Deep learning-automatic honey variety identification

Al Wardi, Al Yaqeen Salim Mohammed and Al Husaini, Mohammed Abdulla Salim and Al Husaini, Yousuf Nasser and Abdulghafor, Rawad and Habaebi, Mohamed Hadi and Al Masqari, Yasir Abdallah Khalifa (2026) Deep learning-automatic honey variety identification. In: 2026 IEEE 5th International Multidisciplinary Conference on Engineering Technology (IMCET), 15-17 April 2026, Beirut - Lebanon.

PDF - Published Version Restricted to Repository staff only Download (1MB) | Request a copy

Abstract

Honey authentication plays a vital role in ensuring food quality and helping consumers know more about their products. Traditional methods like chemical analysis and melissopalynology are trusted but can be slow, expensive, and require expert skills. To make things easier and faster, this study introduces a deep learning approach that automatically identifies honey varieties using and RGB images. A dataset of 5000 honey samples was collected from six locally relevant classes: Sidr honey, Samr honey, Abu Tuwaiq honey, multi-flower (Alzuhor) honey, date honey, and adulterated honey (Not Honey). Using a ResNet-50 neural network trained with the Adam optimizer and categorical cross-entropy, we split our dataset into 70% for training, 15% for validation, and 15% for testing, ensuring reliable evaluation with 10-fold cross-validation. The results are quite promising: our model achieved an impressive average accuracy of 99.80%, along with high sensitivity, specificity, precision, F1-score, and AUC values, all indicating excellent classification performance. The confusion matrix shows that different honey types are well distinguished, with only a few misclassifications among visually similar samples. When tested on new, unseen samples, the model provided confidence scores ranging from 86.5% to perfect 100%. The obtained results indicate that that deep learning image analysis can offer a powerful, non-invasive method for identifying honey varieties. Moving forward, we plan to expand our dataset and explore combining multiple data types for even better results.

Item Type:	Proceeding Paper (Plenary Papers)
Uncontrolled Keywords:	Honey Variety Classification, Deep Learning, RGB Imaging, Convolutional Neural Networks, Food Authentication.
Subjects:	T Technology > TK Electrical engineering. Electronics Nuclear engineering > TK7800 Electronics. Computer engineering. Computer hardware. Photoelectronic devices > TK7885 Computer engineering
Kulliyyahs/Centres/Divisions/Institutes (Can select more than one option. Press CONTROL button):	Kulliyyah of Engineering > Department of Electrical and Computer Engineering Kulliyyah of Engineering
Depositing User:	Dr. Mohamed Hadi Habaebi
Date Deposited:	19 May 2026 15:52
Last Modified:	19 May 2026 15:52
Queue Number:	2026-05-Q3453
URI:	http://irep.iium.edu.my/id/eprint/129062

Actions (login required)

View Item