Scholarlink

Diseases in pumpkin plants have enormous agricultural productivity threat, which causes economic losses. These diseases also lead to concern in nutritional safety. The identification of pumpkin diseases for effective disease management is required in an accurate and efficient manner. The research helps to detect and classify the pumpkin plant disease so that this threat to farmers and their crops can be minimized. Initially, a basic model CNN is implemented to deter-mine the basic performance with an accuracy of only 81.81%. For improved results, transfer learning is implemented by using pre-educated models such as ResNet50, EfficientNetB0, DenseNet121, and VGG16 in which EfficientNetB0 performed the best with an accuracy of 92.00%. Further, an attention mechanism was used called Dynamic Spatial Attention (DSA) along with pretrained models to improve the detection performance. This DSA integrated model focused more on important areas of the image to refine and improve the detection accuracies, which are observed in all the models integrated with DSA with best accuracy of 96.15% given by EfficientNetB0 + DSA. This study may help farmers and other agricultural professionals to efficiently treat pumpkin diseases and contribute to further development in automated disease detection.

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

Cucurbita Pepo Disease Classification Through Attention-Enhanced Deep Learning Models

  • 章节

  • 发表日期：

    2026年

查看全文
  • Sweeti Sah,
  • Shweta Sharma,
  • Alok Jaiswal,
  • Prabhat Kumar,
  • Hansi Beri,
  • Balaji Vuppala,
  • Sachi Nandan Mohanty

摘要

Diseases in pumpkin plants have enormous agricultural productivity threat, which causes economic losses. These diseases also lead to concern in nutritional safety. The identification of pumpkin diseases for effective disease management is required in an accurate and efficient manner. The research helps to detect and classify the pumpkin plant disease so that this threat to farmers and their crops can be minimized. Initially, a basic model CNN is implemented to deter-mine the basic performance with an accuracy of only 81.81%. For improved results, transfer learning is implemented by using pre-educated models such as ResNet50, EfficientNetB0, DenseNet121, and VGG16 in which EfficientNetB0 performed the best with an accuracy of 92.00%. Further, an attention mechanism was used called Dynamic Spatial Attention (DSA) along with pretrained models to improve the detection performance. This DSA integrated model focused more on important areas of the image to refine and improve the detection accuracies, which are observed in all the models integrated with DSA with best accuracy of 96.15% given by EfficientNetB0 + DSA. This study may help farmers and other agricultural professionals to efficiently treat pumpkin diseases and contribute to further development in automated disease detection.