检索结果-内蒙古大学图书馆

MILG:Realistic lip-sync video generation with audio-modulated image inpainting

Visual Informatics 2024年第3期8卷 71-81页

作者： Han Bao Xuhong Zhang Qinying Wang Kangming Liang Zonghui Wang Shouling Ji Wenzhi Chen School of Software Technology Zhejiang UniversityHangzhouChina College of Computer Science Zhejiang UniversityHangzhouChina College of Engineering Zhejiang UniversityHangzhouChina

Existing lip synchronization(lip-sync)methods generate accurately synchronized mouths and faces in a generated ***,they still confront the problem of artifacts in regions of non-interest(RONI),e.g.,background and other parts of a face,which decreases the overall visual *** solve these problems,we innovatively introduce diverse image inpainting to lip-sync *** propose Modulated Inpainting Lip-sync GAN(MILG),an audio-constraint inpainting network to predict synchronous *** utilizes prior knowledge of RONI and audio sequences to predict lip shape instead of image generation,which can keep the RONI ***,we integrate modulated spatially probabilistic diversity normalization(MSPD Norm)in our inpainting network,which helps the network generate fine-grained diverse mouth movements guided by the continuous audio ***,to lower the training overhead,we modify the contrastive loss in lipsync to support small-batch-size and few-sample *** experiments demonstrate that our approach outperforms the existing state-of-the-art of image quality and authenticity while keeping lip-sync.

关键词： Lip-sync Image inpainting Face generation Modulated SPD normalization

来源：评论

学校读者我要写书评

暂无评论

Leveraging user's preference and social circle for personalized recommendation via matrix factorization with sub-linear convergence rate

引用

Journal of Intelligent and Fuzzy Systems 2024年第1-2期47卷 1-13页

作者： Qiu, Laixiang Zhou, Wang Tian, Ying Ul Haq, Amin Ahmad, Sultan School of Computer and Software Engineering Xihua University Chengdu China Intelligent Policing Key Laboratory of Sichuan Province Sichuan Police College LuZhou China School of Computer Science and Engineering University of Electronic Science and Technology of China Chengdu China Department of Computer Science College of Computer Engineering and Sciences Prince Sattam Bin Abdulaziz University Alkharj Saudi Arabia

Recommender systems aim to filter information effectively and recommend useful sources to match users' requirements. However, the exponential growth of information in recent social networks may cause low prediction accuracy for recommendation systems. This article proposes a unified personalized recommendation architecture referred to as PSRec, which incorporates user preference and social relationship into matrix factorization framework. Specifically, PSRec generates two collections for textual reviews and contextual information respectively, and performs preference learning for each user via the Latent Dirichlet Allocation topic model. Moreover, PSRec exploits the inner relations within the social circle for recommendation, including direct trust relationship and indirect trust relationship. Additionally, it's certificated that PSRec can converge at a sub-linear rate via theoretical analysis. Experimental results over DoubanMovie, CiaoDVDs and Yelp demonstrate the superiority of the proposed PSRec, which can achieve significant improvements and provide much better user experience while compared with other benchmark models. © 2024 - IOS Press. All rights reserved.

关键词： Matrix factorization

来源：评论

学校读者我要写书评

暂无评论

CEMOP: Enhancing MOBA match outcome predictions by understanding teammates and opponents effects 3

CEMOP: Enhancing MOBA match outcome predictions by understan...

引用

3rd International Symposium on computer Applications and Information Systems, ISCAIS 2024

作者： Zhao, Jiangfeng College of Computer Science and Software Engineering Shenzhen University Shenzhen518057 China

