Recognising plant leaf diseases from images is difficult because symptoms of different diseases often look alike, and expert inspection is not always available to growers. This paper presents a hybrid quantum deep learning framework that predicts both the plant species and the disease of a leaf image in a single pass. An ImageNet-pretrained MobileNetV2 backbone extracts a high-dimensional feature vector, which a small projection reduces to four values that drive a four-qubit variational quantum circuit built from angle embedding and strongly entangling layers. The resulting Pauli-Z expectation values are concatenated with the classical features and passed to separate plant and disease classification heads. The model is compared with a MobileNetV2-only baseline on an identical stratified train–test split covering six crops and twenty disease and healthy classes. The hybrid model scores higher than the baseline on accuracy, precision, recall and F1-score for both tasks, although the margins are modest and come from a single split. The framework is also deployed as a Flask web application that adds blur and leaf validation, leaf segmentation, Grad-CAM explanations and disease-management guidance. The results are therefore presented as preliminary evidence of the value of a compact quantum feature stage rather than a definitive advantage.
Introduction
The proposed project focuses on developing a Hybrid Quantum-Classical Deep Learning System for Plant Disease Detection and Classification. The main aim is to identify plant species and detect diseases from leaf images using artificial intelligence, deep learning, and quantum machine learning techniques.
Plant diseases significantly affect crop production and quality. Traditionally, farmers identify diseases through visual inspection, which can be time-consuming, subjective, and prone to errors. To overcome these limitations, the proposed system uses image processing and machine learning to automate plant disease identification and provide useful treatment guidance.
Main Objectives
Plant Identification: Identify the plant species from an uploaded leaf image.
Disease Detection: Detect whether the plant is healthy or affected by a disease.
Hybrid Model Development: Combine MobileNetV2 with a four-qubit variational quantum circuit to improve feature extraction.
Explainable AI: Use Grad-CAM heat maps to highlight the image regions responsible for predictions.
Disease Guidance: Provide disease-specific information and recommendations to help users take appropriate action.
Proposed Methodology
The system uses a combination of classical deep learning and quantum machine learning.
Image Preprocessing: Uploaded images are resized to 224 × 224 pixels, converted into RGB format, and normalized.
Image Validation: The system checks image blur and verifies whether the uploaded image contains a leaf.
Leaf Segmentation: Separates the leaf from the background to improve plant identification.
Feature Extraction: MobileNetV2 extracts 1280-dimensional visual features from the input image.
Quantum Feature Enhancement: A four-qubit variational quantum circuit processes additional features and generates four quantum features.
Feature Fusion: Classical and quantum features are combined into a 1284-dimensional feature vector.
Multi-Task Classification: Two separate output layers predict the plant species (6 classes) and disease category (20 classes).
Result Visualization: The application displays prediction labels, confidence scores, Grad-CAM heat maps, and disease-specific guidance.
Dataset and Technologies
The dataset contains six plant species:
Apple
Tomato
Potato
Corn
Grape
Guava
It includes 20 disease and healthy categories, with up to 300 images per class. The dataset is divided into 80% training data and 20% testing data, resulting in 1,200 test images.
The major technologies used include:
MobileNetV2: For efficient image feature extraction.
Variational Quantum Circuit: For quantum-based feature enhancement.
Grad-CAM: For visual explanation of model predictions.
Deep Learning: For plant and disease classification.
Web Application: For uploading images and displaying predictions and recommendations.
Existing System vs. Proposed System
The existing system uses a MobileNetV2-based CNN to classify plant species and diseases.
The proposed system enhances this approach by introducing a four-qubit quantum circuit, image validation, leaf segmentation, explainable AI, and disease-specific recommendations.
Both models use the same dataset split and evaluation procedure to compare the contribution of the quantum component.
Conclusion
This paper presented a MobileNetV2-based multi-task framework that adds the output of a four-qubit variational quantum circuit to the classical features, so that the plant species and the disease are predicted from one leaf image, together with a web application offering input validation, Grad-CAM explanations and disease guidance. Against a MobileNetV2-only baseline on the same split, the hybrid model improved plant accuracy from 95.42% to 96.08% and disease accuracy from 86.00% to 87.58%. These gains are small, come from a single split, and are not separated from the effect of the added dense layers; they indicate promise rather than a proven advantage for the quantum stage. Future work will repeat the experiments over several seeds with statistical testing, add a classical ablation of the quantum branch, explore other circuit designs, extend the dataset with more crops and field-captured images, validate the management advice with agricultural experts, and optimise the model for mobile devices.
References
[1] A. Bhargava et al., “Plant leaf disease detection, classification, and diagnosis using computer vision and artificial intelligence: A review,” IEEE Access, vol. 12, pp. 37443–37469, 2024, doi: 10.1109/ACCESS.2024.3373001.
[2] V. Balafas et al., “Machine learning and deep learning for plant disease classification and detection,” IEEE Access, vol. 11, pp. 114352–114377, 2023, doi: 10.1109/ACCESS.2023.3324722.
[3] M. Sandler, A. Howard, M. Zhu, A. Zhmoginov, and L.-C. Chen, “MobileNetV2: Inverted residuals and linear bottlenecks,” in Proc. IEEE/CVF Conf. Comput. Vis. Pattern Recognit. (CVPR), 2018, pp. 4510–4520, doi: 10.1109/CVPR.2018.00474.
[4] J. Biamonte et al., “Quantum machine learning,” Nature, vol. 549, pp. 195–202, 2017, doi: 10.1038/nature23474.
[5] M. Schuld and N. Killoran, “Quantum machine learning in feature Hilbert spaces,” Phys. Rev. Lett., vol. 122, no. 4, Art. no. 040504, 2019, doi: 10.1103/PhysRevLett.122.040504.
[6] V. Havlí?ek et al., “Supervised learning with quantum-enhanced feature spaces,” Nature, vol. 567, pp. 209–212, 2019, doi: 10.1038/s41586-019-0980-2.
[7] N. Nigar, H. M. Faisal, M. Umer, O. A. Oki, and J. M. Lukose, “Improving plant disease classification with deep-learning-based prediction model using explainable artificial intelligence,” IEEE Access, vol. 12, pp. 100005–100014, 2024, doi: 10.1109/ACCESS.2024.3428553.
[8] A. Oad et al., “Plant leaf disease detection using ensemble learning and explainable AI,” IEEE Access, vol. 12, pp. 156038–156049, 2024, doi: 10.1109/ACCESS.2024.3484574.
[9] A. Daalah, S. Aldousari, S. Alrashed, H. Alenezi, A. Alhajri, and A. Eleyan, “Plant leaf disease classification using fine-tuned pre-trained CNN models,” in Proc. 6th Int. Conf. Bio-engineering for Smart Technologies (BioSMART), 2025, doi: 10.1109/BioSMART66413.2025.11046055.
[10] A. Prommakhot, J. Onshaunjit, W. Ooppakaew, G. Samseemoung, and J. Srinonchat, “Hybrid CNN and transformer-based sequential learning techniques for plant disease classification,” IEEE Access, vol. 13, pp. 122876–122887, 2025, doi: 10.1109/ACCESS.2025.3586285.
[11] R. R. Selvaraju et al., “Grad-CAM: Visual explanations from deep networks via gradient-based localization,” in Proc. IEEE Int. Conf. Comput. Vis. (ICCV), 2017, pp. 618–626, doi: 10.1109/ICCV.2017.74.