检索结果-内蒙古大学图书馆

2025 IEEE/CVF Winter Conference on Applications of computer Vision, WACV 2025

作者： Li, Mingxian Sun, Hao Lei, Yingtie Zhang, Xiaofeng Dong, Yihang Zhou, Yilin Li, Zimeng Chen, Xuhang School of Computer Science and Engineering Huizhou University China The Ohio State University Department of Computer Science and Engineering United States School of Electronic Information and Electrical Engineering Shanghai Jiao Tong University China Shenzhen Institute of Advanced Technology Chinese Academy of Sciences China School of Electronic and Communication Engineering Shenzhen Polytechnic University China

ISBN: (纸本)9798331510831

Document images are often degraded by various stains, significantly impacting their readability and hindering downstream applications such as document digitization and analysis. The absence of a comprehensive stained document dataset has limited the effectiveness of existing document enhancement methods in removing stains while preserving fine-grained details. To address this challenge, we construct StainDoc, the first large-scale, high-resolution (2145 x 2245) dataset specifically designed for document stain removal. StainDoc comprises over 5,000 pairs of stained and clean document images across multiple scenes. This dataset encompasses a diverse range of stain types, severities, and document backgrounds, facilitating robust training and evaluation of document stain removal algorithms. Furthermore, we propose StainRestorer, a Transformer-based document stain removal approach. StainRestorer employs a memory-augmented Transformer architecture that captures hierarchical stain representations at part, instance, and semantic levels via the DocMemory module. The Stain Removal Transformer (SRTransformer) leverages these feature representations through a dual attention mechanism: an enhanced spatial attention with an expanded receptive field, and a channel attention captures channel-wise feature importance. This combination enables precise stain removal while preserving document content integrity. Extensive experiments demonstrate StainRestorer's superior performance over state-of-the-art methods on the Stain-Doc dataset and its variants *** and ***, establishing a new benchmark for document stain removal. Our work highlights the potential of memory-augmented Transformers for this task and contributes a valuable dataset to advance future research. © 2025 IEEE.

关键词：

来源：评论

学校读者我要写书评

暂无评论

LiFi-CFI: Light-weight Fine-grained Hardware CFI Protection for RISC-V 1

LiFi-CFI: Light-weight Fine-grained Hardware CFI Protection ...

引用

1st IEEE International Conference on Design, Test and Technology of Integrated Systems, DTTIS 2023

作者： Roodsari, Mahboobe Sadeghipour Sheikhshoaei, Fatemeh Maunero, Nicolo Prinetto, Paolo Navabi, Zain University of Tehran School of Electrical and Computer Engineering Tehran Iran Imt School for Advanced Studies Lucca Italy Politecnico di Torino Department of Automation and Information Turin Italy

ISBN: (纸本)9798350316957

In today's open-source industry, RISC-V based embedded devices are used in various applications where security is critical. RISC-V processors are vulnerable to many runtime attacks, including powerful code reuse and code injection in jump/return-oriented programming (JOP/ROP), known as control flow attacks (CFA). For this, several security techniques have been proposed to protect embedded devices from CFA, and one of the most secure categories is Control Flow Integrity (CFI). Alternative techniques consider the tradeoff between runtime hardware overhead and security protection. In particular, hardware-based techniques have attracted more attention in recent years because they consume less power and have less runtime overhead, which is critical in embedded systems. Considering these constraints, we have proposed a novel fine-grained technique for jump-oriented programming, where the required memory is replaced with a short Programmable Array Logic (PAL) to reduce runtime and hardware overhead. Also, a low-overhead shadow stack is proposed for return-oriented programming to reduce the required memory capacity compared to the traditional shadow stack implementation by handling the recursive function calls. The proposed Architecture called LiFi-CFI, is a lightweight hardware monitoring solution that can be attached to any IoT-class soft processor. LiFi-CFI not only reduces power consumption and hardware overhead compared to other advanced solutions but also maintains the security guarantees of the main CFI solution. It is a compact solution as a countermeasure with less than 1% hardware resource utilization for Xilinx Artix7 and 0.5% runtime overhead. © 2023 IEEE.

