检索结果-内蒙古大学图书馆

A Quantum Spatial Graph Convolutional Neural Network Model on Quantum Circuits

IEEE TRANSACTIONS ON NEURAL NETWORKS AND learning SYSTEMS 2024年第5期PP卷 8025-8035页

作者： Zheng, Jin Gao, Qing Ogorzalek, Maciej Lu, Jinhu Deng, Yue Beihang Univ Sch Automat Sci & Elect Engn Beijing 100191 Peoples R China Zhongguancun Lab Beijing 100094 Peoples R China Beihang Univ Hangzhou Innovat Inst Hangzhou 310000 Peoples R China Jagiellonian Univ Dept Informat Technol PL-30348 Krakow Poland Beihang Univ Sch Automat Sci & Elect Engn Beijing Peoples R China Beihang Univ Sch Astronaut Beijing 102206 Peoples R China

This article proposes a quantum spatial graph convolutional neural network (QSGCN) model that is implementable on quantum circuits, providing a novel avenue to processing non-Euclidean type data based on the state-of-the-art parameterized quantum circuit (PQC) computing platforms. Four basic blocks are constructed to formulate the whole QSGCN model, including the quantum encoding, the quantum graph convolutional layer, the quantum graph pooling layer, and the network optimization. In particular, the trainability of the QSGCN model is analyzed through discussions on the barren plateau phenomenon. Simulation results from various types of graph data are presented to demonstrate the learning, generalization, and robustness capabilities of the proposed quantum neural network (QNN) model.

关键词： Graph convolutional neural networks (GCNs) quantum neural networks (QNNs) quantum spatial graph convolutional neural network (QSGCN)

来源：评论

学校读者我要写书评

暂无评论

Forest Wildfire Detection from Satellite image Using deep learning

Forest Wildfire Detection from Satellite Image Using Deep Le...

引用

2024 International Conference on Communication, Computing and Internet of Things, IC3IoT 2024

作者： Elizaroshan, D. Raj Kumar, J.S. Karunya Institute Of Technology And Sciences Department Of Computer Science And Engineering Coimbatore India

ISBN: (纸本)9798350352689

This study presents an improved methodology of wildfire detection in the forest using PyTorch, a wise machine learning framework, and the Flask application for real-time monitoring. Focusing on accuracy and speed in detection, the methodology mainly refers to data acquisition, and preprocessing of satellite image datasets in its essence, meanwhile featuring the training of robust models through data augmentation. The experimental results using the PyTorch model, which depends on DenseNet201, have proven its ability to determine patterns and reflect highly dynamic contrasts between vertical and horizontal pans;this is especially true when the principal element of the considered satellite image is diffuse and indeterminate. The framework has been integrated into a user-friendly Flask web app, and it gives outputs of immediate wildfire detection upon image upload. The approach is also shown to deliver better metrics thus a more efficient solution as proven by the experimental results. The study informs environmental technology about introducing an easy-to-apply, applicable, and adaptive tool that may have potential associations with environmental agencies for environmentally sustainable approaches to wildfire management. © 2024 IEEE.

关键词： Environmental technology

来源：评论

学校读者我要写书评

暂无评论

deep learning Based High-Resolution Frequency Estimation for Sparse Radar Range Profiles

Deep Learning Based High-Resolution Frequency Estimation for...

引用

IEEE Radar Conference (RadarConf)

作者： Biswas, Sabyasachi Gurbuz, Ali C. Mississippi State Univ Dept Elect & Comp Engn Mississippi State MS 39762 USA

