Additional comparison results to SinCUT [22], a GAN-based image-to-image translation method trained on a single input pair (Fig. 8).