1 paper
Lin Zhu, Xinbing Wang, Chenghu Zhou +2
Given a style-reference image as the additional image condition, text-to-image diffusion models have demonstrated impressive capabilities in generating images that possess the cont…