CoMoGAN: continuous model-guided image-to-image translation