Scholarlink

This research presents a novel deep-learning architecture intended to classify corn leaf diseases with high accuracy and reliability. Our model integrates advanced techniques such as Depthwise Separable Convolutions (DSC), Squeeze-and-Excitation (SE) Blocks, and Residual Connections (RC) to capture intricate patterns and enhance feature extraction effectively. The dataset used comprises 4 distinct categories: Common Rust, Blight, Gray Leaf Spot, and Healthy, with a total of 4,188 images sourced from the PlantVillage and PlantDoc datasets. The obtained results present the accuracy of the proposed method. The model reached a validation accuracy of 88.31% and a test accuracy of 87.35%. The proposed architecture shows superior performance metrics results by achieving a precision (0.87), recall (0.87), and an F1-score (0.87). Specifically, the model exhibited high precision and recall for the Common Rust and Healthy categories, achieving F1-scores of 0.94 and 0.97. The results emphasize the potential of our approach in precisely diagnosing leaf diseases which is crucial for timely and effective agricultural management.

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

A Hybrid CNN Architecture for Efficient Detection of Maize Plant Diseases

  • 章节

  • 发表日期：

    2025年

查看全文
  • Md. Asaduzzaman,
  • Md. Johir Uddin Bhuiyan,
  • Habiba Akter Rimi,
  • Hashibul Ahsan Shoaib,
  • Md. Raisul Islam

摘要

This research presents a novel deep-learning architecture intended to classify corn leaf diseases with high accuracy and reliability. Our model integrates advanced techniques such as Depthwise Separable Convolutions (DSC), Squeeze-and-Excitation (SE) Blocks, and Residual Connections (RC) to capture intricate patterns and enhance feature extraction effectively. The dataset used comprises 4 distinct categories: Common Rust, Blight, Gray Leaf Spot, and Healthy, with a total of 4,188 images sourced from the PlantVillage and PlantDoc datasets. The obtained results present the accuracy of the proposed method. The model reached a validation accuracy of 88.31% and a test accuracy of 87.35%. The proposed architecture shows superior performance metrics results by achieving a precision (0.87), recall (0.87), and an F1-score (0.87). Specifically, the model exhibited high precision and recall for the Common Rust and Healthy categories, achieving F1-scores of 0.94 and 0.97. The results emphasize the potential of our approach in precisely diagnosing leaf diseases which is crucial for timely and effective agricultural management.