Unsupervised Domain Adaptation excels in transferring predictive models from a labeled source domain to an unlabeled target domain. However, acquiring sufficient source domain samples in specific real-world applicatio...
详细信息
Unsupervised Domain Adaptation excels in transferring predictive models from a labeled source domain to an unlabeled target domain. However, acquiring sufficient source domain samples in specific real-world applications is challenging. To address this issue, previous work introduced few-shot unsupervised domain adaptation to explore domain adaptation methods with lower requirements for the number of source domain samples. Despite progress, they still require a relatively higher number of source domain samples and exhibit significantly lower performance than methods using ample source domain samples. In this paper, we explore amore realistic and challenging scenario for few-shot unsupervised domain adaptation, where the source domain contains only a few samples per category. To extract more information from this limited data, we introduce the vision-language pre-trained model CLIP as the backbone network and propose a prompt-guided prototypealignment network. Specifically, we use category text features obtained from domain-shared soft prompts as class-specific prototypes and align cross-domain image features with these shared prototypes. During the alignment process, to reduce the impact of erroneous information in pseudo-labels, we design a sample weighting method based on truncated Laplace distribution and an alignment method that mines implicit negative information in pseudo-labels based on complementary labels. Ultimately, experiments conducted on several domain adaptation benchmark datasets demonstrate that our method offers significant advantages in scenarios with limited source domain samples and achieves competitive performance compared to unsupervised domain adaptation methods that rely on ample labeled samples.
Unsupervised domain adaptation (UDA) aims to transfer knowledge from a labeled source domain to an unlabeled target domain. However, scarcity of labeled source domain samples in practical scenarios due to high annotat...
详细信息
ISBN:
(纸本)9789819755967;9789819755974
Unsupervised domain adaptation (UDA) aims to transfer knowledge from a labeled source domain to an unlabeled target domain. However, scarcity of labeled source domain samples in practical scenarios due to high annotation costs or difficulty in sample acquisition poses a challenge. Recent work has proposed Few-Shot Unsupervised Domain Adaptation (FUDA) to address this challenge, achieving promising results through cross-domain self-supervised learning. However, their performance still significantly falls below that of UDA methods utilizing ample labeled source domain samples, primarily due to the scarcity of supervision, which hampers the model's ability to learn feature representations and classifiers that generalize well in the target domain. To address this issue, we introduce the visual-language pre-trained model CLIP as the backbone network and propose a Prompt-Guided Multi-prototypealignment framework (PMPA) for FUDA. By learning soft prompts containing diverse semantic information and aligning both source and target domains to multiple sets of class prototypes, PMPA achieves higher performance on the target domain with fewer source domain samples. Extensive experiments on OfficeHome, VisDA-2017, and Mini-DomainNet datasets demonstrate that our approach significantly outperforms previous stateof-the-art FUDAmethods and achieves comparable performance toUDAmethods utilizing ample source domain samples.
暂无评论