International journal of intelligent engineering and systems
M-HQNet: A Hybrid Quantum-classical Convolutional Neural Network with PSO-based Feature Selection for Multiclass Classification of South Indian Mango Leaf Diseases
Thaseentaj Shaik, PravinRamdasKshirsagar, Vijaya Krishna Akula, Kunda Suresh Babu, C Venkata Subbaiah
Early and accurate identification of mango leaf diseases is vital for safeguarding crop health, particularly in tropical agricultural regions like South India.This study presents M-HQNet, a novel Hybrid Quantum-Classical Convolutional Neural Network framework designed for the multiclass classification of mango leaf diseases.The pipeline begins with image preprocessing and data augmentation to improve model generalization.A pretrained MobileNetV2 is employed to perform deep feature extraction from augmented mango leaf images.To reduce dimensionality and align with quantum hardware constraints, Particle Swarm Optimization(PSO) is used, yielding a 4-dimensional feature vector optimized for quantum encoding.These selected features are embedded into a 4-qubit quantum circuit using a ZZFeatureMap and processed through a RealAmplitudes ansatz via angle encoding.The quantum circuit is implemented as a Hybrid-QCNN layer, followed by a Convolutional Neural Network (CNN) based classifier for final multiclass predictions.Experimental results demonstrate high effectiveness, achieving 96.2% accuracy.Comprehensive ablation studies and cross-validation confirm the critical contributions of PSO, quantum encoding, and the hybrid QCNN architecture.
Early and accurate identification of mango leaf diseases is vital for safeguarding crop health, particularly in tropical agricultural regions like South India.This study presents M-HQNet, a novel Hybrid Quantum-Classical Convolutional Neural Network framework designed for the multiclass classification of mango leaf diseases.The pipeline begins with image preprocessing and data augmentation to improve model generalization.A pretrained MobileNetV2 is employed to perform deep feature extraction from augmented mango leaf images.To reduce dimensionality and align with quantum hardware constraints, Particle Swarm Optimization(PSO) is used, yielding a 4-dimensional feature vector optimized for quantum encoding.These selected features are embedded into a 4-qubit quantum circuit using a ZZFeatureMap and processed through a RealAmplitudes ansatz via angle encoding.The quantum circuit is implemented as a Hybrid-QCNN layer, followed by a Convolutional Neural Network (CNN) based classifier for final multiclass predictions.Experimental results demonstrate high effectiveness, achieving 96.2% accuracy.Comprehensive ablation studies and cross-validation confirm the critical contributions of PSO, quantum encoding, and the hybrid QCNN architecture.
Plant leaf diseases are a critical threat to global food security, with early detection complicated by high inter-class similarity, environmental variations, and the demand for scalable multiclass classification across diverse crops. To address these challenges, we propose P-HQNN, a lightweight yet robust 4-qubit Hybrid Quantum–Classical Convolutional Neural Network designed for near-term quantum devices. The model uses EfficientNet-B0 to extract deep feature, and Particle Swarm Optimization (PSO) with Random Forest ranking is used to reduce the 1280-dimensional embeddings into four highly discriminative features that can be represented using quantum encoding. These features are next coded as a 4-qubit quantum circuit of ZZFeatureMap and RealAmplitudes ansatz and a hybrid quantum layer is formed whose outputs are concatenated with a CNN classifier to arrive at final predictions. This optimization strategy of features representation based on quantum awareness balances high-dimensional representation and qubit constraints, and is proved to be more stable in 4-qubit encoding and better in comparison to higher-qubit encoding under noisy conditions. Experimental evaluation on the “New Plant Diseases Dataset” achieved 98.2% accuracy, 98.5% precision, 97.7% recall, 98% F1-score, and a macro ROC-AUC of 0.985, confirming that low-qubit hybrid models can deliver state-of-the-art plant disease classification while ensuring practicality and scalability for real-world agricultural applications on near-term quantum computing platforms.
Reduced crop losses and guaranteeing agricultural productivity depend on the early and accurate detection of plant diseases. This work offers a Hybrid Quantum-Classical HQC-CNN, or convolutional in nature neural network that uses image analysis to determine various diseases of apple leaves. The model is executed using the PennyLane a system which blends quantum computing techniques with conventional deep computing methods. In the classical segment, layers for collaboration and convolution were used to extract significant spatial features from normalised and resized leaf images.These features are mapped into quantum states by Angle Embedding, and intricate patterns are revealed by StronglyEntanglingLayers processing them inside an electromagnetic circuit. The quantum information layer's output is sent to a softmax classifier, which divides the input into different disease groups using Pauli-Z expectation value measurements. The framework has been examined and developed with a publicly available dataset of apple leaf images, and it demonstrated strong performance on several evaluation parameters, such as the F1-s remembering, reliability, and quality. These findings show how well the model can identify a range of leaf diseases and point to how it could help precision farming and the broader field of quantum improved neural networks.
