Scholarlink

Most plant diseases exhibit visual symptoms, and the approach used nowadays by an expert plant pathologist to assess the condition by visual inspection of diseased plant leaves. The reality that manually diagnosing diseases is time-consuming, and the effectiveness of the disease detection is related to the pathologist’s abilities, makes it an ideal application to diagnose using computer-aided systems. Instead of relying on traditional machine learning techniques that require manual feature extraction for optimal results, this research utilizes deep convolutional neural network models like VGG19, ResNet50, DenseNet201, and EfficientNetB7. These models eliminate the need for preprocessing and offer effective classification capabilities when identifying plant diseases from pictures of plant leaves. The PlantVillage dataset was utilized to train models and correctly categorize them into 38 different categories. Models based on VGG19, ResNet50, DenseNet201 and EfficientNetB7 are suggested, with test accuracy of 96.62%, 98.62%, 99.38%, and 98.63%, respectively.

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

An Efficient Plant Disease Classification Using Transfer Learning

  • 会议论文

  • 发表日期：

    2025年

查看全文
  • Lakshmi Narayana Chintala,
  • G. Sriram,
  • Venkata Ramana Kondapalli

摘要

Most plant diseases exhibit visual symptoms, and the approach used nowadays by an expert plant pathologist to assess the condition by visual inspection of diseased plant leaves. The reality that manually diagnosing diseases is time-consuming, and the effectiveness of the disease detection is related to the pathologist’s abilities, makes it an ideal application to diagnose using computer-aided systems. Instead of relying on traditional machine learning techniques that require manual feature extraction for optimal results, this research utilizes deep convolutional neural network models like VGG19, ResNet50, DenseNet201, and EfficientNetB7. These models eliminate the need for preprocessing and offer effective classification capabilities when identifying plant diseases from pictures of plant leaves. The PlantVillage dataset was utilized to train models and correctly categorize them into 38 different categories. Models based on VGG19, ResNet50, DenseNet201 and EfficientNetB7 are suggested, with test accuracy of 96.62%, 98.62%, 99.38%, and 98.63%, respectively.