检索结果-内蒙古大学图书馆

International Conference on Control, Automation and systems ( ICCAS)

作者： Seungwoo Song Gilsu Yeo Hyun-Woo Kim Myungjin Cho Min-Chul Lee Graduate School of Computer Science and Systems Engineering Kyushu Institute of Technology Iizuka City Fukuoka Japan School of ICT Robotics and Mechanical Engineering IITC Hankyoung National University Anseong Republic of Koera

ISBN: (数字)9788993215380

ISBN: (纸本)9798331517939

In this paper, we propose improved three-dimensional (3D) photon counting imaging using singular value decomposition (SVD). In conventional 3D photon counting imaging can visualize 3D image under photon-starved conditions. However, the conventional method has problems that it may not visualize 3D image under extremely photon-starved conditions and the quality of visualized image may be degraded. To solve this problem, we apply SVD to image before applying the photon counting integral imaging in this paper. To verify feasibility of our method, we implement experiments and show the experimental results. In addition, for numerical comparison, the image quality assessment methods such as structural similarity index map (SSIM) and blind/referenceless image spatial quality evaluator (BRISQUE) are calculated.

关键词： Image quality Visualization Three-dimensional displays Image resolution Noise Indexes Image reconstruction Photonics Singular value decomposition Meteorology

来源：评论

学校读者我要写书评

暂无评论

ITSVA: Toward 6G-Enabled Vision Analytics over Integrated Terrestrial-Satellite Network

ITSVA: Toward 6G-Enabled Vision Analytics over Integrated Te...

引用

Satellite Computing (Satellite), IEEE International Conference on

作者： Miao Zhang Jiaxing Li Jianxin Shi Yifei Zhu Lei Zhang Hengzhi Wang School of Computing Science Simon Fraser University Canada College of Computer Science Nankai University China UM-SJTU Joint Institute Shanghai Jiao Tong University China College of Computer Science and Software Engineering Shenzhen University China

The sixth-generation (6G) mobile communications system is expected to integrate the terrestrial and low Earth orbit satellite networks (LSN) to provide seamless global Internet service. This will create new opportunities for edge-assisted mobile vision analytics (MVA), which offloads frames over networks to edge servers for analysis, thereby overcoming the local computational resource constraints. With the integrated terrestrial and LSN (ITLSN), edge-assisted MVA can reach its full potential in remote and maritime areas. Nevertheless, the proximity of LEO satellites to the Earth is a double-edged sword. It offers benefits in latency and data rates but also brings challenges like frequent satellite handovers and volatile channel conditions. To demystify the in-the-wild performance of ITLSN, we carry out large-scale measurements with a major LSN service provider. The measurement results confirm the highly asymmetric and dynamic network performance of today's ITLSN, which can present non-trivial challenges for MVA frame offloading. We thus propose an ITLSN -adaptive MVA offloading framework, IT SVA, to address the inherent dynamics brought by network conditions, video content, and the offloading strategy. Extensive trace-driven simulation experiments are further conducted to verify the effectiveness of ITSVA.

关键词：

来源：评论

学校读者我要写书评

暂无评论

Blockchain Anchored Federated Learning and Tokenized Traceability for Sustainable Food Supply Chains 4

Blockchain Anchored Federated Learning and Tokenized Traceab...

引用

4th International Conference on Ubiquitous Computing and Intelligent Information systems, ICUIS 2024

作者： Chaganti, Koushik Reddy Kumar, B. Naveen Gutta, Pruthvi Krishna Elicherla, Sivananda Lahari Reddy Nagesh, C. Raghavendar, K. University of the Cumberlands Dept. of Computer and Information Systems KY United States Vignan Institute of Technology and Science Department of Information Technology Hyderabad India Tesla Dept of It AustinTX United States Dayananda Sagar University Dept of Cse Bangalore India Srinivasa Ramanujan Institute of Technology Dept of Cse Anantapur India Teegala Krishna Reddy Engineering College Dept of Cse Hyderabad India

