Sınırlı Veriyle Bitki Hastalığı Tespiti: MAML++ ve ResNet Mimarileri ile Etkin Bir Meta-Öğrenme Yaklaşımı

Öz

Bitki hastalıklarının tespiti, küresel gıda güvenliği için kritik öneme sahipken, bu alandaki derin öğrenme modelleri genellikle büyük etiketli veri kümelerine ihtiyaç duyar. Tarımsal uygulamalarda bu tür verilerin toplanmasındaki zorluklar, veri kısıtlı senaryolarda çalışabilen etkin yöntemlere olan ihtiyacı artırmaktadır. Bu çalışma, az sayıda örnekle öğrenme problemine çözüm olarak MAML++ meta-öğrenme algoritmasını ResNet-18, ResNet-34 ve ResNet-50 gibi farklı derinlikteki mimarilerle entegre etmektedir. PlantVillage veri kümesinin elma hastalıkları alt kümesinde yapılan deneylerde, model performansları 2-way/4-way görevler ve 1/3/5-shot konfigürasyonları altında değerlendirilmiştir. Sonuçlar, destek örneği (shot) sayısının artmasının doğruluğu artırdığını, ancak daha derin ağların her zaman daha iyi sonuç vermediğini göstermiştir. Özellikle daha hafif bir mimari olan ResNet-18, 2-way 5-shot senaryosunda %92,53 doğruluk oranıyla en yüksek performansı sergilemiş; daha derin olan ResNet-50 modeline yakın bir başarıyı daha düşük kayıp değeriyle elde etmiştir. Bu bulgular, veri kısıtlı tarımsal uygulamalarda, MAML++ gibi meta-öğrenme yaklaşımlarının hafif modellerle birleştirilmesinin hem verimli hem de yüksek performanslı bir çözüm sunduğunu ortaya koymaktadır.

Anahtar Kelimeler

• Meta-öğrenme
• Az örnekle öğrenme
• ResNet
• PlantVillage
• Bitki hastalıklarının tespiti

Efficient Plant Disease Detection with Limited Data: A Meta-Learning Approach Using MAML++ and ResNet Architectures

Öz

ABSTRACT: Plant disease detection is critical for global food security, yet deep learning models often depend on large annotated datasets that are scarce in agriculture. This scarcity necessitates efficient methods that can perform well under data-limited conditions. This study addresses the few-shot learning challenge by integrating the MAML++ meta-learning algorithm with ResNet backbones of varying depths: ResNet-18, ResNet-34, and ResNet-50. Using the apple disease subset of the PlantVillage dataset, we evaluated model performance across 2-way/4-way tasks and 1/3/5-shot configurations. Our findings reveal that while more support samples (shots) improved accuracy, network depth did not linearly correlate with better performance. Notably, the lightweight ResNet-18 architecture achieved the highest accuracy of 92.53% in the 2-way 5-shot scenario, matching the performance of the deeper ResNet-50 but with a lower loss value. These results demonstrate that combining meta-learning approaches like MAML++ with lighter architectures offers an efficient and high-performing solution for data-constrained agricultural applications.

Anahtar Kelimeler

• Meta-learning
• Few-shot learning
• ResNet
• PlantVillage
• Plant disease detection

Kaynakça

1. Alom, M. Z., Taha, T. M., Yakopcic, C., Westberg, S., Sidike, P., Nasrin, M. S., … Asari, V. K. (2019). A State-of-the-Art Survey on Deep Learning Theory and Architectures. Electronics 2019, Vol. 8, Page 292, 8(3), 292. doi:10.3390/ELECTRONICS8030292
2. Alpsalaz, F., Özüpak, Y., Aslan, E. ve Uzel, H. (2025). Classification of maize leaf diseases with deep learning: Performance evaluation of the proposed model and use of explicable artificial intelligence. Chemometrics and Intelligent Laboratory Systems, 262, 105412. doi:10.1016/J.CHEMOLAB.2025.105412
3. Antoniou, A., Storkey, A. ve Edwards, H. (2018). How to train your MAML. 7th International Conference on Learning Representations, ICLR 2019. https://arxiv.org/abs/1810.09502v3 adresinden erişildi.
4. Arnal Barbedo, J. G. (2019). Plant disease identification from individual lesions and spots using deep learning. Biosystems Engineering, 180, 96-107. doi:10.1016/J.BIOSYSTEMSENG.2019.02.002
5. Aslan, E. ve ÖZÜPAK, Y. (2024). Diagnosis And Accurate Classification of Apple Leaf Diseases Using Vision Transformers. Computer and Decision Making: An International Journal, 1, 1-12. doi:10.59543/COMDEM.V1I.10039
6. Chen, L., Cui, X., Li, W., Zandonadi, R. P., Braz, R. ve Botelho, A. (2021). Meta-Learning for Few-Shot Plant Disease Detection. Foods 2021, Vol. 10, Page 2441, 10(10), 2441. doi:10.3390/FOODS10102441
7. Dosovitskiy, A., Beyer, L., Kolesnikov, A., Weissenborn, D., Zhai, X., Unterthiner, T., … 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 adresinden erişildi.
8. Ferentinos, K. P. (2018). Deep learning models for plant disease detection and diagnosis. Computers and Electronics in Agriculture, 145, 311-318. doi:10.1016/J.COMPAG.2018.01.009

