Scholarlink

<p>Potatoes are a crucial global vegetable, valued for their diversity and high demand, making them a significant commercial crop. In Algeria, potatoes are widely consumed and vital to agriculture. Algerian farmers face challenges in swiftly identifying leaf diseases that can cause extensive damage if not detected early. To address this issue, we employed a dataset of potato leaf images from Kaggle and applied various deep learning architectures to achieve high classification accuracy through advanced algorithms. Our study focuses on two prevalent potato diseases: early blight and late blight. The dataset includes 2152 images: 2000 of diseased leaves and 152 of healthy leaves. We assessed several deep learning models, including InceptionV3, DenseNet121, VGG16 and Vision Transformer (ViT). The ViT exhibited superior performance, achieving the highest test accuracy of 97.68% and demonstrating optimal performance metrics. This comparative study offers farmers precise and timely disease detection, enhancing crop protection through sophisticated mathematical and computational methods.</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

Mathematical modeling and optimization of vision transformers for disease classification in agronomic imaging

  • 期刊论文

  • 发表日期：

    2025年12月27日

查看全文
  • Soumia Benbakreti,
  • Samir Benbakreti,
  • Kadda Benyahia,
  • Ahmed Roumane,
  • Abdelkrim Ouhab,
  • Mohamed Benouis

摘要

Potatoes are a crucial global vegetable, valued for their diversity and high demand, making them a significant commercial crop. In Algeria, potatoes are widely consumed and vital to agriculture. Algerian farmers face challenges in swiftly identifying leaf diseases that can cause extensive damage if not detected early. To address this issue, we employed a dataset of potato leaf images from Kaggle and applied various deep learning architectures to achieve high classification accuracy through advanced algorithms. Our study focuses on two prevalent potato diseases: early blight and late blight. The dataset includes 2152 images: 2000 of diseased leaves and 152 of healthy leaves. We assessed several deep learning models, including InceptionV3, DenseNet121, VGG16 and Vision Transformer (ViT). The ViT exhibited superior performance, achieving the highest test accuracy of 97.68% and demonstrating optimal performance metrics. This comparative study offers farmers precise and timely disease detection, enhancing crop protection through sophisticated mathematical and computational methods.