TPL-DA: A Novel Threshold-Free Pseudolabel Learning Framework for Domain Adaptive Semantic Segmentation of High-Resolution Remote Sensing Images
Semantic segmentation techniques for remote sensing scene understanding have significantly advanced, enhancing the refined Earth observation. However, most methods highly depend on extensive annotated data, leading to performance deterioration in complex high-resolution remote sensing cross-domain s...
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/10758236/ |
| Tags: |
Add Tag
No Tags, Be the first to tag this record!
|
| _version_ | 1850053511587299328 |
|---|---|
| author | Yan Ren Jie Long Xiaowen Gao Ming Zhang Guoqing Liu Nan Su |
| author_facet | Yan Ren Jie Long Xiaowen Gao Ming Zhang Guoqing Liu Nan Su |
| author_sort | Yan Ren |
| collection | DOAJ |
| description | Semantic segmentation techniques for remote sensing scene understanding have significantly advanced, enhancing the refined Earth observation. However, most methods highly depend on extensive annotated data, leading to performance deterioration in complex high-resolution remote sensing cross-domain scenes, where variations in image conditions and environments are prevalent. Domain adaptive semantic segmentation (DASS) has been proposed to mitigate the reliance on dense and costly annotations, typically using stagewise training. This article addresses three key challenges in existing DASS methods: 1) insufficient warmup training, limiting potential performance gains; 2) rigid pseudolabel threshold settings in self-training (ST) result in performance bottlenecks; 3) entropy-based prediction bias alone fails to effectively identify high-confidence noise early in ST. To address these issues, we propose a novel threshold-free pseudolabel learning framework, TPL-DA. During the warmup stage, we introduce a multiview bidirectional consistency learning mechanism within a teacher–student architecture. This mechanism employs a bias-free data augmentation strategy, fostering consistent bidirectional predictions in teacher–student networks, thereby enhancing domain generalization and feature robustness. Our multiscale context-enhanced prediction module further amplifies this. In the ST stage, we propose a dynamic threshold-free pseudolabel learning strategy that utilizes well-aligned feature prototypes in the feature space to guide pseudolabel generation in the probability space, eliminating the threshold constraints. In addition, we model uncertainty using relative entropy and incorporate it into the optimization objective to manage high-confidence noise. Extensive experiments on the LoveDA, Potsdam, and Vaihingen datasets demonstrate that TPL-DA consistently outperforms existing methods and popular benchmarks, significantly enhancing DASS performance across diverse cross-domain scenes. |
| format | Article |
| id | doaj-art-e034f4d7eac74ff18fe35f7505e99c44 |
| institution | DOAJ |
| 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-e034f4d7eac74ff18fe35f7505e99c442025-08-20T02:52:30ZengIEEEIEEE Journal of Selected Topics in Applied Earth Observations and Remote Sensing1939-14042151-15352025-01-01181926194510.1109/JSTARS.2024.350207510758236TPL-DA: A Novel Threshold-Free Pseudolabel Learning Framework for Domain Adaptive Semantic Segmentation of High-Resolution Remote Sensing ImagesYan Ren0https://orcid.org/0000-0001-6369-2313Jie Long1https://orcid.org/0009-0007-2466-2497Xiaowen Gao2https://orcid.org/0009-0001-4256-0159Ming Zhang3https://orcid.org/0000-0003-2638-9002Guoqing Liu4https://orcid.org/0009-0007-5746-960XNan Su5https://orcid.org/0009-0002-3056-4950School of Automation and Electrical Engineering, Inner Mongolia University of Science and Technology, Baotou, Inner Mongolia, ChinaSchool of Automation and Electrical Engineering, Inner Mongolia University of Science and Technology, Baotou, Inner Mongolia, ChinaSchool of Automation and Electrical Engineering, Inner Mongolia University of Science and Technology, Baotou, Inner Mongolia, ChinaSchool of Automation and Electrical Engineering, Inner Mongolia University of Science and Technology, Baotou, Inner Mongolia, ChinaSchool of Automation and Electrical Engineering, Inner Mongolia University of Science and Technology, Baotou, Inner Mongolia, ChinaSchool of Automation and Electrical Engineering, Inner Mongolia University of Science and Technology, Baotou, Inner Mongolia, ChinaSemantic segmentation techniques for remote sensing scene understanding have significantly advanced, enhancing the