检索结果-内蒙古大学图书馆

Few-Shot Referring Video Single- and Multi-Object Segmentation Via Cross-Modal Affinity with Instance Sequence Matching

引用

International Journal of computer Vision 2025年 1-19页

作者： Liu, Heng Li, Guanghui Gao, Mingqi Zhen, Xiantong Zheng, Feng Wang, Yang School of Computer Science and Technology Anhui University of Technology Maxiang Road Ma’anshan243032 China Department of Computer Science and Engineering Southern University of Science and Technology Xueyuan Avenue Shenzhen518055 China School of Computer Science and Information Engineering Hefei University of Technology Feicui Road Hefei230601 China

Referring Video Object Segmentation (RVOS) aims to segment specific objects in videos based on the provided natural language descriptions. As a new supervised visual learning task, achieving RVOS for a given scene requires a substantial amount of annotated data. However, only minimal annotations are usually available for new scenes in realistic scenarios. Another practical problem is that, apart from a single object, multiple objects of the same category coexist in the same scene. Both of these issues may significantly reduce the performance of existing RVOS methods in handling real-world applications. In this paper, we propose a simple yet effective model to address these issues by incorporating a newly designed cross-modal affinity (CMA) module based on a Transformer architecture. The CMA module facilitates the establishment of multi-modal affinity over a limited number of samples, allowing the rapid acquisition of new semantic information while fostering the model’s adaptability to diverse scenarios. Furthermore, we extend our FS-RVOS approach to multiple objects through a new instance sequence matching module over CMA, which filters out all object trajectories with similarity to language features that exceed a matching threshold, thereby achieving few-shot referring multi-object segmentation (FS-RVMOS). To foster research in this field, we establish a new dataset based on currently available datasets, which covers many scenarios in terms of single-object and multi-object data, hence effectively simulating real-world scenes. Extensive experiments and comparative analyses underscore the exceptional performance of our proposed FS-RVOS and FS-RVMOS methods. Our method consistently outperforms existing related approaches through practical performance evaluations and robustness studies, achieving optimal performance on metrics across diverse benchmark tests. © The Author(s), under exclusive licence to Springer science+Business Media, LLC, part of Springer Nature 2025.

关键词： Modal analysis

来源：评论

学校读者我要写书评

暂无评论

Federated Deep Reinforcement Learning for Multimodal Content Caching in Edge-Cloud Networks

引用

IEEE Transactions on Network science and engineering 2025年第3期12卷 2188-2201页

作者： Feng, Weijia Zuo, Xinyu Zhang, Ruojia Zhu, Yichen Wang, Chenyang Guo, Jia Sun, Chuan Tianjin Normal University School of Computer and Information Engineering Tianjin300387 China Shenzhen University College of Computer Science and Software Engineering Shenzhen518132 China Shenzhen University Guangdong Laboratory of Artificial Intelligence and Digital Economy Shenzhen518132 China Chongqing Normal University School of Computer and Information Science Chongqing401333 China

Edge caching presents a promising avenue for mitigating backbone network congestion by strategically caching frequently accessed content at the network periphery. As most current edge caching solutions are designed for single-modal content requests, they cannot deal with the increasing volume of multi-modal content requests. In this study, we investigate the issue of multimodal content caching in edge-cloud networks. Firstly, we establish a heterogeneous edge-cloud network adept at caching multimodal content proximate to end-users to facilitate expeditious content delivery. By leveraging latent representations of multimodal content, we identify distinct user request modalities for multimodal content. Subsequently, we formulate caching replacement operations as a Markov Decision Process (MDP) aimed at minimizing user-content access latency. Moreover, we propose a decentralized multimodal content caching framework at the network edge based on federated deep reinforcement learning. This framework affords distributed decision-making and learning capabilities, thereby alleviating the strain on centralized resources and augmenting caching efficacy. To demonstrate the efficacy of our proposed framework, we conduct comprehensive experiments utilizing the Noah-Wukong dataset. Experimental results provide evidence that our framework reduces average latency by up to 10% compared to traditional methods, highlighting its proficiency in enhancing cache performance in edge-cloud networks. © 2013 IEEE.

