Sparse point annotations for remote sensing image segmentation
Abstract In the realm of deep learning, fine-grained semantic segmentation of Remote Sensing Images (RSIs) requires densely annotated pixel samples. However, acquiring such precise labels for training often incurs substantial financial and human costs. While point annotations are easier to acquire t...
Saved in:
| Main Authors: | , , , , , |
|---|---|
| Format: | Article |
| Language: | English |
| Published: |
Nature Portfolio
2025-07-01
|
| Series: | Scientific Reports |
| Subjects: | |
| Online Access: | https://doi.org/10.1038/s41598-025-12969-6 |
| Tags: |
Add Tag
No Tags, Be the first to tag this record!
|
| _version_ | 1849234497843757056 |
|---|---|
| author | Sixian Chan Wangjie Zhou Yanjing Lei Chao Li Jie Hu Feng Hong |
| author_facet | Sixian Chan Wangjie Zhou Yanjing Lei Chao Li Jie Hu Feng Hong |
| author_sort | Sixian Chan |
| collection | DOAJ |
| description | Abstract In the realm of deep learning, fine-grained semantic segmentation of Remote Sensing Images (RSIs) requires densely annotated pixel samples. However, acquiring such precise labels for training often incurs substantial financial and human costs. While point annotations are easier to acquire than pixel-wise annotations, they lack detailed contour information and spatial coverage. To address this issue, we propose the Point-Based Expand Network (PENet) for Remote Sensing Semantic Segmentation (RSSS). PENet leverages dynamic label expansion guided by high-dimensional semantic feature similarity. To compensate for missing structural cues, a dedicated Segment Anything Model (SAM) branch generates supplementary point-based pseudo-labels that help recover object boundaries and sizes. These SAM-generated labels are then used as anchors in the pseudo-generation branch, which dynamically expands supervision signals by evaluating feature-space similarities. This synergistic mechanism allows PENet to progressively refine semantic labels despite sparse supervision. To better capture spatial information across channels, we integrate the Efficient Multi-scale Attention (EMA) module, which enables dynamic label adjustment and enhances self-supervised learning. We validate the effectiveness of the proposed framework through extensive experiments on the Potsdam and Vaihingen datasets. The results demonstrate the strong potential of point annotations for effective and scalable semantic segmentation of RSIs. |
| format | Article |
| id | doaj-art-e186cc2da8c04b16948b538f0bd4ad35 |
| institution | Kabale University |
| issn | 2045-2322 |
| language | English |
| publishDate | 2025-07-01 |
| publisher | Nature Portfolio |
| record_format | Article |
| series | Scientific Reports |
| spelling | doaj-art-e186cc2da8c04b16948b538f0bd4ad352025-08-20T04:03:07ZengNature PortfolioScientific Reports2045-23222025-07-0115111710.1038/s41598-025-12969-6Sparse point annotations for remote sensing image segmentationSixian Chan0Wangjie Zhou1Yanjing Lei2Chao Li3Jie Hu4Feng Hong5The College of Computer Science and Technology at Zhejiang University of TechnologyThe College of Computer Science and Technology at Zhejiang University of TechnologyThe College of Computer Science and Technology at Zhejiang University of TechnologyZhijiang College, Zhejiang University of TechnologyThe Key Laboratory of Intelligent Informatics for Safety & Emergency of Zhejiang Province at Wenzhou UniversitySchool of Information Science and Technology, Zhejiang Shuren UniversityAbstract In the realm of deep learning, fine-grained semantic segmentation of Remote Sensing Images (RSIs) requires densely annotated pixel samples. However, acquiring such precise labels for training often incurs substantial financial and human costs. While point annotations are easier to acquire than pixel-wise annotations, they lack detailed contour information and spatial coverage. To address this issue, we propose the Point-Based Expand Network (PENet) for Remote Sensing Semantic Segmentation (RSSS). PENet leverages dynamic label expansion guided by high-dimensional semantic feature similarity. To compensate for missing structural cues, a dedicated Segment Anything Model (SAM) branch generates supplementary point-based pseudo-labels that help recover object boundaries and sizes. These SAM-generated labels are then used as anchors in the pseudo-generation branch, which dynamically expands supervision signals by evaluating feature-space similarities. This synergistic mechanism allows PENet to progressively refine semantic labels despite sparse supervision. To better capture spatial information across channels, we integrate the Efficient Multi-scale Attention (EMA) module, which enables dynamic label adjustment and enhances self-supervised learning. We validate the effectiveness of the proposed framework through extensive experiments on the Potsdam and Vaihingen datasets. The results demonstrate the strong potential of point annotations for effective and scalable semantic segmentation of RSIs.https://doi.org/10.1038/s41598-025-12969-6Point labelSemantic segmentationWeakly supervisionRemote sensing image |
| spellingShingle | Sixian Chan Wangjie Zhou Yanjing Lei Chao Li Jie Hu Feng Hong Sparse point annotations for remote sensing image segmentation Scientific Reports Point label Semantic segmentation Weakly supervision Remote sensing image |
| title | Sparse point annotations for remote sensing image segmentation |
| title_full | Sparse point annotations for remote sensing image segmentation |
| title_fullStr | Sparse point annotations for remote sensing image segmentation |
| title_full_unstemmed | Sparse point annotations for remote sensing image segmentation |
| title_short | Sparse point annotations for remote sensing image segmentation |
| title_sort | sparse point annotations for remote sensing image segmentation |
| topic | Point label Semantic segmentation Weakly supervision Remote sensing image |
| url | https://doi.org/10.1038/s41598-025-12969-6 |
| work_keys_str_mv | AT sixianchan sparsepointannotationsforremotesensingimagesegmentation AT wangjiezhou sparsepointannotationsforremotesensingimagesegmentation AT yanjinglei sparsepointannotationsforremotesensingimagesegmentation AT chaoli sparsepointannotationsforremotesensingimagesegmentation AT jiehu sparsepointannotationsforremotesensingimagesegmentation AT fenghong sparsepointannotationsforremotesensingimagesegmentation |