关键词： Embedded systems

来源：评论

学校读者我要写书评

暂无评论

Privacy-Preserving Federated Learning through Clustered Sampling on Fine-Tuning Distributed non-iid Large Language Models 21

Privacy-Preserving Federated Learning through Clustered Samp...

引用

21st IEEE International Symposium on Parallel and Distributed Processing with Applications, 13th IEEE International Conference on Big Data and Cloud Computing, 16th IEEE International Conference on Social Computing and Networking and 13th International Conference on Sustainable Computing and Communications, ISPA/BDCloud/SocialCom/SustainCom 2023

作者： Yun, Sheng Bhuiyan, Zakirul Alam Sadi, Md. Taufiq Al Hasib Su, Shen Fordham University Department of Computer and Information Sciences New York10458 United States North South University Department of Electrical & Computer Engineering Dhaka Bangladesh Guangzhou University Cyberspace Institute of Advanced Technology Guangzhou China

ISBN: (纸本)9798350329223

Recently, Large Language Models (LLMs) have been a phenomenal trend in the Artificial intelligence field. However, training and fine-tuning can be challenging because of privacy concerns and limited computing resources. Federated Learning (FL) has emerged as a novel machine learning framework offering privacy protection. The challenges in applying FL to real-world applications include dealing with heterogeneous data, poor client updates, and client selection. This paper introduces Privacy-preserving Federated Learning through Clustered Sampling on LLMs (FCLM), a framework that clusters models by their distribution similarity. It helps the model group similar models to improve text data heterogeneity handling and privacy concerns in distributed machine-learning environments. The FCLM framework is implemented and evaluated using popular Language models and text data. The framework shows a robust performance over the heterogeneous text data, which can further extend to the use of more complex LLMs. © 2023 IEEE.

关键词： Privacy-preserving techniques

来源：评论

学校读者我要写书评

暂无评论

On-site Testing and Preventive Maintenance for a Fire Automation System Employing PSO, FA and ANN Algorithms 1

On-site Testing and Preventive Maintenance for a Fire Automa...

引用

1st IEEE International Conference on Emerging Research in Computational Science, ICERCS 2023

作者： Rubini, B. Kalaiarasi, G. Vasuki, S. Pradeep Kumar, S. Murugan, Suganiya Arul Stephen, C. Vels Institute of Science Technology and Advanced Studies Department of Electrical and Electronics Engineering Pallavaram Chennai India Vignan Foundation for Science Technology and Research Deemed to Be University Department of Advanced Computer Science and Engineering Andhrapradesh Guntur India M. Kumarasamy College of Engineering Department of Electronics and Communication Engineering Tamilnadu Karur India Srm Institute of Science and Technology Department of Computing Technologies Kaattankulathur Chennai India Vels Institute of Science Technology and Advanced Studies Department of Electronics and Communication Engineering Pallavaram Chennai India

ISBN: (纸本)9798350359756

engineering elements influencing in commercial and government buildings fire safety rendition buildings have been deliberated. For both new and existing buildings, all variables such as structure of building, damage due to fire from prevention compartments, Time of incidents evacuation routes should be control leverage on people's evacuation routes, controlled, and additional standard fire protection equipment are imposed. This page outlines the classes of diesel fire pumps, batteries, and pumping systems, besides the elements prompting fire rendition. Onsite diesel generator pump must be connected to a hydrant and the water pumping pressure range must be customized to 8 bar using a pressure gauge. The substantive intent of this concert is to reinforce firefighter safety. Only experienced technicians and training professionals in fire guard are responsible for this part. To simulate the battery tightness position, an extra-critical condition has been considered. Owing to the position of the fire safety properties and battery reaction in buildings, the Firefly and Particle Swarm optimization algorithms have been exerted on the iterations with objective values used for analysis in the iterative process using Matlab 2021 software. Two well-known algorithms are compared to see which provides the best results. The ANN serves as a tool for analyzing and optimizing the pump's rendition in maintaining an adequate water supply for fire emergencies. © 2023 IEEE.

