Scholarlink

<p>MuraiDetect is introduced as a comprehensive dataset specifically created to facilitate the restoration of damages in Indian temple murals. Murals, an integral part of India’s rich cultural heritage, are susceptible to various forms of deterioration over time, necessitating efficient methods for their preservation and restoration. This dataset is meticulously compiled from a diverse collection of mural images sourced from various temples across the state, of Tamil Nadu in India, capturing a wide range of styles, subjects, and historical contexts. The objective of the given research is to create a custom dataset, MuraiDetect for the restoration of damaged murals. The work includes a collection of damaged murals and automatic detection of damaged regions thereby creating binary masks to leverage the restoration process by proposing an improved version of Mask R-CNN (region-based convolutional neural network). It includes a special attention mechanism called Convolutional Block Attention Module (CBAM) to prioritize certain spatial regions of the feature maps that are more likely to contain damaged regions. This attention mechanism helps the Region Proposal Network (RPN) of Mask R-CNN to focus on relevant areas of the image, thereby improving proposal generation. Lastly, the performance of the proposed work is evaluated using precision-recall (PR) curve thus achieving the superior performance (92.66%) than existing studies.</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

MuraiDetect: a comprehensive dataset for automatic damage detection and restoration in Indian murals using improved mask R-CNN

  • 期刊论文

  • 发表日期：

    2025年11月11日

查看全文
  • G. Sumathi,
  • M. Uma Devi

摘要

MuraiDetect is introduced as a comprehensive dataset specifically created to facilitate the restoration of damages in Indian temple murals. Murals, an integral part of India’s rich cultural heritage, are susceptible to various forms of deterioration over time, necessitating efficient methods for their preservation and restoration. This dataset is meticulously compiled from a diverse collection of mural images sourced from various temples across the state, of Tamil Nadu in India, capturing a wide range of styles, subjects, and historical contexts. The objective of the given research is to create a custom dataset, MuraiDetect for the restoration of damaged murals. The work includes a collection of damaged murals and automatic detection of damaged regions thereby creating binary masks to leverage the restoration process by proposing an improved version of Mask R-CNN (region-based convolutional neural network). It includes a special attention mechanism called Convolutional Block Attention Module (CBAM) to prioritize certain spatial regions of the feature maps that are more likely to contain damaged regions. This attention mechanism helps the Region Proposal Network (RPN) of Mask R-CNN to focus on relevant areas of the image, thereby improving proposal generation. Lastly, the performance of the proposed work is evaluated using precision-recall (PR) curve thus achieving the superior performance (92.66%) than existing studies.