9. Finn, C., Abbeel, P. ve 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 adresinden erişildi.
10. He, K., Zhang, X., Ren, S. ve Sun, J. (2016). Deep Residual Learning for Image Recognition. http://image-net.org/challenges/LSVRC/2015/ adresinden erişildi.
11. Hospedales, T., Antoniou, A., Micaelli, P. ve Storkey, A. (2022). Meta-Learning in Neural Networks: A Survey. IEEE Transactions on Pattern Analysis and Machine Intelligence, 44(9), 5149-5169. doi:10.1109/TPAMI.2021.3079209
12. Hughes, David. P. ve 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 adresinden erişildi.
13. Kamilaris, A. ve Prenafeta-Boldú, F. X. (2018). Deep learning in agriculture: A survey. Computers and Electronics in Agriculture, 147, 70-90. doi:10.1016/J.COMPAG.2018.02.016
14. Li, J., Feng, Q., Yang, J., Zhang, J. ve Yang, S. (2025). Few-shot crop disease recognition using sequence- weighted ensemble model-agnostic meta-learning. Frontiers in Plant Science, 16, 1615873. doi:10.3389/FPLS.2025.1615873/BIBTEX
15. Mohanty, S. P., Hughes, D. P. ve Salathé, M. (2016). Using deep learning for image-based plant disease detection. Frontiers in Plant Science, 7(September), 215232. doi:10.3389/FPLS.2016.01419/BIBTEX
16. Özüpak, Y., Alpsalaz, F., Aslan, E. ve Uzel, H. (2025). Hybrid deep learning model for maize leaf disease classification with explainable AI. New Zealand Journal of Crop and Horticultural Science, 53(5), 2942-2964. doi:10.1080/01140671.2025.2519570
17. Paszke, A., Gross, S., Massa, F., Lerer, A., Bradbury Google, J., Chanan, G., … Chintala, S. (2019). PyTorch: An Imperative Style, High-Performance Deep Learning Library. Advances in Neural Information Processing Systems, 32.
18. Raghu, A., Raghu, M., Bengio, S. ve Vinyals, O. (2019). Rapid Learning or Feature Reuse? Towards Understanding the Effectiveness of MAML. 8th International Conference on Learning Representations, ICLR 2020. https://arxiv.org/pdf/1909.09157 adresinden erişildi.
19. Ravi, S. ve Larochelle, H. (2022, 21 Temmuz). Optimization as a Model for Few-Shot Learning.
20. Rawat, W. ve Wang, Z. (2017). Deep convolutional neural networks for image classification: A comprehensive review. Neural Computation, 29(9), 2352-2449. doi:10.1162/NECO_A_00990
21. Rezaei, M., Diepeveen, D., Laga, H., Jones, M. G. K. ve Sohel, F. (2024). Plant disease recognition in a low data scenario using few-shot learning. Computers and Electronics in Agriculture, 219, 108812. doi:10.1016/J.COMPAG.2024.108812
22. Savary, S., Willocquet, L., Pethybridge, S. J., Esker, P., McRoberts, N. ve Nelson, A. (2019). The global burden of pathogens and pests on major food crops. Nature Ecology and Evolution, 3(3), 430-439. doi:10.1038/S41559-018-0793-Y;SUBJMETA=1143,2808,449,631,706;KWRD=AGRICULTURE,GEOGRAPHY,PLANT+SCIENCES
23. Sun, J., Cao, W., Fu, X., Ochi, S. ve Yamanaka, T. (2024). Few-shot learning for plant disease recognition: A review. Agronomy Journal, 116(3), 1204-1216. doi:10.1002/AGJ2.21285;JOURNAL:JOURNAL:14350645;PAGE:STRING:ARTICLE/CHAPTER
24. Too, E. C., Yujian, L., Njuki, S. ve Yingchun, L. (2019). A comparative study of fine-tuning deep learning models for plant disease identification. Computers and Electronics in Agriculture, 161, 272-279. doi:10.1016/J.COMPAG.2018.03.032
25. Wang, Y., Yao, Q., Kwok, J. T. ve Ni, L. M. (2021). Generalizing from a Few Examples: A Survey on Few-shot Learning. ACM Computing Surveys, 53(3). doi:10.1145/3386252;WEBSITE:WEBSITE:DL-SITE;REQUESTEDJOURNAL:JOURNAL:CSUR;TAXONOMY:TAXONOMY:ACM-PUBTYPE;PAGEGROUP:STRING:PUBLICATION
26. Yao, H., Wang, Y., Wei, Y., Zhao, P., Mahdavi, M., Lian, D. ve Finn, C. (2021). Meta-learning with an Adaptive Task Scheduler. Advances in Neural Information Processing Systems, 9, 7497-7509. https://arxiv.org/pdf/2110.14057 adresinden erişildi.