kaz29Вестник Казахстанско-Британского технического университетаHerald of the Kazakh-British Technical University1998-66882959-8109Казахстанско-Британский Технический Университет10.55452/1998-6688-2025-22-3-149-160kaz29-2112Research ArticleКОМПЬЮТЕРНЫЕ НАУКИCOMPUTER SCIENCEАНАЛИЗ ЭФФЕКТИВНОСТИ ТРАНСФОРМЕРОВ И СВЕРТОЧНЫХ НЕЙРОСЕТЕЙ ДЛЯ КЛАССИФИКАЦИИ ЗАБОЛЕВАНИЙ РИСАCOMPARATIVE ANALYSIS OF THE EFFECTIVENESS OF TRANSFORMER AND CONVOLUTIONAL NEURAL NETWORK ARCHITECTURES FOR AUTOMATIC CLASSIFICATION OF RICE LEAF DISEASEShttps://orcid.org/0009-0008-5800-9960ДжураевД. Б.JurayevD. B.

бакалавр

г. Алматы

Bachelor student 

 Almaty 

doninadirov@mail.ruhttps://orcid.org/0000-0003-3853-8896УалиеваИ. М.UaliyevaI. M.

к.ф.-м.н., ассоциированный профессор

г. Алматы

Cand.Phys.-Math.Sc., Associate Professor

Almaty

i.ualiyeva@gmail.comhttps://orcid.org/0000-0002-1141-7595АкжаловаА. Ж.AkzhalovaA. Zh.

доктор PhD, профессор

г. Алматы 

PhD, professor

Almaty

a.akzhalova@kbtu.kzКазахский национальный университет им. аль-ФарабиКазахстанAl-Farabi Kazakh National UniversityKazakhstanКазахстанско-Британский технический университетКазахстанKazakh-British Technical UniversityKazakhstan202527092025223149160Copyright © Джураев Д.Б., Уалиева И.М., Акжалова А.Ж., 20252025Джураев Д.Б., Уалиева И.М., Акжалова А.Ж.Jurayev D.B., Ualiyeva I.M., Akzhalova A.Z.Данная работа распространяется под лицензией Creative Commons Attribution 4.0.This work is licensed under a Creative Commons Attribution 4.0 License.https://vestnik.kbtu.edu.kz/jour/article/view/2112

В данной статье представлен сравнительный анализ современных архитектур нейросетей, сверточных нейронных сетей (CNN) и трансформеров для автоматической диагностики заболеваний листьев риса. В рамках экспериментов были обучены и протестированы модели DenseNet121, ResNet, Vision Transformer (ViT) и MaxViT, после чего проведена их оценка по точности и вычислительной эффективности. Исследование выполнено на основе крупномасштабного датасета, включающего реальные изображения больных и здоровых листьев риса, что делает результаты актуальными для сельскохозяйственной науки и практики. Эксперименты включали оптимизацию гиперпараметров, применение методов аугментации данных, а также использование функций потерь и методов регуляризации с целью повышения обобщающей способности моделей. Для оценки качества использовались метрики точности классификации, F1-мера, а также показатели вычислительной эффективности, такие как время предсказания и объем потребляемых ресурсов. Полученные результаты показали, что модели на основе трансформеров, в частности MaxViT, достигают точности до 94,10%. Это связано с их способностью эффективно моделировать как локальные, так и глобальные признаки изображений за счет механизмов внимания и глубокой контекстуализации. В то же время CNN-архитектуры, такие как DenseNet121 и ResNet, демонстрируют высокую скорость работы и устойчивость в условиях ограниченных вычислительных ресурсов.

This article presents a comparative analysis of modern neural network architectures, convolutional neural networks (CNNs) and transformers, for the automatic diagnosis of rice leaf diseases. In the experiments, DenseNet121, ResNet, Vision Transformer (ViT), and MaxViT models were trained and tested, followed by their evaluation in terms of accuracy and computational efficiency. The study was conducted on a large-scale dataset containing real images of healthy and diseased rice leaves, which makes the results highly relevant for agricultural science and practice. The experiments included hyperparameter optimization, application of data augmentation techniques, and the use of loss functions and regularization methods to improve the generalization ability of the models. The evaluation metrics comprised classification accuracy, F1-score, as well as computational efficiency indicators such as prediction time and resource consumption. The results showed that transformer-based models, particularly MaxViT, achieve accuracy of up to 94.10%. This is attributed to their ability to effectively capture both local and global image features through attention mechanisms and deep contextualization. At the same time, CNN architectures such as DenseNet121 and ResNet demonstrate high processing speed and robustness under limited computational resources.