ISBN: (纸本)9798331529635

The management of food supply chains relates to just about the most imperative issues concerning sustainability, quality, and traceability, where existing methodologies usually fail to meet the demands caused by data privacy, lack of transparency, and inefficiencies concerning collaboration among stakeholders. This compromises data security, and thus there arises single points of failure. Conversely, traditional traceability frameworks fail to incite wide-scale participation and lack real-time accuracy. Having these limitations in mind, we introduce a hybrid framework consisting of three advanced methodologies: Decentralized Federated Learning with Blockchain Anchoring (DFLBA), Dynamic Tokenized Traceability Framework (DTTF), and Adaptive Supply Chain Graph Neural Networks (ASCGNN) process. DFLBA would support privacy-preserving machine learning as such that collaborative training of yield prediction and crop monitoring models by stakeholders without revealing raw data will be feasible. Blockchain anchoring ensures logs of training updates are tamper-proof, thus increasing trust and transparency. DTTF introduces tokenized incentives that enhance traceability in real-time: every product has a digital twin represented by blockchain-anchored tokens updated by stakeholders. Smart contracts enforce the compliance that will improve participation and lifecycle data accuracy. ASCGNN models the food supply chain as a dynamic graph, leveraging graph neural networks to capture stakeholder interdependencies, optimize operations, and detect fraud. Blockchain integration ensures reliability and transparency by securing edge weights. The suggested framework has demonstrated good impact, like 95% model accuracy for yield prediction, 99% traceability coverage, and 92% prediction accuracy for bottlenecks of the supply chain. It provides a scalable, privacy-preserving, and transparent solution that enhances the efficiency of food supply chains, and positively impacts stakeholders in a

关键词： Federated learning

来源：评论

学校读者我要写书评

暂无评论

Guardian - Smart Assistant Tool for Visually Impaired People

Guardian - Smart Assistant Tool for Visually Impaired People

引用

Advancements in Computing (ICAC), International Conference on

作者： C.K. Amarasinghe R.D.S.P. Pinto K.N. Sudusinghe Department of Computer Science and Software Engineering Sri Lanka Institute of Information Technology Malabe Sri Lanka

ISBN: (纸本)9798350398106

At present, with the advancement of technology, various devices and solutions have been found to aid the visually impaired community (VI Community). Even with the countless technological breakthroughs, yet they face many problems performing the most basic functions in daily life. Identifying the objects, they use daily, identifying a person, whether it’s someone they know or not, and their emotions, and reading a text information displayed anywhere, without the assistance from another person are the basic issues we deal with and try to resolve using a tool consisting of a pair of spectacles with an inbuilt camera that is integrated with a mobile application. The inbuilt camera will capture the image of a text containing label, an object, or a person which will be then detected, analyzed, and recognized and will be converted to Speech using Google TTS engine and produced through the headphones giving the output to the user. Tesseract OCR, YOLO algorithm, and TensorFlow models have been used for each feature of the tool respectively. This tool will be very beneficial to a blind person as it mitigates them the frustration caused by being incapable of performing daily activities without any assistance from another person.

关键词： Performance evaluation Text recognition Computational modeling Optical character recognition Sociology Speech recognition Cameras

来源：评论

学校读者我要写书评

暂无评论

Attackers Can Do Better: Over- and Understated Factors of Model Stealing Attacks

Attackers Can Do Better: Over- and Understated Factors of Mo...

