Transforming Pistachio Classification: The Power of ResNet-50 and SVM Integration

In this study, a novel approach for Pistachio image classification is proposed, combining the feature extraction capabilities of the ResNet-50 convolutional neural network (CNN) with the classification power of a Support Vector Machine (SVM). The model was trained and evaluated on a dataset of 2148 Pistachio images, using 30 epochs to ensure optimal learning. The training process demonstrated significant improvements in accuracy during the initial 20 epochs, reaching a final training accuracy of 92.64% and a validation accuracy of 97.21%. The model’s loss values also exhibited steady decreases, indicating strong generalization to unseen data. The primary novelty of this research lies in the effective integration of deep feature extraction through ResNet-50 with SVM classification, providing high accuracy for agricultural image classification tasks.This approach demonstrates the potential for achieving robust Pistachio classification with fewer epochs, offering an efficient alternative to traditional CNN classifiers. Future research could explore larger datasets and alternative hybrid methods to further enhance classification performance. Additionally, testing the model’s applicability to other agricultural products could validate its generalizability across various domains.