Leaf Condition Classification Model of Apple Tree Leaves Using Machine Learning

Abstract

Throughout time, agriculture has faced the constant challenge of determining whether plants are sick or healthy, which often leads to the application of incorrect measures. For this reason, a study was conducted to discern the health of apple plants from the analysis of their leaves. For this purpose, 5052 images were collected from websites such as GitHub and Kaggle, classifying them into healthy leaves and diseased leaves. Several image processing techniques were applied, such as grayscale, edge detection, and Salt and pepper noise filtering. Subsequently, several learning models were developed, such as Support Vector Machine, Logistic Regression, MobileNet, and Vision Transformer. The results were evaluated using metrics such as Accuracy, Precision, Recall, and F1 score. Among all models, the Vision Transformer algorithm proved to be the most effective, with superior metrics: 96.02% Accuracy and 96.02% Recall. This approach offers promising prospects for accurate identification of apple plant health status, providing a valuable tool for decision-making in agriculture. © The Author(s), under exclusive license to Springer Nature Switzerland AG 2024.