Scholarlink

This research presents a comprehensive study on the development of an efficient and accurate plant disease prediction system using transfer learning models. The objective is to provide a reliable and accessible solution for farmers and agronomists to identify and address plant diseases in a timely manner, leading to improved crop yields and sustainable farming practices. The study analyzes five transfer learning models: Efficientnet_v2, Inception_v3, Mobilenet_v2, Resnet_v2, and Nasnet. Among these models, Efficientnet_v2, Resnet_v2, and Mobilenet_v2 demonstrate superior performance and are selected for further analysis. The research evaluates different experimental conditions: Global Model, Crop-Specific Approach, Disease-Specific Approach, and Plant-Family Based Approach. The Plant-Family Based Approach, focusing on the Nightshade plant family, exhibits superior accuracy in disease detection. An ensemble model combining three transfer learning models is developed using the Plant-Family Based Approach. The ensemble model shows exceptional performance and 94% accuracy. Extensive testing and validation using diverse datasets demonstrate the system's high accuracy and efficiency in detecting diseases in various plant species. Future work includes expanding the dataset, fine-tuning hyperparameters, enabling real-time disease monitoring, and collaborating with agricultural experts to incorporate domain knowledge.

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 Disease Detection with Transfer Learning and Ensemble Techniques Based on Plant Families

  • 会议论文

  • 发表日期：

    2026年

查看全文
  • A. Mohamed Aslam Sujah,
  • Asanka P. Sayakkara,
  • H. B. P. Sandani

摘要

This research presents a comprehensive study on the development of an efficient and accurate plant disease prediction system using transfer learning models. The objective is to provide a reliable and accessible solution for farmers and agronomists to identify and address plant diseases in a timely manner, leading to improved crop yields and sustainable farming practices. The study analyzes five transfer learning models: Efficientnet_v2, Inception_v3, Mobilenet_v2, Resnet_v2, and Nasnet. Among these models, Efficientnet_v2, Resnet_v2, and Mobilenet_v2 demonstrate superior performance and are selected for further analysis. The research evaluates different experimental conditions: Global Model, Crop-Specific Approach, Disease-Specific Approach, and Plant-Family Based Approach. The Plant-Family Based Approach, focusing on the Nightshade plant family, exhibits superior accuracy in disease detection. An ensemble model combining three transfer learning models is developed using the Plant-Family Based Approach. The ensemble model shows exceptional performance and 94% accuracy. Extensive testing and validation using diverse datasets demonstrate the system's high accuracy and efficiency in detecting diseases in various plant species. Future work includes expanding the dataset, fine-tuning hyperparameters, enabling real-time disease monitoring, and collaborating with agricultural experts to incorporate domain knowledge.