· ai
[Paper] Canvas-to-Image: Compositional Image Generation with Multimodal Controls
While modern diffusion models excel at generating high-quality and diverse images, they still struggle with high-fidelity compositional and multimodal control, ...