Scholarlink

Detecting durian diseases is crucial for improving durian quality and economic value in agriculture. This project aims to develop a durian disease recognition system that identifies specific diseases on durian leaves and husks through image analysis. The system can recognize multiple durian diseases, Durian Leaf and Husk Diseases. A diverse image dataset containing samples of various diseases and healthy durians has been prepared to train and test the durian disease recognition system, sourced from Kaggle and Roboflow. The dataset includes six categories of images: durian leaves infected with Algal Leaf Spot, Leaf Blight, Leaf Spot, healthy durian leaves, durian husks infected with Phytophthora, and healthy durian husks. Using this diverse dataset, the system can learn the unique visual characteristics of each disease, enabling accurate recognition. This work used four models for comparison, including Convolutional Neural Networks (CNN) and three pre-trained models: VGG16, Inception, and AlexNet. Experimental results indicate that the Inception model performed the best, achieving an accuracy of 97.4%. The proposed Inception model effectively classifies durian diseases, providing robust support for preventing disease spread and improving durian quality.

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

Durian Fruit Diseases Detection Using Convolutional Neural Networks (CNN) and Pre-trained Models

  • 会议论文

  • 发表日期：

    2025年

查看全文
  • Ahmed Abed Mohammed,
  • Hasan Alqaseer,
  • Putra Sumari,
  • Mustafa M. Abd Zaid

摘要

Detecting durian diseases is crucial for improving durian quality and economic value in agriculture. This project aims to develop a durian disease recognition system that identifies specific diseases on durian leaves and husks through image analysis. The system can recognize multiple durian diseases, Durian Leaf and Husk Diseases. A diverse image dataset containing samples of various diseases and healthy durians has been prepared to train and test the durian disease recognition system, sourced from Kaggle and Roboflow. The dataset includes six categories of images: durian leaves infected with Algal Leaf Spot, Leaf Blight, Leaf Spot, healthy durian leaves, durian husks infected with Phytophthora, and healthy durian husks. Using this diverse dataset, the system can learn the unique visual characteristics of each disease, enabling accurate recognition. This work used four models for comparison, including Convolutional Neural Networks (CNN) and three pre-trained models: VGG16, Inception, and AlexNet. Experimental results indicate that the Inception model performed the best, achieving an accuracy of 97.4%. The proposed Inception model effectively classifies durian diseases, providing robust support for preventing disease spread and improving durian quality.