Abstract Crop and weed detection is an essential task in precision agriculture which is typically approached through classical machine learning, and more recently, through quantum machine learning (QML). This research compares classical and quantum approaches for crop-weed distinctions in aerial field photography. A convolutional neural network (CNN) stands as a classical baseline, and two QML models, a quantum support vector classifier and a variational quantum classifier (VQC) are utilized with a publicly accessible annotated dataset. The models are evaluated in terms of classification accuracy, inference velocity, and computational complexity. The CNN achieves a peak test accuracy of approximately 94%, which surpasses the QML approaches yielding 85%–89% accuracies. fiCNN achieves substantially faster inference, 10 ms per image and hundreds of milliseconds for quantum models due to circuit execution and simulation cost. While worse at predicting, QML models still consume considerably fewer trainable parameters (tens compared to over one million for the CNN), which may signal benefits on model compactness. The results show that classical deep learning is the most effective and reliable approach for crop and weed detection in 2025, while QML models represent a promising future direction and will become more competitive as the computing hardware and algorithms advance.
Abstract Crop and weed detection is an essential task in precision agriculture which is typically approached through classical machine learning, and more recently, through quantum machine learning (QML). This research compares classical and quantum approaches for crop-weed distinctions in aerial field photography. A convolutional neural network (CNN) stands as a classical baseline, and two QML models, a quantum support vector classifier and a variational quantum classifier (VQC) are utilized with a publicly accessible annotated dataset. The models are evaluated in terms of classification accuracy, inference velocity, and computational complexity. The CNN achieves a peak test accuracy of approximately 94%, which surpasses the QML approaches yielding 85%–89% accuracies. fiCNN achieves substantially faster inference, 10 ms per image and hundreds of milliseconds for quantum models due to circuit execution and simulation cost. While worse at predicting, QML models still consume considerably fewer trainable parameters (tens compared to over one million for the CNN), which may signal benefits on model compactness. The results show that classical deep learning is the most effective and reliable approach for crop and weed detection in 2025, while QML models represent a promising future direction and will become more competitive as the computing hardware and algorithms advance.
Classification of crops and weeds is an important aspect in precision agriculture, as it aids in providing a means for efficient weed control and reduced herbicide applications to achieve high crop productivity. In this paper, a quantum-classical hybrid deep learning approach is presented for accurate identification of crops and weeds from sorghum images using the dataset Sorghum Crop Weed Dataset_Classification. In this hybrid framework, a convolutional neural network (CNN) for feature extraction in the spatial domain, along with a variational quantum circuit (VQC) to provide better feature representation and classification abilities in higher-dimensional quantum space, is used. This allows for effective differentiation of crop and weed instances that appear alike in different agricultural conditions. Data preprocessing methods like resizing, normalization, and data augmentation (rotation, flipping, and brightness adjustment) are performed to ensure uniformity in data and increase the generalizability of the trained model. According to experimental evaluations conducted using traditional metrics, the designed model attains an accuracy of 97.3%, precision of 96.9%, recall of 97.2%, F1-score of 97.1%, and a training loss of 0.09. The results show a significant increase from baseline approaches such as CNN (92.5% accuracy), ResNet (94.0% accuracy), and transfer learning (95.0% accuracy). In addition, the designed model performs well when tested under challenging situations encountered in the real-world agricultural environment, such as varying illuminations, partial occlusion, and background noises. In summary, the above results prove the success of combining classical deep learning and quantum feature learning to enhance classification performance.
This study investigates the effectiveness of a hybrid quantum-classical framework for plant disease detection, with a focus on improving classification accuracy compared to conventional approaches. Image-based datasets of crop diseases were processed and encoded into variational quantum circuits (VQCs), which were then integrated with classical machine learning classifiers. The proposed model achieved an overall classification accuracy of 95.2%, outperforming a baseline convolutional neural network that recorded 91.4% accuracy on the same dataset. Performance evaluation was carried out using multiple metrics including precision, recall, and F1-score, with the hybrid approach showing consistent gains across all categories. Furthermore, the model demonstrated reduced overfitting and improved robustness when tested under noisy and high-dimensional input conditions. These results highlight the potential of quantum-assisted machine learning to significantly enhance crop disease detection, offering a scalable and accurate pathway for sustainable agriculture and precision farming.
