Scholarlink

<p>Buddha head is a valuable cultural heritage of human civilization. Detection and counting of Buddha’s head is a very important work. Especially in the case of a large number of Buddha heads and dense, detection and counting is very necessary. This work is helpful for cultural relics conservators to quickly obtain the number of Buddha heads and carry out the next protection and repair work. However, the detection and counting of Buddha’s head usually rely on visual inspection, which is very time-consuming and labour-intensive in dense cases. In this paper, we design a new YOLOv8 network with an efficient multi-scale attention module to detect and count dense Buddha heads. Our research data comes from high-precision images: sculptures of Buddha heads in the Wanfo Hall in Fangshan District, Beijing. Through data enhancement and transformation, we have 3000 images for training. The final mean accuracy value (mAP) reached 0.92. and found that the detection effect on broken Buddha heads is still considerable. To the best of our knowledge, this paper is the first work to apply an improved YOLO deep learning model to address the target detection and counting of Buddha head, which has achieved good results.</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

Dense Buddha head object detection and counting YOLOv8 network based on multi-scale attention and data augmentation fusion

  • 期刊论文

  • 出版模式：

    开放获取

  • 发表日期：

    2025年2月22日

查看全文
  • Yang Li,
  • Yalun Wang,
  • Dong Sui,
  • Maozu Guo

摘要

Buddha head is a valuable cultural heritage of human civilization. Detection and counting of Buddha’s head is a very important work. Especially in the case of a large number of Buddha heads and dense, detection and counting is very necessary. This work is helpful for cultural relics conservators to quickly obtain the number of Buddha heads and carry out the next protection and repair work. However, the detection and counting of Buddha’s head usually rely on visual inspection, which is very time-consuming and labour-intensive in dense cases. In this paper, we design a new YOLOv8 network with an efficient multi-scale attention module to detect and count dense Buddha heads. Our research data comes from high-precision images: sculptures of Buddha heads in the Wanfo Hall in Fangshan District, Beijing. Through data enhancement and transformation, we have 3000 images for training. The final mean accuracy value (mAP) reached 0.92. and found that the detection effect on broken Buddha heads is still considerable. To the best of our knowledge, this paper is the first work to apply an improved YOLO deep learning model to address the target detection and counting of Buddha head, which has achieved good results.