Aplikasi Klasifikasi Penyakit Pada Tanaman Jambu Air Menggunakan Metode CNN

Abstract

Diseases in water apple plants can significantly reduce the quality and quantity of the harvest, thus an accurate early classification system is needed. This research aims to develop an automatic classification system to assist in the early diagnosis of diseases in water apple plants. This system uses a Convolutional Neural Network (CNN) model with the MobileNetV2 architecture optimized through fine-tuning techniques. The dataset was obtained from Kaggle, Roboflow, and direct data collection, with a distribution of 80 percent training data, 10 percent validation data, and 10 percent testing data. The model was trained to recognize nine different conditions from images of leaves and fruits, then integrated into a web application. Testing results showed very good performance, with a final accuracy of 94 percent and a balanced F1-Score of 93 percent on unseen test data. However, the model faces challenges in distinguishing diseases with high visual similarity, especially in the class of leaves with brown spots. Overall, this research successfully produced an effective and accurate classification system. The developed application has high practical potential as an early diagnosis aid to improve plant health management and reduce the potential for crop loss.