Unsupervised Remote Sensing Domain Adaptation Method with Adversarial Network and Auxiliary Task
An important prerequisite when annotating the remote sensing images by machine learning is that there are enough training samples for training, but labeling the samples is very time-consuming. In this paper, we solve the problem of unsupervised learning with small sample size in remote sensing image...
Main Authors: | , , , |
---|---|
Format: | Article |
Language: | zho |
Published: |
Surveying and Mapping Press
2017-12-01
|
Series: | Acta Geodaetica et Cartographica Sinica |
Subjects: | |
Online Access: | http://html.rhhz.net/CHXB/html/2017-12-1969.htm |
id |
doaj-8fb530356e154d3590700f102831f9b7 |
---|---|
record_format |
Article |
spelling |
doaj-8fb530356e154d3590700f102831f9b72020-11-24T22:28:05ZzhoSurveying and Mapping PressActa Geodaetica et Cartographica Sinica1001-15951001-15952017-12-0146121969197710.11947/j.AGCS.2017.201702912017120291Unsupervised Remote Sensing Domain Adaptation Method with Adversarial Network and Auxiliary TaskXU Suhui0MU Xiaodong1ZHANG Xiongmei2CHAI Dong3Department of Information Engineering, Rocket Force Engineering University, Xi'an 710025, ChinaDepartment of Information Engineering, Rocket Force Engineering University, Xi'an 710025, ChinaDepartment of Information Engineering, Rocket Force Engineering University, Xi'an 710025, ChinaBeijing Aeronautical Technology Research Center, Beijing 100076, ChinaAn important prerequisite when annotating the remote sensing images by machine learning is that there are enough training samples for training, but labeling the samples is very time-consuming. In this paper, we solve the problem of unsupervised learning with small sample size in remote sensing image scene classification by domain adaptation method. A new domain adaptation framework is proposed which combines adversarial network and auxiliary task. Firstly, a novel remote sensing scene classification framework is established based on deep convolution neural networks. Secondly, a domain classifier is added to the network, in order to learn the domain-invariant features. The gradient direction of the domain loss is opposite to the label loss during the back propagation, which makes the domain predictor failed to distinguish the sample's domain. Lastly, we introduce an auxiliary task for the network, which augments the training samples and improves the generalization ability of the network. The experiments demonstrate better results in unsupervised classification with small sample sizes of remote sensing images compared to the baseline unsupervised domain adaptation approaches.http://html.rhhz.net/CHXB/html/2017-12-1969.htmremote sensing imagescene classificationdomain adaptationdeep convolutional neural networkadversarial networkmulti-task learning |
collection |
DOAJ |
language |
zho |
format |
Article |
sources |
DOAJ |
author |
XU Suhui MU Xiaodong ZHANG Xiongmei CHAI Dong |
spellingShingle |
XU Suhui MU Xiaodong ZHANG Xiongmei CHAI Dong Unsupervised Remote Sensing Domain Adaptation Method with Adversarial Network and Auxiliary Task Acta Geodaetica et Cartographica Sinica remote sensing image scene classification domain adaptation deep convolutional neural network adversarial network multi-task learning |
author_facet |
XU Suhui MU Xiaodong ZHANG Xiongmei CHAI Dong |
author_sort |
XU Suhui |
title |
Unsupervised Remote Sensing Domain Adaptation Method with Adversarial Network and Auxiliary Task |
title_short |
Unsupervised Remote Sensing Domain Adaptation Method with Adversarial Network and Auxiliary Task |
title_full |
Unsupervised Remote Sensing Domain Adaptation Method with Adversarial Network and Auxiliary Task |
title_fullStr |
Unsupervised Remote Sensing Domain Adaptation Method with Adversarial Network and Auxiliary Task |
title_full_unstemmed |
Unsupervised Remote Sensing Domain Adaptation Method with Adversarial Network and Auxiliary Task |
title_sort |
unsupervised remote sensing domain adaptation method with adversarial network and auxiliary task |
publisher |
Surveying and Mapping Press |
series |
Acta Geodaetica et Cartographica Sinica |
issn |
1001-1595 1001-1595 |
publishDate |
2017-12-01 |
description |
An important prerequisite when annotating the remote sensing images by machine learning is that there are enough training samples for training, but labeling the samples is very time-consuming. In this paper, we solve the problem of unsupervised learning with small sample size in remote sensing image scene classification by domain adaptation method. A new domain adaptation framework is proposed which combines adversarial network and auxiliary task. Firstly, a novel remote sensing scene classification framework is established based on deep convolution neural networks. Secondly, a domain classifier is added to the network, in order to learn the domain-invariant features. The gradient direction of the domain loss is opposite to the label loss during the back propagation, which makes the domain predictor failed to distinguish the sample's domain. Lastly, we introduce an auxiliary task for the network, which augments the training samples and improves the generalization ability of the network. The experiments demonstrate better results in unsupervised classification with small sample sizes of remote sensing images compared to the baseline unsupervised domain adaptation approaches. |
topic |
remote sensing image scene classification domain adaptation deep convolutional neural network adversarial network multi-task learning |
url |
http://html.rhhz.net/CHXB/html/2017-12-1969.htm |
work_keys_str_mv |
AT xusuhui unsupervisedremotesensingdomainadaptationmethodwithadversarialnetworkandauxiliarytask AT muxiaodong unsupervisedremotesensingdomainadaptationmethodwithadversarialnetworkandauxiliarytask AT zhangxiongmei unsupervisedremotesensingdomainadaptationmethodwithadversarialnetworkandauxiliarytask AT chaidong unsupervisedremotesensingdomainadaptationmethodwithadversarialnetworkandauxiliarytask |
_version_ |
1725747960840454144 |