ISBN: (纸本)9798350329216;9798350329209

Frequency estimation is a pivotal process in many signal-processing applications. Generating radar range profiles for linear frequency modulated radar systems is such a case where spectral analysis is used to estimate target ranges. Conventional methods like fast Fourier transform (FFT) are the golden standard in frequency estimation, despite its Rayleigh resolution limit and high sidelobe levels. To address such limitations this paper introduces HRFreqNet;a deep neural network (DNN) architecture for high-resolution frequency estimation from 1D complex time domain data consisting of multiple frequency components. Our deep learning (DL) architecture consists of an auto-encoder block to improve signal-to-noise ratio (SNR), a frequency estimation block to learn frequency transformations to generate pseudo frequency representations(FR), and finally, a 1D-UNET block to reconstruct high-resolution FR. Experimental results on synthetically generated data show enhanced performance in terms of resolution, estimation accuracy, and ability to suppress noise. Achieved range profiles are also sparser with lower sidelobe levels. The proposed HRFreqNet is evaluated over both synthetic and experimental real-world radar data and it is observed that accurate, sparse, high-resolution range profiles are obtained compared to existing approaches.

关键词： Stretch processing deep learning Frequency Estimation Auto-encoder Unet Radar Range profile Sparsity

来源：评论

学校读者我要写书评

暂无评论

4K real time image to image Translation Network With Transformers

引用

IEEE ACCESS 2022年 10卷 73057-73067页

作者： Shibasaki, Kei Fukuzaki, Shota Ikehara, Masaaki Keio Univ Fac Sci & Technol Dept Elect & Comp Engn Yokohama Kanagawa 2238522 Japan

CNNs have traditionally been applied in computer vision. Recently, applying Transformer networks, originally a technique in natural language processing, to computer vision has received much attention and produced superior results. However, Transformers and their derivation have drawbacks that the computational cost and memory usage increase rapidly with the image resolution. In this paper, we propose the Laplacian Pyramid Translation Transformer (LPTT) for image to image translation. The Laplacian Pyramid Translation Network, a previous study of this work, creates Laplacian pyramid of the input images and processes each component with CNNs. However, LPTT transforms the high-frequency components with CNNs and the low-frequency components with Axial Transformer blocks. LPTT can have Transformer's expressive power while reducing the computational cost and memory usage. LPTT significantly improves the quality of generated images and inference speed for high-resolution images over conventional methods. LPTT is the first network with a Transformer that can perform practical inference in real time on 4K resolution images. LPTT can also process 8K images in real time depending on the model conditions and the performance of the GPU. The ablation study in this paper suggests that even when processing high-resolution images, the performance is improved while maintaining the inference speed by computing the low-resolution component with a Transformer. LPTT improves PSNR value by 0.41 dB in MIT-Adobe FiveK dataset. The greater the number of layers in the Laplacian pyramid, the greater the improvement of LPTT over the Laplacian Pyramid Translation Network.

关键词： Transformers Laplace equations image resolution Computational efficiency Transforms Task analysis Tensors deep learning image to image translation Laplacian pyramid photo retouching transformer

来源：评论

学校读者我要写书评

暂无评论

Enhancing Gemstone Classification Accuracy: A deep learning Approach Using CNN, VGG-19, and Xception 1

Enhancing Gemstone Classification Accuracy: A Deep Learning ...

引用

1st International Conference on Innovations in Communications, Electrical and Computer Engineering, ICICEC 2024

作者： Nadimpalli, Mani Salya Kumar Baleboyina, Karthik Madhu Kiran Vundavalli, Yasaswi Durga Tummalapenta, Tej Tarun Kamepalli, Sujatha Vfstr Deemed to be University School of Computing Guntur India

ISBN: (纸本)9798350376517

Gemstones demand has been steadily on the rise in recent years, driven by various factors. Notably, the soaring popularity of colored gemstones in jewelry designs and the pervasive influence of social media platforms significantly contribute to this trend. A key development in the gemstone market is the growing consumer preference for ethically sourced stones. As consumers become more socially aware, they actively seek out gemstones that are mined and processed with a strong commitment to environmental sustainability and ethical practices. Some gemstones may look alike but their properties, price, rarity may vary, so wrong certification or recognition of the gemstone can be a problem for both dealers and customers. To avoid this there are pre built models like CNN to classify images automatically. In this paper traditional CNN, VGG-19, Xception models are implemented and trained on a large gemstone dataset containing 87 classes. The main aim is to increase the accuracy of the models which can become potential source to classify a gemstone in real time. Finally, it was proved that Xception network is best suited for this task with an accuracy about 85.21 percentage. This marks the rise of deep learning models in the gemstones field © 2024 IEEE.

