Scholarlink

<p>Agriculture plays a pivotal role in global economic growth, yet it faces significant challenges from pests and crop diseases. Early detection is crucial for preventing large-scale crop losses and ensuring food security. This study introduces an adaptive hybrid transformer model, Swin-HViT, which integrates the strengths of a vision transformer (ViT) and a Swin transformer to accurately predict crop diseases. While ViT captures global image features, the Swin Transformer excels at extracting fine-grained local details. To effectively integrate global and local feature, an adaptive fusion mechanism based on a trainable MLP is introduced, enabling dynamic feature weighting within the hybrid architecture. Evaluated on two benchmark datasets, Corn and PlantDoc, our model achieved accuracies of 99.03% and 83.07%, respectively, surpassing recent works. Here, we demonstrate the effectiveness of combining complementary transformer architectures to improve disease identification in diverse agricultural settings. The code, data and the hybrid model are available at <a href="https://github.com/hema2107/Swin-HViT">https://github.com/hema2107/Swin-HViT</a>.</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

Swin-HViT for accurate crop disease prediction using an adaptive hybrid transformer model

  • 期刊论文

  • 出版模式：

    开放获取

  • 发表日期：

    2026年5月20日

查看全文
  • Hemalatha Gunasekaran,
  • N. R. Wilfred Blessing,
  • Naveen Vijayakumar Watson,
  • B. Hariharan,
  • Angelin Gladys Jesudoss,
  • C. G. Anupama

摘要

Agriculture plays a pivotal role in global economic growth, yet it faces significant challenges from pests and crop diseases. Early detection is crucial for preventing large-scale crop losses and ensuring food security. This study introduces an adaptive hybrid transformer model, Swin-HViT, which integrates the strengths of a vision transformer (ViT) and a Swin transformer to accurately predict crop diseases. While ViT captures global image features, the Swin Transformer excels at extracting fine-grained local details. To effectively integrate global and local feature, an adaptive fusion mechanism based on a trainable MLP is introduced, enabling dynamic feature weighting within the hybrid architecture. Evaluated on two benchmark datasets, Corn and PlantDoc, our model achieved accuracies of 99.03% and 83.07%, respectively, surpassing recent works. Here, we demonstrate the effectiveness of combining complementary transformer architectures to improve disease identification in diverse agricultural settings. The code, data and the hybrid model are available at https://github.com/hema2107/Swin-HViT.