Penerapan Deep Learning Dengan Algoritma CNN Untuk Klasifikasi Varietas Beras Menggunakan Arsitektur MobileNetV3

Rice is one of the staple foods consumed by most people, especially in Asia. Manual identification of rice varieties requires considerable time and specialized skills. To overcome this, this study proposes the use of MobileNetV3 architecture in developing a rice variety classification model. The dataset used consists of seven types of rice, namely Arborio, Basmati, Ispala, Jasmine, Karacadag, Konawe, and Konjac, with a total of 700 samples obtained from Kaggle and direct documentation. The preprocessing stages in this study include resizing, normalization, conversion to RGB format, and feature extraction. The data was then divided with a ratio of 70% for training, 20% for validation, and 10% for testing. The model was developed using transfer learning method and evaluated based on accuracy, precision, recall, and F1-score metrics. The results showed that MobileNetV3-Large achieved an accuracy rate of 99%, which was superior to the VGG19 model that only achieved 96% accuracy. Thus, MobileNetV3-Large proved to have good performance in classifying rice varieties.