关键词： CNN deep learning Models Gemstones image Augmentation VGG-19 Xception

来源：评论

学校读者我要写书评

暂无评论

deep learning Based Proactive Optimization for Mobile LiFi Systems With Channel Aging

引用

IEEE TRANSACTIONS ON COMMUNICATIONS 2024年第6期72卷 3543-3557页

作者： Arfaoui, Mohamed Amine Ghrayeb, Ali Assi, Chadi Qaraqe, Marwa InterDigital Canada Ltee Montreal PQ H3A 3G4 Canada Texas A&M Univ Qatar Elect & Comp Engn ECE Dept Doha Qatar Concordia Univ Concordia Inst Informat Syst Engn CIISE Montreal PQ H3G 1M8 Canada Hamad Bin Khalifa Univ Qatar Fdn Coll Sci & Engn Doha Qatar

This paper investigates the channel aging problem of mobile light-fidelity (LiFi) systems. In the LiFi physical layer, the majority of the optimization problems for mobile users are non-convex and require the use of dual decomposition or heuristics techniques. Such techniques are based on iterative algorithms, and often cause a high processing delay at the physical layer. Hence, the obtained solutions are rendered sub-optimal since the LiFi channels are evolving. In this paper, a proactive-optimization (PO) approach that can alleviate the LiFi channel aging problem is proposed. The core idea is to design a long-short-term-memory (LSTM) network that is capable of predicting posterior positions and orientations of mobile users, which can be then used to predict their channel coefficients. Consequently, the obtained channel coefficients can be exploited to derive near-optimal transmission-schemes prior to the intended service-time, which enables real-time service. Through various simulations, the performance of the designed LSTM model is evaluated in terms of prediction error and inference complexity, as well as its application in a practical LiFi optimization problem.

关键词： Light fidelity Optimization Three-dimensional displays Estimation Wireless communication Aging Channel estimation LiFi LSTM mobile users prediction proactive optimization random orientation

来源：评论

学校读者我要写书评

暂无评论

MFP-YOLO: a multi-scale feature perception network for CT bone metastasis detection

引用

MEDICAL & BIOLOGICAL ENGINEERING & COMPUTING 2024年第3期63卷 629-640页

作者： Lu, Wenrui Zhang, Wei Liu, Yanyan Xu, Lingyun Fan, Yimeng Meng, Zhaowei Jia, Qiang Tianjin Univ Sch Microelect Tianjin 300072 Peoples R China Nankai Univ Coll Elect Informat & Opt Engn Tianjin 300072 Peoples R China Tianjin Med Univ Gen Hosp Dept Nucl Med Tianjin 300052 Peoples R China

Bone metastasis is one of the most common forms of metastasis in the late stages of malignancy. The early detection of bone metastases can help clinicians develop appropriate treatment plans. CT images are essential for diagnosing and assessing bone metastases in clinical practice. However, early bone metastasis lesions occupy a small part of the image and display variable sizes as the condition progresses, which adds complexity to the detection. To improve diagnostic efficiency, this paper proposes a novel algorithm-MFP-YOLO. Building on the YOLOv5 algorithm, this approach introduces a feature extraction module capable of capturing global information and designs a new content-aware feature pyramid structure to improve the network's capability in processing lesions of varying sizes. Moreover, this paper innovatively applies a transformer-structure decoder to bone metastasis detection. A dataset comprising 3921 CT images was created specifically for this task. The proposed method outperforms the baseline model with a 5.5% increase in precision and a 7.7% boost in recall. The experimental results indicate that this method can meet the needs of bone metastasis detection tasks in real scenarios and provide assistance for medical diagnosis.

