检索结果-内蒙古大学图书馆

MA-MFIF: When misaligned multi-focus Image fusion meets deep homography estimation

Multimedia Tools and Applications 2025年第12期84卷 10877-10898页

作者： Zhao, Baojun Luo, Fei Fuentes, Joel Ding, Weichao Gu, Chunhua School of Information Science and Engineering East China University of Science and Technology Shanghai200237 China Shanghai Key Laboratory of Computer Software Evaluating and Testing Shanghai China Department of Computer Science and Information Technologies Universidad del Bio-Bio Chillán3780000 Chile

Multi-focus image fusion is a technique that combines multiple out-of-focus images to enhance the overall image quality. It has gained significant attention in recent years, thanks to the advancements in deep learning. However, one of the persistent challenges in this field is the processing of misaligned data, which can negatively impact the fusion results. To overcome this problem, a novel fusion framework with pre-registration is proposed for the fusion of misaligned multi-focus images. For pre-registration, content-aware deep homography estimation is used, which performs transfer learning on a real multi-focus image dataset to adapt to registration under defocused conditions. For fusion, a fusion module with dual-branch feature interaction is utilized to avoid invalid feature fusion and trained on real light field dataset to achieve better fusion performance. Qualitative and quantitative experimental results show that the proposed method has a 2-3 percentage point improvement in multiple evaluation metrics compared to existing advanced registration and fusion methods, and a maximum improvement of 4.83 percentage points in fusion performance when tested independently on the Lytro dataset. Additionally, We find that the value of the Qcv metric is greatly influenced by the alignment status of the input images, leading to its inability to reflect the fusion quality of aligned images. © The Author(s), under exclusive licence to Springer Science+Business Media, LLC, part of Springer Nature 2024.

关键词： Image enhancement

来源：评论

学校读者我要写书评

暂无评论

Enhancing Performance of Machine Learning Models in Healthcare: An Analytical Framework for Assessing and Improving Data Quality 5th

Enhancing Performance of Machine Learning Models in Healthca...

引用

5th International Conference on Advances in information Communication technology and Computing, AICTC 2024

作者： Al-Hgaish, Areen Atan, Rodziah Yaakob, Razali Osman, Mohd Hafeez Department of Software Engineering and Information Systems Faculty of Computer Science and Information Technology Universiti Putra Malaysia Serdang44330 Malaysia Department of Computer Science Faculty of Computer Science and Information Technology Universiti Putra Malaysia Serdang44330 Malaysia

ISBN: (纸本)9789819761050

The utilization of Data-Driven Machine Learning (DDML) models in the healthcare sector poses unique challenges due to the crucial nature of clinical decision-making and its impact on patient outcomes. A primary concern in this domain is the data quality that forms the basis for these models, which is essential for their effectiveness. While numerous studies are dedicated to leveraging Machine Learning (ML) to enhance patient care, a substantial gap exists in understanding and evaluating the quality of training data before training ML models. This issue is exacerbated by the lack of well-defined standards to ensure that data characteristics align with the specific goals of ML in healthcare. To address this, our study introduces a conceptual three-dimensional framework focusing on data accuracy, completeness, and consistency aimed at assessing and improving the quality of training data. This framework is designed to enhance the performance, predictability, and interpretability of healthcare-specific ML models and reduce biases, thus bridging current research gaps and laying the groundwork for future developments in ML-driven healthcare systems. This paper presents the initial step for our proposed Data Quality Framework (DQF), which addresses the three dimensions mentioned above. It discusses its potential implications and applications in healthcare data, and its main goal is to provide a comprehensive overview of the proposed DQF, illustrating its significance in advancing healthcare data quality and ultimately improving patient outcomes. © The Author(s), under exclusive license to Springer Nature Singapore Pte Ltd. 2025.

关键词： Data accuracy

来源：评论

学校读者我要写书评

暂无评论

MHAN: Bottleneck Fusion Model Based on Hybrid Attention Network for Multimodal Emotion Recognition

引用

Journal of Shanghai Jiaotong University (Science) 2025年 1-8页

作者： Li, Kai Huang, Ya Zhong, Gang Nurmemet, Yolwas Wushouer, Silamu School of Software Xinjiang University Urumqi830091 China College of Information Science and Engineering Xinjiang University Urumqi830017 China Xinjiang Multilingual Information Technology Laboratory Xinjiang University Urumqi830017 China

