Scholarlink

<p>Soybean (<i>Glycine max</i> L. Merrill) is an important crop that makes a significant contribution to the world food production and biofuel industries. However, diseases such as bacterial blight, brown spots, and powdery mildew have a serious impact on soybean yield and quality. This study proposes a novel Soybean Quantum Network with Swin Transformer (SoyQNet-SwinT) model based on deep learning and a quantum-inspired attention mechanism to accurately classify soybean leaf diseases. The model overcomes the shortcomings of existing methods by improving feature learning through quantum-inspired self-attention and the Swin Transformer architecture. The Quantum-Enhanced SoyQNet-SwinT model integrates Convolutional Neural Networks (CNNs), a quantum-inspired attention mechanism and Swin Transformer blocks. A dataset of soybean leaf images and data augmentation techniques was used for training to improve model generalisation. The model was evaluated on the Auburn soybean disease image (ASDID) dataset. Gradient-weighted Class Activation Mapping (Grad-CAM), Local Interpretable Model-agnostic Explanations (LIME), and SHapley Additive exPlanations (SHAP) were applied to improve model interpretability, enabling interpretation of the image regions contributing to model predictions. The Accuracy, F1-score, and recall were 94%, 93.7%, and 94.5%, respectively, for the ASDID. The model achieved 98% and 97% accuracy on PlantVillage and AI2018, respectively, outperforming the compared baseline models, including MobileNetV2 and InceptionV7. The random-shifting technique improved precision by 1.5%. Future research will focus on improving performance for subtle disease types, such as Yellow Mosaic, and on reducing computational overhead through more efficient implementations of quantum-inspired attention mechanisms.</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

A quantum-enhanced self-attention model with swin transformer for soybean leaf disease classification with explainable AI

  • 期刊论文

  • 发表日期：

    2026年6月15日

查看全文
  • Shiva Mehta,
  • Vinay Kukreja

摘要

Soybean (Glycine max L. Merrill) is an important crop that makes a significant contribution to the world food production and biofuel industries. However, diseases such as bacterial blight, brown spots, and powdery mildew have a serious impact on soybean yield and quality. This study proposes a novel Soybean Quantum Network with Swin Transformer (SoyQNet-SwinT) model based on deep learning and a quantum-inspired attention mechanism to accurately classify soybean leaf diseases. The model overcomes the shortcomings of existing methods by improving feature learning through quantum-inspired self-attention and the Swin Transformer architecture. The Quantum-Enhanced SoyQNet-SwinT model integrates Convolutional Neural Networks (CNNs), a quantum-inspired attention mechanism and Swin Transformer blocks. A dataset of soybean leaf images and data augmentation techniques was used for training to improve model generalisation. The model was evaluated on the Auburn soybean disease image (ASDID) dataset. Gradient-weighted Class Activation Mapping (Grad-CAM), Local Interpretable Model-agnostic Explanations (LIME), and SHapley Additive exPlanations (SHAP) were applied to improve model interpretability, enabling interpretation of the image regions contributing to model predictions. The Accuracy, F1-score, and recall were 94%, 93.7%, and 94.5%, respectively, for the ASDID. The model achieved 98% and 97% accuracy on PlantVillage and AI2018, respectively, outperforming the compared baseline models, including MobileNetV2 and InceptionV7. The random-shifting technique improved precision by 1.5%. Future research will focus on improving performance for subtle disease types, such as Yellow Mosaic, and on reducing computational overhead through more efficient implementations of quantum-inspired attention mechanisms.