SavedAdd your journal

Journals / Iğdır Üniversitesi Fen Bilimleri Enstitüsü Dergisi / 2025 / Cilt: 15 Sayı: 4

A Robust Apple Disease Classification Approach Using Vision Transformer-Based Meta-Learning

Görüntü Dönüştürücü Tabanlı Meta-Öğrenme ile Gürbüz Bir Elma Hastalığı Sınıflandırma Yaklaşımı

Metehan Gunde(IĞDIR ÜNİVERSİTESİ, IĞDIR TEKNİK BİLİMLER MESLEK YÜKSEKOKULU)0000-0003-3323-7525
Gültekin Işık(IĞDIR ÜNİVERSİTESİ, MÜHENDİSLİK FAKÜLTESİ)0000-0003-3037-5586
Pages
1193–1205
Publication Date
2025-12-01
🏛️Bibliographic metadata and full text access for this article are provided via DergiPark.
DergiPark Record ↗
✨

AI Summary & Key Takeaways

AI Insight
⚡ TL;DR(Too Long; Didn't Read)

These findings indicate that, despite MAML achieving better classification performance, MAML++ is more consistent and robust.

📌 3 Key Takeaways

🎯
Objective & Hypothesis:Few-shot learning has emerged as an important solution for agricultural classification problems, where labelled data is often scarce.
📊
Key Findings & Data:The performance of these approaches was evaluated in 2-way 5-shot and 4-way 1-shot scenarios.
💡
Conclusion & Impact:These findings indicate that, despite MAML achieving better classification performance, MAML++ is more consistent and robust.
🔬Design:Modeling & Simulation
📍Scope:Comprehensive Sample
⚙️Variables:Az sayıda örnekle öğrenme · Meta öğrenme · MAML

Abstract

Few-shot learning has emerged as an important solution for agricultural classification problems, where labelled data is often scarce. This study proposes the integration of the Vision Transformer (ViT) model with Model-Agnostic Meta-Learning (MAML and MAML++) algorithms to classify apple leaf diseases. The performance of these approaches was evaluated in 2-way 5-shot and 4-way 1-shot scenarios. The MAML algorithm achieved its best performance in the 2-way 5-shot scenario (96.27% accuracy, 0.1299 loss), while its worst performance was observed in the 4-way 1-shot scenario (88.80% accuracy, 0.2884 loss), resulting in a 7.47% accuracy gap. In contrast, the MAML++ algorithm demonstrated better consistency, with a smaller accuracy gap of 7.13% between its best (90.73% accuracy, 0.3580 loss) and worst (83.60% accuracy, 0.5401 loss) performances. These findings indicate that, despite MAML achieving better classification performance, MAML++ is more consistent and robust. By combining the feature extraction capabilities of Vision Transformers with meta-learning algorithms, this study presents a new approach for accurate and reliable disease classification under real-world conditions. The results demonstrate that the proposed approach has significant potential in agricultural applications, particularly in scenarios with limited labelled data.

Özet

Az örnekle öğrenme, etiketlenmiş verilerin kısıtlı olduğu tarımsal sınıflandırma problemleri için önemli bir çözüm olarak ortaya çıkmıştır. Bu çalışma elma yapraklarındaki hastalıkları sınıflandırmak amacıyla Görüntü Dönüştürücü modeli ile Model-Bağımsız Meta-Öğrenme algoritmalarını (MAML ve MAML++) birlikte kullanmayı önermektedir. Bu yaklaşımların performansı, 2-way 5-shot ve 4-way 1-shot senaryolarında değerlendirilmiştir. MAML algoritması, 2-way 5-shot senaryosunda en iyi performansını (96,27% doğruluk, 0,1299 kayıp) elde ederken, 4-way 1-shot senaryosunda en kötü performansını (88,80% doğruluk, 0,2884 kayıp) sergilemiş ve bu da %7,47'lik bir doğruluk farkı yaratmıştır. Buna karşılık, MAML++ algoritması, en iyi (90,73% doğruluk, 0,3580 kayıp) ve en kötü (83,60% doğruluk, 0,5401 kayıp) performansları arasında %7,13'lük daha küçük bir doğruluk farkı ile daha iyi bir tutarlılık sergilemiştir. Bu bulgular, MAML'ın daha iyi sınıflandırma performansına rağmen MAML++'ın daha tutarlı ve daha dayanıklı olduğunu göstermektedir. Bu çalışma, Görüntü Dönüştürücü modellerinin özellik çıkarım yeteneklerini meta-öğrenme algoritmalarıyla birleştirerek gerçek dünya koşullarında doğru ve güvenilir hastalık sınıflandırması için yeni bir yaklaşım sunmaktadır. Sonuçlar, geliştirilen yaklaşımın tarımsal uygulamalarda, özellikle etiketlenmiş verilerin kısıtlı olduğu senaryolarda önemli bir potansiyele sahip olduğunu göstermektedir.

