Scholarlink

<p>Plant leaf disease and treatment control are critical factors that affect crop yield and quality. Advanced artificial intelligence (AI) technologies, increasingly popular in various computer vision fields, can address complex problems in food sciences and agriculture. We propose a&#xa0;novel deep learning-based diagnostic framework using AI algorithms for fruit leaf disease classification. The system employs the deep learning algorithm ResNet101 with transfer learning to classify 29&#xa0;classes, including&#xa0;22 diseased and seven healthy classes. To enhance generalization and to mitigate overfitting, five-fold cross-validation is used for model training. Evaluated on a&#xa0;standard dataset, the proposed approach achieves an average accuracy of 98.25% for plant leaf disease detection and an area under the curve (AUC) value of 99.06%, outperforming state-of-the-art methods. Additionally, the model is flexible, and its performance is validated on unseen field-collected data of fruits and vegetables. This advanced AI-based model can be developed into software as a&#xa0;service (SaaS) to assist growers in monitoring crop health efficiently, reducing the excessive use of agrochemicals, and protecting the environment.</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

Multi-stage Deep Learning Model-Based Plant Leaf Disease Diagnosis for Sustainable Crop Health Monitoring

  • 期刊论文

  • 发表日期：

    2025年8月19日

查看全文
  • Mehdi Hassan,
  • Safdar Ali,
  • Jin Young Kim,
  • Gwang-Hyun Yu,
  • Syed Fahad Tahir

摘要

Plant leaf disease and treatment control are critical factors that affect crop yield and quality. Advanced artificial intelligence (AI) technologies, increasingly popular in various computer vision fields, can address complex problems in food sciences and agriculture. We propose a novel deep learning-based diagnostic framework using AI algorithms for fruit leaf disease classification. The system employs the deep learning algorithm ResNet101 with transfer learning to classify 29 classes, including 22 diseased and seven healthy classes. To enhance generalization and to mitigate overfitting, five-fold cross-validation is used for model training. Evaluated on a standard dataset, the proposed approach achieves an average accuracy of 98.25% for plant leaf disease detection and an area under the curve (AUC) value of 99.06%, outperforming state-of-the-art methods. Additionally, the model is flexible, and its performance is validated on unseen field-collected data of fruits and vegetables. This advanced AI-based model can be developed into software as a service (SaaS) to assist growers in monitoring crop health efficiently, reducing the excessive use of agrochemicals, and protecting the environment.