Consistent text-to-image generation via scene de-contextualization

DiffMRe-assign community