Unsupervised Domain Adaptation Using Generative Adversarial Networks for Semantic Segmentation of Aerial Images

Segmenting aerial images is of great potential in surveillance and scene understanding of urban areas. It provides a mean for automatic reporting of the different events that happen in inhabited areas. This remarkably promotes public safety and traffic management applications. After the wide adoptio...

Full description

Bibliographic Details
Main Authors:	Bilel Benjdira, Yakoub Bazi, Anis Koubaa, Kais Ouni
Format:	Article
Language:	English
Published:	MDPI AG 2019-06-01
Series:	Remote Sensing
Subjects:	convolutional neural networks semantic segmentation aerial imagery domain adaptation gener ative adversarial networks
Online Access:	https://www.mdpi.com/2072-4292/11/11/1369

id	doaj-25a8ef698f5d475ab3cb79f97fd0948b
record_format	Article
spelling	doaj-25a8ef698f5d475ab3cb79f97fd0948b2020-11-24T21:56:52ZengMDPI AGRemote Sensing2072-42922019-06-011111136910.3390/rs11111369rs11111369Unsupervised Domain Adaptation Using Generative Adversarial Networks for Semantic Segmentation of Aerial ImagesBilel Benjdira0Yakoub Bazi1Anis Koubaa2Kais Ouni3Robotics and internet of things Laboratory, College of Computer and Information Sciences, Prince Sultan University, Riyadh 11586, Saudi ArabiaComputer Engineering Department, College of Computer and Information Sciences, King Saud University, Riyadh 11543, Saudi ArabiaPrince Sultan University, Saudi Arabia/Gaitech Robotics, China/CISTER, INESC-TEC, ISEP, Polytechnic Institute of Porto, 4200-465 Porto, PortugalResearch Laboratory Smart Electricity & ICT, SEICT, LR18ES44, National Engineering School of Carthage, University of Carthage, Carthage 1054, TunisiaSegmenting aerial images is of great potential in surveillance and scene understanding of urban areas. It provides a mean for automatic reporting of the different events that happen in inhabited areas. This remarkably promotes public safety and traffic management applications. After the wide adoption of convolutional neural networks methods, the accuracy of semantic segmentation algorithms could easily surpass 80% if a robust dataset is provided. Despite this success, the deployment of a pretrained segmentation model to survey a new city that is not included in the training set significantly decreases accuracy. This is due to the domain shift between the source dataset on which the model is trained and the new target domain of the new city images. In this paper, we address this issue and consider the challenge of domain adaptation in semantic segmentation of aerial images. We designed an algorithm that reduces the domain shift impact using generative adversarial networks (GANs). In the experiments, we tested the proposed methodology on the International Society for Photogrammetry and Remote Sensing (ISPRS) semantic segmentation dataset and found that our method improves overall accuracy from 35% to 52% when passing from the Potsdam domain (considered as source domain) to the Vaihingen domain (considered as target domain). In addition, the method allows efficiently recovering the inverted classes due to sensor variation. In particular, it improves the average segmentation accuracy of the inverted classes due to sensor variation from 14% to 61%.https://www.mdpi.com/2072-4292/11/11/1369convolutional neural networkssemantic segmentationaerial imagerydomain adaptationgener ative adversarial networks
collection	DOAJ
language	English
format	Article
sources	DOAJ
author	Bilel Benjdira Yakoub Bazi Anis Koubaa Kais Ouni
spellingShingle	Bilel Benjdira Yakoub Bazi Anis Koubaa Kais Ouni Unsupervised Domain Adaptation Using Generative Adversarial Networks for Semantic Segmentation of Aerial Images Remote Sensing convolutional neural networks semantic segmentation aerial imagery domain adaptation gener ative adversarial networks
author_facet	Bilel Benjdira Yakoub Bazi Anis Koubaa Kais Ouni
author_sort	Bilel Benjdira
title	Unsupervised Domain Adaptation Using Generative Adversarial Networks for Semantic Segmentation of Aerial Images
title_short	Unsupervised Domain Adaptation Using Generative Adversarial Networks for Semantic Segmentation of Aerial Images
title_full	Unsupervised Domain Adaptation Using Generative Adversarial Networks for Semantic Segmentation of Aerial Images
title_fullStr	Unsupervised Domain Adaptation Using Generative Adversarial Networks for Semantic Segmentation of Aerial Images
title_full_unstemmed	Unsupervised Domain Adaptation Using Generative Adversarial Networks for Semantic Segmentation of Aerial Images
title_sort	unsupervised domain adaptation using generative adversarial networks for semantic segmentation of aerial images
publisher	MDPI AG
series	Remote Sensing
issn	2072-4292
publishDate	2019-06-01
description	Segmenting aerial images is of great potential in surveillance and scene understanding of urban areas. It provides a mean for automatic reporting of the different events that happen in inhabited areas. This remarkably promotes public safety and traffic management applications. After the wide adoption of convolutional neural networks methods, the accuracy of semantic segmentation algorithms could easily surpass 80% if a robust dataset is provided. Despite this success, the deployment of a pretrained segmentation model to survey a new city that is not included in the training set significantly decreases accuracy. This is due to the domain shift between the source dataset on which the model is trained and the new target domain of the new city images. In this paper, we address this issue and consider the challenge of domain adaptation in semantic segmentation of aerial images. We designed an algorithm that reduces the domain shift impact using generative adversarial networks (GANs). In the experiments, we tested the proposed methodology on the International Society for Photogrammetry and Remote Sensing (ISPRS) semantic segmentation dataset and found that our method improves overall accuracy from 35% to 52% when passing from the Potsdam domain (considered as source domain) to the Vaihingen domain (considered as target domain). In addition, the method allows efficiently recovering the inverted classes due to sensor variation. In particular, it improves the average segmentation accuracy of the inverted classes due to sensor variation from 14% to 61%.
topic	convolutional neural networks semantic segmentation aerial imagery domain adaptation gener ative adversarial networks
url	https://www.mdpi.com/2072-4292/11/11/1369
work_keys_str_mv	AT bilelbenjdira unsuperviseddomainadaptationusinggenerativeadversarialnetworksforsemanticsegmentationofaerialimages AT yakoubbazi unsuperviseddomainadaptationusinggenerativeadversarialnetworksforsemanticsegmentationofaerialimages AT aniskoubaa unsuperviseddomainadaptationusinggenerativeadversarialnetworksforsemanticsegmentationofaerialimages AT kaisouni unsuperviseddomainadaptationusinggenerativeadversarialnetworksforsemanticsegmentationofaerialimages
_version_	1725856749471137792

Unsupervised Domain Adaptation Using Generative Adversarial Networks for Semantic Segmentation of Aerial Images

Similar Items