Mango leaf disease classification using a hybrid model of ResNet50 and quantum machine learning

Variation in plant health, particularly due to diseases, significantly impacts agricultural productivity and food security. Traditionally, disease detection in crops like mango relies on expert visual inspections, but automated systems leveraging advanced computational techniques are emerging as efficient alternatives. This research addresses the challenge of limited discriminative power in traditional visual inspection methods by developing a next-generation, hybrid deep learning and quantum machine learning (QML) framework for automated mango leaf disease classification. The proposed system leverages ResNet50, a 50-layer residual convolutional neural network pre-trained on ImageNet, combined with a Quantum Support Vector Machine (QSVM) utilising quantum kernel methods under classical simulation, to enable selective detection and accurate identification of eight disease categories from complex leaf image patterns. The system is evaluated across three configurations—standalone ResNet50, ResNet50 with classical SVM, and ResNet50 with QSVM—using the MangoLeafBD dataset of 4,000 mango leaf images, split into training and test sets via stratified sampling (80/20) to preserve class distribution across eight disease categories: Anthracnose, Bacterial Canker, Cutting Weevil, Die Back, Gall Midge, Healthy, Powdery Mildew, and Sooty Mould. Experimental results demonstrate a test accuracy of 0.989 ± 0.005 (mean ± std. across five independent runs) for the hybrid ResNet50-QSVM model, outperforming both standalone ResNet50 (0.953 ± 0.012) and ResNet50 with a classical SVM (0.981 ± 0.008). Additionally, the integration of quantum kernel methods under classical simulation yielded competitive performance on non-linear feature separation, particularly for visually similar disease pairs. The findings highlight the viability of implementing quantum-inspired kernel methods within a hybrid classical-quantum framework to improve the accuracy and reliability of automated mango leaf disease detection, offering a scalable foundation for further investigation in real-world crop monitoring applications.