🔖

Cite This Article

Gunde, M., Işık, G. (2025). A Robust Apple Disease Classification Approach Using Vision Transformer-Based Meta-Learning. Iğdır Üniversitesi Fen Bilimleri Enstitüsü Dergisi, Cilt: 15 Sayı: 4, 1193–1205. https://doi.org/10.21597/jist.1641204

📄 Full Text PDF Viewer

📄

PDF görüntüleyici hazırlanıyor...

References (20)

  1. Akhtar, F., Partheeban, N., Daniel, A., Sriramulu, S., Mehra, S., & Gupta, N. (2021). Plant Disease Detection based on Deep Learning Approach. 2021 International Conference on Advance Computing and Innovative Technologies in Engineering, ICACITE 2021, 74–77. https://doi.org/10.1109/ICACITE51222.2021.9404647
  2. Andrew, J., Eunice, J., Popescu, D. E., Chowdary, M. K., & Hemanth, J. (2022). Deep Learning-Based Leaf Disease Detection in Crops Using Images for Agricultural Applications. Agronomy, 12(10). https://doi.org/10.3390/agronomy12102395
  3. Ball, J. (2021). Few-Shot Learning for Image Classification of Common Flora. http://arxiv.org/abs/2105.03056 Bracino, A. A., Concepcion, R. S., Bedruz, R. A. R., Dadios, E. P., & Vicerra, R. R. P. (2020). Development of a Hybrid Machine Learning Model for Apple (Malus domestica) Health Detection and Disease Classification. 2020 IEEE 12th International Conference on Humanoid, Nanotechnology, Information Technology, Communication and Control, Environment, and Management (HNICEM), 1–6. https://doi.org/10.1109/HNICEM51456.2020.9400139
  4. Chakraborty, S., Paul, S., & Rahat-Uz-Zaman, M. (2021). Prediction of Apple Leaf Diseases Using Multiclass Support Vector Machine. 2021 2nd International Conference on Robotics, Electrical and Signal Processing Techniques (ICREST), 147–151. https://doi.org/10.1109/ICREST51555.2021.9331132
  5. Dosovitskiy, A., Beyer, L., Kolesnikov, A., Weissenborn, D., Zhai, X., Unterthiner, T., Dehghani, M., Minderer, M., Heigold, G., Gelly, S., Uszkoreit, J., & Houlsby, N. (2020). An Image is Worth 16x16 Words: Transformers for Image Recognition at Scale. ICLR 2021 - 9th International Conference on Learning Representations. https://arxiv.org/abs/2010.11929v2
  6. Figueroa-Flores, C., & San-Martin, P. (2023). Few-Shot Learning for Image Classification of Common Flora. ArXiv, abs/2105.03056. https://doi.org/10.3389/FPLS.2023.1211490 Finn, C., Abbeel, P., & Levine, S. (2017). Model-Agnostic Meta-Learning for Fast Adaptation of Deep Networks. 34th International Conference on Machine Learning, ICML 2017, 3, 1856–1868. https://arxiv.org/abs/1703.03400v3 Hughes, David. P., & Salathe, M. (2015). An open access repository of images on plant health to enable the development of mobile disease diagnostics. https://arxiv.org/abs/1511.08060v2
  7. Işık, G. (2023). Improving Plant Disease Recognition Through Gradient-Based Few-shot Learning with Attention Mechanisms. Journal of the Institute of Science and Technology, 13(3), 1482–1495. https://doi.org/10.21597/JIST.1283491
  8. Işık, G., & Paçal, İ. (2024). Few-shot classification of ultrasound breast cancer images using meta-learning algorithms. Neural Computing and Applications, 36(20), 12047–12059. https://doi.org/10.1007/S00521-024-09767-Y/TABLES/7