关键词： Federated learning

来源：评论

学校读者我要写书评

暂无评论

Adversarial multi-image steganography via texture evaluation and multi-scale image enhancement

引用

Multimedia Tools and Applications 2025年第9期84卷 5793-5823页

作者： Li, Fengyong Li, Longwei Zeng, Yishu Yu, Jiang Qin, Chuan College of Computer Science and Technology Shanghai University of Electric Power Shanghai201306 China School of Information and Computer Shanghai Business School Shanghai201400 China School of Optical-Electrical and Computer Engineering University of Shanghai for Science and Technology Shanghai200092 China

Multi-image steganography refers to a data-hiding scheme where a user tries to hide confidential messages within multiple images. Different from the traditional steganography which only requires the security of an individual image, multi-image steganography considers an overall security for a batch of images. However, existing multi-image steganography all faces a nontrivial problem: how to optimally allocate payload into multiple images to guarantee the security of batch images. To address this problem, this paper proposes an adversarial multi-image steganographic scheme. A multi-scale texture evaluation mechanism is firstly calculated to determine the embeddable cover images. Subsequently, a series of multi-scale filters are introduced to enhance the image content, which can be used to guide the optimal payload assignment of each image. Furthermore, an adversarial embedding mechanism is designed by dynamically adjusting the random gradient mapping of batch images, and finally achieving secure multi-image steganography. Our proposed scheme can optimize the overall steganographic security performance of multiple images, while ensuring the anti-steganalysis capability of a single image. Extensive experiments demonstrate that our scheme can achieve superior performance for multi-image steganography over different large-scale image sets, and outperforms state-of-the art schemes in terms of both single-image and multi-image anti-steganalysis capability. © The Author(s), under exclusive licence to Springer science+Business Media, LLC, part of Springer Nature 2024.

关键词： Embeddings

来源：评论

学校读者我要写书评

暂无评论

Single-level Discrete Two Dimensional Wavelet Transform Based Multiscale Deep Learning Framework for Two-Wheeler Helmet Classification

EAI Endorsed Transactions on Industrial Networks and Intelli...

引用

EAI Endorsed Transactions on Industrial Networks and Intelligent systems 2025年第2期12卷 1-13页

作者： Annadurai, Amrutha Prusty, Manas Ranjan Pandey, Trilok Nath Patra, Subhra Rani School of Computer Science and Engineering Vellore Institute of Technology Chennai India Centre for Cyber Physical Systems Vellore Institute of Technology Chennai India Information Systems and Operations Management University of Texas at Arlington Texas 76019 United States

INTRODUCTION: A robust method is proposed in this paper to detect helmet usage in two-wheeler riders to enhance road safety. OBJECTIVES: This involves a custom made dataset that contains 1000 images captured under diverse real-world scenarios, including variations in helmet size, colour, and lighting conditions. This dataset has two classes namely with helmet and without helmet. METHODS: The proposed helmet classification approach utilizes the Multi-Scale Deep Convolutional Neural Network (CNN) framework cascaded with Long Short-Term Memory (LSTM) network. Initially the Multi-Scale Deep CNN extracts modes by applying Single-level Discrete 2D Wavelet Transform (dwt2) to decompose the original images. In particular, four different modes are used for segmenting a single image namely approximation, horizontal detail, vertical detail and diagonal detail. After feeding the segmented images into a Multi-Scale Deep CNN model, it is cascaded with an LSTM network. RESULTS: The proposed model achieved accuracies of 99.20% and 95.99% using both 5-Fold Cross-Validation (CV) and Hold-out CV methods, respectively. CONCLUSION: This result was better than the CNN-LSTM, dwt2-LSTM and a tailor made CNN model. © 2025 A. Annadurai et al.

