Scholarlink

Rice is a pivotal crop serving the majority of the population and is vulnerable to diseases that can reduce yields by 20–40%. The early identification of these illnesses can have a beneficial effect on the crop. Consequently, producers must be well- aware of the identification of these diseases visually. Even with their best efforts, farmers could never inspect the huge farmlands on a daily basis. An efficient approach for analyzing and classifying rice leaves illness based on deep transfer learning has been presented in this study. An improved method based on pre-trained MobileNets is used in the revised strategy. This method can correctly identify and diagnose six different classes of rice plants. The dataset of rice crop images has been taken from Kaggle repository. Data has been preprocessed and balanced using SMOTE approach. These classes include healthy, narrow brown spot, leaf scald, brown spot, bacterial leaf blight, and brown blast. Training accuracy for the MobileNet model was [99.69%] while testing accuracy was [98.57%]. Other performance matrix on the testing set included [98.60%] F1-score, [98.60%] recall, [98.60%] precision. Comparison done with a base VGG19 model shows that MobileNet significantly performed well, achieving higher scores across all +. When tested against comparable methods in the existing literature this modified strategy generated noticeably superior results.

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

Transfer Learning-Based Pre-trained Model for Classification and Detection of Rice Leaf Disease

  • 会议论文

  • 发表日期：

    2025年

查看全文
  • Niharika Sharma,
  • Jaimala Jha

摘要

Rice is a pivotal crop serving the majority of the population and is vulnerable to diseases that can reduce yields by 20–40%. The early identification of these illnesses can have a beneficial effect on the crop. Consequently, producers must be well- aware of the identification of these diseases visually. Even with their best efforts, farmers could never inspect the huge farmlands on a daily basis. An efficient approach for analyzing and classifying rice leaves illness based on deep transfer learning has been presented in this study. An improved method based on pre-trained MobileNets is used in the revised strategy. This method can correctly identify and diagnose six different classes of rice plants. The dataset of rice crop images has been taken from Kaggle repository. Data has been preprocessed and balanced using SMOTE approach. These classes include healthy, narrow brown spot, leaf scald, brown spot, bacterial leaf blight, and brown blast. Training accuracy for the MobileNet model was [99.69%] while testing accuracy was [98.57%]. Other performance matrix on the testing set included [98.60%] F1-score, [98.60%] recall, [98.60%] precision. Comparison done with a base VGG19 model shows that MobileNet significantly performed well, achieving higher scores across all +. When tested against comparable methods in the existing literature this modified strategy generated noticeably superior results.