引用

Secure and Trustworthy Machine Learning (SaTML), IEEE Conference on

作者： Daryna Oliynyk Rudolf Mayer Andreas Rauber CDL AsTra Faculty of Computer Science University of Vienna Vienna Austria SBA Research Vienna Austria Faculty of Informatics TU Wien Institute of Information Systems Engineering Vienna Austria

ISBN: (数字)9798331517113

ISBN: (纸本)9798331517120

Machine learning (ML) models were shown to be vulnerable to model stealing attacks, which lead to intellectual property infringement. Among other attack methods, substitute model training is an all-encompassing attack applicable to any machine learning model whose behaviour can be approximated from input-output queries. Whereas previous works mainly focused on improving the performance of substitute models by, e.g. developing a new substitute training method, there have been only limited ablation studies that try to understand the impact the strength of an attacker has on the substitute model's performance. As a result, different authors came to diverse, sometimes contradicting, conclusions. In this work, we exhaustively examine the ambivalent influence of different factors resulting from varying the attacker's capabilities and knowledge on a substitute training attack. Our findings suggest that some of the factors that have been considered important in the past are, in fact, not that influential; instead, we discover new correlations between attack conditions and success rate. In particular, we demonstrate that better-performing target models enable higher-fidelity attacks and explain the intuition behind this phenomenon. Further, we propose to shift the focus from the complexity of target models toward the complexity of their learning tasks. Therefore, for the substitute model, rather than aiming for a higher architecture complexity, we suggest focusing on getting data of higher complexity and an appropriate architecture. Finally, we demonstrate that even in the most limited data-free scenario, there is no need to overcompensate weak knowledge with unrealistic capabilities in the form of millions of queries. Our results often exceed or match the performance of previous attacks that assume a stronger attacker, suggesting that these stronger attacks are likely endangering a model owner's intellectual property to a significantly higher degree than shown until now.

关键词： Training Correlation Query processing Focusing Intellectual property Data collection Data models Adversarial machine learning Complexity theory Tuning

来源：评论

学校读者我要写书评

暂无评论

科学研究的第五范式——以智能驱动的材料设计为例

引用

engineering 2023年第5期24卷 126-137,I0003,I0004页

作者： Can Leng Zhuo Tang Yi-Ge Zhou Zean Tian Wei-Qing Huang Jie Liu Keqin Li Kenli Li Science and Technology on Parallel and Distributed Processing Laboratory National University of Defense TechnologyChangsha 410073China Laboratory of Software Engineering for Complex Systems National University of Defense TechnologyChangsha 410073China National Supercomputing Center in Changsha Changsha 410082China College of Computer Science and Electronic Engineering Hunan UniversityChangsha 410082China Institute of Chemical Biology and Nanomedicine State Key Laboratory of Chemo/Biosensing and ChemometricsCollege of Chemistry and Chemical EngineeringHunan UniversityChangsha 410082China Department of Applied Physics School of Physics and ElectronicsHunan UniversityChangsha 410082China Department of Computer Science State University of New YorkNew PaltzNY 12561USA

科学正在进入一个新时代——第五范式——它被认为是知识整合到不同领域的主要特征,是基于无所不在的机器学习系统的计算社区中智能驱动的工作。在此,我们通过在天河一号超级计算机系统上构建的催化材料专门设计的典型平台案例,生动地... 详细信息

科学正在进入一个新时代——第五范式——它被认为是知识整合到不同领域的主要特征,是基于无所不在的机器学习系统的计算社区中智能驱动的工作。在此,我们通过在天河一号超级计算机系统上构建的催化材料专门设计的典型平台案例,生动地阐明了第五范式的本质,旨在促进第五范式在其他领域的培养。第五范式平台主要包括模型自动构建(原始数据提取)、指纹自动构建(神经网络特征选择)以及跨学科知识串联的重复迭代(“火山图”)。与分解一起进行的是对迭代中实现的体系结构的性能评估。通过讨论,第五范式的智能驱动平台可以极大地简化和改进研究中极其繁琐和具有挑战性的工作,并通过补偿机器学习中缺少样本和替代一些由于计算资源不足而导致的数值计算来实现数值计算与机器学习之间的相互反馈,从而加速探索过程。在数据驱动的学科中,跨学科专家的协同作用和对动态数据需求的急剧增长仍然是一个挑战。我们相信,对第五范式平台的一瞥可以为其在其他领域的应用铺平道路。

