Skip to the content.


Domain adaptation and generalization is crucial for a wide range of vision tasks, such as image classification, semantic segmentation, or video recognition, thus laying the basis for everyday applications like smart arrangement of image collections on our smartphones, web-scale image search in our brower or monitor public safety. Today, the predominant approach for domain adaptation and generalization that learn domain-invariant visual representations is adversarial learning. which specifically aims at aligning the visual representations of different domains. Moreover, self-supervised learning is also applied to domain adaptation and generalization, which is the promising direction.

In this tutorial we will provide an in-depth coverage of the recent advances in domain adaptation and generalization and also applications and future directions. Specifically, the tutorial will cover the following subjects: (1) background on domain adaptation, (2) recent advances in visual domain adaptation, (3) domain adaptation vs. domain generalization, (4) recent advances in visual domain generalization and finally, (5) we provide applications and future perspectives of domain adaptation and generalization.



Ronghang Zhu
University of Georgia
Sheng Li
University of Virginia
Xiang Yu
Prime Video, Amazon