DGFamba: Learning Flow Factorized State Space for Visual Domain Generalization

2025-04-10

Proceedings of the AAAI Conference on Artificial Intelligence

AI Summary

The core challenge in vision domain generalization lies in severe domain shifts induced by inter-domain style variations, while semantic content remains stable. To address this, we propose a style-content disentanglement framework built upon selective state space models (SSMs). Our method introduces flow decomposition into SSMs for the first time, constructing a differentiable implicit flow mapping and a probabilistic path alignment mechanism to explicitly separate style and content representations in latent space. By incorporating style-augmented embeddings and flow factorization, we ensure domain-invariant state embeddings. Grounded on VMamba, our approach enables robust zero-shot transfer from source domains to arbitrary unseen target domains. Extensive experiments on standard benchmarks—including PACS and Office-Home—demonstrate state-of-the-art performance, with substantial improvements in cross-domain classification accuracy.

Technology Category

• Computer Vision: Diffusion Models for Vision
• Machine Learning: Transfer, Domain Adaptation, Multi-Task Learning
• Application Domains: Humanities & Computational Social Science

Application Category

• Search and Retrieval-Augmented AI: Vertical and domain-specific search
• Web Mining and Content Analysis: Large pretrained models with web data
• Semantics and Knowledge: Methods to enhance, augment, integrate or synergize semantic models such as knowledge graphs and LLMs

Abstract

Domain generalization aims to learn a representation from the source domain, which can be generalized to arbitrary unseen target domains. A fundamental challenge for visual domain generalization is the domain gap caused by the dramatic style variation whereas the image content is stable. The realm of selective state space, exemplified by VMamba, demonstrates its global receptive field in representing the content. However, the way exploiting the domain-invariant property for selective state space is rarely explored. In this paper, we propose a novel Flow Factorized State Space model, dubbed as DGFamba, for visual domain generalization. To maintain domain consistency, we innovatively map the style-augmented and the original state embeddings by flow factorization. In this latent flow space, each state embedding from a certain style is specified by a latent probability path. By aligning these probability paths in the latent space, the state embeddings are able to represent the same content distribution regardless of the style differences. Extensive experiments conducted on various visual domain generalization settings show its state-of-the-art performance.

Problem

• Addresses domain gap in visual domain generalization
• Explores domain-invariant properties for selective state space
• Aligns probability paths to maintain content distribution consistency

Innovation

• Flow Factorized State Space model
• Map style-augmented embeddings via flow factorization
• Align probability paths in latent space