A potential path forward is Quantum Machine Learning (QML), which aims to leverage quantum computing in conjunction with classical machine learning to enhance computing efficiency and the expressiveness of models. In this paper, two different quantum classifiers - Variational Quantum Classifier (VQC) and Quantum Kernel Support Vector Machine (QSVM) - are compared with three classical classifiers as baseline classifiers - Logistic Regression, Support Vector Machine (SVM), and a Multi-Layer Perceptron (MLP) - on the Breast Cancer Wisconsin dataset. The quantum circuits were created in the PennyLane framework and simulated on a classical backend. However, in terms of accuracy, classical Logistic Regression performed better with an accuracy of 97.8%, classical SVM and QSVM with an accuracy of 95.6% each, although the Quantum VQC achieved a lower accuracy of 88.9% and had a recall of 100% for the benign class, though it correctly identified only 12 of the 17 malignant cases (a malignant-class recall of approximately 70.6%). The drawback of quantum models is the higher training time; however, since the quantum circuit needs to be classically simulated, the quantum SVM took 23.29 seconds compared to less than 0.01 seconds for the classical linear models. These results indicate that for small structured datasets, classifiers based on quantum computing have not yet surpassed well-tuned classical counterparts. In some respects (e.g., benign-class recall), they perform competitively, though not on malignant-class recall, where the VQC in particular performed worse than the classical baselines, which is worth further investigation on real quantum computers.
Quantum machine learning (QML) is receiving increasing attention as a potential tool for addressing the computational challenges inherent in agricultural remote sensing, where multimodal Earth observational satellite, soil, and climate datasets must be processed at scale. We review recent developments in quantum–classical hybrid modeling for core remote-sensing tasks, including crop type classification, yield estimation, cloud detection, and time series forecasting. Researchers typically combine classical feature extractors with shallow variational circuits or quantum kernels to accommodate the limited qubit counts and noise levels of contemporary quantum devices. We summarize performance trends reported in the literature, emphasizing the role of classical preprocessing, compact quantum circuits, and rigorous benchmarking against strong classical baselines. We consolidate the reviewed studies in a structured comparison covering task, dataset, qubit count, encoding, execution platform, and reported performance relative to classical baselines. The current hardware and data limitations that restrict current QML applications associated with state preparation, circuit depth, and encoding of agricultural datasets are discussed. Near-term prospects for QML in agriculture can involve small-scale hybrid models for multi-class crop mapping and short horizon sequence prediction, while longer-term opportunities may encompass quantum accelerated image segmentation, data fusion, and large scale optimization. Current evidence indicates that QML performance at best matches strong classical baselines, with the most plausible near-term utility arising in narrow settings such as severely label-limited learning and QUBO-structured optimization. Future gains will depend jointly on hardware maturation, data encoding strategies, algorithm design, and integration with classical AI systems.
Breast cancer is rated as one of the top leading diagnosed and deadliest cancers affecting women worldwide and causing high mortality rate among cancers. With the data-driven applications, hybrid quantum-classical machine learning models have evolved as significant alternatives for handling complex feature spaces with the growing interest in quantum computing. This paper presents a systematic comparative study of classical and quantum machine learning approaches for breast cancer classification using the Wisconsin Breast Cancer Dataset. In this architecture we have preprocessed classical feature vectors using standard normalized techniques and parallelly quantum states are mapped through data embedding techniques which relies on parameterized rotational gates. In this work we have implemented all quantum models using the Pennylane library, resulting in hybrid quantum–classical optimization on near-term quantum simulators. The comparison results in multiple classical and quantum models. In classical we have Support Vector Machine (SVM), K-Nearest Neighbors (KNN) and Convolutional Neural Networks (CNN), which are compared with the quantum-based models such as the Variational Quantum Classifier (VQC), Quantum Support Vector Machine (QSVM) and Quantum Convolutional Neural Network (QCNN). In QSVM, we have performed dimensionality reduction using Principal Component Analysis (PCA) before to quantum embedding, and quantum-generated features are used to train a classical linear SVM. The evaluation metrics results in accuracy, precision, recall, F1-score, ROC curves and confusion matrices. The classical models achieves the accuracy in the range of 94%-96%, whereas competitive performance of quantum models demonstrate depending on the embedding strategy and circuit configuration. In the implemented quantum models, the variational quantum classifier achieved the highest classification accuracy of 96.06%, surpassing other evaluated models. These results demonstrate quantum embedding combined with variational quantum circuits can effectively capture discriminative patterns in medical datasets, indicating the implementation of quantum machine learning models for breast cancer diagnosis with the present quantum simulating platforms.
This paper presents a systematic empirical benchmarking of quantum machine learning (QML) applications within the under-explored domain of agricultural image classification. The authors implement and evaluate four hybrid quantum-classical architectures—including variational quantum classifiers (VQC), quanvolutional neural networks (QNN), and a ResNet-50 combined with a Quantum Support Vector Machine (QSVM)—against a purely classical EfficientNet baseline. The architectures are tested on three agricultural disease datasets: potato tuber disease, jujube insect pests, and paddy leaf rust. The experimental findings reveal that the ResNet-50+QSVM hybrid model achieves the optimal balance of classification accuracy and training stability, reaching a 93.4% mean accuracy. Crucially, the hybrid QSVM demonstrates extreme sample efficiency by requiring 41% fewer labeled training samples than the classical EfficientNet baseline to cross a 90% accuracy threshold, showing the capability of quantum feature maps in low-data regimes.