1 paper
Xin Lin, Haodong Li, Zhifei Zhang +5
Controllable text-to-image diffusion models can often follow the global layout of spatial conditions, yet still violate fine-grained structures such as object boundaries, thin cont…