Scholarlink

<p>Plant diseases caused by pathogens such as fungi, bacteria, and viruses that impact crop health in conjunction with an environmental stressor impact agricultural production. In this case study we propose a strong plant disease detection framework that uses machine-learning-based pre-processing, Generative Adversarial Network (GAN)-based augmentation, and a similar segmentation method. First, low-quality plant images will be enhanced to increase the quality and resolution of the images before using GAN to augment the dataset. MCT in conjunction with lesion based is used to segment the plant images. All performance evaluation was done across ten epochs of training using accuracy, precision, recall, and F1-score to measure performance. At Epoch 1 the model obtained 65% accuracy, 62% precision, 60% recall, and 61% F1-score. By Epoch 10, the model performances at 97% accuracy, 96% precision, 95% recall, and 95.5% F1-score indicating a robust learning curve. Likewise, the loss value progressed from 0.45 to 0.07 indicating a sustainable learning curve and effective training and adaptation of the model. Clearly the results indicated that the current system has significant capabilities to improve plant disease recognition while providing a dependable and efficient tool for continuously monitoring plant health and dealing with cropping disease in a precision agriculture context.</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

Plant Disease Estimation and Classification with the Generative Adversarial Network Based on Image Synthesis Using Segmentation Model Classification

  • 期刊论文

  • 发表日期：

    2025年10月3日

查看全文
  • Nagam Anjineyulu,
  • S. Nirmalraj

摘要

Plant diseases caused by pathogens such as fungi, bacteria, and viruses that impact crop health in conjunction with an environmental stressor impact agricultural production. In this case study we propose a strong plant disease detection framework that uses machine-learning-based pre-processing, Generative Adversarial Network (GAN)-based augmentation, and a similar segmentation method. First, low-quality plant images will be enhanced to increase the quality and resolution of the images before using GAN to augment the dataset. MCT in conjunction with lesion based is used to segment the plant images. All performance evaluation was done across ten epochs of training using accuracy, precision, recall, and F1-score to measure performance. At Epoch 1 the model obtained 65% accuracy, 62% precision, 60% recall, and 61% F1-score. By Epoch 10, the model performances at 97% accuracy, 96% precision, 95% recall, and 95.5% F1-score indicating a robust learning curve. Likewise, the loss value progressed from 0.45 to 0.07 indicating a sustainable learning curve and effective training and adaptation of the model. Clearly the results indicated that the current system has significant capabilities to improve plant disease recognition while providing a dependable and efficient tool for continuously monitoring plant health and dealing with cropping disease in a precision agriculture context.