关键词：机器学习自动构建天河一号数据驱动神经网络动态数据知识整合科学研究

来源：评论

学校读者我要写书评

暂无评论

A Survey of Medical Vision-and-Language Applications and Their Techniques

arXiv

引用

arXiv 2024年

作者： Chen, Qi Zhao, Ruoshan Wang, Sinuo Phan, Vu Minh Hieu van den Hengel, Anton Verjans, Johan Liao, Zhibin To, Minh-Son Xia, Yong Chen, Jian Xie, Yutong Wu, Qi Australian Institute for Machine Learning University of Adelaide Australia School of Software Engineering South China University of Technology China South Australia Medical Imaging Australia School of Computer Science & Engineering Northwestern Polytechnical University China

Medical vision-and-language models (MVLMs) have attracted substantial interest due to their capability to offer a natural language interface for interpreting complex medical data. Their applications are versatile and have the potential to improve diagnostic accuracy and decision-making for individual patients while also contributing to enhanced public health monitoring, disease surveillance, and policy-making through more efficient analysis of large data sets. MVLMS integrate natural language processing with medical images to enable a more comprehensive and contextual understanding of medical images alongside their corresponding textual information. Unlike general vision-and-language models trained on diverse, non-specialized datasets, MVLMs are purpose-built for the medical domain, automatically extracting and interpreting critical information from medical images and textual reports to support clinical decision-making. Popular clinical applications of MVLMs include automated medical report generation, medical visual question answering, medical multimodal segmentation, diagnosis and prognosis and medical image-text retrieval. Here, we provide a comprehensive overview of MVLMs and the various medical tasks to which they have been applied. We conduct a detailed analysis of various vision-and-language model architectures, focusing on their distinct strategies for cross-modal integration/exploitation of medical visual and textual features. We also examine the datasets used for these tasks and compare the performance of different models based on standardized evaluation metrics. Furthermore, we highlight potential challenges and summarize future research trends and directions. The full collection of papers and codes is available at: https://***/MVLMs. © 2024, CC BY.

关键词： Clinical research

来源：评论

学校读者我要写书评

暂无评论

Health-Care Monitoring of Patient using CNN based Model in Internet of Things

Health-Care Monitoring of Patient using CNN based Model in I...

引用

Applied Intelligence and Sustainable Computing (ICAISC), International Conference on

作者： B D Parameshachari K.L Hemalatha Javad Rahebi Mohammed Rashad Baker B.P Upendra Roy Department of Electronics and Communications Engineering Nitte Meenakshi Institute of Technology Bengaluru India Department of Information Science and Engineering Sri Krishna Institute of Technology Bangalore India Department of Software engineering istanbul topkapi university Department of Software College of Computer Science and Information Technology University of Kirkuk Kirkuk Iraq Department of Electronics and Communication Engineering Channabasaveshwara Institute of Technology Gubbi Tumkur India

There is rising public concern about exposure to radiofrequency (RF) electromagnetic fields (EMF) as more and more wireless communications become concentrated in everyday living surroundings. There are a number of obstacles, primarily originating from infrastructure expenses, but recent technology breakthroughs are shifting attention to Internet of Things (IoT) devices that enable automatic and continuous real-time EMF monitoring. The Internet of Things (IoT) has made the world a better place by expanding the capabilities of telemedicine and allowing for more precise remote monitoring of patients. Improved healthcare facilities for both patients and doctors/hospitals are only one way that IoT is making a significant impact in the medical field. Five main parts make up the proposed system: patient data gathering, primary report creation, hospital care, pharmacy care, and diagnostics. This study presents an Internet of Things (IoT)-based healthcare scheme accomplished of real-time nursing of energetic signs and environmental factors for discrete patients. Five sensors, including a heart-rate monitor, a body temperature monitor, a room temperature monitor, a carbon monoxide monitor, and a carbon dioxide monitor, are utilised to collect data from the hospital setting. Medical personnel receive patient updates through a portal and use this data for further diagnosis and care. The success of the system demonstrates that the built prototype is ideal for healthcare monitoring. This study has investigated unique approaches to the automatic construction of convolutional neural network (CNN) topologies, and it has also developed a new method for diagnosis. The results of the trials prove that the recommended algorithm achieves improved than the state-of-the-art techniques.