Multimodal emotion recognition (MER) offers a more comprehensive and accurate understanding of human emotions by analyzing and integrating information from various modalities. However, previous work may have yet to fully consider the emotion feature information within and between modalities, which is the most significant challenge in this domain. Therefore, a novel multimodal hybrid attention network (MHAN) is proposed to achieve more effective information aggregation between text and audio. MHAN includes three main modules, i.e., hybrid encoder block (HEB), multi-head cross-attention (MCA) block, and bottleneck fusion (BF) block. Specifically, the HEB can extract advanced feature representation from text and audio sequences. Moreover, the MCA block can capture common features between text and audio to support emotion recognition. To filter out some redundant information, the BF block is designed to better extract the complementary and most relevant information between text and audio modalities. To validate the effectiveness of MHAN, we have conducted extensive experiments on the IEMOCAP and MELD datasets, and the experimental results demonstrate that MHAN outperforms existing state-of-the-art methods and better interprets subjective human emotions in the applications. (Figure presented.) © Shanghai Jiao Tong University 2025.

关键词： A bottleneck fusion (BF) multi-head cross-attention (MCA) multimodal emotion recognition (MER) multimodal hybrid attention network (MHAN) TP183

来源：评论

学校读者我要写书评

暂无评论

U-net of joint spatial domains with multi-scale atrous convolution for rectal image segmentation

引用

Multimedia Tools and Applications 2025年 1-17页

作者： Rao, Yunbo Gao, Li Zeng, Shaoning Shao, Tingting Sun, Jihong School of Information and Software Engineering University of Electronic Science and Technology of China Chengdu611731 China Yangtze Delta Region Institute University of Electronic Science and Technology of China Huzhou313000 China Zhejiang University School of Medicine Zhejiang Hangzhou310058 China

Medical image segmentation is very important for the diagnosis of related diseases. To reduce the labeling work of related medical images, numerous models based on U-Net have been proposed to achieve automatic segmentation of target regions. However, most of these models are only trained in one coordinate system, ignoring the joint effects of different spatial coordinate systems. In addition, most of the encoding modules in these models do not pay attention to multi-scale spatial information. Our proposed solution for the aforementioned challenges involves using U-Net model with joint spatial domains and multi-scale encoding module, which enables us to segment rectal image better. The model includes a self-designed multi-layer dilated convolution encoding module named AIR (Atrous Inception Residual Block), to achieve a multi-scale content fusion. Besides this, it utilizes the center point and polar coordinates to realize attention mechanism and rotation invariance. Furthermore, retraining the output of polar coordinate network with Cartesian coordinate system realizes the translation invariance of segmentation. Compared with the commonly used medical segmentation models, the dice coefficient of our model is improved by about 2% on our in-house rectal dataset. © The Author(s), under exclusive licence to Springer Science+Business Media, LLC, part of Springer Nature 2025.

关键词： Image segmentation

来源：评论

学校读者我要写书评

暂无评论

SA-ResNet:An Intrusion Detection Method Based on Spatial Attention Mechanism and Residual Neural Network Fusion

引用

Computers, Materials & Continua 2025年第5期83卷 3335-3350页

作者： Zengyu Cai Yuming Dai Jianwei Zhang Yuan Feng School of Computer Science and Technology Zhengzhou University of Light IndustryZhengzhou450066China College of Software Engineering Zhengzhou University of Light IndustryZhengzhou450066China Faculty of Information Engineering Xuchang Vocational Technical CollegeXuchang461000China School of Electronic Information Zhengzhou University of Light IndustryZhengzhou450066China

The rapid development and widespread adoption of Internet technology have significantly increased Internet traffic,highlighting the growing importance of network *** Detection Systems(IDS)are essential for safeguarding network *** address the low accuracy of existing intrusion detection models in identifying network attacks,this paper proposes an intrusion detection method based on the fusion of Spatial Attention mechanism and Residual Neural Network(SA-ResNet).Utilizing residual connections can effectively capture local features in the data;by introducing a spatial attention mechanism,the global dependency relationships of intrusion features can be extracted,enhancing the intrusion recognition model’s focus on the global features of intrusions,and effectively improving the accuracy of intrusion *** proposed model in this paper was experimentally verified on theNSL-KDD *** experimental results showthat the intrusion recognition accuracy of the intrusion detection method based on SA-ResNet has reached 99.86%,and its overall accuracy is 0.41% higher than that of traditional Convolutional Neural Network(CNN)models.

