Implementasi Algoritma Convolutional Neural Network (CNN) dalam Klasifikasi Jenis Serangga Hama pada Gudang Beras Menggunakan Model MobilenetV3-Large Berbasis Website

Abstract

Pest insects pose a significant challenge in rice storage warehouses, leading to a decrease in both the quality and quantity of rice during storage. The traditional manual identification of pest insects is often inefficient, time-consuming, and lacks accuracy. This study aims to develop a web-based pest insect classification system for rice storage warehouses using the Convolutional Neural Network (CNN) algorithm with the MobileNetV3-Large architecture. The model was trained using 800 datasets per class, with the Adam optimizer (learning rate 0.0001) and 20 training epochs. The experimental results demonstrate that the model achieved an accuracy of 95% without overfitting. Evaluation through a confusion matrix revealed consistent performance with balanced precision, recall, and F1-score across all classes. The proposed system includes a user-friendly web interface that allows users to upload pest insect images for classification. Image preprocessing steps, such as resizing, normalization, and augmentation, ensure the data is suitable for the MobileNetV3-Large model. The classification results are displayed to users in both textual and visual formats, enabling quick and accurate identification. Tests conducted on real-world images confirmed the system's reliability and robustness. This solution enables efficient pest insect identification, facilitating faster decision-making for pest control, and supports maintaining the quality and quantity of rice during storage.