ISBN: (纸本)9781510681316

Multiplayer Online Battle Arena (MOBA) games currently dominate the esports landscape, offering a concrete and vivid embodiment for team comparisons, where accurately predicting the winning team is both important and challenging. Recent studies often rely on the static features of team members to evaluate a team’s capabilities and predict their likelihood of winning. However, such approaches fail to consider the critical game-contextual influences of interactions with teammates and opponents on individuals. During a match, an individual’s performance is consistently influenced by teammates and opponents, leading to variations in individual features across different games. Moreover, the measurement of team capabilities also needs improvement, as the single scalar rating employed is too simplistic for sophisticated MOBA games. To this end, we propose a novel Context-Enhanced Match Outcome Prediction model (CEMOP), which incorporates the effect of both teammates and opponents for a better measurement of team capabilities. Specifically, we designed a match context encoder for MOBA games to evaluates the impact of teammates and opponents on individuals separately and embeds the information into the individual representations. After that the single team member’s representations which contain the influence of teammates and opponents are aggregated based on their importance to create a comprehensive representation of the team capabilities, enables CEMOP to adaptively adjust team evaluations. Experimental results from three real-world MOBA games clearly verify the effectiveness of CEMOP compared with several state-of-the-art methods. © 2024 SPIE.

关键词： Data mining

来源：评论

学校读者我要写书评

暂无评论

Rainfall Process Similarity Algorithm Research Based on Deep Learning by Using Precipitation Images 7

Rainfall Process Similarity Algorithm Research Based on Deep...

引用

7th International Conference on computer Information science and Application Technology, CISAT 2024

作者： Yu, Yufeng Wu, Panqi Cai, Difei Liu, Weifu Wang, Xuhang Hohai University College of Computer Science and Software Engineering Nanjing210098 China

ISBN: (纸本)9798350375107

Rainfall is the main cause of flood disasters, and analyzing its features plays a crucial role in preventing flood disasters. How to extract rainfall process features and conduct rainfall similarity analysis is a challenging work. Based on the similarity search method for daily precipitation images, this paper proposes a rainfall process similarity search method, which extracts rainfall duration, cumulative precipitation, temporal distribution, spatial distribution and center movement trajectory as the features of rainfall process, and defines a set-weighting method based on the Grey Wolf Optimization algorithm (IGWO) to weight and fuse the five extracted features as the similarity measure of rainfall process. The similarity search results of rainfall process images on the Jialing Basin illustrate that the method proposed in this paper can efficiently and accurately identify similar rainfall processes in historical scenes, thereby providing a new technical support for intelligent rainfall-flood forecasting. © 2024 IEEE.

关键词： Rain

来源：评论

学校读者我要写书评

暂无评论

DeSAM2: Detection-Driven SAM2 for Hyperspectral Object Tracking 14

DeSAM2: Detection-Driven SAM2 for Hyperspectral Object Track...

引用

14th Workshop on Hyperspectral Imaging and Signal Processing: Evolution in Remote Sensing, WHISPERS 2024

作者： He, Wangquan Ren, Qi Zhang, Hongtao Mo, Ziqian Li, Nanying Xu, Meng Jia, Sen College of Computer Science and Software Engineering Shenzhen University Shenzhen518060 China

ISBN: (纸本)9798331513139

Hyperspectral object tracking (HOT) captures subtle object features, enabling precise identification and tracking in complex backgrounds. However, the high-dimensional nature of data and rapid object changes in dynamic scenes make accurate and robust tracking challenging. This paper proposes a detection-driven Segment Anything Model 2 (SAM2) for HOT (named DeSAM2), which enhances tracking accuracy and robustness through the collaboration of initial prompt tracking and a detection-based self-prompt auxiliary mechanism. Specifically, The method first employs a key band selection strategy to ensure that the spectral difference between object and background is maximized in complex backgrounds. Then, initial prompt tracking is used to acquire initial position of the object, and this process uses SAM2 as a base scheme to maintain high accuracy tracking. Furthermore, to address missed and false detections in initial tracking results within complex scenes, a detection-based self-prompt auxiliary mechanism is designed to further refine and correct tracking, thereby enhancing the adaptability and continuity of the model in dynamic environments. Experimental results on the datasets provided by the HOT Challenge 2024 indicate that the proposed method demonstrates superior performance compared to other advanced trackers. © 2024 IEEE.

关键词： Object detection

来源：评论

学校读者我要写书评

暂无评论

VR-Recon: Visual Refiner for Fine-Grained Human Reconstruction 4

VR-Recon: Visual Refiner for Fine-Grained Human Reconstructi...