关键词： deep learning Bone metastasis detection YOLOv5 Multi-scale Transformer

来源：评论

学校读者我要写书评

暂无评论

Boosted dipper throated optimization algorithm-based Xception neural network for skin cancer diagnosis: An optimal approach

引用

HELIYON 2024年第5期10卷 e26415页

作者： Tang, Xiaofei Sheykhahmad, Fatima Rashid Univ Sci & Technol Liaoning Sch Comp Sci & Software Engn Anshan 114051 Liaoning Peoples R China Islamic Azad Univ Ardabil Branch Ardebil Iran Islamic Univ Coll Tech Engn Najaf Iraq

Skin cancer is a prevalent form of cancer that necessitates prompt and precise detection. However, current diagnostic methods for skin cancer are either invasive, time-consuming, or unreliable. Consequently, there is a demand for an innovative and efficient approach to diagnose skin cancer that utilizes non-invasive and automated techniques. In this study, a unique method has been proposed for diagnosing skin cancer by employing an Xception neural network that has been optimized using Boosted Dipper Throated Optimization (BDTO) algorithm. The Xception neural network is a deep learning model capable of extracting high-level features from skin dermoscopy images, while the BDTO algorithm is a bio-inspired optimization technique that can determine the optimal parameters and weights for the Xception neural network. To enhance the quality and diversity of the images, the ISIC dataset is utilized, a widely accepted benchmark system for skin cancer diagnosis, and various image preprocessing and data augmentation techniques were implemented. By comparing the method with several contemporary approaches, it has been demonstrated that the method outperforms others in detecting skin cancer. The method achieves an average precision of 94.936%, an average accuracy of 94.206%, and an average recall of 97.092% for skin cancer diagnosis, surpassing the performance of alternative methods. Additionally, the 5-fold ROC curve and error curve have been presented for the data validation to showcase the superiority and robustness of the method.

关键词： Skin cancer Melanoma Xception neural network Boosted dipper throated optimization image processing Diagnosis

来源：评论

学校读者我要写书评

暂无评论

Multilevel Multimodal Framework for Automatic Collateral Scoring in Brain Stroke

引用

IEEE ACCESS 2024年 12卷 33730-33748页

作者： Raj, Rishi Pruthviraja, Dayananda Gupta, Ayush Mathew, Jimson Kannath, Santhosh Kumar Prakash, Adity Rajan, Jeny Indian Inst Technol Patna Dept Comp Sci & Engn Patna 801106 India Manipal Acad Higher Educ Manipal Inst Technol Bengaluru Dept Informat Technol Manipal 576104 Karnataka India Sree Chitra Tirunal Inst Med Sci & Technol Dept Imaging Sci & Intervent Radiol Thiruvananthapuram 695011 Kerala India Banaras Hindu Univ Inst Med Sci Dept Radiodiag & Imaging Varanasi 221005 India Natl Inst Technol Karnataka Dept Comp Sci & Engn Surathkal 575025 India

In patients with ischemic brain stroke, collateral circulation plays a crucial role in selecting patients suitable for endovascular therapy. The presence of well-developed collaterals improves the patient's chances of recovery. In clinical practice, the presence of collaterals is diagnosed on a Computed Tomography Angiography scan. The radiologist grades it on the basis of subjective visual assessment, which is prone to interobserver and intraobserver variability. Computer-based methods of collateral assessment face the challenge of non-uniform scan volume, leading to manual selection of slices, meaning that the most imperative slices have to be manually selected by the radiologist. This paper proposes a multilevel multimodal hierarchical framework for automated collateral scoring. Specifically, we propose deploying a Convolutional Neural Network for image selection based on the visibility of collaterals and a multimodal model for comparing the occluded and contralateral sides of the brain for collateral scoring. We also generate a patient-level prediction by integrating automated machine learning in the proposed framework. While the proposed multimodal predictor contributes to Artificial Intelligence, the proposed end-to-end framework is an application in engineering. The proposed framework has been trained and tested on 116 patients, with five-fold cross-validation, achieving an accuracy of 91.17% for multi-class collateral scores and 94.118% for binary class collateral scores. The proposed multimodal predictor achieved a weighted F1 score of 0.86 and 0.95 on multi-class and binary-class collateral scores, respectively. The proposed framework is fast, efficient, and scalable for real-world deployments. Automated evaluation of collaterals with attention maps for explainability would complement radiologists' efforts. Code for the proposed framework is available

