SavedAdd your journal

Journals / Bozok Journal of Engineering and Architecture / 2026 / Cilt: 5 Sayı: 1

Comparison of EfficientNet, VGG16, and YOLO models in deep learning-based classification of hazelnut quality.

Fındık kalitesinin derin öğrenme tabanlı sınıflandırılmasında EfficientNet, VGG16 ve YOLO modellerinin karşılaştırılması

Özgür Tomak(Giresun Üniversitesi)0000-0003-2993-6913
Pages
63–75
Publication Date
2026-06-29
🏛️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)

According to the results obtained, the EfficientNet model achieved an accuracy of 0.9718 in the independent test set, the VGG16 model achieved an accuracy of 0.9835, and the YOLO model achieved an accuracy of 0.9987.

📌 3 Key Takeaways

🎯
Objective & Hypothesis:The need for fast, objective, and reliable classification systems in the quality control of agricultural products is increasing day by day.
📊
Key Findings & Data:In this study, three different deep learning-based classification models EfficientNet, VGG16, and YOLO were compared for hazelnut quality classification.
💡
Conclusion & Impact:According to the results obtained, the EfficientNet model achieved an accuracy of 0.9718 in the independent test set, the VGG16 model achieved an accuracy of 0.9835, and the YOLO model achieved an accuracy of 0.9987.
🔬Design:Modeling & Simulation
📍Scope:Comprehensive Sample
⚙️Variables:Derin Öğrenme · Fındık Kalite Sınıflandırması · EfficientNet

Abstract

The need for fast, objective, and reliable classification systems in the quality control of agricultural products is increasing day by day. In this study, three different deep learning-based classification models EfficientNet, VGG16, and YOLO were compared for hazelnut quality classification. The dataset consisted of three classes, poor quality hazelnuts, shelled hazelnuts, and good quality hazelnuts, totaling 15,770 images. In the experimental process, the dataset was divided into training, validation, and independent test subsets; thus, model performances were evaluated on previously unseen test data. According to the results obtained, the EfficientNet model achieved an accuracy of 0.9718 in the independent test set, the VGG16 model achieved an accuracy of 0.9835, and the YOLO model achieved an accuracy of 0.9987. When the confusion matrices were examined, it was observed that in the EfficientNet and VGG16 models, errors were mainly concentrated between the poor quality and good quality hazelnut classes, while only a very limited number of misclassifications occurred in the YOLO model. The findings show that all three models were successful in hazelnut quality classification, but the highest performance in terms of both overall accuracy and class-specific error minimization was achieved by the YOLO model. In conclusion, it has been demonstrated that deep learning-based approaches can be effectively used in automated hazelnut quality control, and that YOLO-based classification, in particular, offers a powerful solution to this problem.

Özet

Tarımsal ürünlerin kalite kontrolünde hızlı, nesnel ve güvenilir sınıflandırma sistemlerine duyulan ihtiyaç her geçen gün artmaktadır. Bu çalışmada, fındık kalite sınıflandırması için derin öğrenme tabanlı üç farklı yaklaşım olan EfficientNet, VGG16 ve YOLO tabanlı sınıflandırma modeli karşılaştırılmıştır. Veri kümesi, kötü fındık, iç fındık ve iyi fındık olmak üzere üç sınıftan ve toplam 15.770 görüntüden oluşmaktadır. Deneysel süreçte veri kümesi eğitim, doğrulama ve bağımsız test alt kümelerine ayrılmış; böylece model performansları daha önce görülmemiş test verileri üzerinde değerlendirilmiştir. Elde edilen sonuçlara göre EfficientNet modeli bağımsız test kümesinde 0.9718 doğruluk, VGG16 modeli 0.9835 doğruluk ve YOLO modeli 0.9987 doğruluk elde etmiştir. Karmaşıklık matrisleri incelendiğinde, EfficientNet ve VGG16 modellerinde hataların ağırlıklı olarak kötü fındık ile iyi fındık sınıfları arasında yoğunlaştığı, YOLO modelinde ise yalnızca çok sınırlı sayıda yanlış sınıflandırma gerçekleştiği görülmüştür. Bulgular, üç modelin de fındık kalite sınıflandırmasında başarılı olduğunu, ancak hem genel doğruluk hem de sınıf bazında hata minimizasyonu açısından en yüksek performansın YOLO modeli tarafından sağlandığını göstermektedir. Sonuç olarak, derin öğrenme tabanlı yaklaşımların otomatik fındık kalite kontrolünde etkili biçimde kullanılabileceği ve özellikle YOLO tabanlı sınıflandırmanın bu problem için güçlü bir çözüm sunduğu ortaya konmuştur.

