Learning Domain Generalized Remote Sensing Image Segmentation by Multiscale Instance Disentanglement

Remote sensing image segmentation is a fundamental task in Earth observation. Rapid development has been made in the past decade owing to the deep learning techniques. Most of the existing methods assume that the training and inference remote sensing images hold the identical and independent distrib...

Full description

Saved in:
Bibliographic Details
Main Authors: Jie Luo, Tianwen Luo, Maoyang Wang, Linyi Li, Wen Zhang, Lingkui Meng
Format: Article
Language:English
Published: IEEE 2025-01-01
Series:IEEE Journal of Selected Topics in Applied Earth Observations and Remote Sensing
Subjects:
Online Access:https://ieeexplore.ieee.org/document/10999045/
Tags: Add Tag
No Tags, Be the first to tag this record!
_version_ 1850231985499275264
author Jie Luo
Tianwen Luo
Maoyang Wang
Linyi Li
Wen Zhang
Lingkui Meng
author_facet Jie Luo
Tianwen Luo
Maoyang Wang
Linyi Li
Wen Zhang
Lingkui Meng
author_sort Jie Luo
collection DOAJ
description Remote sensing image segmentation is a fundamental task in Earth observation. Rapid development has been made in the past decade owing to the deep learning techniques. Most of the existing methods assume that the training and inference remote sensing images hold the identical and independent distribution. However, in practical cases, the pretrained segmentation model has to deal with remote sensing images that have not seen before. In this article, we advance this bottleneck by introducing the domain generalization (DG) paradigm to remote sensing image segmentation. It does not need to involve any images from the target domain during training, compared with the existing domain adaptation (DA) paradigm. This article focuses on the key challenges from the various instance size and dramatic image appearance from cross-domain remote sensing images, and proposes a novel multiscale instance disentanglement learning scheme. Technically, it consists of three key components, namely, multiscale instance encoding (MSIE), domain-invariant instance disentanglement (DII), and domain generalized semantic decoding (SD). The proposed MSIE module incorporates both depth convolution and multiscale representing, so as to learn robust semantic representation despite the cross-domain scale variation. The DII module, on the other hand, helps eliminate the impact of cross-domain styles. Finally, the domain generalized SD fuses the style-invariant representation and outputs the per-pixel semantic prediction. Extensive experiments on four cross-domain settings show its effectiveness over existing DA and DG methods from both remote sensing and computer vision community, yielding an mean intersection over union improvement of 3.3% and 6.0% on the Postdam and Vaihigen unseen target domain over the second best, respectively.
format Article
id doaj-art-55c96c2e6dfc43c4b2cb1120f55419c7
institution OA Journals
issn 1939-1404
2151-1535
language English
publishDate 2025-01-01
publisher IEEE
record_format Article
series IEEE Journal of Selected Topics in Applied Earth Observations and Remote Sensing
spelling doaj-art-55c96c2e6dfc43c4b2cb1120f55419c72025-08-20T02:03:19ZengIEEEIEEE Journal of Selected Topics in Applied Earth Observations and Remote Sensing1939-14042151-15352025-01-0118135381355310.1109/JSTARS.2025.356885310999045Learning Domain Generalized Remote Sensing Image Segmentation by Multiscale Instance DisentanglementJie Luo0https://orcid.org/0009-0006-6847-6939Tianwen Luo1Maoyang Wang2Linyi Li3https://orcid.org/0000-0002-2185-8407Wen Zhang4https://orcid.org/0000-0001-5576-5973Lingkui Meng5https://orcid.org/0000-0002-4097-3080School of Remote Sensing and Information Engineering, Wuhan University, Wuhan, ChinaGuiZhou Water and Power Survey-Design Institute Company Ltd., Guiyang, ChinaGuiZhou Water and Power Survey-Design Institute Company Ltd., Guiyang, ChinaSchool of Remote Sensing and Information Engineering, Wuhan University, Wuhan, ChinaSchool of Remote Sensing and Information Engineering, Wuhan University, Wuhan, ChinaSchool of Remote Sensing and Information Engineering, Wuhan University, Wuhan, ChinaRemote sensing image segmentation is a fundamental task in Earth observation. Rapid development has been made in the past decade owing to the deep learning techniques. Most of the existing methods assume that the training and inference remote sensing images hold the identical and independent distribution. However, in practical cases, the pretrained segmentation model has to deal with remote sensing images that have not seen before. In this article, we advance this bottleneck by introducing the domain generalization (DG) paradigm to remote sensing image segmentation. It does not need to involve any images from the target domain during training, compared with the existing domain adaptation (DA) paradigm. This article focuses on the key challenges from the various instance size and dramatic image appearance from cross-domain remote sensing images, and proposes a novel multiscale instance disentanglement learning scheme. Technically, it consists of three key components, namely, multiscale instance encoding (MSIE), domain-invariant instance disentanglement (DII), and domain generalized semantic decoding (SD). The proposed MSIE module incorporates both depth convolution and multiscale representing, so as to learn robust semantic representation despite the cross-domain scale variation. The DII module, on the other hand, helps eliminate the impact of cross-domain styles. Finally, the domain generalized SD fuses the style-invariant representation and outputs the per-pixel semantic prediction. Extensive experiments on four cross-domain settings show its effectiveness over existing DA and DG methods from both remote sensing and computer vision community, yielding an mean intersection over union improvement of 3.3% and 6.0% on the Postdam and Vaihigen unseen target domain over the second best, respectively.https://ieeexplore.ieee.org/document/10999045/Domain generalization (DG)instance disentanglementmultiscale representationremote sensing image segmentation
spellingShingle Jie Luo
Tianwen Luo
Maoyang Wang
Linyi Li
Wen Zhang
Lingkui Meng
Learning Domain Generalized Remote Sensing Image Segmentation by Multiscale Instance Disentanglement
IEEE Journal of Selected Topics in Applied Earth Observations and Remote Sensing
Domain generalization (DG)
instance disentanglement
multiscale representation
remote sensing image segmentation
title Learning Domain Generalized Remote Sensing Image Segmentation by Multiscale Instance Disentanglement
title_full Learning Domain Generalized Remote Sensing Image Segmentation by Multiscale Instance Disentanglement
title_fullStr Learning Domain Generalized Remote Sensing Image Segmentation by Multiscale Instance Disentanglement
title_full_unstemmed Learning Domain Generalized Remote Sensing Image Segmentation by Multiscale Instance Disentanglement
title_short Learning Domain Generalized Remote Sensing Image Segmentation by Multiscale Instance Disentanglement
title_sort learning domain generalized remote sensing image segmentation by multiscale instance disentanglement
topic Domain generalization (DG)
instance disentanglement
multiscale representation
remote sensing image segmentation
url https://ieeexplore.ieee.org/document/10999045/
work_keys_str_mv AT jieluo learningdomaingeneralizedremotesensingimagesegmentationbymultiscaleinstancedisentanglement
AT tianwenluo learningdomaingeneralizedremotesensingimagesegmentationbymultiscaleinstancedisentanglement
AT maoyangwang learningdomaingeneralizedremotesensingimagesegmentationbymultiscaleinstancedisentanglement
AT linyili learningdomaingeneralizedremotesensingimagesegmentationbymultiscaleinstancedisentanglement
AT wenzhang learningdomaingeneralizedremotesensingimagesegmentationbymultiscaleinstancedisentanglement
AT lingkuimeng learningdomaingeneralizedremotesensingimagesegmentationbymultiscaleinstancedisentanglement