Scholarlink

This research paper investigates the optimization of Convolutional Neural Network (CNN) models and their optimizers for leaf disease classification, including the development of a custom CNN model. The study systematically evaluates various pretrained CNN architectures, focusing on popular models from the ResNet and MobileNet families, along with optimization algorithms such as Adam, Stochastic Gradient Descent, AdaDelta, and RMSprop. Through extensive experiments and comparative analyses, the research identifies the most effective model-optimizer pairings based on performance metrics like accuracy, Precision, Recall, and F1 Score. Among the combinations evaluated, the ResNet-152 V2 model with the Adam optimizer emerged as the top performer, achieving remarkable results: 99.12% accuracy, 99.21% Precision, 99.04% Recall, and 99.17% F1 Score.Additionally, the Custom CNN model demonstrated competitive results, with 89.12% accuracy, 87.65% Precision, 88.34% Recall, and 86.09% F1 Score.This research highlights the effectiveness of the ResNet-152 V2 model paired with the Adam optimizer for plant disease diagnosis, significantly enhancing accuracy in automated leaf disease classification and paving the way for more advanced agricultural management solutions.

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

Exploring CNN Model Optimization Strategies for Plant Disease Classification

  • 会议论文

  • 发表日期：

    2025年

查看全文
  • Abhishek Dwivedi,
  • Lalit Birla,
  • Chiranjit Pal,
  • Arvinder Kaur

摘要

This research paper investigates the optimization of Convolutional Neural Network (CNN) models and their optimizers for leaf disease classification, including the development of a custom CNN model. The study systematically evaluates various pretrained CNN architectures, focusing on popular models from the ResNet and MobileNet families, along with optimization algorithms such as Adam, Stochastic Gradient Descent, AdaDelta, and RMSprop. Through extensive experiments and comparative analyses, the research identifies the most effective model-optimizer pairings based on performance metrics like accuracy, Precision, Recall, and F1 Score. Among the combinations evaluated, the ResNet-152 V2 model with the Adam optimizer emerged as the top performer, achieving remarkable results: 99.12% accuracy, 99.21% Precision, 99.04% Recall, and 99.17% F1 Score.Additionally, the Custom CNN model demonstrated competitive results, with 89.12% accuracy, 87.65% Precision, 88.34% Recall, and 86.09% F1 Score.This research highlights the effectiveness of the ResNet-152 V2 model paired with the Adam optimizer for plant disease diagnosis, significantly enhancing accuracy in automated leaf disease classification and paving the way for more advanced agricultural management solutions.