Challenges
Datasets
Workspaces
Papers
Discussions
Leaderboard
Log inSign up
Challenges
Datasets
Workspaces
Papers
Discussions
Leaderboard
Blog
Job Board
Q3AS

© 2026 Aqora Quantum S.A.S.

TermsPrivacyLegal Notice

arXiv:2609.14025

Quantum-Gated LiteSSD: A Parameter-Efficient Lightweight Hybrid Quantum-Classical Framework for Forward-Looking Sonar Object Detection

Niloy Kumar Mondal, Poulomi Sarker Puja

Published Sep 12, 2026• 0 citations
arXivHTMLPDF

Abstract

Forward-looking sonar object detection is essential for underwater perception, yet deployment on embedded platforms requires highly compact models. To address this challenge, we explore quantum computing and introduce Quantum-Gated LiteSSD, a parameter-efficient hybrid quantum--classical detector that reformulates QuCNet-style multi-circuit quantum processing as an identity-centered channel-gating mechanism for spatial feature modulation. Experiments on the Marine Debris Watertank dataset and UATD forward-looking sonar benchmarks demonstrate an effective parameter--accuracy trade-off. The proposed detector achieves 90.84% $\mathrm{mAP}_{50}$ on Watertank with approximately $62\times$ fewer parameters than YOLO26s and $164.3\times$ fewer parameters than SSD-VGG16. On UATD, the model achieves 70.37% $\mathrm{mAP}_{50}$ with only 0.150M parameters, making it approximately $4.1\times$ smaller than SSGA-YOLO while retaining meaningful multi-class detection capability.

DOI
10.48550/arxiv.2609.14025

Related papers

Research and proposal of a hybrid quantum artificial intelligence model for classifying marine mammal signals based on passive sonar principles

Hoang Bach Nhat, Nguyen Duc Hai, Pham Thanh Thuong +5 more

The convergence between artificial intelligence (AI) and quantum computing is ushering in a new era for problems requiring high levels of computational complexity, particularly in digital signal processing. This paper presents an overview and evaluation of the development trends in hybrid Quantum-AI models, focusing on applications for the recognition and classification of underwater acoustic signals—a challenging domain due to instability, high noise levels, and data scarcity. The study proposes a Hybrid Quantum-CNN model, utilizing CNN for feature extraction and dimensionality reduction, combined with a Variational Quantum Classifier for classification in Hilbert space. The results of the proposed model are compared with an equivalent CNN network, demonstrating that HQC achieves higher accuracy (an increase of 3,5%), while excelling in efficiency: faster convergence speed, and superior generalization capabilities on the same real underwater acoustic dataset. These findings highlight the benefits of quantum approaches not only in accuracy but also in constructing more efficient and robust models; hybrid models open promising avenues for applying quantum machine learning to complex signal reception systems in practical.

Journal of Military Science and Technology•Dec 31, 2025•0 citations

QLiteNet: A Lightweight Quantum-Enhanced Hybrid Architecture for Cancer Image Classification

Quang Nhan Hoang, Duc Nguyen Huu Dang, Tri Nguyen Thai +1 more

Accurate cancer diagnosis from biomedical images is vital for clinical decision-making, yet many deep learning models require substantial computational resources. Although quantum machine learning (QML) can improve feature representation, it often struggles with scalability and complexity on large image datasets. To address these challenges, we propose QLiteNet, a lightweight quantum-classical hybrid architecture that achieves competitive accuracy while remaining parameter-efficient. QLiteNet employs a streamlined convolutional backbone with a quantum module for enhanced feature separation across complex class boundaries. Experiments demonstrate that QLiteNet uses approximately 0.66 million parameters while achieving 96.52% accuracy on the BRISC and 95.39% on the BreaKHis datasets. M Accordingly, these results highlight QLiteNet as a parameter-efficient option for cancer image classification.

International Conference on Green Technology and Sustainable Development (GTSD)•Jul 1, 2026•0 citations

CQ-CNN: A lightweight hybrid classical–quantum convolutional neural network for Alzheimer’s disease detection using 3D structural brain MRI

