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

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.
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
StatePublished - 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

Network performance
Semantics
Neural networks
Experiments

Cite this

Romijnders, R. R. F. M., Meletis, P., & Dubbelman, G. (2019). A domain agnostic normalization layer for unsupervised adversarial domain adaptation. In 2019 IEEE Winter Conference on Applications of Computer Vision (WACV) (pp. 1866-1875). Piscataway: Institute of Electrical and Electronics Engineers. DOI: 10.1109/WACV.2019.00203
Romijnders, R.R.F.M. ; Meletis, Panagiotis ; Dubbelman, Gijs. / A domain agnostic normalization layer for unsupervised adversarial domain adaptation. 2019 IEEE Winter Conference on Applications of Computer Vision (WACV). Piscataway : Institute of Electrical and Electronics Engineers, 2019. pp. 1866-1875
@inproceedings{16c8ca1f454b42c4995e4176b3aabb49,
title = "A domain agnostic normalization layer for unsupervised adversarial domain adaptation",
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.",
author = "R.R.F.M. Romijnders and Panagiotis Meletis and Gijs Dubbelman",
year = "2019",
doi = "10.1109/WACV.2019.00203",
language = "English",
pages = "1866--1875",
booktitle = "2019 IEEE Winter Conference on Applications of Computer Vision (WACV)",
publisher = "Institute of Electrical and Electronics Engineers",
address = "United States",

}

Romijnders, RRFM, Meletis, P & Dubbelman, G 2019, A domain agnostic normalization layer for unsupervised adversarial domain adaptation. in 2019 IEEE Winter Conference on Applications of Computer Vision (WACV). Institute of Electrical and Electronics Engineers, Piscataway, pp. 1866-1875, 2019 IEEE Winter Conference on Applications of Computer Vision (WACV), Waikoloa Village, United States, 7/01/19. DOI: 10.1109/WACV.2019.00203

A domain agnostic normalization layer for unsupervised adversarial domain adaptation. / Romijnders, R.R.F.M.; Meletis, Panagiotis; Dubbelman, Gijs.

2019 IEEE Winter Conference on Applications of Computer Vision (WACV). Piscataway : Institute of Electrical and Electronics Engineers, 2019. p. 1866-1875.

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

TY - GEN

T1 - A domain agnostic normalization layer for unsupervised adversarial domain adaptation

AU - Romijnders,R.R.F.M.

AU - Meletis,Panagiotis

AU - Dubbelman,Gijs

PY - 2019

Y1 - 2019

N2 - 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.

AB - 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.

U2 - 10.1109/WACV.2019.00203

DO - 10.1109/WACV.2019.00203

M3 - Conference contribution

SP - 1866

EP - 1875

BT - 2019 IEEE Winter Conference on Applications of Computer Vision (WACV)

PB - Institute of Electrical and Electronics Engineers

CY - Piscataway

ER -

Romijnders RRFM, Meletis P, Dubbelman G. A domain agnostic normalization layer for unsupervised adversarial domain adaptation. In 2019 IEEE Winter Conference on Applications of Computer Vision (WACV). Piscataway: Institute of Electrical and Electronics Engineers. 2019. p. 1866-1875. Available from, DOI: 10.1109/WACV.2019.00203