Fine-Grained Style Alignment and Class Balance for Unsupervised Domain Adaptation in Remote Sensing Image Segmentation
Unsupervised domain adaptation (UDA) is an effective method for addressing the domain shift issue between the source and target domains in high-resolution remote sensing images classification. However, existing methods still face significant challenges when dealing with substantial style discrepanci...
Saved in:
| Main Authors: | , , , |
|---|---|
| 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/11050970/ |
| Tags: |
Add Tag
No Tags, Be the first to tag this record!
|
| _version_ | 1849430999718428672 |
|---|---|
| author | Yousheng Xu Weiji Wang Wei Yao Shengzhou Xu |
| author_facet | Yousheng Xu Weiji Wang Wei Yao Shengzhou Xu |
| author_sort | Yousheng Xu |
| collection | DOAJ |
| description | Unsupervised domain adaptation (UDA) is an effective method for addressing the domain shift issue between the source and target domains in high-resolution remote sensing images classification. However, existing methods still face significant challenges when dealing with substantial style discrepancies and class imbalance. For instance, the widely used adaptive instance normalization (AdaIN) method reduces the distribution difference by transforming the style of source domain images to match that of the target domain. However, when confronted with large style variations, this method often leads to the loss of fine-grained details such as color, texture, and contrast, negatively impacting the transfer performance of the model. In addition, class imbalance causes the model to be biased toward dominant classes, resulting in decreased overall classification accuracy. To address these issues, this article proposes two innovative modules. First, we introduce the SelfAdaIN module, which precisely controls the style transformation process through an adaptive convolutional generation mechanism. This module effectively prevents the loss of fine details found in traditional AdaIN methods, achieving fine-grained style alignment between the source and target domains. Second, we propose a class balance loss function that adjusts the weight distribution of target domain samples to mitigate the influence of dominant classes and enhance the focus on rare categories, thereby alleviating the issue of class imbalance. Extensive experiments on four cross-domain remote sensing datasets demonstrate that the proposed methods significantly improve classification accuracy. |
| format | Article |
| id | doaj-art-e133de344de44c92b14e0a6c4fef5b72 |
| institution | Kabale University |
| 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-e133de344de44c92b14e0a6c4fef5b722025-08-20T03:27:47ZengIEEEIEEE Journal of Selected Topics in Applied Earth Observations and Remote Sensing1939-14042151-15352025-01-0118166631667910.1109/JSTARS.2025.358294911050970Fine-Grained Style Alignment and Class Balance for Unsupervised Domain Adaptation in Remote Sensing Image SegmentationYousheng Xu0https://orcid.org/0009-0008-8992-6315Weiji Wang1https://orcid.org/0009-0002-7385-5432Wei Yao2https://orcid.org/0000-0001-7488-5997Shengzhou Xu3https://orcid.org/0000-0003-4285-9588School of Computer Science, South-Central Minzu University, Wuhan, ChinaSchool of Computer Science, South-Central Minzu University, Wuhan, ChinaSchool of Computer Science, South-Central Minzu University, Wuhan, ChinaSchool of Computer Science, South-Central Minzu University, Wuhan, ChinaUnsupervised domain adaptation (UDA) is an effective method for addressing the domain shift issue between the source and target domains in high-resolution remote sensing images classification. However, existing methods still face significant challenges when dealing with substantial style discrepancies and class imbalance. For instance, the widely used adaptive instance normalization (AdaIN) method reduces the distribution difference by transforming the style of source domain images to match that of the target domain. However, when confronted with large style variations, this method often leads to the loss of fine-grained details such as color, texture, and contrast, negatively impacting the transfer performance of the model. In addition, class imbalance causes the model to be biased toward dominant classes, resulting in decreased overall classification accuracy. To address these issues, this article proposes two innovative modules. First, we introduce the SelfAdaIN module, which precisely controls the style transformation process through an adaptive convolutional generation mechanism. This module effectively prevents the loss of fine details found in traditional AdaIN methods, achieving fine-grained style alignment between the source and target domains. Second, we propose a class balance loss function that adjusts the weight distribution of target domain samples to mitigate the influence of dominant classes and enhance the focus on rare categories, thereby alleviating the issue of class imbalance. Extensive experiments on four cross-domain remote sensing datasets demonstrate that the proposed methods significantly improve classification accuracy.https://ieeexplore.ieee.org/document/11050970/Class balance loss (CBL)high-resolution (HR) remote sensing image (RSI) classificationSelfAdaINstyle transferunsupervised domain adaptation (UDA) |
| spellingShingle | Yousheng Xu Weiji Wang Wei Yao Shengzhou Xu Fine-Grained Style Alignment and Class Balance for Unsupervised Domain Adaptation in Remote Sensing Image Segmentation IEEE Journal of Selected Topics in Applied Earth Observations and Remote Sensing Class balance loss (CBL) high-resolution (HR) remote sensing image (RSI) classification SelfAdaIN style transfer unsupervised domain adaptation (UDA) |
| title | Fine-Grained Style Alignment and Class Balance for Unsupervised Domain Adaptation in Remote Sensing Image Segmentation |
| title_full | Fine-Grained Style Alignment and Class Balance for Unsupervised Domain Adaptation in Remote Sensing Image Segmentation |
| title_fullStr | Fine-Grained Style Alignment and Class Balance for Unsupervised Domain Adaptation in Remote Sensing Image Segmentation |
| title_full_unstemmed | Fine-Grained Style Alignment and Class Balance for Unsupervised Domain Adaptation in Remote Sensing Image Segmentation |
| title_short | Fine-Grained Style Alignment and Class Balance for Unsupervised Domain Adaptation in Remote Sensing Image Segmentation |
| title_sort | fine grained style alignment and class balance for unsupervised domain adaptation in remote sensing image segmentation |
| topic | Class balance loss (CBL) high-resolution (HR) remote sensing image (RSI) classification SelfAdaIN style transfer unsupervised domain adaptation (UDA) |
| url | https://ieeexplore.ieee.org/document/11050970/ |
| work_keys_str_mv | AT youshengxu finegrainedstylealignmentandclassbalanceforunsuperviseddomainadaptationinremotesensingimagesegmentation AT weijiwang finegrainedstylealignmentandclassbalanceforunsuperviseddomainadaptationinremotesensingimagesegmentation AT weiyao finegrainedstylealignmentandclassbalanceforunsuperviseddomainadaptationinremotesensingimagesegmentation AT shengzhouxu finegrainedstylealignmentandclassbalanceforunsuperviseddomainadaptationinremotesensingimagesegmentation |