SciTePress - Affiliation Profile

Research.Publish.Connect.

Title:
ISBN:
Year:
Acronym:
Subject:

Advanced Search Proceedings Search

If you're looking for an exact phrase use quotation marks on text fields.

Title:
Author:
Affiliation:
Subject:

Advanced Search Papers Search

If you're looking for an exact phrase use quotation marks on text fields.

Name:
Affiliation:
Country:
Conference:
Subject:

Advanced Search Authors Search

If you're looking for an exact phrase use quotation marks on text fields.

Name:
Country:
Subject:

Advanced Search Affiliations Search

If you're looking for an exact phrase use quotation marks on text fields.

Proceedings

Proceedings Search

Title:
ISBN:
Year:
Acronym:
Subject:

Advanced Search Proceedings Search

If you're looking for an exact phrase use quotation marks on text fields.

Papers

Papers Search

Title:
Author:
Affiliation:
Subject:

Advanced Search Papers Search

If you're looking for an exact phrase use quotation marks on text fields.

Authors

Authors Search

Name:
Affiliation:
Country:
Conference:
Subject:

Advanced Search Authors Search

If you're looking for an exact phrase use quotation marks on text fields.

Advanced Search

Affiliation

Univ. Lyon, École Centrale de Lyon, CNRS, INSA Lyon, Univ. Claude Bernard Lyon 1, Univ. Louis Lumière Lyon 2, LIRIS, UMR5205, 69134 Ecully, France

Context (please use complete words only)

Field

Sort by

1 Papers

Learning Less Generalizable Patterns for Better Test-Time Adaptation

Topics: Deep Learning for Visual Understanding ; Transfer Learning; Domain Adaptation

Authors: Thomas Duboudin ¹ , Emmanuel Dellandréa ¹ , Corentin Abgrall ² , Gilles Hénaff ² and Liming Chen ¹

Affiliations: ¹ Univ. Lyon, École Centrale de Lyon, CNRS, INSA Lyon, Univ. Claude Bernard Lyon 1, Univ. Louis Lumière Lyon 2, LIRIS, UMR5205, 69134 Ecully, France , ² Thales LAS France SAS, 78990 Élancourt, France

Digital Object Identifier: 10.5220/0011893800003417

Proceedings of VISAPP 2023/VISIGRAPP, ISBN: 978-989-758-634-7

Quick Abstract

Deep neural networks often fail to generalize outside of their training distribution, particularly when only a single data domain is available during training. While test-time adaptation has yielded encouraging results in this setting, we argue that to reach further improvements, these approaches should be combined with training procedure modifications aiming to learn a more diverse set of patterns. Indeed, test-time adaptation methods usually have to rely on a limited representation because of the shortcut learning phenomenon: only a subset of the available predictive patterns is learned with standard training. In this paper, we first show that the combined use of existing training-time strategies and test-time batch normalization, a simple adaptation method, does not always improve upon the test-time adaptation alone on the PACS benchmark. Furthermore, experiments on Office-Home show that very few training-time methods improve upon standard training, with or without test-time batch normalization. Therefore, we propose a novel approach that mitigates the shortcut learning behavior by having an additional classification branch learn less predictive and generalizable patterns. Our experiments show that our method improves upon the state-of-the-art results on both benchmarks and benefits the most to test-time batch normalization.