refined Earth observation. However, most methods highly depend on extensive annotated data, leading to performance deterioration in complex high-resolution remote sensing cross-domain scenes, where variations in image conditions and environments are prevalent. Domain adaptive semantic segmentation (DASS) has been proposed to mitigate the reliance on dense and costly annotations, typically using stagewise training. This article addresses three key challenges in existing DASS methods: 1) insufficient warmup training, limiting potential performance gains; 2) rigid pseudolabel threshold settings in self-training (ST) result in performance bottlenecks; 3) entropy-based prediction bias alone fails to effectively identify high-confidence noise early in ST. To address these issues, we propose a novel threshold-free pseudolabel learning framework, TPL-DA. During the warmup stage, we introduce a multiview bidirectional consistency learning mechanism within a teacher–student architecture. This mechanism employs a bias-free data augmentation strategy, fostering consistent bidirectional predictions in teacher–student networks, thereby enhancing domain generalization and feature robustness. Our multiscale context-enhanced prediction module further amplifies this. In the ST stage, we propose a dynamic threshold-free pseudolabel learning strategy that utilizes well-aligned feature prototypes in the feature space to guide pseudolabel generation in the probability space, eliminating the threshold constraints. In addition, we model uncertainty using relative entropy and incorporate it into the optimization objective to manage high-confidence noise. Extensive experiments on the LoveDA, Potsdam, and Vaihingen datasets demonstrate that TPL-DA consistently outperforms existing methods and popular benchmarks, significantly enhancing DASS performance across diverse cross-domain scenes.https://ieeexplore.ieee.org/document/10758236/Domain adaptive semantic segmentation (DASS)high-resolution remote sensing (HRRS)self-training (ST)thresholduncertainty estimation |
| spellingShingle | Yan Ren Jie Long Xiaowen Gao Ming Zhang Guoqing Liu Nan Su TPL-DA: A Novel Threshold-Free Pseudolabel Learning Framework for Domain Adaptive Semantic Segmentation of High-Resolution Remote Sensing Images IEEE Journal of Selected Topics in Applied Earth Observations and Remote Sensing Domain adaptive semantic segmentation (DASS) high-resolution remote sensing (HRRS) self-training (ST) threshold uncertainty estimation |
| title | TPL-DA: A Novel Threshold-Free Pseudolabel Learning Framework for Domain Adaptive Semantic Segmentation of High-Resolution Remote Sensing Images |
| title_full | TPL-DA: A Novel Threshold-Free Pseudolabel Learning Framework for Domain Adaptive Semantic Segmentation of High-Resolution Remote Sensing Images |
| title_fullStr | TPL-DA: A Novel Threshold-Free Pseudolabel Learning Framework for Domain Adaptive Semantic Segmentation of High-Resolution Remote Sensing Images |
| title_full_unstemmed | TPL-DA: A Novel Threshold-Free Pseudolabel Learning Framework for Domain Adaptive Semantic Segmentation of High-Resolution Remote Sensing Images |
| title_short | TPL-DA: A Novel Threshold-Free Pseudolabel Learning Framework for Domain Adaptive Semantic Segmentation of High-Resolution Remote Sensing Images |
| title_sort | tpl da a novel threshold free pseudolabel learning framework for domain adaptive semantic segmentation of high resolution remote sensing images |
| topic | Domain adaptive semantic segmentation (DASS) high-resolution remote sensing (HRRS) self-training (ST) threshold uncertainty estimation |
| url | https://ieeexplore.ieee.org/document/10758236/ |
| work_keys_str_mv | AT yanren tpldaanovelthresholdfreepseudolabellearningframeworkfordomainadaptivesemanticsegmentationofhighresolutionremotesensingimages AT jielong tpldaanovelthresholdfreepseudolabellearningframeworkfordomainadaptivesemanticsegmentationofhighresolutionremotesensingimages AT xiaowengao tpldaanovelthresholdfreepseudolabellearningframeworkfordomainadaptivesemanticsegmentationofhighresolutionremotesensingimages AT mingzhang tpldaanovelthresholdfreepseudolabellearningframeworkfordomainadaptivesemanticsegmentationofhighresolutionremotesensingimages AT guoqingliu tpldaanovelthresholdfreepseudolabellearningframeworkfordomainadaptivesemanticsegmentationofhighresolutionremotesensingimages AT nansu tpldaanovelthresholdfreepseudolabellearningframeworkfordomainadaptivesemanticsegmentationofhighresolutionremotesensingimages |