Scholarlink

<p>A basic task in computer vision is image classification, which involves examining patterns and features to group images into specified labels. Images of strawberry and cherry leaves from the PlantVillage dataset are used to differentiate between diseased and healthy leaves. Feature extraction was performed through a fine-tuned VGG16 model, extracting relevant high-level characteristics from images of both datasets by unfreezing the higher convolutional layers and replacing the fully connected layer so that the network adapts to the plant leaf dataset. The extracted characteristics were then individually fed into several classification methods, such as <i>K-Nearest Neighbours (k-NN), Random Forests (RF), Support Vector Machines (SVM), and Gradient Boosting</i> etc. To discover the suitable algorithm for the task, the performance of classifiers was evaluated using multiple evaluation measures. Therefore, the objectives of the research paper are threefold: firstly, to review the literature on ongoing image classification methods and how they are used for agricultural datasets. Secondly, to build and propose a framework that integrates conventional machine learning classifiers with deep learning-based feature extraction. Finally, to implement the proposed framework into practice and evaluate it using two standard datasets. With classifiers like SVM and Logistic Regression (LR), the results showed that combining deep learning with machine learning produced high effectiveness consistently, with an accuracy of 99.60%. On the other hand, classifiers such as Naïve Bayes and k-NN proclaimed lower accuracy for both datasets, emphasising the variation in algorithm performance.</p>

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

A Comparative Analysis of Plant Leaf Disease Detection Combining Deep Learning and Machine Learning Models

  • 期刊论文

  • 发表日期：

    2026年5月6日

查看全文
  • Smriti Bansal,
  • Supreet Thapar,
  • Anil Kumar

摘要

A basic task in computer vision is image classification, which involves examining patterns and features to group images into specified labels. Images of strawberry and cherry leaves from the PlantVillage dataset are used to differentiate between diseased and healthy leaves. Feature extraction was performed through a fine-tuned VGG16 model, extracting relevant high-level characteristics from images of both datasets by unfreezing the higher convolutional layers and replacing the fully connected layer so that the network adapts to the plant leaf dataset. The extracted characteristics were then individually fed into several classification methods, such as K-Nearest Neighbours (k-NN), Random Forests (RF), Support Vector Machines (SVM), and Gradient Boosting etc. To discover the suitable algorithm for the task, the performance of classifiers was evaluated using multiple evaluation measures. Therefore, the objectives of the research paper are threefold: firstly, to review the literature on ongoing image classification methods and how they are used for agricultural datasets. Secondly, to build and propose a framework that integrates conventional machine learning classifiers with deep learning-based feature extraction. Finally, to implement the proposed framework into practice and evaluate it using two standard datasets. With classifiers like SVM and Logistic Regression (LR), the results showed that combining deep learning with machine learning produced high effectiveness consistently, with an accuracy of 99.60%. On the other hand, classifiers such as Naïve Bayes and k-NN proclaimed lower accuracy for both datasets, emphasising the variation in algorithm performance.