Optimized Deep Learning Model for Pomegranate Disease Detection: A Convolutional Neural Network Long Short-Term Memory Approach

Abstract

Pomegranate is a high-value fruit globally recognized for its nutritional benefits and applications in traditional medicine and cosmetics. India is a key player in the global pomegranate market, but the industry faces challenges such as diseases that affect crop productivity and economic losses for farmers. This study proposes a novel approach to pomegranate disease detection using a hybrid Convolutional Neural Network (CNN) and Long Short-Term Memory (LSTM) model. The proposed model leverages CNNs for effective feature extraction and LSTMs for sequential data handling, achieving superior performance compared to traditional methods and other deep learning techniques. Experimental results demonstrate high accuracy, recall, precision, and F1 score. The Proposed model achieved an accuracy of 98.53% and loss of 0.0677. The study also explores the limitations of transfer learning approaches such as VGG16 and ResNet50, and larger models like AlexNet, which did not perform well in this context. The findings suggest that the hybrid CNN-LSTM model offers a scalable and adaptable solution for agricultural disease detection, with potential applications for various crops.