Çilek Çeşidi Sınıflandırmasında Az Örnekli Öğrenme: Prototipik Ağ Tabanlı Bir Yaklaşım

Öz

Farklı çilek türlerinin sınıflandırılması, tarım ve kalite kontrolünde kritik bir problem olmakla birlikte, büyük ölçekli etiketli veri setlerine duyulan ihtiyaç nedeniyle geleneksel derin öğrenme çözümleri için önemli bir engel oluşturmaktadır. Bu çalışmada, yalnızca 103 etiketli eğitim görüntüsü içeren Mendeley Strawberry Grading Dataset kullanılarak dört farklı ticari çilek çeşidini (A, B, C ve Jumbo) sınıflandırmak amacıyla Prototipik Ağlara dayalı bir Az Örnekli Öğrenme (FSL) modeli uygulanmıştır. Isınma ön-eğitimi, diferansiyel öğrenme oranı stratejisi ve 4-yollu 5-örnek episodik eğitimin bir araya getirilmesiyle özgün bir eğitim boru hattı oluşturulmuş; ResNet50 ve EfficientNet-B0 omurgaları 5-katlı tabakalı çapraz doğrulama protokolü altında karşılaştırmalı olarak değerlendirilmiştir. ResNet50 sırasıyla %99.00 test doğruluğu, %99.11 makro F1 skoru ve %98.62 Cohen Kappa değeri elde ederken, EfficientNet-B0 %98.00, %98.33 ve %97.30 değerlerine ulaşmıştır. Her iki model de %99.80 AUC değeri üretmiştir. GradCAM++ kullanılarak gerçekleştirilen yorumlanabilirlik analizi, her iki ağın da dikkatini meyvelerin ilgili bölgelerine yoğunlaştırdığını ortaya koymuştur. Bu çalışma, metrik tabanlı az örnekli öğrenmenin, tarımsal görüntülere ait veri setlerinin yetersiz olduğu durumlarda geleneksel derin öğrenme yöntemlerine başarılı bir alternatif olarak kullanılabileceğini kanıtlamıştır.

Anahtar Kelimeler

• Az örnekli öğrenme
• Prototipik ağlar
• Metrik tabanlı öğrenme
• Hassas tarım
• Çilek çeşidi sınıflandırması.

Etik Beyan

Bu makale için etik kurul onayına gerek yoktur. Bu makale ile ilgili olarak herhangi bir kişi/kurumla çıkar çatışması bulunmamaktadır.

Few-Shot Learning for Strawberry Variety Classification: A Prototypical Network-Based Approach

Öz

Classification of different types of strawberries is an essential problem in agriculture and quality control; however, it presents a challenge to traditional deep learning solutions because of the need for large-scale labeled datasets. In this research, a few-shot learning (FSL) model based on Prototypical Networks was applied to classify four different commercial types of strawberries (A, B, C, and Jumbo), utilizing the Mendeley Strawberry Grading Dataset that includes only 103 labeled images for training. A unique pipeline for training an FSL model was created through a combination of warm-up pre-training, differential learning rate strategy, and 4-way 5-shot episodic training; two backbones (ResNet50 and EfficientNet-B0) were compared under a 5-fold stratified cross-validation. While ResNet50 obtained test accuracy, macro F1-score, and Cohen's Kappa equal to 99.00%, 99.11%, and 98.62% respectively, EfficientNet-B0 achieved 98.00%, 98.33%, and 97.30%. Both models produced an AUC value of 99.80%. Interpretability analysis utilizing GradCAM++ showed that both networks focused on relevant areas of the fruits. This research proved that metric-based few-shot learning can be used successfully as an alternative solution to traditional deep learning methods in case of scarce datasets of agricultural images.

Anahtar Kelimeler

• : Few-shot learning
• Prototypical networks
• Metric-based learning
• Precision agriculture
• Strawberry variety classification.

Etik Beyan

There is no need for an ethics committee approval for this article. There is no conflict of interest with any person/institution for this article.

