Scholarlink

<p>Apples are among the most widely consumed fruits globally due to their numerous health benefits. However, their production is significantly impacted by various leaf diseases. Accurate and timely detection of these diseases is challenging due to the similarities between healthy and diseased leaves, as well as issues like image blurring, clutter, and varying conditions. This paper presents E‑AppleNet, an advanced approach based on EfficientNetV2, which incorporates attention mechanisms and additional dense layers at the end of the model structure to enhance disease classification. To tackle class imbalance and prevent model overfitting, we utilize transfer learning and multi-class focal loss. Our model, tested on the PlantVillage dataset, a&#xa0;complex repository with real-world images, achieves a&#xa0;remarkable 99% accuracy. Additionally, we generated the heatmap visualizations, which confirm the model’s robustness in handling diverse image distortions. This method provides a&#xa0;reliable solution for automated apple disease detection, with potential benefits for precision agriculture.</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

E-AppleNet: An Enhanced Deep Learning Approach for Apple Fruit Leaf Disease Classification

  • 期刊论文

  • 发表日期：

    2025年1月3日

查看全文
  • Ameen Banjar,
  • Ali Javed,
  • Marriam Nawaz,
  • Hussain Dawood

摘要

Apples are among the most widely consumed fruits globally due to their numerous health benefits. However, their production is significantly impacted by various leaf diseases. Accurate and timely detection of these diseases is challenging due to the similarities between healthy and diseased leaves, as well as issues like image blurring, clutter, and varying conditions. This paper presents E‑AppleNet, an advanced approach based on EfficientNetV2, which incorporates attention mechanisms and additional dense layers at the end of the model structure to enhance disease classification. To tackle class imbalance and prevent model overfitting, we utilize transfer learning and multi-class focal loss. Our model, tested on the PlantVillage dataset, a complex repository with real-world images, achieves a remarkable 99% accuracy. Additionally, we generated the heatmap visualizations, which confirm the model’s robustness in handling diverse image distortions. This method provides a reliable solution for automated apple disease detection, with potential benefits for precision agriculture.