Pomegranate Disease Detection Using CNN-LSTM Hybrid Model

Abstract

Pomegranate is a high-value fruit globally recognized for its nutritionalbenefits and applications in traditional medicine and cosmetics. India is akey player in the global pomegranate market, but the industry faceschallenges such as diseases that affect crop productivity and economic lossesfor farmers. This study proposes a novel approach to pomegranate diseasedetection using a hybrid Convolutional Neural Network (CNN) and Longshort- Term Memory (LSTM) model. The proposed model leverages CNNsfor effective feature extraction and LSTMs for sequential data handling,achieving superior performance compared to traditional methods and otherdeep learning techniques. Experimental results demonstrate high accuracy,recall, precision, and F1 score. The Proposed model achieved an accuracy of98.53% and loss of 0.0677. The study also explores the limitations oftransfer learning approaches such as VGG16 and ResNet50, and largermodels like AlexNet, which did not perform well in this context. Thefindings suggest that the hybrid CNN-LSTM model offers a scalable andadaptable solution for agricultural disease detection, with potentialapplications for various crops.