Scholarlink

Agriculture, a bedrock of human civilization, not only sustains economies but also feeds nations. However, the vitality of crops is often undermined by diseases that afflict plant leaves and crops, stunting their growth. The timely and accurate identification and categorization of these diseases are imperative to curbing their spread and minimizing crop damage. The research explores the effectiveness of transfer learning models in the critical domain of plant disease detection. Utilizing the expansive Plant Disease Expert Dataset featuring 58 diverse classes, this research employs image augmentation to mitigate model bias towards dominant classes. The study rigorously assesses five notable transfer learning models: DenseNet169, MobileNetV2, InceptionResNetV2, DenseNet201, and EfficientNetB6. Notably, EfficientNetB6 stands out as the top performer, achieving an impressive testing accuracy of 99.11%. The other transfer learning models also demonstrated commendable performance.

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

Comparative Evaluation of Transfer Learning Models for Accurate Plant Disease Detection

  • 会议论文

  • 发表日期：

    2025年

查看全文
  • Amol Satsangi,
  • Shaurya Jain,
  • Sajal Satsangi

摘要

Agriculture, a bedrock of human civilization, not only sustains economies but also feeds nations. However, the vitality of crops is often undermined by diseases that afflict plant leaves and crops, stunting their growth. The timely and accurate identification and categorization of these diseases are imperative to curbing their spread and minimizing crop damage. The research explores the effectiveness of transfer learning models in the critical domain of plant disease detection. Utilizing the expansive Plant Disease Expert Dataset featuring 58 diverse classes, this research employs image augmentation to mitigate model bias towards dominant classes. The study rigorously assesses five notable transfer learning models: DenseNet169, MobileNetV2, InceptionResNetV2, DenseNet201, and EfficientNetB6. Notably, EfficientNetB6 stands out as the top performer, achieving an impressive testing accuracy of 99.11%. The other transfer learning models also demonstrated commendable performance.