🔖

Cite This Article

Tomak, Ö. (2026). Comparison of EfficientNet, VGG16, and YOLO models in deep learning-based classification of hazelnut quality.. Bozok Journal of Engineering and Architecture, Cilt: 5 Sayı: 1, 63–75. https://doi.org/10.70700/bjea.1920773

📄 Full Text PDF Viewer

📄

PDF görüntüleyici hazırlanıyor...

References (13)

  1. V. Hemamalini, S. Rajarajeswari, S. Nachiyappan, ve ekibi, “Food quality inspection and grading using efficient image segmentation and machine learning-based system,” J. Food Qual., vol. 2022, Art. no. 5262294, 2022, doi: 10.1155/2022/5262294.
  2. Z. Li, F. Liu, W. Yang, ve ekibi, “A survey of convolutional neural networks: Analysis, applications, and prospects,” IEEE Trans. Neural Netw. Learn. Syst., vol. 33, no. 12, pp. 6999–7019, Dec. 2022, doi: 10.1109/TNNLS.2021.3084827.
  3. A. Krizhevsky, I. Sutskever, and G. E. Hinton, “ImageNet classification with deep convolutional neural networks,” in Advances in Neural Information Processing Systems, vol. 25, 2012, pp. 1097–1105.
  4. K. Simonyan and A. Zisserman, “Very deep convolutional networks for large-scale image recognition,” in Proc. 3rd Int. Conf. Learn. Represent. (ICLR), 2015, pp. 1–14.
  5. M. Tan and Q. Le, “EfficientNet: Rethinking model scaling for convolutional neural networks,” in Proc. 36th Int. Conf. Mach. Learn. (ICML), Long Beach, CA, USA, Jun. 2019, pp. 6105–6114.
  6. J. Redmon, S. Divvala, R. Girshick, ve ekibi, “You only look once: Unified, real-time object detection,” in Proc. IEEE Conf. Comput. Vis. Pattern Recognit. (CVPR), Las Vegas, NV, USA, 2016, pp. 779–788, doi: 10.1109/CVPR.2016.91.
  7. I. Sa, Z. Ge, F. Dayoub, ve ekibi, “DeepFruits: A fruit detection system using deep neural networks,” Sensors, vol. 16, no. 8, Art. no. 1222, 2016, doi: 10.3390/s16081222.
  8. E. Güneş, “Hazelnuts,” Mendeley Data, V2, 2022, doi: 10.17632/dvvx6kst3f.2.
  9. Z. Ünal and H. Aktaş, “Classification of hazelnut kernels with deep learning,” Postharvest Biol. Technol., vol. 197, Art. no. 112225, Mar. 2023, doi: 10.1016/j.postharvbio.2022.112225.
  10. H. Aktaş, T. Kızıldeniz, and Z. Ünal, “Classification of pistachios with deep learning and assessing the effect of various datasets on accuracy,” J. Food Meas. Charact., vol. 16, no. 3, pp. 1983–1996, 2022, doi: 10.1007/s11694-022-01313-5.
  11. D. Abidin, “Veri artırma stratejilerine sistematik bakış [A systematic overview of data augmentation strategies],” in Bilgisayar Bilimleri ve Mühendisliği: Teori, Metodoloji ve Uygulama, A. Şentürk, Ed. Ankara, Türkiye: Platanus Publishing, 2025, pp. 35–64, doi: 10.5281/zenodo.17396194.
  12. R. Khanam and M. Hussain, “YOLOv11: An overview of the key architectural enhancements,” arXiv preprint arXiv:2410.17725, 2024.
  13. C. M. Bishop, Pattern Recognition and Machine Learning. New York, NY, USA: Springer, 2006.