CutDiffusion: A Simple, Fast, Cheap, and Strong Diffusion Extrapolation Method
This work addresses the need for efficient and affordable high-resolution image generation in diffusion models, representing an incremental improvement over existing patch-wise extrapolation methods.
The paper tackles the problem of adapting low-resolution diffusion models to higher resolutions by proposing CutDiffusion, a tuning-free method that simplifies and accelerates the process, achieving strong generation performance with fewer inference patches and lower GPU costs.
Transforming large pre-trained low-resolution diffusion models to cater to higher-resolution demands, i.e., diffusion extrapolation, significantly improves diffusion adaptability. We propose tuning-free CutDiffusion, aimed at simplifying and accelerating the diffusion extrapolation process, making it more affordable and improving performance. CutDiffusion abides by the existing patch-wise extrapolation but cuts a standard patch diffusion process into an initial phase focused on comprehensive structure denoising and a subsequent phase dedicated to specific detail refinement. Comprehensive experiments highlight the numerous almighty advantages of CutDiffusion: (1) simple method construction that enables a concise higher-resolution diffusion process without third-party engagement; (2) fast inference speed achieved through a single-step higher-resolution diffusion process, and fewer inference patches required; (3) cheap GPU cost resulting from patch-wise inference and fewer patches during the comprehensive structure denoising; (4) strong generation performance, stemming from the emphasis on specific detail refinement.