Scholarlink

<p>Various pathogens cause wheat leaf fungal diseases that are challenging to control, resulting in crop yield and quality losses. Hence, early and accurate recognition of wheat leaf fungal diseases is essential for effective disease management. Existing models perform well in controlled environments with uniform backgrounds, but they difficult to deliver accurate results. To improve accuracy of wheat leaf fungal disease detection, this study proposes the CaiT-YOLOv9s-CBAM model, which is based on class attention in image transformer (CaiT) based object detection You Look Only Once (YOLOv9s), and the Convolutional Block Attention Mechanism (CBAM) model. The dataset of wheat leaf images infected with different fungal pathogens is used to train model. The CaiT helps to extracts high-quality features from wheat leaf images by processing them into patches. Combination of YOLOv9s and CBAM helps to recognize and localize fungal disease areas on wheat leaves whether these areas are small or overlap while CBAM extract most relevant disease features by minimizing background noise in each patch. The performance of the proposed model was compared to the available state-of-the-art detection YOLOv4, YOLOv5s, and Faster-RCNN models. The recognition accuracy of the proposed model was 98.6%, with a precision of 97.6%, recall of 98.5%, and a mAP of 97.95% higher than other versions of the YOLOv9. In addition, it shows strong robustness for fungal diseases recognition in complex and overlap regions. CaiT-YOLOv9s-CBAM model can accurately and efficiently detect wheat leaf fungal diseases, with better detection performance than existing models.</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

CaiT-YOLOv9s-CBAM Deep Learning Model for Wheat Fungal Diseases: In-Depth Multifaceted Approach for Feature Refinement, Localization and Recognition

  • 期刊论文

  • 发表日期：

    2026年3月12日

查看全文
  • Deepak Kumar,
  • Vinay Kukreja,
  • Mohammed Al-Sarem,
  • Vivek Bhardwaj

摘要

Various pathogens cause wheat leaf fungal diseases that are challenging to control, resulting in crop yield and quality losses. Hence, early and accurate recognition of wheat leaf fungal diseases is essential for effective disease management. Existing models perform well in controlled environments with uniform backgrounds, but they difficult to deliver accurate results. To improve accuracy of wheat leaf fungal disease detection, this study proposes the CaiT-YOLOv9s-CBAM model, which is based on class attention in image transformer (CaiT) based object detection You Look Only Once (YOLOv9s), and the Convolutional Block Attention Mechanism (CBAM) model. The dataset of wheat leaf images infected with different fungal pathogens is used to train model. The CaiT helps to extracts high-quality features from wheat leaf images by processing them into patches. Combination of YOLOv9s and CBAM helps to recognize and localize fungal disease areas on wheat leaves whether these areas are small or overlap while CBAM extract most relevant disease features by minimizing background noise in each patch. The performance of the proposed model was compared to the available state-of-the-art detection YOLOv4, YOLOv5s, and Faster-RCNN models. The recognition accuracy of the proposed model was 98.6%, with a precision of 97.6%, recall of 98.5%, and a mAP of 97.95% higher than other versions of the YOLOv9. In addition, it shows strong robustness for fungal diseases recognition in complex and overlap regions. CaiT-YOLOv9s-CBAM model can accurately and efficiently detect wheat leaf fungal diseases, with better detection performance than existing models.