Scholarlink

<p>Living conservation of architectural heritage relies on in-depth exploration of its cultural connotations and formation mechanisms. Catholic churches built before 1949 in the Sichuan-Chongqing region are crucial witnesses of Sino-Western cultural integration but lack systematic and precise stylistic classification and influencing factor analysis. This study introduced deep learning for automatic facade style classification and refined feature extraction, combined with Geodetector for quantitative analysis of spatial distribution influencing factors. Results showed: (1) ResNet50 and ViT achieved 96.22% and 95.32% classification accuracy, respectively; (2) Eight stylistic types were identified, featuring Chinese plaques as dominant localized elements in Western-style churches and arched doorways as predominant Western elements in Chinese-style churches; (3) Buddhist temple distribution with a q-value of 0.3096 and diocesan case number with a q-value of 0.3378 were the primary influencing factors for Chinese-style and Western-style churches, respectively. This study supports living conservation and value interpretation of Sino-Western integrated architectural 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

Church facade styles and influencing factors in Sichuan-Chongqing region based on deep learning

  • 期刊论文

  • 出版模式：

    开放获取

  • 发表日期：

    2026年5月1日

查看全文
  • Qin Lai,
  • Beibei Zhang,
  • Xinkai Li,
  • Huizhen Zhang,
  • Yunzhang Li

摘要

Living conservation of architectural heritage relies on in-depth exploration of its cultural connotations and formation mechanisms. Catholic churches built before 1949 in the Sichuan-Chongqing region are crucial witnesses of Sino-Western cultural integration but lack systematic and precise stylistic classification and influencing factor analysis. This study introduced deep learning for automatic facade style classification and refined feature extraction, combined with Geodetector for quantitative analysis of spatial distribution influencing factors. Results showed: (1) ResNet50 and ViT achieved 96.22% and 95.32% classification accuracy, respectively; (2) Eight stylistic types were identified, featuring Chinese plaques as dominant localized elements in Western-style churches and arched doorways as predominant Western elements in Chinese-style churches; (3) Buddhist temple distribution with a q-value of 0.3096 and diocesan case number with a q-value of 0.3378 were the primary influencing factors for Chinese-style and Western-style churches, respectively. This study supports living conservation and value interpretation of Sino-Western integrated architectural heritage.