Scholarlink

<p>The interpretation of traditional architectural styles usually relies solely on the architecture itself, lacking a collaborative interpretation of multi-level features. In this paper, a hierarchical framework is proposed using multi-source data. First, the natural environment is analyzed through terrain, water, and vegetation using a digital elevation model and remote sensing images. Secondly, architectural features are analyzed by using the proposed dual-branch network model and Grad-CAM method with remote sensing images and online architectural landscape images. Thirdly, architectural landscapes are extracted by using semantic segmentation model and T-SNE method with online architectural landscape images. The hierarchical framework systematically reveals the inherent coupling relationship among natural environment, architectural features, and architectural landscapes from multiple levels. Based on traditional architecture dataset, the proposed dual-branch model achieves the highest average value of recall (89.73%), precision (90.37%), and F1-score (89.60%). The framework can provide technical support for the protection of Chinese traditional architectural cultural heritage.</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

Identifying and interpreting traditional architectural style characteristics based on deep learning

  • 期刊论文

  • 出版模式：

    开放获取

  • 发表日期：

    2026年7月10日

查看全文
  • Changyao Chen,
  • Shaodan Li,
  • Qi Yao,
  • Yincui Hu,
  • Ziyue Fan

摘要

The interpretation of traditional architectural styles usually relies solely on the architecture itself, lacking a collaborative interpretation of multi-level features. In this paper, a hierarchical framework is proposed using multi-source data. First, the natural environment is analyzed through terrain, water, and vegetation using a digital elevation model and remote sensing images. Secondly, architectural features are analyzed by using the proposed dual-branch network model and Grad-CAM method with remote sensing images and online architectural landscape images. Thirdly, architectural landscapes are extracted by using semantic segmentation model and T-SNE method with online architectural landscape images. The hierarchical framework systematically reveals the inherent coupling relationship among natural environment, architectural features, and architectural landscapes from multiple levels. Based on traditional architecture dataset, the proposed dual-branch model achieves the highest average value of recall (89.73%), precision (90.37%), and F1-score (89.60%). The framework can provide technical support for the protection of Chinese traditional architectural cultural heritage.