Implementation of Machine Learning for Disease Detection in Tomato Plants Using Convolutional Neural Networks

Abstract

Diseases in tomato plants can be highly detrimental to tomato farmers, with common afflictions such as begomovirus, blight, and spider mites posing significant challenges. The implementation of machine learning offers a promising solution to address these issues and mitigate the financial losses caused by such diseases. This study aims to evaluate the effectiveness of machine learning in detecting plant diseases using Convolutional Neural Networks (CNN). The data used in this implementation was obtained from public datasets available on Kaggle and real-time data collected directly from tomato farms in Kadudampit, Sukabumi Regency. The Kaggle dataset contains 4,800 images categorized into three classes: begomovirus, blight, and spider mites. Meanwhile, the real-time dataset comprises 450 images, also divided into the same three classes. The performance of the machine learning model was tested using different datasets, with accuracy measured through a confusion matrix. The results showed that the machine learning model trained on the public dataset achieved the highest accuracy of 97%. The model trained on a combination of the public and real-time datasets achieved an accuracy of 94%, while the model trained solely on the real-time dataset achieved an accuracy of 80%. A machine learning model is considered effective if its accuracy exceeds 75%. Therefore, based on the three tests conducted, it can be concluded that the machine learning models demonstrated a good level of accuracy in detecting diseases in tomato plants