Source-Free Domain Adaptation with Diffusion-Guided Source Data Generation

This paper introduces a novel approach to leverage the generalizability capability of Diffusion Models for Source-Free Domain Adaptation (DM-SFDA). Our proposed DM-SFDA method involves fine-tuning a pre-trained text-to-image diffusion model to generate source domain images using features from the target images to guide the diffusion process. Specifically, the pre-trained diffusion model is fine-tuned to generate source samples that minimize entropy and maximize confidence for the pre-trained source model. We then apply established unsupervised domain adaptation techniques to align the generated …

capability cs.ai cs.cv cs.lg data diffusion diffusion model diffusion models domain domain adaptation features fine-tuning free generate guide image image diffusion images novel paper process source data text text-to-image