Scholarlink

<p>Satellite imagery segmentation plays a crucial role in environmental monitoring and sustainable development. However, the domain gap and high annotation cost are two major challenges that limit system performance in practical environments. To address these issues, this paper proposes a confidence-aware semisupervised domain adaptation method for satellite imagery segmentation. Unlike conventional approaches that directly incorporate labeled target data into the segmentation loss, our method leverages labeled target data to refine the confidence of pseudolabels generated by a teacher model. This strategy alleviates a key bottleneck in semisupervised domain adaptation, where the model may overfit a small set of labeled data. Furthermore, we introduce a sparsity constraint to select compact features that generalize robustly across both domains. The proposed method is evaluated on the LoveDA, SyntheWorld, and Open Earth Map datasets in two scenarios. The experimental results demonstrate that using only 5% labeled target data enables our approach to achieve 98.04% and 93.78% of fully supervised performance on the LoveDA and Open Earth Map datasets, respectively. Moreover, qualitative experiments show that our method does not overfit the training data and yields reasonable results. The code for this paper is available at <a href="https://github.com/Vo-Linh/SF-CAN.git">https://github.com/Vo-Linh/SF-CAN.git</a>.</p>

错误：搜索内容不能为空，请输入英文关键词
错误：关键词超出字数限制，请精简
高级检索

Stochastic and confidence-aware network (SCAN)-based semisupervised domain adaptation for satellite imagery segmentation

  • 期刊论文

  • 出版模式：

    开放获取

  • 发表日期：

    2026年7月3日

查看全文
  • Manh-Hung Nguyen,
  • Van-Linh Vo,
  • Long-Thien Bui,
  • Chi-Cuong Vu,
  • Chin-Chung Huang

摘要

Satellite imagery segmentation plays a crucial role in environmental monitoring and sustainable development. However, the domain gap and high annotation cost are two major challenges that limit system performance in practical environments. To address these issues, this paper proposes a confidence-aware semisupervised domain adaptation method for satellite imagery segmentation. Unlike conventional approaches that directly incorporate labeled target data into the segmentation loss, our method leverages labeled target data to refine the confidence of pseudolabels generated by a teacher model. This strategy alleviates a key bottleneck in semisupervised domain adaptation, where the model may overfit a small set of labeled data. Furthermore, we introduce a sparsity constraint to select compact features that generalize robustly across both domains. The proposed method is evaluated on the LoveDA, SyntheWorld, and Open Earth Map datasets in two scenarios. The experimental results demonstrate that using only 5% labeled target data enables our approach to achieve 98.04% and 93.78% of fully supervised performance on the LoveDA and Open Earth Map datasets, respectively. Moreover, qualitative experiments show that our method does not overfit the training data and yields reasonable results. The code for this paper is available at https://github.com/Vo-Linh/SF-CAN.git.