关键词： Fires

来源：评论

学校读者我要写书评

暂无评论

Catalyzing EEG signal analysis: unveiling the potential of machine learning-enabled smart K nearest neighbor outlier detection

引用

International Journal of Information Technology (Singapore) 2024年 1-12页

作者： Aymen, Abid Khediri, Salim El Thaljaoui, Adel Miladi, Moahmed Kachouri, Abdennaceur Computer & amp Embedded Systems Lab National Engineering School of Sfax University of Sfax Sfax Tunisia Department of Information Technology College of Computer Qassim University Buraydah Saudi Arabia Department of Computer Science Faculty of Science of Gafsa University of Gafsa Gafsa Tunisia Preparatory Institute for Engineering Studies of Gafsa Gafsa Tunisia Department of Computer Science and Information College of Science at Zulfi Majmaah University Al-Majmaah 11952 Saudi Arabia Higher Institute of Computer Science and Multimedia of Sfax University of Sfax Sfax Tunisia Advanced Fluid Dynamics Energetics and Environment Lab National Engineering School of Sfax University of Sfax Sfax Tunisia

Electroencephalogram (EEG) data are susceptible to artifacts, such as lapses in concentration or poor imagination, which can significantly impact the accuracy of disease diagnosis in e-health applications. To mitigate this issue, the use of machine learning (ML) and potentially artificial intelligence (AI) solutions to accurately identify outliers becomes crucial. Unlike many AI methods that incorporate unnecessary or redundant input variables, our study focuses on detecting anomalous values in EEG data through the K nearest neighbor (KNN) process and Euclidean distance metric. Our proposed unsupervised non-parametric algorithm, known as the smart KNN outlier detector (SKOD), eliminates the need for initial parameter configurations such as the number of neighbors (K), while achieving high performance. Evaluation of SKOD using real EEG data from 140 trials demonstrated sensitivity and specificity exceeding 60%, with nearly perfect accuracy in detecting outliers reaching close to 100%. © Bharati Vidyapeeth's institute of computer Applications and Management 2024.

关键词： Data analysis EEG signal Healthcare Knowledge-based systems Machine learning Outlier detection

来源：评论

学校读者我要写书评

暂无评论

On Security-Constrained Bidding of Shared Transportation Hubs for Renewable Fuels and Refined Oil Participating Within Energy, Regulation and Pipeline Logistics Markets 7

On Security-Constrained Bidding of Shared Transportation Hub...

引用

7th IEEE Student Conference on Electric Machines and Systems, SCEMS 2024

作者： Wang, Shengshi Li, Hongyi Huang, Danji Li, Kun Cui, Shichang Ai, Xiaomeng Fang, Jiakun Wen, Jinyu School of Electrical and Electronic Engineering Huazhong University of Science and Technology State Key Laboratory of Advanced Electromagnetic Technology Wuhan China School of Engineering Cardiff University Cardiff United Kingdom Iowa State University Department of Electrical and Computer Engineering Ames United States University of Macau Department of Electrical and Computer Engineering China PetroChina ShenZhen New Energy Research Institute Co. Ltd. Shenzhen China

ISBN: (纸本)9798350366945

This paper investigates the security-constrained bidding strategy of the emerging shared transportation hubs for renewable fuels and refined oil (STH-RRs) participating within energy, regulation and pipeline logistics markets. Firstly, to excavate more flexible capability for the power system as well as to increase the revenue of the STH-RR operation company, the model for STH-RR participating in regulation market is established. Secondly, the security-constrained two-stage bidding model is first proposed for STH-RRs to balance the revenue across energy, regulation and pipeline logistics markets. Particularly, to guarantee secure operation of STH-RRs and meet the delivery demand, the rigorous multi-product transportation constraints are considered in both stages. Case studies based on the modified real-world multi-product pipelines in South China validate the proposed method, with the revenue and turnover increase of 14% and 4.7%, respectively, compared with conventional methods. © 2024 IEEE.

