Multi-Class Classification of Rice Leaf Diseases Using MobileNetV2 Architecture with Transfer Learning
Journal of Applied Informatics and Computing · 24.04.2026 · 10 (2) · 1896-1904
DOI: 10.30871/jaic.v10i2.12499 ↗Kurz gefasst
Fachartikel in der Zeitschrift Journal of Applied Informatics and Computing, online erschienen am 24.04.2026. Verfasst von Muhammad Bisri Mustofa und 2 weiteren Autor:innen. Der Volltext ist frei zugänglich (Open Access, CC BY-SA 4.0).
Abstract (Originalsprache)
Rice leaf diseases are a major cause of yield loss and remain a persistent challenge in agriculture. Conventional diagnosis through visual inspection is subjective and time-consuming, necessitating an accurate and efficient automated system for early detection. This study develops and evaluates a multi-class rice leaf disease classifier based on digital images using MobileNetV2 with transfer learning and fine-tuning of the last 20 layers. The dataset comprises 6,180 images distributed equally across four classes brown spot, leaf blast, leaf blight, and normal with an imbalance ratio of 1.00:1, split into 70% training, 15% validation, and 15% testing using stratified random sampling. The methodology incorporates image normalization, adaptive data augmentation, and class weighting as preventive regularization strategies. The model achieves 97.20% overall test accuracy, with precision of 97.30%, recall of 97.09%, F1-score of 97.19%, and AUC of 99.72%, converging optimally at epoch 25 across 30 training epochs. Per-class accuracy ranges from 96.12% (brown spot) to 97.84% (leaf blast and normal), with misclassifications primarily occurring between visually similar disease classes. With a compact model size of 9.89 MB (FP32) reducible to 2.71 MB via TFLite conversion, and an average inference time of 144.80 ms per image, the proposed model demonstrates high efficiency and strong generalization, making it well-suited for deployment on mobile devices as a practical tool for real-time early rice disease detection in field settings.
Zitieren
APA
Mustofa, M. B., Arifin, M., & Zulfia, S. (2026). Multi-Class Classification of Rice Leaf Diseases Using MobileNetV2 Architecture with Transfer Learning. Journal of Applied Informatics and Computing, 10(2), 1896-1904. https://doi.org/10.30871/jaic.v10i2.12499
Vancouver
Mustofa MB, Arifin M, Zulfia S. Multi-Class Classification of Rice Leaf Diseases Using MobileNetV2 Architecture with Transfer Learning. Journal of Applied Informatics and Computing. 2026;10(2):1896-1904. doi:10.30871/jaic.v10i2.12499
BibTeX
@article{Mustofa2026,
author = {Mustofa, Muhammad Bisri and Arifin, Muhammad and Zulfia, Shofiatuz},
title = {Multi-Class Classification of Rice Leaf Diseases Using MobileNetV2 Architecture with Transfer Learning},
journal = {Journal of Applied Informatics and Computing},
year = {2026},
volume = {10},
number = {2},
pages = {1896--1904},
doi = {10.30871/jaic.v10i2.12499}
}
EUVIMED