глубокое обучениесверточные нейронные сетиVision TransformerMaxViTдиагностика растенийрисклассификация изображенийdeep learningconvolutional neural networksVision TransformerMaxViTplant disease diagnosisriceimage classificationReferencesAhad, M.T., Li, Y., Song, B., Bhuiyan, T. Comparison of CNN-based deep learning architectures for rice diseases classification. Artificial Intelligence in Agriculture, 9, 22–35 (2023). https://doi.org/10.1016/j.aiia.2023.07.001.Ahad, M.T., Li, Y., Song, B., Bhuiyan, T. Comparison of CNN-based deep learning architectures for rice diseases classification. Artificial Intelligence in Agriculture, 9, 22–35 (2023). https://doi.org/10.1016/j.aiia.2023.07.001.Sethy, P.K., Barpanda, N.K., Rath, A.K., Behera, S.K. Deep feature based rice leaf disease identification using support vector machine. Computers and Electronics in Agriculture, 175, 105527 (2020). https://doi.org/10.1016/j.compag.2020.105527.Sethy, P.K., Barpanda, N.K., Rath, A.K., Behera, S.K. Deep feature based rice leaf disease identification using support vector machine. Computers and Electronics in Agriculture, 175, 105527 (2020). https://doi.org/10.1016/j.compag.2020.105527.Simhadri, C.G., Kondaveeti, H.K., Vatsavayi, V.K., Mitra, A., Ananthachari, P. Deep learning for rice leaf disease detection: A systematic literature review on emerging trends, methodologies and techniques. Information Processing in Agriculture (2024, May). https://doi.org/10.1016/j.inpa.2024.04.006.Simhadri, C.G., Kondaveeti, H.K., Vatsavayi, V.K., Mitra, A., Ananthachari, P. Deep learning for rice leaf disease detection: A systematic literature review on emerging trends, methodologies and techniques. Information Processing in Agriculture (2024, May). https://doi.org/10.1016/j.inpa.2024.04.006.Dosovitskiy, A., Beyer, L., Kolesnikov, A., Weissenborn, D., Zhai, X., Unterthiner, T. et al. An image is worth 16x16 words: Transformers for image recognition at scale. arXiv preprint (2020). URL: https://arxiv.org/abs/2010.11929.Dosovitskiy, A., Beyer, L., Kolesnikov, A., Weissenborn, D., Zhai, X., Unterthiner, T. et al. An image is worth 16x16 words: Transformers for image recognition at scale. arXiv preprint (2020). URL: https://arxiv.org/abs/2010.11929.Haque, M.A., Pastor-Escuredo, D., Brigui, I., Kesswani, N., Bordoloi, S., Ray, A. K. Rice disease identification using Vision Transformer (ViT) based network. The Future of Artificial Intelligence and Robotics (Cham: Springer, 2024), pp. 732–741. https://doi.org/10.1007/978-3-031-60935-0_63.Haque, M.A., Pastor-Escuredo, D., Brigui, I., Kesswani, N., Bordoloi, S., Ray, A. K. Rice disease identification using Vision Transformer (ViT) based network. The Future of Artificial Intelligence and Robotics (Cham: Springer, 2024), pp. 732–741. https://doi.org/10.1007/978-3-031-60935-0_63.Mhaned, A., Salma, M., El Haji, M., Jamal, B. Plant disease detection using vision transformers. International Journal of Electrical and Computer Engineering, 15 (2), 2334–2344 (2025). https://doi.org/10.11591/ijece.v15i2.Mhaned, A., Salma, M., El Haji, M., Jamal, B. Plant disease detection using vision transformers. International Journal of Electrical and Computer Engineering, 15 (2), 2334–2344 (2025). https://doi.org/10.11591/ijece.v15i2.Zhang, H. Attention-based feature enhancement for rice leaf disease recognition. Proc. 2nd Int. Conf. Artificial Intelligence and Automation in High-Performance Computing (AIAHPC 2022), pp. 41–49. https://doi.org/10.1117/12.2641832.Zhang, H. Attention-based feature enhancement for rice leaf disease recognition. Proc. 2nd Int. Conf. Artificial Intelligence and Automation in High-Performance Computing (AIAHPC 2022), pp. 41–49. https://doi.org/10.1117/12.2641832.Tharani, P., Baranidharan, B. A hybrid ViT-CNN model premeditated for rice leaf disease identification. International Journal of Computational Methods and Experimental Measurements, 12 (1), 35–43 (2024). https://doi.org/10.18280/ijcmem.120104.Tharani, P., Baranidharan, B. A hybrid ViT-CNN model premeditated for rice leaf disease identification. International Journal of Computational Methods and Experimental Measurements, 12 (1), 35–43 (2024). https://doi.org/10.18280/ijcmem.120104.Mustofa, S., Munna, M., Emon, Y., Rabbany, G., Ahad, M. A comprehensive review on plant leaf disease detection using deep learning // arXiv preprint (2023). URL: https://arxiv.org/abs/2308.14087.Mustofa, S., Munna, M., Emon, Y., Rabbany, G., Ahad, M. A comprehensive review on plant leaf disease detection using deep learning // arXiv preprint (2023). URL: https://arxiv.org/abs/2308.14087.Li, Z., Liu, F., Yang, W., Peng, S., Zhou, J. A survey of convolutional neural networks: Analysis, applications, and prospects. IEEE Transactions on Neural Networks and Learning Systems (2021), pp. 1–21. https://doi.org/10.1109/TNNLS.2021.3084827.Li, Z., Liu, F., Yang, W., Peng, S., Zhou, J. A survey of convolutional neural networks: Analysis, applications, and prospects. IEEE Transactions on Neural Networks and Learning Systems (2021), pp. 1–21. https://doi.org/10.1109/TNNLS.2021.3084827.Tu, Z., Talebi, H., Zhang, H., Yang, F., Milanfar, P., Bovik, A., Li, Y. MaxViT: Multi-axis vision transformer. arXiv preprint (2022). URL: https://arxiv.org/abs/2204.01697.Tu, Z., Talebi, H., Zhang, H., Yang, F., Milanfar, P., Bovik, A., Li, Y. MaxViT: Multi-axis vision transformer. arXiv preprint (2022). URL: https://arxiv.org/abs/2204.01697.Ahad, Md & Li, Yan & Song, Bo & Bhuiyan, Touhid. Comparison of CNN-based deep learning architectures for rice diseases classification. Artificial Intelligence in Agriculture, 9, 22–35 (2023). https://doi.org/ 10.1016/j.aiia.2023.07.001.Ahad, Md & Li, Yan & Song, Bo & Bhuiyan, Touhid. Comparison of CNN-based deep learning architectures for rice diseases classification. Artificial Intelligence in Agriculture, 9, 22–35 (2023). https://doi.org/ 10.1016/j.aiia.2023.07.001.

The authors declare that there are no conflicts of interest present.