关键词： Pipelines

来源：评论

学校读者我要写书评

暂无评论

Deep Learning PNN Based Fault Monitoring System for Three Phase Industrial Drive System 7

Deep Learning PNN Based Fault Monitoring System for Three Ph...

引用

7th International Conference on Circuit Power and Computing Technologies, ICCPCT 2024

作者： Chempavathy, B. Raju, K. David Mani, P.K. Vijayalakshmi, S. Janaki, N. Karthikeyan, D. New Horizon College of Engineering Department of Computer Science and Engineering Bangalore India Malla Reddy Engineering College for Women Department of Electrical and Electronics Engineering Hyderabad India Vel Tech Multi Tech Dr. Rangarajan Dr. Sakunthala Engineering College Department of Electrical and Electronics Engineering Chennai India K. Ramakrishnan College of Technology Department of Electrical and Electronics Engineering Tiruchirappalli Tamil Nadu Samayapuram India Technology and Advanced Studies Vels Institute of Science Department of Electrical and Electronics Engineering Tamilnadu Chennai India SRM Institute of Science and Technology Department of Electrical and Electronics Engineering Chennai India

ISBN: (纸本)9798350372816

Asynchronous motors are widely used in a variety of sectors. Even though induction motors are strong and dependable, they are susceptible to a variety of problems. Faults in induction motors can result in horrible events such as operator injuries, manufacturing disruption, and raw material loss. As a result, defect identification has become more crucial in induction motor management. Among the different problems that can occur in the motor, bearing failure is a severe issue that can cause catastrophic damage to the machine if not detected at an early stage of the fault. As consequently, the functioning of bushings in induction machinery must be checked on a regular basis. This research proposes a novel method for analyzing speed using discrete wavelet transform (DWT) and identifying bearing difficulties using Probabilistic Neural Network (PNN) optimized by improved salp swarm algorithm. The stator electricity is examined and classified when an induction motor is operating under varied loading conditions with good and troublesome bearings. Scale-Invariant Feature Transform (SIFT) is a feature extraction method that condenses visual content into a set of points that may be used to find similar patterns in other images. The proposed PNN classifier can classify different types of bearing faults, and the empirical findings validate the created approach. PNN-based motor bearing recognizing defects and diagnosis outperforms SVM and ANN classifiers in this regard. The Shortest path accuracy of the PNN Classifier is 98.5% respectively. © 2024 IEEE.

关键词： Discrete wavelet transforms

来源：评论

学校读者我要写书评

暂无评论

Recognition of Deep Fake Voice Acoustic using Ensemble Bagging Model

Recognition of Deep Fake Voice Acoustic using Ensemble Baggi...

引用

Electronics and Sustainable Communication Systems (ICESC), 2020 International Conference on

作者： Shaik Abdul Yunus Basha K.Ulaga Priya Department of Computer Science and Engineering Vels Institute of Science Technology & Advanced Studies

ISBN: (数字)9798350379945

ISBN: (纸本)9798350379952

As artificial speech generates technological evolution; fake voice information has become an increasingly prevalent avenue for fraud. Numerous investigations have been carried out into machine learning techniques, signal processing techniques, and the Automatic Speaker Verification System to address such issues. Since there are increasing methods to create fake information, techniques must be developed to determine when audio recordings employed as electronic proof are corrupted. Such recordings contain the potential to be manipulated. Hence, this research work identified the data regarding audio systems as either fake or real by i) extracting features from audio file metadata using oversampling techniques and ii) Developing an Ensemble Bagging Model appropriate for analyzing input audio signal followed by classification of fake (fraudulent) and real (authentic) audio. Our experimental results reveal that the ensemble bagging model attained an accuracy of 99.5%, precision of 99%, 99.6% recall, and 99.9% Fl-measure in the prediction of the audio system compared with traditional approaches.

