Scholarlink

<p>Agriculture plays a vital role in global food security, yet plant diseases remain a major challenge, causing significant yield losses and economic instability. Timely and accurate detection of plant diseases, especially leaf diseases, is essential for mitigating these effects. Traditional methods of disease detection, which are based on visual inspection, are often slow and prone to errors. In this study, we propose an advanced deep learning-based approach that combines YOLOv8 (you only look once version 8) for plant disease classification with EigenCAM for interpretability. YOLOv8, a state-of-the-art convolutional neural network, excels in real-time object detection, offering high speed and accuracy. EigenCAM enhances model transparency by generating class activation maps that allow farmers and agronomists to visually understand which areas of a plant leaf image influence the model’s decision-making process. We evaluate our model on four distinct plant disease datasets: the PlantVillage, Potato Disease Leaf, Apple Tree Leaf Disease, and Plant Disease datasets, which cover a variety of plant species and disease conditions. The results demonstrate that our YOLOv8-based model achieves high classification accuracy, surpassing 98% in disease detection. Furthermore, the integration of EigenCAM not only provides valuable insights into model decision-making but also improves user trust and the practical applicability of the system. This combination of high performance and interpretability makes the proposed approach a promising solution for precision agriculture, enabling early disease detection, reducing crop losses, and supporting informed decision-making for crop management.</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

A hybrid YOLOv8-EigenCAM model for accurate and interpretable plant disease classification

  • 期刊论文

  • 发表日期：

    2026年2月9日

查看全文
  • Hossam Magdy Balaha

摘要

Agriculture plays a vital role in global food security, yet plant diseases remain a major challenge, causing significant yield losses and economic instability. Timely and accurate detection of plant diseases, especially leaf diseases, is essential for mitigating these effects. Traditional methods of disease detection, which are based on visual inspection, are often slow and prone to errors. In this study, we propose an advanced deep learning-based approach that combines YOLOv8 (you only look once version 8) for plant disease classification with EigenCAM for interpretability. YOLOv8, a state-of-the-art convolutional neural network, excels in real-time object detection, offering high speed and accuracy. EigenCAM enhances model transparency by generating class activation maps that allow farmers and agronomists to visually understand which areas of a plant leaf image influence the model’s decision-making process. We evaluate our model on four distinct plant disease datasets: the PlantVillage, Potato Disease Leaf, Apple Tree Leaf Disease, and Plant Disease datasets, which cover a variety of plant species and disease conditions. The results demonstrate that our YOLOv8-based model achieves high classification accuracy, surpassing 98% in disease detection. Furthermore, the integration of EigenCAM not only provides valuable insights into model decision-making but also improves user trust and the practical applicability of the system. This combination of high performance and interpretability makes the proposed approach a promising solution for precision agriculture, enabling early disease detection, reducing crop losses, and supporting informed decision-making for crop management.