Uni-DAD: Unified Distillation and Adaptation of Diffusion Models for Few-step Few-shot Image Generation

Researchers introduce Uni-DAD, a unified approach that combines diffusion model distillation and adaptation into a single pipeline for efficient few-shot image generation. The method achieves comparable quality to state-of-the-art methods while requiring less than 4 sampling steps, addressing the computational cost issues of traditional diffusion models.

• →Uni-DAD combines diffusion model distillation and adaptation in a single stage, eliminating the need for complex two-stage pipelines.
• →The method uses dual-domain distribution matching and multi-head GAN loss to preserve source knowledge while adapting to new domains.
• →Uni-DAD achieves state-of-the-art quality with less than 4 sampling steps, significantly improving computational efficiency.
• →The approach shows superior performance in few-shot image generation and subject-driven personalization benchmarks.
• →The unified pipeline reduces design complexity while maintaining or improving quality and diversity compared to existing methods.