关键词： Deepfakes Accuracy Audio systems Speech recognition Metadata Feature extraction Audio recording Real-time systems Fraud Bagging

来源：评论

学校读者我要写书评

暂无评论

A hybrid technique for selection and minimization of test cases in regression testing

引用

Recent Advances in computer Science and Communications 2021年第2期14卷 477-488页

作者： Singh, Leena Singh, Shailendra N. Dawra, Sudhir Department of Computer Science and Engineering Amity University Noida India Department of Computer Application Tecnia Institute of Advanced Studies New Delhi India

Background: In today’s era, modifications in a software is a common requirement by customers. When changes are made to existing software, re-testing of all the test cases is required to ensure that the newly introduced changes do not have any unwanted effect on the behavior of the software. However, re-testing of all the test cases would not only be time consuming but also expensive. Therefore, there is a need for a technique that reduces the number of tests to be performed. Regression testing is one of the ways to reduce the number of test cases. Selection technique is one such method which seeks to identify the test cases that are relevant to some set of recent changes. Objective: It is evident that most of the studies have used different selection techniques and have focused only on one parameter for achieving reduced test suite size without compromising the performance of regression testing. However, to the best of our knowledge, no study has taken two or more parameters of cov-erage, and/or execution time in a single testing. This paper presents a hybrid technique that combines both regression test selection using slicing technique and minimization of test cases using modified firefly algorithm with combination of parameters coverage and execution time in a single testing. Methods: A hybrid technique has been described that combines both selection and minimization. Selection of test cases is based upon slicing technique while minimization is done using firefly algorithm. Hybrid technique selects and minimizes the test suite using information on statement coverage and execution time. Results: The proposed technique gives 43.33% much superior result as compared to the other hybrid approach in terms of significantly reduced number of test cases. It shows that the resultant test cases were effective enough to cover 100% of the statements, for all the programs. The proposed technique was also tested on four different programs namely Quadratic, Triangle, Next day, Co

关键词： Testing

来源：评论

学校读者我要写书评

暂无评论

Generalized LRS Estimator for Min-Entropy Estimation

Generalized LRS Estimator for Min-Entropy Estimation

引用

作者： Woo, Jiheon Yoo, Chanhee Kim, Young-Sik Cassuto, Yuval Kim, Yongjune Department of Electrical Engineering Pohang37673 Korea Republic of Department of Electrical Engineering and Computer Science Daegu42988 Korea Republic of Chosun University Department of Information and Communication Engineering Gwangju61452 Korea Republic of Technion-Israel Institute of Technology Viterbi Department of Electrical and Computer Engineering Haifa32000 Israel Institute for Convergence Research and Education in Advanced Technology Yonsei University Seoul03722 Korea Republic of

The min-entropy is a widely used metric to quantify the randomness of generated random numbers, which measures the difficulty of guessing the most likely output. It is difficult to accurately estimate the min-entropy of a non-independent and identically distributed (non-IID) source. Hence, NIST Special Publication (SP) 800-90B adopts ten different min-entropy estimators and then conservatively selects the minimum value among these ten min-entropy estimates. Among these estimators, the longest repeated substring (LRS) estimator estimates the collision entropy instead of the min-entropy by counting the number of repeated substrings. Since the collision entropy is an upper bound on the min-entropy, the LRS estimator inherently provides overestimated outputs. In this paper, we propose two techniques to estimate the min-entropy of a non-IID source accurately. The first technique resolves the overestimation problem by translating the collision entropy into the min-entropy. Next, we generalize the LRS estimator by adopting the general Rényi entropy instead of the collision entropy (i.e., Rényi entropy of order two). We show that adopting a higher order can reduce the variance of min-entropy estimates. By integrating these techniques, we propose a generalized LRS estimator that effectively resolves the overestimation problem and provides stable min-entropy estimates. Theoretical analysis and empirical results support that the proposed generalized LRS estimator improves the estimation accuracy significantly, which makes it an appealing alternative to the LRS estimator. © 2005-2012 IEEE.

关键词： Frequency estimation

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：