A domain agnostic normalization layer for unsupervised adversarial domain adaptation

R.R.F.M. Romijnders, Panagiotis Meletis, Gijs Dubbelman

Research output: Chapter in Book/Report/Conference proceedingConference contributionAcademicpeer-review

5 Citations (Scopus)

Abstract

We propose a normalization layer for unsupervised domain adaption in semantic scene segmentation. Normalization layers are known to improve convergence and generalization and are part of many state-of-the-art fully-convolutional neural networks. We show that conventional normalization layers worsen the performance of current Unsupervised Adversarial Domain Adaption (UADA), which is a method to improve network performance on unlabeled data sets and the focus of our research. Therefore, we propose a novel Domain Agnostic Normalization layer and thereby unlock the benefits of normalization layers for unsupervised adversarial domain adaptation. In our evaluation, we adapt from the synthetic GTA5 data set to the real Cityscapes data set, a common benchmark experiment, and surpass the state-of-the-art. As our normalization layer is domain agnostic at test time, we furthermore demonstrate that UADA using Domain Agnostic Normalization improves performance on unseen domains, specifically on Apolloscape and Mapillary.
Original languageEnglish
Title of host publication2019 IEEE Winter Conference on Applications of Computer Vision (WACV)
Place of PublicationPiscataway
PublisherInstitute of Electrical and Electronics Engineers
Pages1866-1875
Number of pages10
ISBN (Electronic)978-1-7281-1975-5
DOIs
Publication statusPublished - 4 Mar 2019
Event2019 IEEE Winter Conference on Applications of Computer Vision (WACV) - Waikoloa Village, United States
Duration: 7 Jan 201911 Jan 2019

Conference

Conference2019 IEEE Winter Conference on Applications of Computer Vision (WACV)
CountryUnited States
CityWaikoloa Village
Period7/01/1911/01/19

Fingerprint Dive into the research topics of 'A domain agnostic normalization layer for unsupervised adversarial domain adaptation'. Together they form a unique fingerprint.

Cite this