X_{real}$) into sketch images (Y), while the inverse Generator network $G_{YX}$ reconstructs the sketch images (Y) back into generated face images ($X_{fake}$).