Scholarlink

<p>Apple, a globally revered fruit with diverse cultivars, often falls victim to various leaf diseases that compromise crop yield and quality. Precise and timely disease identification is crucial for effective orchard management and timely intervention. Recent years has witnessed the exponential growth of Artificial Intelligence, Machine Learning and its subsequent applications to various diversified applications including image processing. This paper intends to utilise the potential of Deep Learning to transform apple leaf disease detection and classification. In this work, the best Convolutional Neural Network pretrained model for identification of apple leaf disease is found using rich visual dataset of 3,200 apple leaf images, meticulously curated to encompass four distinct categories: healthy leaves, those afflicted by apple scab, black rot, and cedar rust. Experimental results revealed that Residual Neural Network achieved highest validation accuracy of 98.9%, significantly exceeding the performance of benchmark models like Visual Geometry Group 19 (96%), InceptionV3 (97%), EfficientNet (97.5%), Visual Geometry Group 16 (98.1%), MobileNet (98.2%) and the traditional baseline Convolutional Neural Network (95%). The extensive experimentation also revealed that an 80:20 training-to-validation split yields the highest classification accuracy. The exceptional accuracy achieved by Residual Neural Network powered model, coupled with the robustness of the image dataset, opens doors for the development of advanced orchard management tools with real-time identification and combating harmful leaf diseases. This can potentially revolutionize agricultural practices, minimizing crop losses and ensuring sustainable apple production.</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

An integrated approach to apple leaf disease detection: leveraging convolutional neural networks for accurate diagnosis

  • 期刊论文

  • 发表日期：

    2025年3月22日

查看全文
  • Dasari Rohith,
  • Praneet Saurabh,
  • Dhananjay Bisen

摘要

Apple, a globally revered fruit with diverse cultivars, often falls victim to various leaf diseases that compromise crop yield and quality. Precise and timely disease identification is crucial for effective orchard management and timely intervention. Recent years has witnessed the exponential growth of Artificial Intelligence, Machine Learning and its subsequent applications to various diversified applications including image processing. This paper intends to utilise the potential of Deep Learning to transform apple leaf disease detection and classification. In this work, the best Convolutional Neural Network pretrained model for identification of apple leaf disease is found using rich visual dataset of 3,200 apple leaf images, meticulously curated to encompass four distinct categories: healthy leaves, those afflicted by apple scab, black rot, and cedar rust. Experimental results revealed that Residual Neural Network achieved highest validation accuracy of 98.9%, significantly exceeding the performance of benchmark models like Visual Geometry Group 19 (96%), InceptionV3 (97%), EfficientNet (97.5%), Visual Geometry Group 16 (98.1%), MobileNet (98.2%) and the traditional baseline Convolutional Neural Network (95%). The extensive experimentation also revealed that an 80:20 training-to-validation split yields the highest classification accuracy. The exceptional accuracy achieved by Residual Neural Network powered model, coupled with the robustness of the image dataset, opens doors for the development of advanced orchard management tools with real-time identification and combating harmful leaf diseases. This can potentially revolutionize agricultural practices, minimizing crop losses and ensuring sustainable apple production.