Model Convolutional Neural Network berbasis Data Lapangan untuk Deteksi Penyakit Daun Padi

Abstrak

Early detection of rice leaf diseases is an important component in maintaining agricultural productivity, especially since rice is a major commodity in Indonesia. However, research on Convolutional Neural Networks (CNN) still uses homogeneous laboratory datasets, which are less than optimal when applied to varying field conditions. This study aims to develop a CNN-based system for classifying rice leaf diseases using field images. This type of research is development using a waterfall model. These stages consist of needs analysis with pre-processing in the form of resizing and augmentation, design, and implementation in a web application. Testing was conducted using a black box to examine the functionality of the system. Our findings show that the system is capable of classifying leaf diseases well based on the field data and functions properly. The training results show an accuracy of 94.64%, precision of 0.93, recall of 0.90, and F1-score of 0.91, with the best performance in the blast and blight classes. Classification errors mainly occurred in the tungro class due to its visual similarity to blast in the early stages of infection. The developed system is capable of performing automatic diagnosis through image uploads and demonstrates learning stability based on the accuracy and loss curves. This research contributes by providing an end-to-end rice leaf disease detection system based on accurate, adaptive, and ready-to-use field data to support direct plant health monitoring in agricultural environments.

Identifier

• DOI: https://doi.org/10.29408/edumatic.v9i3.32634
• OpenAlex: https://openalex.org/W7124683597