MobileNetV2-Driven Image Recognition for Discriminating Pistachio Varieties: Akuri, Antep, Halebi, Kırmızı, and Tekin
15th International Congress of the Innovative Agricultural Technologies, IAT 2025, Antalya, Türkiye, 15 - 19 Ekim 2025, cilt.805 LNCE, ss.28-43, (Tam Metin Bildiri)
- Yayın Türü: Bildiri / Tam Metin Bildiri
- Cilt numarası: 805 LNCE
- Doi Numarası: 10.1007/978-3-032-15375-3_3
- Basıldığı Şehir: Antalya
- Basıldığı Ülke: Türkiye
- Sayfa Sayıları: ss.28-43
- Anahtar Kelimeler: Convolutional Neural Networks (CNN), Deep Learning, MobileNetV2, Pistachio Classification, Transfer Learning
- Ondokuz Mayıs Üniversitesi Adresli: Evet
Özet
Pistachios are one of the most economically valuable nut crops, with different varieties exhibiting distinct morphological characteristics, quality attributes, and market values. Accurate classification of pistachio cultivars is essential for quality control, authentication, and post-harvest handling in the agricultural supply chain. Traditional methods of identifying pistachio types rely heavily on manual inspection, which is time-consuming, prone to human error, and inefficient for large-scale processing. In this study, we propose an automated and efficient approach for pistachio variety classification using a deep learning model based on the MobileNetV2 architecture. The main objective is to develop a lightweight convolutional neural network (CNN) capable of distinguishing between five common pistachio types Akuri, Antep, Halebi, Kırmızı, and Tekin with high accuracy and minimal computational cost. To achieve this, a dataset consisting of labeled images of the five pistachio types was collected and preprocessed using standard augmentation techniques to enhance variability and improve generalization. Transfer learning was employed by fine-tuning the MobileNetV2 model, pretrained on ImageNet, to adapt it for the specific classification task. The model training process was carefully monitored using early stopping and validation accuracy to prevent overfitting. Experimental results demonstrated the effectiveness of the proposed approach, with the model achieving 93% training accuracy, 95% validation accuracy, and 94% test accuracy. These results indicate strong classification performance and generalizability across unseen data. The findings suggest that MobileNetV2 can serve as a practical and scalable solution for real-world pistachio classification tasks in both industrial and research settings. This study contributes to the growing body of work on AI-driven agricultural applications and highlights the potential of deep learning for improving the efficiency and reliability of crop identification and quality assurance systems.