Classification of Pista Species using Fine-Tuned EfficientNetB3 Transfer Learning Model

Pistachios are popular and essential commercial nuts that have varying aesthetic qualities depending on the variety. Farmers, suppliers, and food processors must accurately classify pistachio varietals to maintain quality standards and tailor marketing campaigns based on consumer preferences. This study employs the EfficientNetB3 transfer learning model with fine-tuning layers to organize images of pistachios. The aim is to create a reliable and precise picture classification system that can tell the difference between the two pistachio types, Kirmizi and Siirt. There are 2148 labelled images in the dataset, which were divided into training, validation, and test batches at random. Transfer learning is carried out using EfficientNetB3, a potent DCNN that was pretrained on a sizable dataset and refined using the pistachio dataset. The EfficientNetB3 transfer learning model produced exceptional outcomes, with a pistachio classification accuracy of 99.5%. The model's high accuracy demonstrates the model's capability to accurately recognize and differentiate between the Siirt and Kirmizi pistachio types. The study shows that even with little datasets, transfer learning and fine-tuning approaches may be used to use deep learning models that have already been trained and adapt them to particular image categorization tasks. The automatic quality evaluation and variety identification of pistachios are made possible by the suggested image classification system, which substantially impacts the food and agriculture industries. Accurate categorization increases the pistachio production process's effectiveness and financial viability by promoting targeted marketing and improving quality control. The findings demonstrate the potential real-world uses of the EfficientNetB3 model in various picture classification tasks while demonstrating an effective method for classifying pistachios.