Scholarlink

The early detection and accurate classification of diseases affecting maize crops remains a significant challenge for farmers, particularly in sub-Saharan Africa. Many farmers still rely on traditional methods to identify diseases, which are often inefficient and inaccurate. To address these challenges, this study investigates the use of classical machine learning models - Decision Tree, Logistic Regression, Random Forest, and eXtreme Gradient Boosting as well as pre-trained Convolutional Neural Network (CNN) models, including VGG16, ResNet50, Xception, and DenseNet201, to classify two common maize leaf diseases: Maize Streak Virus and Fall Armyworm, which are prevalent in Sub-Saharan Africa. The findings reveal that among the classical machine learning models, eXtreme Gradient Boosting achieves the highest accuracy at \(73\%\) , while DenseNet201 outperforms all models with an impressive accuracy of \(98\%\) among the pre-trained CNN models. Overall, the pre-trained CNN models not only deliver superior results but also significantly reduce the time and effort required for training.

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

Classical vs Deep Learning Models: A Study on Maize Leaf Diseases Classification

  • 会议论文

  • 发表日期：

    2025年

查看全文
  • Chimango Nyasulu,
  • Cecilia Tivir,
  • Jonathan Mwaura

摘要

The early detection and accurate classification of diseases affecting maize crops remains a significant challenge for farmers, particularly in sub-Saharan Africa. Many farmers still rely on traditional methods to identify diseases, which are often inefficient and inaccurate. To address these challenges, this study investigates the use of classical machine learning models - Decision Tree, Logistic Regression, Random Forest, and eXtreme Gradient Boosting as well as pre-trained Convolutional Neural Network (CNN) models, including VGG16, ResNet50, Xception, and DenseNet201, to classify two common maize leaf diseases: Maize Streak Virus and Fall Armyworm, which are prevalent in Sub-Saharan Africa. The findings reveal that among the classical machine learning models, eXtreme Gradient Boosting achieves the highest accuracy at \(73\%\) , while DenseNet201 outperforms all models with an impressive accuracy of \(98\%\) among the pre-trained CNN models. Overall, the pre-trained CNN models not only deliver superior results but also significantly reduce the time and effort required for training.