Scholarlink

Nepal is a country with a rich cultural and historical heritage, particularly in its city of Patan. The city boasts a vast array of historical structures, each with its own unique story and historical significance. Determining which CNN model performs best on Nepali historic monuments is challenging. Hence, four models named MobileNetV2, ResNet152V2, Xception, and InceptionV3 were compared. By contrasting all of the model outputs, a comparative study of all the models has been conducted. To train the models, over 11,000 images of 20 different classes, each representing a different heritage site within the Patan Durbar square, were used. Through repeated hyperparameter training and testing, the study found that InceptionV3, MobileNetV2, Resnet152V2, and Xception all obtained high accuracy ratings of 85%, 95%, 96%, and 95%, respectively.

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

Deep Neural Networks for Heritage Monuments Classification: A Case Study in Nepal

  • 会议论文

  • 发表日期：

    2025年

查看全文
  • Sijan Dhungana,
  • Sanjeeb Prasad Panday,
  • Aman Shakya,
  • Basanta Joshi

摘要

Nepal is a country with a rich cultural and historical heritage, particularly in its city of Patan. The city boasts a vast array of historical structures, each with its own unique story and historical significance. Determining which CNN model performs best on Nepali historic monuments is challenging. Hence, four models named MobileNetV2, ResNet152V2, Xception, and InceptionV3 were compared. By contrasting all of the model outputs, a comparative study of all the models has been conducted. To train the models, over 11,000 images of 20 different classes, each representing a different heritage site within the Patan Durbar square, were used. Through repeated hyperparameter training and testing, the study found that InceptionV3, MobileNetV2, Resnet152V2, and Xception all obtained high accuracy ratings of 85%, 95%, 96%, and 95%, respectively.