Kaynakça

1. H. Zheng, G. Wang, and X. Li, “Identifying strawberry appearance quality by vision transformers and support vector machine,” J. Food Process. Eng., vol. 45, no. 10, p. e14132, Oct. 2022.
2. Z. Gao, Y. Shao, G. Xuan, Y. Wang, Y. Liu, and X. Han, “Real-time hyperspectral imaging for the in-field estimation of strawberry ripeness with deep learning,” Artif. Intell. Agric., vol. 4, pp. 31–38, Jan. 2020.
3. X. Zhou, W. S. Lee, Y. Ampatzidis, Y. Chen, N. Peres, and C. Fraisse, “Strawberry Maturity Classification from UAV and Near-Ground Imaging Using Deep Learning,” Smart Agric. Technol., vol. 1, p. 100001, Dec. 2021.
4. M. I. Dinata, S. M. S. Nugroho, and R. F. Rachmadi, “Classification of Strawberry Plant Diseases with Leaf Image Using CNN,” in Proc. Int. Conf. Artif. Intell. Comput. Sci. Technol. (ICAICST), Jun. 2021, pp. 68–72.
5. J. Pan, L. Xia, Q. Wu, Y. Guo, Y. Chen, and X. Tian, “Automatic strawberry leaf scorch severity estimation via faster R-CNN and few-shot learning,” Ecol. Inform., vol. 70, p. 101706, Sep. 2022.
6. M. E. Sahin, U. Özkaya, C. Arisoy, H. İ. Coşar, and H. Ulutaş, “CucuNetCNNs: Application of novel ensemble deep neural networks for classification of cucumber leaf disease,” Ain Shams Eng. J., vol. 16, no. 5, p. 103380, Apr. 2025.
7. H. Ulutas and M. E. Sahin, “Classification of cucumber leaf diseases on images using innovative ensembles of deep neural networks,” J. Electron. Imaging, vol. 32, no. 5, p. 053040, Oct. 2023.
8. J. Y. Choi, K. Seo, J. S. Cho, and K. D. Moon, “Applying convolutional neural networks to assess the external quality of strawberries,” J. Food Compos. Anal., vol. 102, p. 104071, Sep. 2021.

9. K. Aghamohammadesmaeilketabforoosh, S. Nikan, G. Antonini, and J. M. Pearce, “Optimizing Strawberry Disease and Quality Detection with Vision Transformers and Attention-Based Convolutional Neural Networks,” Foods, vol. 13, no. 12, p. 1869, Jun. 2024.
10. Y. Wang, Q. Yao, J. T. Kwok, and L. M. Ni, “Generalizing from a Few Examples: A Survey on Few-shot Learning,” ACM Comput. Surv., vol. 53, no. 3, May 2021.
11. L. Li, J. Li, C. Lv, Y. Yuan, and B. Zhao, “Maize residue segmentation using Siamese domain transfer network,” Comput. Electron. Agric., vol. 187, p. 106261, Aug. 2021.
12. J. Snell, K. Swersky, and R. Zemel, “Prototypical Networks for Few-shot Learning,” in Adv. Neural Inf. Process. Syst., vol. 30, 2017.
13. K. He, X. Zhang, S. Ren, and J. Sun, “Deep residual learning for image recognition,” in Proc. IEEE Conf. Comput. Vis. Pattern Recognit. (CVPR), Dec. 2016, pp. 770–778.
14. M. Tan and Q. V. Le, “EfficientNet: Rethinking Model Scaling for Convolutional Neural Networks,” in Proc. 36th Int. Conf. Mach. Learn. (ICML), vol. 97, pp. 6105–6114, 2019. [Online]. Available: https://proceedings.mlr.press/v97/tan19a.html.
15. N. Ragu and J. Teo, “Object detection and classification using few-shot learning in smart agriculture: A scoping mini review,” Front. Sustain. Food Syst., vol. 6, p. 1039299, Jan. 2023.
16. A. Hayat, F. Morgado-Dias, T. Choudhury, T. P. Singh, and K. Kotecha, “FruitVision: A deep learning based automatic fruit grading system,” Open Agric., vol. 9, no. 1, Jan. 2024.
17. S. Yang, Q. Feng, J. Zhang, W. Yang, W. Zhou, and W. Yan, “From laboratory to field: cross-domain few-shot learning for crop disease identification in the field,” Front. Plant Sci., vol. 15, p. 1434222, Dec. 2024.
18. B. D. SI UMN, Strawberry Grading Dataset, vol. 1, 2025.
19. A. Chattopadhay, A. Sarkar, P. Howlader, and V. N. Balasubramanian, “Grad-CAM++: Generalized gradient-based visual explanations for deep convolutional networks,” in Proc. IEEE Winter Conf. Appl. Comput. Vis. (WACV), 2018, pp. 839–847.
20. R. Kohavi, “A Study of Cross-Validation and Bootstrap for Accuracy Estimation and Model Selection,” in Proc. 14th Int. Joint Conf. Artif. Intell. (IJCAI), pp. 1137–1143, 1995. [Online]. Available: http://robotics.stanford.edu/~ronnyk
21. C. Shorten and T. M. Khoshgoftaar, “A survey on image data augmentation for deep learning,” J. Big Data, vol. 6, no. 1, p. 60, Jul. 2019.
22. R. R. Selvaraju, M. Cogswell, A. Das, R. Vedantam, D. Parikh, and D. Batra, “Grad-CAM: Visual Explanations from Deep Networks via Gradient-Based Localization,” in Proc. IEEE Int. Conf. Comput. Vis. (ICCV), Dec. 2017, pp. 618–626.
23. X. Song et al., “Agricultural Image Processing: Challenges, Advances, and Future Trends,” Appl. Sci., vol. 15, no. 16, 2025.
24. J. Li et al., “Large Language Models and Foundation Models in Smart Agriculture: Basics, Opportunities, and Challenges,” Aug. 2023. [Online]. Available: https://arxiv.org/pdf/2308.06668.