Closed kinredon closed 3 months ago
Thank you for your interest on our work! Since our model is targeting on domain-to-domain mapping, there is no explicit component for controlling the degree of stylization. However, as we have already shown in our manuscript, we can obtain weakly-stylized output when using lower NFE (e.g. NFE=1), and fully stylized (fully translated) output when using higher NFE.
For example, I can provide a trade-off parameter to control the degree of style transfer to the target domain style.