Scholarlink

This paper focuses on comparing the effectiveness of four different neural network architectures in the classification of plant health, based on their leaf images, utilizing the PlantVillage dataset specialized in tomatoes. The architectures under investigation include AlexNet, GoogLeNet, VGG16, and Vision Transformer. The evaluation has been conducted under two main configurations: colour images and grayscale images. Additionally, the study compares the performance achieved using the Adam and SGD optimizers. The results of the experiments highlight significant differences in classification effectiveness among the architectures studied, indicating the potential for further optimization through different image configurations and optimizers. This study aims to provide deeper insights into selecting the appropriate neural network architecture for effective plant health recognition, with direct applications in agriculture and crop monitoring.

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

Application of Selected Machine Learning Models in Leaf-Based Plant Health Assessment

  • 会议论文

  • 发表日期：

    2025年

查看全文
  • Jakub Filipek,
  • Marek Pawlicki,
  • Ryszard Choraś,
  • Rafał Kozik,
  • Aleksandra Pawlicka,
  • Michał Choraś

摘要

This paper focuses on comparing the effectiveness of four different neural network architectures in the classification of plant health, based on their leaf images, utilizing the PlantVillage dataset specialized in tomatoes. The architectures under investigation include AlexNet, GoogLeNet, VGG16, and Vision Transformer. The evaluation has been conducted under two main configurations: colour images and grayscale images. Additionally, the study compares the performance achieved using the Adam and SGD optimizers. The results of the experiments highlight significant differences in classification effectiveness among the architectures studied, indicating the potential for further optimization through different image configurations and optimizers. This study aims to provide deeper insights into selecting the appropriate neural network architecture for effective plant health recognition, with direct applications in agriculture and crop monitoring.