关键词： Convolutional neural networks

来源：评论

学校读者我要写书评

暂无评论

AFuNet: an attention-based fusion network to classify texts in a resource-constrained language

引用

Neural Computing and Applications 2025年第9期37卷 6725-6748页

作者： Hossain, Md. Rajib Hoque, Mohammed Moshiul Dewan, M. Ali Akber Hoque, Enamul Siddique, Nazmul Department of Computer Science & Engineering Chittagong University of Engineering & Technology Chittagong4349 Bangladesh School of Computing and Information Systems Faculty of Science and Technology Athabasca University AthabascaABT9S 3A3 Canada School of Information Technology York University TorontoONM3J 1P3 Canada School of Computing Engineering and Intelligent Systems Ulster University Coleraine United Kingdom

In the era of widespread Internet use and extensive social media interaction, the digital realm is accumulating vast amounts of unstructured text data. This unstructured data often contain undesirable information, necessitating time-consuming manual classification efforts. An intelligent text classification system capable of automatically categorizing digitized texts based on semantic meaning is crucial. However, this task is particularly challenging for low-resource languages like Bengali due to a shortage of annotated corpora, issues with out-of-vocabulary words, lack of domain-specific hyperparameter tuning, limited ability to extract generalized text features, and class imbalances within the corpus. AFuNet: an attention-based fusion network to classify texts in a resource-constrained language. AFuNet undergoes a comprehensive four-phase experimental process, including baseline model evaluation and hyperparameter tuning, late fusion and model selection, attention-based early fusion and model identification, and an ablation study with impact analysis. Fine-tuned based on five Bengali text classification corpora, AFuNet achieves impressive accuracies: 96.60 ± 0.2 (BTCC11), 85.37 ± 0.2 (OSBC), 97.35 ± 0.2 (BARD), 93.74 ± 0.2 (IndicNLP), and 96.51 ± 0.2 (ProthomAlo). In comparison with previous state-of-the-art models on these corpora, AFuNet demonstrates significant accuracy improvements ranging from 0.54% to 4.49%, showcasing its effectiveness in advancing text classification capabilities for the Bengali language. © The Author(s), under exclusive licence to Springer-Verlag London Ltd., part of Springer Nature 2025.

关键词： Semantics

来源：评论

学校读者我要写书评

暂无评论

Privacy-Preserving Statistical Analysis With Low Redundancy Over Task-Relevant Microdata

引用

IEEE Transactions on information Forensics and Security 2025年 20卷 4382-4395页

作者： Zhao, Jingcheng Xue, Kaiping Xue, Yingjie Li, Meng Zhu, Bin Yuan, Shaoxian University of Science and Technology of China School of Cyber Science and Technology Anhui Hefei230027 China Hefei University of Technology School of Computer Science and Information Engineering Anhui Hefei230601 China

Privacy-preserving statistical analysis enables the data center to analyze datasets from multiple data owners, extracting valuable insights while safeguarding privacy. However, the observation of microdata involvement in various analysis tasks within the data center can indirectly lead to privacy breaches. For instance, when the data center observes microdata involved in a disease-related task, it may reveal information about the corresponding user’s disease. Existing schemes process the entire dataset for each analysis task to prevent privacy breaches, resulting in significant redundancy overhead due to the large amount of task-irrelevant data involved in processing. In this paper, we propose FDC, which can protect privacy and effectively reduce the redundancy overhead. It frees the data center from huge redundancy overhead. Specifically, we propose a co-design of local differential privacy and multiparty computation with preprocessing by the data owner. This design enables the data center to process only task-relevant and LDP noise-induced microdata instead of the entire dataset while maintaining analysis results without accuracy loss. In some scenarios where preprocessing by the data owner is unfeasible, we present a data center-assisted method to complete preprocessing within the data center. Additionally, we design and optimize a secure shuffle protocol within this method. Finally, we implement and evaluate FDC using the aggregation task as a baseline. With different proportions of task-relevant microdata, experimental results show that the runtime of FDC is 2\sim 11 x faster than existing schemes on LAN and 2\sim 22 x on WAN, and the communication overhead is up to 3\sim 153 x lower. © 2005-2012 IEEE.