Potato leaf diseases pose a serious challenge to agricultural sustainability and food security in regions such as West Bengal, India, where potato cultivation plays a vital socio-economic role. Although classical deep learning models demonstrate strong performance, they often suffer from overfitting, limited generalization, and heavy reliance on large labeled datasets, while manual inspection remains time-consuming and error-prone. This work proposes a hybrid Quantum Convolutional Neural Network (QCNN) integrated with 4-qubit parameterized quantum circuits (PQCs) for accurate classification of five major potato leaf disease categories, including nutrient deficiency, fungal, and viral infections. The model incorporates quantum activation layers, quantum softmax, and a focused loss function to enhance non-linear feature representation and multi-class separability. Grad-CAM is employed to improve the interpretability of predictions. Experiments conducted on a diverse, field-collected dataset from Bardhaman, Bankura, and Purulia districts achieve a classification accuracy of 99.94%, demonstrating superior generalization, robustness, and explainability compared to classical approaches. This study highlights the practical potential of quantum-enhanced AI for precision agriculture.
Medical and agricultural sector needs to make timely and accurate diagnosis of the disease in order to sustain crops and keep patients alive. Although conventional deep learning models can be also applicable to provide good performance, they are also typified with high-computational and scaling along with high-dimensionality agricultural and medical images. In a bid to overcome these challenges, the current paper proposes a Hybrid Quantum -Classical Artificial Intelligence (HQCAI) system that should be used in dual-domain image classification, namely, plant leaf disease classification and oesophageal cancer classification. The proposed architecture implements a conventional Convolutional Neural Network (CNN) to obtain spatial features of high quality and quantum-enhanced classifiers to optimize the decision boundaries relying on quantum superposition and parallelism with Variational Quantum Circuits (VQCs) and Quantum Support Vector Machines (QSVMs). It presents a common knowledge representation layer in order to help in the effective storage of characteristics, cross-domain learning and intelligent inference between heterogeneous data sets. Experiments are carried out with the help of the data collection. The proposed hybrid model gives a better classification rate of 98.42, with precision of 98.11, recall of 97.94 and F1-score of 98.02 compared to CNN and CNN-SVM models in the state of the art. Convergence and performance of the system is also better when noises are present. The results support the practical role of quantum-enhanced visual classification and describe the prospects of hybrid quantum classical intelligence of the agricultural diagnostics and medical diagnostics systems in the future.
Accurate image classification is vital in smart agriculture, where early detection of plant diseases can boost crop yield and sustainability. While traditional Convolutional Neural Networks (CNNs) perform well, they face computational and scalability issues as dataset sizes grow. To address this, we propose a quantum-enhanced deep learning framework that combines the Quantum Fourier Transform (QFT) for image preprocessing with a Quantum Convolutional Neural Network (QCNN) for classification. QFT transforms images into the frequency domain, highlighting disease-related features and reducing background noise, while QCNN leverages quantum principles like superposition and entanglement for faster and more efficient feature learning. We evaluated four architectures—classical CNN, QCNN, CNN+QFT, and the hybrid QCNN+QFT—on coffee leaf datasets containing 2,000, 5,000, and 12,000 images. Across all dataset sizes, QCNN+QFT consistently achieved the highest accuracy and lowest training times. On the largest dataset, it reached 98.87% accuracy in 25.46 seconds per epoch, compared to the CNN’s 85.13% in 638.39 seconds. For 5,000 images, it achieved 94.78% in 11.30 seconds, outperforming CNN’s 83.42% in 215.57 seconds. Even on 2,000 images, it scored 90.06% in 2.12 seconds versus CNN’s 79.23% in 67.48 seconds. While QCNN and CNN+QFT also improved over the classical CNN, neither matched the hybrid’s superior balance of speed and accuracy, demonstrating its potential for scalable, real-time plant disease detection.
Medical and agricultural sector needs to make timely and accurate diagnosis of the disease in order to sustain crops and keep patients alive. Although conventional deep learning models can be also applicable to provide good performance, they are also typified with high-computational and scaling along with high-dimensionality agricultural and medical images. In a bid to overcome these challenges, the current paper proposes a Hybrid Quantum -Classical Artificial Intelligence (HQCAI) system that should be used in dual-domain image classification, namely, plant leaf disease classification and oesophageal cancer classification. The proposed architecture implements a conventional Convolutional Neural Network (CNN) to obtain spatial features of high quality and quantum-enhanced classifiers to optimize the decision boundaries relying on quantum superposition and parallelism with Variational Quantum Circuits (VQCs) and Quantum Support Vector Machines (QSVMs). It presents a common knowledge representation layer in order to help in the effective storage of characteristics, cross-domain learning and intelligent inference between heterogeneous data sets. Experiments are carried out with the help of the data collection. The proposed hybrid model gives a better classification rate of 98.42, with precision of 98.11, recall of 97.94 and F1-score of 98.02 compared to CNN and CNN-SVM models in the state of the art. Convergence and performance of the system is also better when noises are present. The results support the practical role of quantum-enhanced visual classification and describe the prospects of hybrid quantum classical intelligence of the agricultural diagnostics and medical diagnostics systems in the future.