Exploring Deep Learning Approaches for Walnut Phenotype Variety Classification

Abstract

The efficient classification of agricultural commodities like walnuts is crucial for assessing quality and managing the supply chain. This scholarly article analyses various deep learning and data science methods for walnut fruit classification. For this purpose, first, a dataset comprising images of walnuts from Chandler, Fernor, Howard, and Oguzlar varieties was collected. Two different experiments were conducted. In the first experiment, only deep learning methods were used as classifiers. In this experiment, InceptionV3 demonstrated the highest classification accuracy, followed by VGG-19 and VGG-16. In the second experiment, deep learning algorithms were used for feature extraction, followed by support vector machine (SVM), logistic regression (LR), and k-nearest neighbor (k-NN) algorithms for classification. These models resulted in an improvement in overall success rates. The most effective classification was achieved with the InceptionV3 and LR combination, achieving the highest success rate. These results highlight the efficacy of deep learning methodologies in swiftly and accurately classifying agricultural products based on visual information, indicating the potential to strengthen classification systems within the agricultural sector.