Title: Deep Convolutional Neural Networks for Automated Classification of Date Fruit Varieties
Authors: Husam Abd Rahim Eleyan , Samy S. Abu-Naser
Volume: 10
Issue: 2
Pages: 28-34
Publication Date: 2026/02/28
Abstract:
Accurate identification of date fruit varieties is essential for quality control, market standardization, and automated agricultural processing. This study presents a deep learning-based framework for the multi-class classification of nine date fruit varieties using transfer learning. A convolutional neural network built upon the pre-trained VGG16 architecture was fine-tuned to extract discriminative visual features from RGB images and perform robust variety classification. The dataset consists of thousands of labeled images representing diverse date types, and comprehensive preprocessing procedures, including normalization, resizing, and data augmentation-were applied to improve model generalization and mitigate overfitting. The proposed model was evaluated using accuracy, loss, and F1-score metrics to ensure balanced performance assessment across classes. Experimental results demonstrate stable convergence behavior and reliable generalization, achieving approximately 69% validation accuracy in a challenging nine-class classification task. The findings confirm that transfer learning significantly enhances feature representation capability in agricultural image analysis while reducing training complexity. This work contributes to the development of intelligent vision-based systems for automated date fruit identification and provides a scalable foundation for smart agriculture applications, post-harvest automation, and digital quality assessment.