Domain Generalization for Semantic Segmentation: A Survey
This is an incremental survey that addresses the challenge of domain generalization for semantic segmentation, which is critical for applications like biomedicine and automated driving, by summarizing and comparing existing research to guide future work.
This survey tackles the problem of domain generalization for semantic segmentation, where deep neural networks must perform well on unseen target domains without access to them, and it reviews existing approaches, identifies a paradigm shift towards foundation-model-based methods, and provides an extensive performance comparison showing their significant influence.
The generalization of deep neural networks to unknown domains is a major challenge despite their tremendous progress in recent years. For this reason, the dynamic area of domain generalization (DG) has emerged. In contrast to unsupervised domain adaptation, there is no access to or knowledge about the target domains, and DG methods aim to generalize across multiple different unseen target domains. Domain generalization is particularly relevant for the task semantic segmentation which is used in several areas such as biomedicine or automated driving. This survey provides a comprehensive overview of the rapidly evolving topic of domain generalized semantic segmentation. We cluster and review existing approaches and identify the paradigm shift towards foundation-model-based domain generalization. Finally, we provide an extensive performance comparison of all approaches, which highlights the significant influence of foundation models on domain generalization. This survey seeks to advance domain generalization research and inspire scientists to explore new research directions.