关键词： Intrusion detection deep learning residual neural network spatial attention mechanism

来源：评论

学校读者我要写书评

暂无评论

Against Mobile Collusive Eavesdroppers: Cooperative Secure Transmission and Computation in UAV-Assisted MEC Networks

引用

IEEE Transactions on Mobile Computing 2025年第6期24卷 5280-5297页

作者： Zhao, Mingxiong Wang, Zirui Guo, Kun Zhang, Rongqian Quek, Tony Q. S. Ministry of Education Engineering Research Center of Integration and Application of Digital Learning Technology Beijing100039 China Ministry of Education Engineering Research Center of Cyberspace Kunming650504 China Yunnan University of Finance and Economics Yunnan Key Laboratory of Service Computing Kunming650221 China Yunnan University Yunnan Key Laboratory of Software Engineering Kunming650091 China East China Normal University Shanghai Key Laboratory of Multidimensional Information Processing School of Communications and Electronics Engineering Shanghai200241 China Yunnan University School of Information Science and Engineering Kunming650500 China Singapore University of Technology and Design Information Systems Technology and Design 487372 Singapore National Pilot School of Software Yunnan University Kunming China

In Unmanned Aerial Vehicle (UAV)-assisted Mobile Edge Computing (MEC) networks, the security of transmission faces significant challenges due to the vulnerabilities of line-of-sight links and potential eavesdropping on two-hop links. This paper addresses these challenges with an innovative Cooperative Secure Transmission and Computation strategy (CSTC), specifically engineered for time-slotted UAV-assisted MEC networks plagued by mobile collusive eavesdroppers. These eavesdroppers significantly bolster their interception capabilities through coordinated and optimized movements, escalating the security threats. To neutralize these risks, the proposed CSTC employs the UAV and remote devices as helper nodes to emit jamming signals, thereby thwarting eavesdropping activities, while simultaneously facilitating the efficient relay of users' tasks to the base station for advanced processing. The CSTC aims to maximize the sum Secrecy Transmission Rate (STR) satisfying task latency constraints. It involves a joint optimization of UAV trajectory, jamming beamformers, transmit power, and data offloading strategy to expedite task transmission. Additionally, a real-time computation scheduling approach is developed based on a newly defined metric, the Urgency Degree of Users (UDoU), to enhance task processing efficiency. Our extensive simulations validate that the CSTC not only elevates the sum STR but also consistently meets latency constraints, demonstrating its robustness against advanced mobile eavesdropping techniques. © 2002-2012 IEEE.

关键词： Jamming

来源：评论

学校读者我要写书评

暂无评论

Cross-graph Knowledge Exchange for Personalized Response Generation in Dialogue Systems

引用

IEEE Internet of Things Journal 2025年第10期12卷 13769-13789页

作者： Dong, Yuezhou Qin, Ke Ke, Pei Liang, Shuang Luo, Guangchun National Key Laboratory of Intelligent Collaborative Computing China University of Electronic Science and Technology of China School of Computer Science and Engineering Chengdu China University of Electronic Science and Technology of China National Key Laboratory of Intelligent Collaborative Computing Chengdu China University of Electronic Science and Technology of China School of Information and Software Engineering Chengdu China

Recent advancements in language models have greatly improved dialogue systems, but they still face challenges in generating personalized responses that are consistent with the user's persona and dialogue context. Existing approaches typically model dialogue context and persona information together in a unified manner, but they lack fine-grained differentiation between the two, leading to inaccurate user modeling. This misalignment hinders the ability of dialogue systems to produce personalized responses. In this work, we propose CKE (Cross-Graph Knowledge Exchange), a novel algorithm designed to enhance personalized response generation in dialogue systems. CKE constructs separate dialogue user graphs for each party in the dialogue, representing both their dialogue context and persona information. These graphs are then utilized to perform cross-graph structured knowledge aggregation, where the aggregation is under supervised by both its own and the other party's persona and dialogue context, providing richer, more accurate representations. Furthermore, CKE introduces a hybrid prompt template that combines both discrete and continuous elements, improving the language model's ability to leverage graph-structured information. The experimental results demonstrate that CKE significantly outperforms existing baseline methods in generating more coherent, contextually appropriate, and personalized responses in dialogue systems. © 2014 IEEE.

关键词： Knowledge graph

来源：评论

学校读者我要写书评

暂无评论

Combined Distance Multi-Target Tracking Method of Improved TransCenter

引用

Computer engineering and Applications 2025年第9期61卷 177-185页

作者： Zhao, Haitao Yue, Xi Tang, Dan Cai, Bo School of Software Engineering Chengdu University of Information Technology Chengdu610225 China Sichuan Province Engineering Technology Research Center of Support Software of Informatization Application Chengdu610225 China

