arXivDaily arXiv每日学术速递 周一至周五更新
arXiv周末暂无论文更新,休息一下吧,周末愉快~~
arXiv 2407.21311cs.CVcs.AIcs.LG
收藏

基于自监督视觉Transformer与协同跨域对齐的高效无监督域适应

Efficient unsupervised domain adaptation via self-supervised vision transformer and synergistic cross-domain alignment

University of Windsor(温莎大学);Queens University(女王大学)

Ali Abedi, Q. M. Jonathan Wu, Ning Zhang, Farhad Pourpanah

AI总结:

提出EUDA框架,以冻结的DINOv2为特征提取器,结合SDAL损失,在多数据集上实现高效无监督域适应,可训练参数减少42%至99.7%,适配资源受限环境。

摘要:

无监督域适应(UDA)旨在缓解域偏移问题,其中带标注的源数据分布与未标注的目标数据分布存在差异。尽管近期取得了进展,现有方法通常依赖对大型骨干模型进行微调,这会导致计算成本高昂,并限制其在资源受限环境中的可扩展性。这一局限性凸显了对参数高效方法的需求,此类方法需在降低训练复杂度的同时保持强性能。诸如DINOv2之类的自监督基础模型提供了高可迁移的表征,引发了无需全量微调即可实现有效域适应的疑问。为解决该问题,我们提出了高效无监督域适应(EUDA),这是一种参数高效的框架,它将冻结的DINOv2骨干用作特征提取器,仅更新轻量瓶颈层和分类头。我们还采用了协同域对齐损失(SDAL),它结合了交叉熵(CE)与最大均值差异(MMD),以同时促进判别式学习和跨域对齐。在Office-Home、Office-31、VisDA-2017和DomainNet上的实验结果表明,EUDA在不同域复杂度下均达到了具有竞争力的性能,同时可训练参数数量减少了42%至99.7%。这些结果表明,所提方法适用于资源受限和分布式环境。

英文摘要:

Unsupervised domain adaptation (UDA) aims to mitigate domain shift, where the distribution of labeled source data differs from that of unlabeled target data. Despite recent advances, existing methods often rely on fine-tuning large backbone models, which leads to high computational cost and limits scalability in resource-constrained environments. This limitation highlights the need for parameter-efficient approaches that maintain strong performance with reduced training complexity. Self-supervised foundation models such as DINOv2 provide highly transferable representations and raise the question of whether effective domain adaptation can be achieved without full fine-tuning. To address this question, we propose Efficient Unsupervised Domain Adaptation (EUDA), a parameter-efficient framework that leverages a frozen DINOv2 backbone as a feature extractor and updates only a lightweight bottleneck and classification head. We also adopt a synergistic domain alignment loss (SDAL), which combines cross-entropy (CE) and maximum mean discrepancy (MMD) to promote both discriminative learning and cross-domain alignment. Experimental results on Office-Home, Office-31, VisDA-2017, and DomainNet demonstrate that EUDA achieves competitive performance across diverse domain complexities, while reducing the number of trainable parameters by 42 to 99.7%. These results show the suitability of the proposed method for resource-constrained and distributed environments.

机构、总结、中文摘要均由AI分析而来,可能存在错误,请以论文正文为准。

↑