引用

4th IEEE International Conference on Digital Twins and Parallel Intelligence, DTPI 2024

作者： Xie, Hong Zhu, Zibin Zhang, Xiaoyan Shen Zhen University College of Computer Science & Software Engineering Shenzhen China

ISBN: (纸本)9798350349252

To reconstruct 3D human figures with fine-grained details from sparse views, this paper proposes a novel framework called VR-Recon, which is based on a viewpoint refiner. In this framework, we first decouple geometric features from multi-scale feature learning by integrating an MSG image encoder, aiming to extract high-quality multi-scale geometric features. Next, we employ a viewpoint refiner to accurately capture critical fine-grained geometric information in each independent view space, thereby avoiding the weighted average fusion of features and eliminating the influence of invalid information. This method is trained using a consistent cosine loss and a cascaded view-space geometric loss. The former encourages the MSG to accurately perceive fine geometric changes from the image, while the latter helps the viewpoint refiner discern critical information in different view spaces. Experimental results demonstrate that, compared to state-of-the-art methods, VR-Recon exhibits robust performance in capturing fine geometric details. © 2024 IEEE.

关键词： Feature extraction

来源：评论

学校读者我要写书评

暂无评论

Dual-Enhanced High-Order Self-Learning Tensor Singular Value Decomposition for Robust Principal Component Analysis

IEEE Transactions on Artificial Intelligence

引用

IEEE Transactions on Artificial Intelligence 2024年第7期5卷 3564-3578页

作者： Xu, Honghui Fang, Chuangjie Wang, Renfang Chen, Shengyong Zheng, Jianwei Zhejiang University of Technology College of Computer Science and Technology Hangzhou310023 China Zhejiang Wanli University College of Big Data and Software Engineering Ningbo315200 China Tianjin University of Technology College of Computer Sciences and Engineering Tianjin300384 China

Recently, tensor singular value decomposition (TSVD) within high-order (Ho) algebra framework has shed new light on tensor robust principal component analysis (TRPCA) problem. However, HoTSVD lacks flexibility in handling the hidden correlations along different modes of large-scale multidimensional data. Moreover, the utilization of fixed or data-independent transformations in HoTSVD may result in suboptimality. For a relief, we propose a dual-enhanced self-learning TSVD along all modes to address computational flaws and learn a lossless transformation that induces a lower average-rank tensor. Specifically, we multiply the learnable semiorthogonal matrices obtained through Tucker compression with the original tensor along all modes, thus obtaining a core tensor with more inherent low rankness. Building upon this foundation, a new TNN is introduced by generalizing HoTSVD to Mode-k TSVD, followed by the facilitation to the core tensor, achieving dual-enhancement. Moreover, a reweighting scheme is imposed on the Mode-k HoTSVD to learn the global low-rank correlation and provide an efficient numerical solution. Finally, an alternating direction method of multipliers (ADMM)-based algorithm is developed as a solver. Experimental results on several types of multidimensional visual data, including light field images (LFI) and color videos, demonstrate the superiority of the proposal over previous state-of-the-art methods. © 2020 IEEE.

关键词： Discrete Fourier transforms

来源：评论

学校读者我要写书评

暂无评论

Human Action Recognition in Multi-Level Convolutional Temporal Attention Network

Human Action Recognition in Multi-Level Convolutional Tempor...

引用

2025 IEEE International Conference on Acoustics, Speech, and Signal Processing, ICASSP 2025

作者： Huang, Qian Chen, ZhongQi Li, Chang Qian, WeiWen College of Computer Science and Software Engineering Hohai University NanJing China

ISBN: (纸本)9798350368741

Human Action Recognition (HAR) has widespread applications in areas such as human-computer interaction, elderly care, and home healthcare. However, current sensor-based HAR faces challenges of low fine-grained recognition performance and difficulty in distinguishing similar actions. To solve this problem, this paper proposes a model based on Multilevel Convolutional Time Series Attention Network (MCTSANet). By Multi-ResCNN to pay attention to different levels of features and using Time Series Attention (TSA) to pay attention to the more important data in the channel, so as to improve the ability of confusable action *** on three public datasets show that the proposed method outperforms state-of-the-art sensor-based HAR approaches. © 2025 IEEE.

