Scholarlink

<p>Guava fruit diseases including anthracnose and fruit fly infestation having a substantial impact on crop productivity and quality, imposing a number of problems for farmers. Existing methods rely on classic image processing techniques, which are often inaccurate and unsuitable for complex illness patterns. In this study, deep learning techniques such as EfficientNet, Vision Transformer (ViT), Swin Transformer, ResNet50, and ConvNeXt are used to categorize images into multiclass classification problem i.e. healthy, Anthracnose-affected and fruit fly-affected guava fruits. ConvNeXt had the highest accuracy of the models examined, with 93.1% on the training set and 92.5% on the testing set, closely followed by Swin Transformer. The findings show that transformer-based models outperform standard CNNs in detecting subtle illness trends. This study demonstrates the feasibility of using advanced vision transformers for real-time disease detection, paving the road for smart farming solutions and sustainable agricultural practices.</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

Optimized Deep Learning Framework for Plant Disease Detection Using Fine-Tuned Vision Transformers and CNN Model

  • 期刊论文

  • 发表日期：

    2026年5月20日

查看全文
  • Surbhi Vijh,
  • Chin-Shiuh Shieh,
  • Vishal Jain,
  • Mong-Fong Horng

摘要

Guava fruit diseases including anthracnose and fruit fly infestation having a substantial impact on crop productivity and quality, imposing a number of problems for farmers. Existing methods rely on classic image processing techniques, which are often inaccurate and unsuitable for complex illness patterns. In this study, deep learning techniques such as EfficientNet, Vision Transformer (ViT), Swin Transformer, ResNet50, and ConvNeXt are used to categorize images into multiclass classification problem i.e. healthy, Anthracnose-affected and fruit fly-affected guava fruits. ConvNeXt had the highest accuracy of the models examined, with 93.1% on the training set and 92.5% on the testing set, closely followed by Swin Transformer. The findings show that transformer-based models outperform standard CNNs in detecting subtle illness trends. This study demonstrates the feasibility of using advanced vision transformers for real-time disease detection, paving the road for smart farming solutions and sustainable agricultural practices.