关键词： Data centers

来源：评论

学校读者我要写书评

暂无评论

ASL-OOD:Hierarchical Contextual Feature Fusion with Angle-Sensitive Loss for Oriented Object Detection

引用

computers, Materials & Continua 2025年第2期82卷 1879-1899页

作者： Kexin Wang Jiancheng Liu Yuqing Lin Tuo Wang Zhipeng Zhang Wanlong Qi Xingye Han Runyuan Wen Northwest Institute of Mechanical and Electrical Engineering Xianyang712099China School of Information Engineering Chang’an UniversityXi’an710064China School of Computer Science and Technology Xidian UniversityXi’an710071China

Detecting oriented targets in remote sensing images amidst complex and heterogeneous backgrounds remains a formidable challenge in the field of object *** frameworks for oriented detection modules are constrained by intrinsic limitations,including excessive computational and memory overheads,discrepancies between predefined anchors and ground truth bounding boxes,intricate training processes,and feature alignment *** overcome these challenges,we present ASL-OOD(Angle-based SIOU Loss for Oriented Object Detection),a novel,efficient,and robust one-stage framework tailored for oriented object *** ASL-OOD framework comprises three core components:the Transformer-based Backbone(TB),the Transformer-based Neck(TN),and the Angle-SIOU(Scylla Intersection over Union)based Decoupled Head(ASDH).By leveraging the Swin Transformer,the TB and TN modules offer several key advantages,such as the capacity to model long-range dependencies,preserve high-resolution feature representations,seamlessly integrate multi-scale features,and enhance parameter *** improvements empower the model to accurately detect objects across varying *** ASDH module further enhances detection performance by incorporating angle-aware optimization based on SIOU,ensuring precise angular consistency and bounding box *** approach effectively harmonizes shape loss and distance loss during the optimization process,thereby significantly boosting detection *** evaluations and ablation studies on standard benchmark datasets such as DOTA with an mAP(mean Average Precision)of 80.16 percent,HRSC2016 with an mAP of 91.07 percent,MAR20 with an mAP of 85.45 percent,and UAVDT with an mAP of 39.7 percent demonstrate the clear superiority of ASL-OOD over state-of-the-art oriented object detection *** findings underscore the model’s efficacy as an advanced solution for challenging remote sensing object detection tasks.

关键词： Oriented object detection transformer deep learning

来源：评论

学校读者我要写书评

暂无评论

A decentralized and privacy-preserving framework for electronic health records using blockchain

引用

Alexandria engineering Journal 2025年 126卷 196-203页

作者： Alahmari, Saad Alshardan, Amal Al-Wesabi, Fahd N. Sorour, Shaymaa Alghushairy, Omar Alsini, Raed Khadidos, Alaa O. Duhayyim, Mesfer Al Department of Computer Science Applied College Northern Border University Arar Saudi Arabia Department of Information Systems College of Computer and Information Sciences Princess Nourah bint Abdulrahman University P.O. Box 84428 Riyadh11671 Saudi Arabia Department of Computer Science Applied College at Mahayil King Khalid University Saudi Arabia Department of Management Information Systems College of Business Administration King Faisal University Al-Ahsa31982 Saudi Arabia Department of Information Systems and Technology College of Computer Science and Engineering University of Jeddah Jeddah21589 Saudi Arabia Department of Information Systems Faculty of Computing and Information Technology King Abdulaziz University Jeddah Saudi Arabia Department of Computer Science College of Computer Engineering and Sciences Prince Sattam bin Abdulaziz University Al-Kharj16273 Saudi Arabia

As healthcare services have become increasingly digitized, Electronic Health Records (EHRs) have become widely adopted, providing seamless data exchange among providers. Conventional EHRs, however, are extremely vulnerable to cyber threats because patients' sensitive data is centralized and transmitted electronically. The paper proposes a decentralized, privacy-preserving framework for managing EHRs on blockchains in order to address these security and privacy concerns. Using cryptographic techniques, such as homomorphic encryption and zero-knowledge proofs, the proposed system enhances security and ensures data integrity. Additionally, the model facilitates scalable, efficient, and secure access to patient records through the integration of cloud-based storage and blockchain. Using smart contracts, we also ensure compliance with healthcare regulations by regulating access control and authentication. As a result of performance evaluations, the proposed approach is demonstrated to be feasible, and the advantages it offers in terms of security, privacy, and efficiency are highlighted. © 2025 The Authors