关键词：

来源：评论

学校读者我要写书评

暂无评论

Single Image 3D Reconstruction Based on Attention Mechanism and Graph Convolution Network 22

Single Image 3D Reconstruction Based on Attention Mechanism ...

引用

8th International Conference on Computing and Artificial Intelligence, ICCAI 2022

作者： Gao, Wei Yu, Liyang Du, Yuanyuan Lu, Songfeng Software Development Center of Industrial and Commercial Bank of China Limited Zhuhai519000 China School of Computer Science and Technology Huazhong University of Science and Technology Wuhan430074 China School of Cyber Science and Engineering Huazhong University of Science and Technology Wuhan430072 China Shenzhen Huazhong University of Science and Technology Research Institute Shenzhen518057 China

ISBN: (纸本)9781450396110

This paper innovatively proposes a channel attention mechanism and graph convolutional network model adapted to 3D reconstruction, and combines the target detection model to construct a neural network that generates a target 3D model from a single RGB image. The model first generates the target 3D Voxels, and further generates a more refined 3D Meshes model through the graph convolutional network. Compared with the control group algorithm, the Pix3D datasetAPmesh index has been improved by 2.8%, which fully proves the effectiveness of the model in single-image 3D reconstruction. The experimental results show that the algorithm has good usability in the 3D reconstruction of actual mesh points. © 2022 Association for Computing Machinery. All rights reserved.

关键词： Convolution

来源：评论

学校读者我要写书评

暂无评论

Attackers Can Do Better: Over- and Understated Factors of Model Stealing Attacks

arXiv

引用

arXiv 2025年

作者： Oliynyk, Daryna Mayer, Rudolf Rauber, Andreas CDL AsTra Faculty of Computer Science University of Vienna Vienna Austria SBA Research Vienna Austria Institute of Information Systems Engineering Faculty of Informatics TU Wien Vienna Austria

Machine learning (ML) models were shown to be vulnerable to model stealing attacks, which lead to intellectual property infringement. Among other attack methods, substitute model training is an all-encompassing attack applicable to any machine learning model whose behaviour can be approximated from input-output queries. Whereas previous works mainly focused on improving the performance of substitute models by, e.g. developing a new substitute training method, there have been only limited ablation studies that try to understand the impact the strength of an attacker has on the substitute model’s performance. As a result, different authors came to diverse, sometimes contradicting, conclusions. In this work, we exhaustively examine the ambivalent influence of different factors resulting from varying the attacker’s capabilities and knowledge on a substitute training attack. Our findings suggest that some of the factors that have been considered important in the past are, in fact, not that influential;instead, we discover new correlations between attack conditions and success rate. In particular, we demonstrate that better-performing target models enable higher-fidelity attacks and explain the intuition behind this phenomenon. Further, we propose to shift the focus from the complexity of target models toward the complexity of their learning tasks. Therefore, for the substitute model, rather than aiming for a higher architecture complexity, we suggest focusing on getting data of higher complexity and an appropriate architecture. Finally, we demonstrate that even in the most limited data-free scenario, there is no need to overcompensate weak knowledge with unrealistic capabilities in the form of millions of queries. Our results often exceed or match the performance of previous attacks that assume a stronger attacker, suggesting that these stronger attacks are likely endangering a model owner’s intellectual property to a significantly higher degree than shown until now. Cop

关键词： Adversarial machine learning

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：