Scholarlink

<p>In India, rice (<i>Oryza sativa</i>) holds paramount importance as a vital food staple, ranking third in global production. With its dense calorie content, rice serves as a dietary cornerstone for over one-fifth of the global population. However, rice production faces significant challenges stemming from both abiotic and biotic factors, notably diseases and pests, which collectively contribute to a substantial 70% loss in crop yield. Early disease detection poses a daunting task for farmers, as timely intervention is crucial to mitigate potential damage. However, manual inspections by agricultural officers or external experts are hindered by resource constraints, making consistent field assessments challenging. To address this, advancements in technology, such as Convolutional Neural Networks (CNNs), offers a promising solution. By training CNN models with 10,407 labelled images and testing with 3469 images, employing transfer learning models DenseNet169-MLP, ResNet152-MLP, InceptionResNetV2-MLP, MobileNetV3-MLP, and EfficientNetB3-MLP, we aim to enhance disease identification accuracy. Specifically, employing a CNN model with Stratified Cross-validation using fastai techniques, we achieve a remarkable accuracy of 94%, surpassing traditional transfer learning models. This approach holds immense potential to disease identification and crop management practices in paddy fields.</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

Advanced techniques in paddy disease detection: boosting classification accuracy with stratified cross-validation

  • 期刊论文

  • 发表日期：

    2025年12月13日

查看全文
  • Elakya R,
  • Abinaya Gopalakrishnan,
  • Aravind Balakrishnan,
  • Suresh K

摘要

In India, rice (Oryza sativa) holds paramount importance as a vital food staple, ranking third in global production. With its dense calorie content, rice serves as a dietary cornerstone for over one-fifth of the global population. However, rice production faces significant challenges stemming from both abiotic and biotic factors, notably diseases and pests, which collectively contribute to a substantial 70% loss in crop yield. Early disease detection poses a daunting task for farmers, as timely intervention is crucial to mitigate potential damage. However, manual inspections by agricultural officers or external experts are hindered by resource constraints, making consistent field assessments challenging. To address this, advancements in technology, such as Convolutional Neural Networks (CNNs), offers a promising solution. By training CNN models with 10,407 labelled images and testing with 3469 images, employing transfer learning models DenseNet169-MLP, ResNet152-MLP, InceptionResNetV2-MLP, MobileNetV3-MLP, and EfficientNetB3-MLP, we aim to enhance disease identification accuracy. Specifically, employing a CNN model with Stratified Cross-validation using fastai techniques, we achieve a remarkable accuracy of 94%, surpassing traditional transfer learning models. This approach holds immense potential to disease identification and crop management practices in paddy fields.