Scholarlink

<p>The agriculture domain encompasses a vast array of subcategories, with plant and leaf diseases being among the most prevalent and impactful challenges. To address this issue, researchers have employed cutting-edge techniques for the prediction, analysis, and classification of leaf diseases. This study proposes a groundbreaking approach that integrates the feature extraction capabilities of the Swin Transformer with advanced mathematical enhancements to traditional classifiers, including Support Vector Machine (SVM), Random Forest (RF), k-Nearest Neighbor (KNN), and Naive Bayes (NB). The novel mathematical formulations applied to these classifiers not only optimize traditional metrics but also enhance the overall system reliability. These advancements enable the system to generalize effectively across diverse datasets, making it a valuable tool for scalable agricultural applications. Our SwinTransSVM model achieved exceptional performance with a 99.06% detection accuracy and 99% classification accuracy, showcasing its robustness in identifying leaf diseases. Similarly, the SwinTransKNN model recorded an impressive 94.92% detection accuracy and 95% classification accuracy. The SwinTransNB and SwinTransRF models also demonstrated reliable performance with detection accuracies of 83.27% and 79.88%, respectively, and corresponding classification accuracies of 83% and 80%. Further, we have implemented SwinTransGNB to compare the proposed models. The SwinTransGNB scored 94% detection accuracy and 93% classification accuracy. Our proposed methodology establishes a significant benchmark in the realm of leaf disease detection and classification, paving the way for more efficient, accurate, and adaptable solutions to revolutionize precision agriculture.</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

SwinTrans Classifiers - A Framework for Multi-leaf and Multi-disease Classification and Detection in Computer Vision

  • 期刊论文

  • 发表日期：

    2025年5月26日

查看全文
  • Madhavi,
  • Thipendra P. Singh,
  • Shailendra Pratap Singh

摘要

The agriculture domain encompasses a vast array of subcategories, with plant and leaf diseases being among the most prevalent and impactful challenges. To address this issue, researchers have employed cutting-edge techniques for the prediction, analysis, and classification of leaf diseases. This study proposes a groundbreaking approach that integrates the feature extraction capabilities of the Swin Transformer with advanced mathematical enhancements to traditional classifiers, including Support Vector Machine (SVM), Random Forest (RF), k-Nearest Neighbor (KNN), and Naive Bayes (NB). The novel mathematical formulations applied to these classifiers not only optimize traditional metrics but also enhance the overall system reliability. These advancements enable the system to generalize effectively across diverse datasets, making it a valuable tool for scalable agricultural applications. Our SwinTransSVM model achieved exceptional performance with a 99.06% detection accuracy and 99% classification accuracy, showcasing its robustness in identifying leaf diseases. Similarly, the SwinTransKNN model recorded an impressive 94.92% detection accuracy and 95% classification accuracy. The SwinTransNB and SwinTransRF models also demonstrated reliable performance with detection accuracies of 83.27% and 79.88%, respectively, and corresponding classification accuracies of 83% and 80%. Further, we have implemented SwinTransGNB to compare the proposed models. The SwinTransGNB scored 94% detection accuracy and 93% classification accuracy. Our proposed methodology establishes a significant benchmark in the realm of leaf disease detection and classification, paving the way for more efficient, accurate, and adaptable solutions to revolutionize precision agriculture.