FruitNetSH: An Efficient Deep Learning Model for Precision Pomegranate Disease Diagnosis

Abstract

The most important thing in minimizing crop losses and maximizing crop yields is accurate and timely diagnosis of diseases in pomegranate fruit. Though the recent deep learning methods have demonstrated encouraging results, issues like field noise, variability in illumination, complexity of the background, and similarity between classes continue to impact on reliable disease detection. The paper introduces FruitNetSH, a powerful deep learning-based system of automated detection of pomegranate fruit disease. The suggested approach uses the concept of transfer learning and involves the fine-tuned convolutional neural network to extract the discriminative visual features of the pomegranate fruit images in a variety of real-field conditions. In order to enhance robustness and generalization, image preprocessing and augmentation methods are included to deal with variations in noise and scarce training diversity. FruitNetSH is tailored to sort healthy and diseased fruits in more than one category of disease with high diagnostic accuracy and low computation costs. Experimental analysis shows that the proposed framework can be better classified metrics with recent state-of-the-art deep learning models. The learning capability of FruitNetSH of enhanced feature learning capability allows reliable localization of disease-specific characteristics, and minimizes misclassification across visually similar disease classes. The framework proposed will provide a viable, scalable solution to intelligent pomegranate disease diagnosis, supporting precision agriculture via timely intervention and enhanced crop management.