EUVIMEDEUVIMED entdecken
Open Access · CC BY-NC 4.0

Feature Alignment and Domain Adversarial Learning for Cross Domain Recognition

Sophie Williams, Lachlan Bennett, Amelia Clarke

INTERNATIONAL JOURNAL OF RESEARCH & SOCIAL SCIENCES · 25.06.2026 · 6 (1) · 205-212

DOI: 10.66238/ijrss.v6i1.107 ↗

Kurz gefasst

Fachartikel in der Zeitschrift INTERNATIONAL JOURNAL OF RESEARCH & SOCIAL SCIENCES, online erschienen am 25.06.2026. Verfasst von Sophie Williams und 2 weiteren Autor:innen. Der Volltext ist frei zugänglich (Open Access, CC BY-NC 4.0).

AbstractBibliografische AngabenZitierenVerwandte Studien

Abstract (Originalsprache)

Deep learning models often lose accuracy when training and test data come from different domains. This problem limits their use in visual tasks with changes in image style, background, lighting, and data source. This study proposes a cross-domain learning method based on feature alignment and domain-adversarial training. A domain discriminator is used to reduce the difference between source and target domains. Maximum mean discrepancy is added to measure and reduce the feature distribution gap. Experiments were conducted on Office-31 and DomainNet under single-source and multi-source settings. Compared with the baseline model, the proposed method improved target-domain accuracy by 8.9%–15.4% and reduced feature-distribution KL divergence by about 36%. In the multi-source setting, generalization performance increased by about 6.3%. These results show that joint feature alignment and domain-adversarial training can reduce domain shift while retaining useful class information. The method provides a practical way to improve target-domain recognition in medical imaging, remote sensing, industrial inspection, and autonomous perception, where labeled target-domain data are often limited.

Zitieren

APA

Williams, S., Bennett, L., & Clarke, A. (2026). Feature Alignment and Domain Adversarial Learning for Cross Domain Recognition. INTERNATIONAL JOURNAL OF RESEARCH & SOCIAL SCIENCES, 6(1), 205-212. https://doi.org/10.66238/ijrss.v6i1.107

Vancouver

Williams S, Bennett L, Clarke A. Feature Alignment and Domain Adversarial Learning for Cross Domain Recognition. INTERNATIONAL JOURNAL OF RESEARCH & SOCIAL SCIENCES. 2026;6(1):205-212. doi:10.66238/ijrss.v6i1.107

BibTeX
@article{Williams2026,
  author = {Williams, Sophie and Bennett, Lachlan and Clarke, Amelia},
  title = {Feature Alignment and Domain Adversarial Learning for Cross Domain Recognition},
  journal = {INTERNATIONAL JOURNAL OF RESEARCH & SOCIAL SCIENCES},
  year = {2026},
  volume = {6},
  number = {1},
  pages = {205--212},
  doi = {10.66238/ijrss.v6i1.107}
}
Kostenlos starten →