Detection of the Quality of Zivzik Pomegranate Grown in Siirt Using Deep Learning Methods

Abstract

This study aims to determine the quality of the Zivzik pomegranate, a fruit unique to the Siirt region whose quality can only be understood by experts engaged in this business with deep learning methods. Since there is no existing database of Zivzik pomegranate, we first visited the Şirvan district of Siirt, where Zivzik pomegranate grows, many times to create a database, and over a thousand pomegranate photographs were taken and labeled. After the Zivzik pomegranate quality dataset was created, the aim was to determine the quality of Zivzik pomegranate using deep learning methods. AlexNet, VGG-16, VGG-19, ResNet, Inception, XCeption, EfficientNet, and MobileNet deep learning models were applied, and the results were evaluated. As a result of the study, the best accuracy value was obtained from the EfficientNetV2 B0 model at 81.83%. In addition to contributing to the scientific literature, our study is expected to contribute positively to the recognition of the Zivzik pomegranate, the regional economy, and the awareness of consumers and producers about agriculture 4.0 applications. © 2024 IEEE.