Kundu, JN and Venkatesh, RM and Venkat, N and Revanur, A and Babu, RV (2020) Class-Incremental Domain Adaptation. In: Lecture Notes in Computer Science (including subseries Lecture Notes in Artificial Intelligence and Lecture Notes in Bioinformatics), 23 - 28 August 2020, Glasgow, pp. 53-69.
|
PDF
ECCV 2020.pdf - Published Version Download (2MB) | Preview |
Abstract
We introduce a practical Domain Adaptation (DA) paradigm called Class-Incremental Domain Adaptation (CIDA). Existing DA methods tackle domain-shift but are unsuitable for learning novel target-domain classes. Meanwhile, class-incremental (CI) methods enable learning of new classes in absence of source training data, but fail under a domain-shift without labeled supervision. In this work, we effectively identify the limitations of these approaches in the CIDA paradigm. Motivated by theoretical and empirical observations, we propose an effective method, inspired by prototypical networks, that enables classification of target samples into both shared and novel (one-shot) target classes, even under a domain-shift. Our approach yields superior performance as compared to both DA and CI methods in the CIDA paradigm.
| Item Type: | Conference Paper |
|---|---|
| Publication: | Lecture Notes in Computer Science (including subseries Lecture Notes in Artificial Intelligence and Lecture Notes in Bioinformatics) |
| Publisher: | Springer Science and Business Media Deutschland GmbH |
| Additional Information: | The copyright for this article belongs to the Authors. |
| Keywords: | Computer vision, Domain adaptation; Target class; Target domain; Training data, Learning systems |
| Department/Centre: | Division of Interdisciplinary Sciences > Computational and Data Sciences |
| Date Deposited: | 23 Jan 2023 10:11 |
| Last Modified: | 23 Jan 2023 10:11 |
| URI: | https://eprints.iisc.ac.in/id/eprint/79264 |
Actions (login required)
 |
View Item |
