Scholarlink

<p>Identifying and segmenting infected plant leaves is essential for early disease detection and crop yield assessment. Such infections are associated with insects, fungi, or bacteria and pose a challenge in the identification process due to the complex history, vague borders, and low-quality images. We provide Self-Attention Super-Resolution GAN (SASRGAN) for low-resolution leaf images enhancement with finer disease patterns visible and segmentation improved. The generator employs residual dense blocks while conserving details and self-attention convolutional layers concentrate on rather peculiar features of the diseases. To enhance classification, we propose a dual attention classification model, which incorporates channel attention as well as spatial attention. For SASRGAN, UQI equals 0.9817 and VIFP equals 0.979. Our model performs Plant Village Dataset with accuracy of 99.78%, precision of 99.57%, recall of 99.18%, and F1-score of 99.42%.</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

Enhancing plant leaf disease classification through self-attention super-resolution GAN and dual attention model

  • 期刊论文

  • 发表日期：

    2025年12月16日

查看全文
  • V. Sasikala,
  • Venkatramaphanikumar Sistla,
  • Deepak Chowdary Edara,
  • Venkata Krishna Kishore Kolli

摘要

Identifying and segmenting infected plant leaves is essential for early disease detection and crop yield assessment. Such infections are associated with insects, fungi, or bacteria and pose a challenge in the identification process due to the complex history, vague borders, and low-quality images. We provide Self-Attention Super-Resolution GAN (SASRGAN) for low-resolution leaf images enhancement with finer disease patterns visible and segmentation improved. The generator employs residual dense blocks while conserving details and self-attention convolutional layers concentrate on rather peculiar features of the diseases. To enhance classification, we propose a dual attention classification model, which incorporates channel attention as well as spatial attention. For SASRGAN, UQI equals 0.9817 and VIFP equals 0.979. Our model performs Plant Village Dataset with accuracy of 99.78%, precision of 99.57%, recall of 99.18%, and F1-score of 99.42%.