In the fields of intelligent driving and video surveillance, multi-target tracking is widely used. However, when the target is blocked and moves non-linearly, the noise generated at this time will reduce the detection and tracking accuracy. At the same time, many correlation matching algorithms do not take this into account the IoU and appearance imbalance situations. In response to the above problems, an improved multi-target tracking network based on TransCenter is proposed. Firstly, wavelet transform is used to process detection features, and then a contextual collaborative selector is designed to alleviate the negative impact of noise by dynamically selecting tracking features and detection features. Then, Kalman filter prediction values and tracking displacements are integrated to improve nonlinear motionand the accuracy of the predicted displacement. Finally, the weight of the combined distance is optimized based on the difference between the IoU distance and the appearance distance, which solves the problem of combined distance failure when high-speed motion and dramatic changes in appearance occur. Experiments are conducted on the BDD100k and DanceTrack datasets. The results show that compared with the ByteTrack algorithm, the HOTA and mMOTA values of the improved network increase by 4.3 and 5.9 percentage points respectively. Compared with TransCenter, the HOTA increases by 7.4 percentage points and has better performance of good flexibility and tracking accuracy. © 2025 Journal of Computer engineering and Applications Beijing Co., Ltd.;Science Press. All rights reserved.

关键词： Kalman filters

来源：评论

学校读者我要写书评

暂无评论

Exploring Backdoor Vulnerabilities of Chat Models 31

Exploring Backdoor Vulnerabilities of Chat Models

引用

31st International Conference on Computational Linguistics, COLING 2025

作者： Yang, Wenkai Hao, Yunzhuo Lin, Yankai Gaoling School of Artificial Intelligence Renmin University of China China School of Information and Software Engineering University of Electronic Science and Technology of China China

ISBN: (纸本)9798891761964

Recent researches have shown that Large Language Models (LLMs) are susceptible to a security threat known as Backdoor Attack. The backdoored model will behave well in normal cases but exhibit malicious behaviours on inputs inserted with a specific backdoor trigger. Current backdoor studies on LLMs predominantly focus on single-turn instruction-tuned LLMs, while neglecting another realistic scenario where LLMs are fine-tuned on multi-turn conversational data to be chat models. Chat models are extensively adopted across various real-world scenarios, thus the security of chat models deserves increasing attention. Unfortunately, we point out that the flexible multi-turn interaction format instead increases the flexibility of trigger designs and amplifies the vulnerability of chat models to backdoor attacks. In this work, we reveal and achieve a novel backdoor attacking method on chat models by distributing multiple trigger scenarios across user inputs in different rounds, and making the backdoor be triggered only when all trigger scenarios have appeared in the historical conversations. Experimental results demonstrate that our method can achieve high attack success rates (e.g., over 90% ASR on Vicuna-7B) while successfully maintaining the normal capabilities of chat models on providing helpful responses to benign user requests. Also, the backdoor cannot be easily removed by the downstream re-alignment, highlighting the importance of continued research and attention to the security concerns of chat models. © 2025 Association for Computational Linguistics.

关键词： Computational linguistics

来源：评论

学校读者我要写书评

暂无评论

SPRSPS: An Effective Preprocessing Framework to Enhance Vulnerability Detection in Deep Learning Models 24

SPRSPS: An Effective Preprocessing Framework to Enhance Vuln...

引用

4th Asia-Pacific Artificial Intelligence and Big Data Forum, AIBDF 2024

作者： Wang, Jiarui Guo, Wensheng Yang, Xia Sun, Haiyong School of Information and Software Engineering University of Electronic Science and Technology of China Sichuan Chengdu China Chengdu Lianan Tech Sichuan Chengdu China

ISBN: (纸本)9798400710865

Current vulnerability detection methods based on deep learning and program slicing techniques are widely used, but the program representations and slicing strategies they employ are not well-suited for this purpose. This paper introduces the SPRSPS (More Suitable Program Representation and Smarter Program Slicing) framework, which standardizes and enhances preprocessing procedures. The approach involves: first, utilizing features of LLVM IR, such as SSA and control/data flow simplification, to construct program representations more suitable for vulnerability detection;second, optimizing graph representations for graph neural networks commonly used in both general and specific vulnerability detection. Compared to previous work, it more clearly defines node and edge features. Finally, it thoroughly investigates the distances between vulnerability manifestation points and root causes, proposing a smarter slicing strategy. Ablation experiments were conducted on the general vulnerability detection model SedSVD and the specific vulnerability detection model GLICE. After replacing their preprocessing methods with the proposed approach, improvements were observed in accuracy and F1 scores. In the SedSVD project, the false positive rate (FPR) decreased by 12%, accuracy increased by 7%, and the F1 score improved by 10%, with a 95% reduction in the number of result lines, indicating more precise vulnerability localization. In the GLICE project, the FPR decreased by 10%, accuracy improved by 3%, and the number of result lines was significantly reduced by 91%. These experiments demonstrate that using the SPRSPS framework allows for more accurate and efficient vulnerability detection. © 2024 Copyright held by the owner/author(s).

关键词： Graph neural networks

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：