Mominul Islam, Mohammad Junayed Hasan, M. R. C. Mahdy

The automatic detection of Alzheimer's disease (AD) using 3D volumetric MRI data is a complex, multi-domain challenge that has traditionally been addressed by training classical convolutional neural networks (CNNs). With the rise of quantum computing and its potential to replace classical systems in the future, there is a growing need to: (i) develop automated systems for AD detection that run on quantum computers, (ii) explore the capabilities of current-generation classical-quantum architectures, and (iii) identify their potential limitations and advantages. To reduce the complexity of multi-domain expertise while addressing the emerging demands of quantum-based automated systems, our contribution in this paper is twofold. First, we introduce a simple preprocessing framework that converts 3D MRI volumetric data into 2D slices. Second, we propose CQ-CNN, a parameterized quantum circuit (PQC)-based lightweight hybrid classical-quantum convolutional neural network that leverages the computational capabilities of both classical and quantum systems. Our experiments on the OASIS-2 dataset reveal a significant limitation in current hybrid classical-quantum architectures, as they face difficulties converging when class images are highly similar, such as between moderate dementia and non-dementia classes of AD, which leads to gradient failure and optimization stagnation. However, when convergence is achieved, the quantum model demonstrates a promising quantum advantage by attaining state-of-the-art accuracy with far fewer parameters than classical models. For instance, our [Formula: see text]-3-qubit model achieves 97.5% accuracy using only 13.7K parameters (0.05 MB), which is 5.67% higher than a classical model with the same parameter count. Nevertheless, our results highlight the need for improved quantum optimization methods to support the practical deployment of hybrid classical-quantum models in AD detection and related medical imaging tasks.

PLoS ONE•Sep 22, 2025•18 citations

Parameter-Efficient Quantum Machine Learning for Multi-Class Brain Tumor Classification

Rahman Abdur, Tae-Hyong Kim

Medical image classification is a critical task in computer-aided diagnostics, but classical deep learning models often demand substantial computational resources and large parameter spaces. This paper investigates the potential of hybrid quantum-classical neural networks to address this challenge in the context of multi-class brain tumor classification from MRI scans. We introduce a custom hybrid quantum machine learning (QML) model and conduct a rigorous comparative analysis against its purely classical counterpart, as well as established architectures such as ResNet-18 and VGG-16. Our findings demonstrate that the proposed hybrid QML model achieves a state-of-the-art test accuracy of 96.58%, matching the performance of the significantly larger ResNet-18 model. Crucially, the hybrid model outperforms its direct classical counterpart, the Classical Benchmark (96.20%), while utilizing approximately 80% fewer parameters. Although simulating quantum circuits results in longer training times, our findings highlight the remarkable parameter efficiency and competitive accuracy of hybrid quantum models, suggesting a promising direction for developing powerful yet lightweight networks for medical diagnostics.

International Conference on Information and Communication Technology Convergence (ICTC)•Oct 14, 2025•1 citation

Quantum-Assisted Faster RCNN Detection and Classification of Small Underwater Objects Using a Hybrid Classical-Quantum Models

Dharshan V, S Markkandan, A. Sugunapriya

Underwater Small object detection and classification are crucial for ecological monitoring, yet traditional algorithms struggle with underwater image complexity, including low visibility and considerable fluctuations in object sizes. This article presents Quantum-Assisted Faster R-CNN Detection and Classification Models specifically suited to meet these problems, leveraging a transfer learning strategy between classical and quantum neural networks. Firstly, standard models like ResNet-50, EfficientNet-B0, and MobileNet-V2 are pretrained for feature extraction on the underwater dataset. Secondly, these models are then integrated with a quantum layer with two entangling layers, giving three Quantum Neural Network (QNN) hybrid architectures within the Faster R-CNN framework. Finally, the UTDAC2020 dataset is used to evaluate the models, having four underwater small object classes (echinus, starfish, holothurian, and scallop) with images standardized to$640 \times 640$pixels to account for variations in resolution and class imbalance. The QNN-ResNet-50 model performs better than other hybrid models, with a validation accuracy of 81.09% and an F1 score of 0.8347. This work demonstrates the possibility of hybrid classical-quantum models as robust solutions for underwater small object detection and classification, paving the way for future advances in demanding visual environments.

International Conference on Communication, Computing and Electronics Systems (ICCCES)•Jan 21, 2026•0 citations

Parameter-Efficient QCNN-Based Signal Detector for GEO Satellite SISO Communications

