Domain generalization results (%) in the low-data regime with a comparison of various models in different settings (fully labeled, DG, SSL, and SSDG), evaluated on miniDomainNet (Clipart: C, Painting: P, Real: R, and Sketch: S). Here, u means utilization of unlabeled data. Results are reported as mean ± standard deviation over 5 runs.

Domain generalization results (%) in the low-data regime with a comparison of various models in different settings (fully labeled, DG, SSL, and SSDG), evaluated on miniDomainNet (Clipart: C, Painting: P, Real: R, and Sketch: S). Here, u means utilization of unlabeled data...

Full description

Saved in:

Bibliographic Details
Main Author:	Sumaiya Zoha (22184571) (author)
Other Authors:	Jeong-Gun Lee (2048551) (author), Young-Woong Ko (22184574) (author)
Published:	2025
Subjects:	Biochemistry Microbiology Cell Biology Space Science Biological Sciences not elsewhere classified limited labeled data labeled source data multiple source domains learn robust representations 90 %</ b 47 %</ b 45 %</ b quality labeled data achieving robust generalization unseen target domains higher class diversity div >< p refining noisy pseudo supervised domain generalization combines adaptive thresholding cat </ b unseen domains domain generalization quality pseudo noisy pseudo supervised learning aware adaptive domain shifts robust semi pacs </ officehome </ minidomainnet </ transfer knowledge superior performance previous approaches often costly fixed thresholds extensive experiments efficient paradigm diverse set challenging problem
Tags:	Add Tag No Tags, Be the first to tag this record!

Similar Items