关键词： Attention Mechanism Deep Learning Human Action Recognition Wearable Sensors

来源：评论

学校读者我要写书评

暂无评论

Environment-Tolerant Trust Opportunity Routing based on Reinforcement Learning for Internet of Underwater Things

引用

IEEE Transactions on Mobile Computing 2025年第07期24卷 6348-6360页

作者： He, Yu Han, Guangjie Hou, Yun Lin, Chuan Hohai University College of Information Science and Engineering Changzhou213200 China Hohai University Key Laboratory of Maritime Intelligent Network Information Technology Ministry of Education Changzhou213200 China Hohai University College of Computer Science and Software Engineering Nanjing211100 China Northeastern University Software College Shenyang110819 China

The Internet of Underwater Things (IoUT) has garnered significant interest due to its potential applications in monitoring underwater environments. However, the unique characteristics of acoustic communication, such as long propagation delays and high attenuation, present considerable obstacles for achieving efficient and dependable data transmission. Opportunistic routing is a crucial technique for enhancing packet delivery ratios by selecting a set of forwarding nodes and utilizing their cooperative forwarding to boost network throughput. Nevertheless, choosing an excessive number of forwarding nodes can lead to wasteful energy usage and extended communication delays. Moreover, the overlooked trustworthiness of forwarded nodes in most research works can undermine the effectiveness of opportunistic routing. Therefore, this study presents a novel trust opportunistic routing scheme that employs reinforcement learning to achieve resilience in constantly changing underwater settings. The combination of reinforcement learning and trust management enables the proposed opportunistic routing scheme to adapt to the unstable underwater environment and unknown malicious attacks. Initially, a method is introduced for measuring environmental fitness by considering multiple trust factors, including communication success rate, data reliability, and location dynamics. The proposed scheme then uses reinforcement learning to develop a reliable opportunistic routing method based on quantified state information. This component employs the obtained state to formulate action strategies and obtains reward values from environmental inputs. The reward update equation integrates these qualities to optimize the deployment of superior action strategies, finally achieving trust opportunistic routing for underwater data collection. Fundamental experimental results demonstrate that the proposed protocol performs exceptionally well in demanding underwater conditions, outperforming existing method

关键词： Reinforcement learning

来源：评论

学校读者我要写书评

暂无评论

Theoretical Analysis of an Adaptive Closeness Centrality-Based Algorithm for Dynamic Optimization of Transportation Networks

Theoretical Analysis of an Adaptive Closeness Centrality-Bas...

引用

2024 International Conference on engineering Management of Communication and Technology, EMCTECH 2024

作者： Mann, Michael Afeka - the Academic College of Engineering School of Software Engineering and Computer Science Tel Aviv Israel

ISBN: (数字)9798331507909

ISBN: (纸本)9798331507909

Purpose: This paper presents a theoretical analysis of the DynaTrans algorithm, a novel approach for dynamic optimization of urban transportation networks. Design/methodology/approach: We introduce an Adaptive Closeness Centrality (ACC) metric and the DynaTrans algorithm, providing formal proofs of correctness, convergence, and efficiency. The analysis employs graph theory, algorithmic complexity theory, and competitive analysis techniques. Findings: We prove that DynaTrans converges to a local optimal state in O(|V|/ϵ) iterations, with each iteration has limited computational and memory requirements. The algorithm achieves solutions within a factor of O(log|V|) from the global optimum, outperforming simple greedy approaches. Practical implications: DynaTrans offers a theoretically sound foundation for real-time traffic management systems, potentially improving urban mobility and reducing congestion. Originality/value: This work introduces a new paradigm for dynamic transportation network optimization, combining adaptive centrality measures with efficient graph algorithms. The rigorous theoretical analysis provides a solid basis for practical implementation and future research in urban traffic management. © 2024 IEEE.

关键词： Traffic congestion

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：