Ki-Chang Tong, Yeong-Gyun Jung, MinHyeok Choi +3 more

Signal detection in Geostationary Earth Orbit (GEO) satellite links can be degraded by Rician fading, receiver noise and imperfect channel state information (CSI). This paper investigates a parameter-count-efficient hybrid quantum-classical detector for GEO satellite Single-Input Single-Output (SISO) BPSK links. The objective is to evaluate whether a compact Quantum Convolutional Neural Network (QCNN) feature-processing block can provide CNN-comparable uncoded bit-detection reliability under a controlled imperfect-CSI software-simulation setup, while keeping the trainable model size small. The channel and SNR normalizations are explicitly stated, and the conventional random Line-of-Sight (LOS) phase in the Rician model is addressed through a phase-invariance argument for the proposed channel-compensated features. Four physics-based diagnostic features are extracted from a channel-compensated observation and encoded into a 4-qubit QCNN. The QCNN contains 2,289 trainable parameters, including 48 quantum-circuit parameters and 2,241 classical post-processing parameters, whereas the same-input CNN baseline contains 12,609 trainable parameters. Simulations from − 10 dB to 20 dB show that the QCNN achieves 89.39% average accuracy, compared with 89.74% for the CNN. As a complementary reliability-oriented metric, a capacity-weighted effective-throughput indicator is also evaluated, where the QCNN achieves an average effective throughput of 2.376 bps/Hz, compared with 2.406 bps/Hz for the CNN. These results suggest that the proposed QCNN can provide a favorable balance between detection reliability, effective-throughput behavior and trainable parameter count. This balance is particularly relevant to isolated GEO satellite receiver environments, where onboard memory, model-update opportunities and processing resources are limited after deployment.