关键词： Biomedical imaging Stroke (medical condition) deep learning Feature extraction Task analysis Numerical models Manuals Medical expert systems Circulatory system Patient monitoring Surgery Convolutional neural networks Medical treatment biomedical image processing medical expert systems stroke (medical condition)

来源：评论

学校读者我要写书评

暂无评论

Evaluating the Energy Efficiency of Few-Shot learning for Object Detection in Industrial Settings 3

Evaluating the Energy Efficiency of Few-Shot Learning for Ob...

引用

IEEE 3rd real-time and Intelligent Edge Computing Workshop (RAGE)

作者： Tsoumplekas, Georgios Li, Vladislav Siniosoglou, Ilias Argyriou, Vasileios Goudos, Sotirios K. Moscholios, Ioannis D. Radoglou-Grammatikis, Panagiotis Sarigiannidis, Panagiotis MetaMind Innovat PC R&D Dept Kozani Greece Kingston Univ Dept Networks & Digital Media Kingston Upon Thames Surrey England Univ Western Macedonia Dept Elect & Comp Engn Kozani Greece Aristotle Univ Thessaloniki Phys Dept Thessaloniki Greece Univ Peloponnese Dept Informat & Telecommun Tripoli Greece K3Y Ltd Dept Res & Dev Sofia 1000 Bulgaria

ISBN: (纸本)9798350363357;9798350363340

In the ever-evolving era of Artificial Intelligence (AI), model performance has constituted a key metric driving innovation, leading to an exponential growth in model size and complexity. However, sustainability and energy efficiency have been critical requirements during deployment in contemporary industrial settings, necessitating the use of data-efficient approaches such as few-shot learning. In this paper, to alleviate the burden of lengthy model training and minimize energy consumption, a finetuning approach to adapt standard object detection models to downstream tasks is examined. Subsequently, a thorough case study and evaluation of the energy demands of the developed models, applied in object detection benchmark datasets from volatile industrial environments, is presented. Specifically, different finetuning strategies, as well as utilization of ancillary evaluation data during training, are examined, and the trade-off between performance and efficiency is highlighted in this low-data regime. Finally, this paper introduces a novel way to quantify this trade-off through a customized Efficiency Factor metric.

关键词： Few-Shot learning Green AI deep learning Model Optimization Object Detection Industrial image Data

来源：评论

学校读者我要写书评

暂无评论

建议与咨询留下您的常用邮箱和电话号码，以便我们向您反馈解决方案和替代方法

时间限定

文献类型

馆藏选择

核心期刊

语言

文献类型

帮助

文字说明：

检索规则说明：

检索范例：

分类表

所选分类

限定检索结果

文献类型

馆藏范围

日期分布

学科分类号

主题

机构

作者

语言

请选择保存的检索档案：

请选择收藏分类：

通借通还

建议与咨询 留下您的常用邮箱和电话号码，以便我们向您反馈解决方案和替代方法

时间限定

文献类型

馆藏选择

核心期刊

语言

文献类型

帮助

文字说明：

检索规则说明：

检索范例：

分类表

所选分类

限定检索结果

文献类型

馆藏范围

日期分布

学科分类号

主题

机构

作者

语言

请选择保存的检索档案： 新增检索档案 确定 取消

请选择收藏分类： 新增自定义分类 确定 取消

通借通还

建议与咨询留下您的常用邮箱和电话号码，以便我们向您反馈解决方案和替代方法

请选择保存的检索档案：

请选择收藏分类：