Klasifikasi Jenis Penyakit Tanaman Padi Menggunakan Arsitektur Mobilnetv2 Berbasis Transfer Learning Di Desa Karangraharja, Kabupaten Bekasi

Abstract

Rice is an important food commodity in Indonesia, but its productivity is often affected by plant diseases that can reduce yields. In Bekasi Regency, particularly Karangraharja Village, rice plant disease identification is still done manually through visual observation of leaves, making it subjective and dependent on farmer experience. This condition has the potential to cause errors in disease classification and management. Technological developments allow the classification of rice disease types to be carried out using image processing. The purpose of this study was to classify rice plant diseases based on leaf images, according to their disease class using transfer learning techniques. The research stages were carried out using the CRISP-DM methodology, which includes business understanding, data understanding, data preparation, modeling, evaluation, and deployment. The image data used amounted to 6.166 rice leaf images. The classification process was carried out using the MobileNetV2 architecture. The best scenario was obtained by the fine-tuning method with a data division of 80% training, 10% validation, and 10% testing, which resulted in an accuracy of 99% on the testing data and 95% on the new data.