Tell2Adapt: A Unified Framework for Source Free Unsupervised Domain Adaptation via Vision Foundation Model

Yulong Shi, Shijie Li, Ziyi Li, Lin Qi.
IEEE/CVF Conference on Computer Vision and Pattern Recognition (CVPR), June 2026. CCF-A, CORE A*

Abstract

Tell2Adapt presents a unified framework for source-free unsupervised domain adaptation (SFUDA) leveraging vision foundation models. The approach enables effective domain adaptation without accessing source data, utilizing the rich representations of large-scale pretrained models to bridge domain gaps across diverse tasks.

Method Overview

The framework exploits the generalization capability of vision foundation models to perform adaptation in a source-free setting. By leveraging the intrinsic knowledge embedded in these models, Tell2Adapt achieves robust cross-domain performance without requiring access to the original training data.

Resources

• Paper (PDF)
• Code (GitHub)

BibTeX

@InProceedings{Shi_2026_CVPR,
    author    = {Shi, Yulong and Li, Shijie and Li, Ziyi and Qi, Lin},
    title     = {Tell2Adapt: A Unified Framework for Source Free Unsupervised Domain Adaptation via Vision Foundation Model},
    booktitle = {Proceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition (CVPR)},
    month     = {June},
    year      = {2026},
    pages     = {6941-6950}
}