Glocal Alignment for Unsupervised Domain Adaptation

Sachin Chhabra, Prabal Bijoy Dutta, Baoxin Li, Hemanth Venkateswara

Research output: Chapter in Book/Report/Conference proceedingConference contribution

Abstract

Traditional unsupervised domain adaptation methods attempt to align source and target domains globally and are agnostic to the categories of the data points. This results in an inaccurate categorical alignment and diminishes the classification performance on the target domain. In this paper, we alter existing adversarial domain alignment methods to adhere to category alignment by imputing category information. We partition the samples based on category using source labels and target pseudo labels and then apply domain alignment for every category. Our proposed modification provides a boost in performance even with a modest pseudo label estimator. We evaluate our approach on 4 popular domain alignment loss functions using object recognition and digit datasets.

Original languageEnglish (US)
Title of host publicationMULL 2021 - Proceedings of the 1st Workshop on Multimedia Understanding with Less Labeling, co-located with ACM MM 2021
PublisherAssociation for Computing Machinery, Inc
Pages45-51
Number of pages7
ISBN (Electronic)9781450386814
DOIs
StatePublished - Oct 24 2021
Event1st Workshop on Multimedia Understanding with Less Labeling, MULL 2021, co-located with ACM MM 2021 - Virtual, Online, China
Duration: Oct 24 2021 → …

Publication series

NameMULL 2021 - Proceedings of the 1st Workshop on Multimedia Understanding with Less Labeling, co-located with ACM MM 2021

Conference

Conference1st Workshop on Multimedia Understanding with Less Labeling, MULL 2021, co-located with ACM MM 2021
Country/TerritoryChina
CityVirtual, Online
Period10/24/21 → …

Keywords

  • category alignment
  • domain alignment
  • local alignment
  • unsupervised domain adaptation

ASJC Scopus subject areas

  • Computational Theory and Mathematics
  • Computer Vision and Pattern Recognition
  • Software
  • Computer Science Applications

Fingerprint

Dive into the research topics of 'Glocal Alignment for Unsupervised Domain Adaptation'. Together they form a unique fingerprint.

Cite this