关键词： Electronic health record

来源：评论

学校读者我要写书评

暂无评论

Enhancing aviation safety: Machine learning for real-time ADS-B injection detection through advanced data analysis

引用

Alexandria engineering Journal 2025年 126卷 262-276页

作者： Rahman, Md. Atiqur Bhuiyan, Touhid Ali, M. Ameer Department of Computer Science and Engineering East West university Dhaka Bangladesh School of Information Technology Washington University of Science and Technology Alexandria United States Department of Computer Science and Engineering Bangladesh University of Business and Technology Dhaka Bangladesh

Airplanes play a critical role in global transportation, ensuring the efficient movement of people and goods. Although generally safe, aviation systems occasionally encounter incidents and accidents that underscore the need for proactive risk management. This study employs machine learning to detect abnormalities in commercial aircraft operations using Automatic Dependent Surveillance–Broadcast (ADS-B) data. Given the growing reliance on ADS-B technology, concerns regarding its susceptibility to security breaches, such as injection attacks, have intensified. To address these vulnerabilities, we propose a robust ADS-B injection detection system. Employing GridSearchCV for model optimization, it effectively identifies and categorizes injection risks. The system's performance, evaluated using the ADS-B Message Injection Attacks Dataset, achieves outstanding results, including a value of 0.9970 for the accuracy, precision, recall, and F1 score. The proposed classifier also demonstrates a higher area under the curve (0.9999), specificity (0.9956), and Cohen's kappa (0.9954) than existing approaches, while achieving a lower log loss (0.0107). This research significantly enhances aviation security by introducing a highly accurate, computationally efficient, and reliable real-time detection model for ADS-B injection attacks, ensuring the integrity and resilience of modern flight control systems. © 2025 The Authors

关键词： Risk management

来源：评论

学校读者我要写书评

暂无评论

RaSA: Robust and Adaptive Secure Aggregation for Edge-Assisted Hierarchical Federated Learning

引用

IEEE Transactions on information Forensics and Security 2025年 20卷 4280-4295页

作者： Wang, Lingling Huang, Mei Zhang, Zhengyin Li, Meng Wang, Jingjing Gai, Keke Qingdao University of Science and Technology School of Information Science and Technology Shandong Qingdao266061 China Hefei University of Technology School of Computer Science and Information Engineering Anhui Hefei230601 China Beijing Institute of Technology School of Cyberspace Science and Technology Beijing100081 China

Secure Aggregation (SA), in the Federated Learning (FL) setting, enables distributed clients to collaboratively learn a shared global model while keeping their raw data and local gradients private. However, when SA is implemented in edge-intelligence-driven FL, the open and heterogeneous environments will hinder model aggregation, slow down model convergence speed, and decrease model generalization ability. To address these issues, we present a Robust and adaptive Secure Aggregation (RaSA) protocol to guarantee robustness and privacy in the presence of non-IID data, heterogeneous system, and malicious edge servers. Specifically, we first design an adaptive weights updating strategy to address the non-IID data issue by considering the impact of both gradient similarity and gradient diversity on the model aggregation. Meanwhile, we enhance privacy protection by preventing privacy leakage from both gradients and aggregation weights. Different from previous work, we address system heterogeneity in the case of malicious attacks, and the malicious behavior from edge servers can be detected by the proposed verifiable approach. Moreover, we eliminate the influence of straggling communication links and dropouts on the model convergence by combining efficient product-coded computing with repetition-based secret sharing. Finally, we perform a theoretical analysis that proves the security of RaSA. Extensive experimental results show that RaSA can ensure model convergence without affecting the generalization ability under non-IID scenarios. Moreover, the decoding efficiency of RaSA achieves 1.33x and 6.4x faster than the state-of-the-art product-coded and one-dimensional coded computing schemes. © 2005-2012 IEEE.

关键词： Differential privacy

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：