  9. Liu, J., & Wang, X. (2021). Plant diseases and pests detection based on deep learning: a review. Plant Methods, 17(1). https://doi.org/10.1186/S13007-021-00722-9
  10. Nagaraju, M., & Chawla, P. (2020). Systematic review of deep learning techniques in plant disease detection. International Journal of System Assurance Engineering and Management, 11(3), 547–560. https://doi.org/10.1007/S13198-020-00972-1/FIGURES/9
  11. Nagaraju, Y., Venkatesh, Swetha, S., & Stalin, S. (2020). Apple and Grape Leaf Diseases Classification using Transfer Learning via Fine-tuned Classifier. 2020 IEEE International Conference on Machine Learning and Applied Network Technologies (ICMLANT), 1–6. https://doi.org/10.1109/ICMLANT50963.2020.9355991
  12. Nurgazin, M., & Tu, N. A. (2023). A Comparative Study of Vision Transformer Encoders and Few-shot Learning for Medical Image Classification. 2023 IEEE/CVF International Conference on Computer Vision Workshops (ICCVW), 2505–2513. https://doi.org/10.1109/ICCVW60793.2023.00265
  13. Pacal, I., & Işık, G. (2024). Utilizing convolutional neural networks and vision transformers for precise corn leaf disease identification. Neural Computing and Applications, 37(4), 2479–2496. https://doi.org/10.1007/S00521-024-10769-Z/TABLES/5
  14. Ravi, S., & Larochelle, H. (2022). Optimization as a Model for Few-Shot Learning. Saleem, M. H., Potgieter, J., & Arif, K. M. (2019). Plant Disease Detection and Classification by Deep Learning. Plants 2019, Vol. 8, Page 468, 8(11), 468. https://doi.org/10.3390/PLANTS8110468
  15. Shao, Y., Wu, W., You, X., Gao, C., & Sang, N. (2023). Improving the Generalization of MAML in Few-Shot Classification via Bi-Level Constraint. IEEE Transactions on Circuits and Systems for Video Technology, 33(7), 3284–3295. https://doi.org/10.1109/TCSVT.2022.3232717
  16. Thapa, R., Zhang, K., Snavely, N., Belongie, S., & Khan, A. (2020). The Plant Pathology Challenge 2020 data set to classify foliar disease of apples. Applications in Plant Sciences, 8(9). https://doi.org/10.1002/APS3.11390
  17. Wu, X., Deng, H., Wang, Q., Lei, L., Gao, Y., & Hao, G. (2023). Meta-learning shows great potential in plant disease recognition under few available samples. The Plant Journal, 114(4), 767–782. https://doi.org/10.1111/TPJ.16176
  18. Xu, J., & Du, Q. (2020). Learning transferable features in meta-learning for few-shot text classification. Pattern Recognit. Lett., 135, 271–278. https://doi.org/10.1016/J.PATREC.2020.05.007
  19. Vezıroglu, E., Pacal, I., & Coşkunçay, A. (2023). Derin Evrişimli Sinir Ağları Kullanılarak Pirinç Hastalıklarının Sınıflandırılması. Journal of the Institute of Science and Technology, 13(2), 792-814. https://doi.org/10.21597/jist.1265769
  20. Ye, H. J., Ming, L., Zhan, D. C., & Chao, W. L. (2024). How to Train Your MAML to Excel in Few-Shot Classification. ArXiv, abs/2106.16245(3), 1425–1440. https://doi.org/10.1109/TPAMI.2022.3160362