RectifID: Personalizing Rectified Flow with Anchored Classifier Guidance
This addresses the need for adaptable personalization in image generation across various domains, offering a training-free solution that is incremental over existing classifier guidance methods.
The paper tackles the problem of customizing diffusion models for identity-preserving image generation from user-provided references without extensive training, achieving flexible personalization using off-the-shelf image discriminators.
Customizing diffusion models to generate identity-preserving images from user-provided reference images is an intriguing new problem. The prevalent approaches typically require training on extensive domain-specific images to achieve identity preservation, which lacks flexibility across different use cases. To address this issue, we exploit classifier guidance, a training-free technique that steers diffusion models using an existing classifier, for personalized image generation. Our study shows that based on a recent rectified flow framework, the major limitation of vanilla classifier guidance in requiring a special classifier can be resolved with a simple fixed-point solution, allowing flexible personalization with off-the-shelf image discriminators. Moreover, its solving procedure proves to be stable when anchored to a reference flow trajectory, with a convergence guarantee. The derived method is implemented on rectified flow with different off-the-shelf image discriminators, delivering advantageous personalization results for human faces, live subjects, and certain objects. Code is available at https://github.com/feifeiobama/RectifID.