International Conference on Computer and Communications Management (ICCCM)•Jul 24, 2026•0 citations
Forward-looking sonar object detection is essential for underwater perception, yet deployment on embedded platforms requires highly compact models. To address this challenge, we explore quantum computing and introduce Quantum-Gated LiteSSD, a parameter-efficient hybrid quantum--classical detector that reformulates QuCNet-style multi-circuit quantum processing as an identity-centered channel-gating mechanism for spatial feature modulation. Experiments on the Marine Debris Watertank dataset and UATD forward-looking sonar benchmarks demonstrate an effective parameter--accuracy trade-off. The proposed detector achieves 90.84% mAP50\mathrm{mAP}_{50}mAP50​ on Watertank with approximately 62×62\times62× fewer parameters than YOLO26s and 164.3×164.3\times164.3× fewer parameters than SSD-VGG16. On UATD, the model achieves 70.37% mAP50\mathrm{mAP}_{50}mAP50​ with only 0.150M parameters, making it approximately 4.1×4.1\times4.1× smaller than SSGA-YOLO while retaining meaningful multi-class detection capability.
The convergence between artificial intelligence (AI) and quantum computing is ushering in a new era for problems requiring high levels of computational complexity, particularly in digital signal processing. This paper presents an overview and evaluation of the development trends in hybrid Quantum-AI models, focusing on applications for the recognition and classification of underwater acoustic signals—a challenging domain due to instability, high noise levels, and data scarcity. The study proposes a Hybrid Quantum-CNN model, utilizing CNN for feature extraction and dimensionality reduction, combined with a Variational Quantum Classifier for classification in Hilbert space. The results of the proposed model are compared with an equivalent CNN network, demonstrating that HQC achieves higher accuracy (an increase of 3,5%), while excelling in efficiency: faster convergence speed, and superior generalization capabilities on the same real underwater acoustic dataset. These findings highlight the benefits of quantum approaches not only in accuracy but also in constructing more efficient and robust models; hybrid models open promising avenues for applying quantum machine learning to complex signal reception systems in practical.
Accurate cancer diagnosis from biomedical images is vital for clinical decision-making, yet many deep learning models require substantial computational resources. Although quantum machine learning (QML) can improve feature representation, it often struggles with scalability and complexity on large image datasets. To address these challenges, we propose QLiteNet, a lightweight quantum-classical hybrid architecture that achieves competitive accuracy while remaining parameter-efficient. QLiteNet employs a streamlined convolutional backbone with a quantum module for enhanced feature separation across complex class boundaries. Experiments demonstrate that QLiteNet uses approximately 0.66 million parameters while achieving 96.52% accuracy on the BRISC and 95.39% on the BreaKHis datasets. M Accordingly, these results highlight QLiteNet as a parameter-efficient option for cancer image classification.
The automatic detection of Alzheimer's disease (AD) using 3D volumetric MRI data is a complex, multi-domain challenge that has traditionally been addressed by training classical convolutional neural networks (CNNs). With the rise of quantum computing and its potential to replace classical systems in the future, there is a growing need to: (i) develop automated systems for AD detection that run on quantum computers, (ii) explore the capabilities of current-generation classical-quantum architectures, and (iii) identify their potential limitations and advantages. To reduce the complexity of multi-domain expertise while addressing the emerging demands of quantum-based automated systems, our contribution in this paper is twofold. First, we introduce a simple preprocessing framework that converts 3D MRI volumetric data into 2D slices. Second, we propose CQ-CNN, a parameterized quantum circuit (PQC)-based lightweight hybrid classical-quantum convolutional neural network that leverages the computational capabilities of both classical and quantum systems. Our experiments on the OASIS-2 dataset reveal a significant limitation in current hybrid classical-quantum architectures, as they face difficulties converging when class images are highly similar, such as between moderate dementia and non-dementia classes of AD, which leads to gradient failure and optimization stagnation. However, when convergence is achieved, the quantum model demonstrates a promising quantum advantage by attaining state-of-the-art accuracy with far fewer parameters than classical models. For instance, our [Formula: see text]-3-qubit model achieves 97.5% accuracy using only 13.7K parameters (0.05 MB), which is 5.67% higher than a classical model with the same parameter count. Nevertheless, our results highlight the need for improved quantum optimization methods to support the practical deployment of hybrid classical-quantum models in AD detection and related medical imaging tasks.
Medical image classification is a critical task in computer-aided diagnostics, but classical deep learning models often demand substantial computational resources and large parameter spaces. This paper investigates the potential of hybrid quantum-classical neural networks to address this challenge in the context of multi-class brain tumor classification from MRI scans. We introduce a custom hybrid quantum machine learning (QML) model and conduct a rigorous comparative analysis against its purely classical counterpart, as well as established architectures such as ResNet-18 and VGG-16. Our findings demonstrate that the proposed hybrid QML model achieves a state-of-the-art test accuracy of 96.58%, matching the performance of the significantly larger ResNet-18 model. Crucially, the hybrid model outperforms its direct classical counterpart, the Classical Benchmark (96.20%), while utilizing approximately 80% fewer parameters. Although simulating quantum circuits results in longer training times, our findings highlight the remarkable parameter efficiency and competitive accuracy of hybrid quantum models, suggesting a promising direction for developing powerful yet lightweight networks for medical diagnostics.
Underwater Small object detection and classification are crucial for ecological monitoring, yet traditional algorithms struggle with underwater image complexity, including low visibility and considerable fluctuations in object sizes. This article presents Quantum-Assisted Faster R-CNN Detection and Classification Models specifically suited to meet these problems, leveraging a transfer learning strategy between classical and quantum neural networks. Firstly, standard models like ResNet-50, EfficientNet-B0, and MobileNet-V2 are pretrained for feature extraction on the underwater dataset. Secondly, these models are then integrated with a quantum layer with two entangling layers, giving three Quantum Neural Network (QNN) hybrid architectures within the Faster R-CNN framework. Finally, the UTDAC2020 dataset is used to evaluate the models, having four underwater small object classes (echinus, starfish, holothurian, and scallop) with images standardized to640×640640 \times 640640×640pixels to account for variations in resolution and class imbalance. The QNN-ResNet-50 model performs better than other hybrid models, with a validation accuracy of 81.09% and an F1 score of 0.8347. This work demonstrates the possibility of hybrid classical-quantum models as robust solutions for underwater small object detection and classification, paving the way for future advances in demanding visual environments.
Signal detection in Geostationary Earth Orbit (GEO) satellite links can be degraded by Rician fading, receiver noise and imperfect channel state information (CSI). This paper investigates a parameter-count-efficient hybrid quantum-classical detector for GEO satellite Single-Input Single-Output (SISO) BPSK links. The objective is to evaluate whether a compact Quantum Convolutional Neural Network (QCNN) feature-processing block can provide CNN-comparable uncoded bit-detection reliability under a controlled imperfect-CSI software-simulation setup, while keeping the trainable model size small. The channel and SNR normalizations are explicitly stated, and the conventional random Line-of-Sight (LOS) phase in the Rician model is addressed through a phase-invariance argument for the proposed channel-compensated features. Four physics-based diagnostic features are extracted from a channel-compensated observation and encoded into a 4-qubit QCNN. The QCNN contains 2,289 trainable parameters, including 48 quantum-circuit parameters and 2,241 classical post-processing parameters, whereas the same-input CNN baseline contains 12,609 trainable parameters. Simulations from − 10 dB to 20 dB show that the QCNN achieves 89.39% average accuracy, compared with 89.74% for the CNN. As a complementary reliability-oriented metric, a capacity-weighted effective-throughput indicator is also evaluated, where the QCNN achieves an average effective throughput of 2.376 bps/Hz, compared with 2.406 bps/Hz for the CNN. These results suggest that the proposed QCNN can provide a favorable balance between detection reliability, effective-throughput behavior and trainable parameter count. This balance is particularly relevant to isolated GEO satellite receiver environments, where onboard memory, model-update opportunities and processing resources are limited after deployment.