检索结果-内蒙古大学图书馆

Neural Rendering and Flow-assisted Unsupervised Multi-view Stereo for Real-time Monocular Tracking and Scene Perception

引用

IEEE Transactions on Automation Science and engineering 2025年

作者： Tong, Kevin W. Cai, Yandong Jie, Yu-Wen Duan, Ya Hou, Yuhong Wu, Edmond Q. Nanjing University of Posts and Telecommunications College of Automation College of Artificial Intelligence Nanjing210023 China Ministry of Education Key Laboratory of Social Computing and Cognitive Intelligence Dalian University of Technology Dalian116024 China Shanxi University School of Automation and Software Engineering Shanxi 237016 China Flight Test Technology and Engineering Center of the Chinese Flight Test Establishment Xian710089 China China Civil Aviation Shanghai Aircraft Airworthiness Certification Center Shanghai200335 China Shanghai Jiao Tong University Department of Computer Science and Engineering Shanghai200240 China

The existing camera tracking and perception methods mainly rely on sparse SLAM, which limits the dense perception ability of the scene and affects the reliability of auxiliary decision-making. Different from this, this work proposes a real-time tracking and unsupervised dense sensing framework. Firstly, the dense depth value of the scene is predicted by unsupervised multi-view stereo to remove the dependence on labeled data. Then, the quality of synthetic pseudo-reference image is quantified according to the predicted depth map and used as a weighted guidance to train the unsupervised model, thus reducing the ambiguity of feature matching in areas such as specular reflection. Moreover, the sparse optical flow of the keyframes is solved by real-time and robust ORB feature matching operator, which assists the high-precision training of unsupervised depth inference model. To increase the prediction accuracy of occluded area, a novel rendering consistency loss via neural radiance fields is designed to constrain the geometric characteristics of object surface. Finally, dense direct image alignment is performed from a global model to improve the tracking robustness, which is incrementally constructed from dense depth prediction. Extensive experiments on synthetic datasets and real datasets validate the effectiveness and practicability of the proposed work, which is an effective supplement to the existing SLAM work. © 2025 IEEE.

关键词： Stereo image processing

来源：评论

学校读者我要写书评

暂无评论

Motor imagery enhances performance beyond the imagined action

引用

Proceedings of the National Academy of Sciences of the United States of America 2025年第20期122卷 e2423642122页

作者： Gippert, Magdalena Shih, Pei-Cheng Heed, Tobias Howard, Ian S. Jamshidi Idaji, Mina Villringer, Arno Sehm, Bernhard Nikulin, Vadim V. Department of Neurology Max Planck Institute for Human Cognitive and Brain Sciences Leipzig 04103 Germany Sony Computer Science Laboratories 141-0022 Tokyo Japan Department of Psychology and Centre for Cognitive Neuroscience University of Salzburg 5020 Salzburg Austria School of Engineering Computing and Mathematics Faculty of Science and Engineering University of Plymouth Plymouth PL4 8AA United Kingdom Machine Learning Group Berlin Institute for the Foundations of Learning and Data 10587 Berlin Germany Machine Learning Group Institute of Software Engineering and Theoretical Computer Science Electrical Engineering and Computer Science Faculty Technical University Berlin 10587 Berlin Germany Department of Neurology Martin Luther University of Halle-Wittenberg Halle (Saale) 06120 Germany

Motor imagery is frequently utilized to improve the performance of specific target movements in sports and rehabilitation. In this study, we show that motor imagery can facilitate learning of not only the imagined target movements but also sequentially linked overt movements. Hybrid sequences comprising imagined and physically executed segments allowed participants to learn specific movement characteristics of the executed segments when they were consistently associated with specific imagined segments. Electrophysiological recordings revealed that the degree of event-related synchronization in the alpha and beta bands during a basic motor imagery task was correlated with imagery-evoked motor learning. Thus, both behavioral and neural evidence indicate that motor imagery's benefits extend beyond the imagined movements, improving performance in linked overt movements. This provides decisive evidence for the functional equivalence of imagined and overt movements and suggests applications for imagery in sports and rehabilitation.

关键词： adaptation force field motor imagery motor learning motor sequence

来源：评论

学校读者我要写书评

暂无评论

The Robustness of Spiking Neural Networks in Communication and its Application towards Network Efficiency in Federated Learning

The Robustness of Spiking Neural Networks in Communication a...

引用

2024 IEEE International Performance, computing, and Communications Conference, IPCCC 2024

作者： Nguyen, Manh V. Zhao, Liang Deng, Bobin Severa, William Xu, Honghui Wu, Shaoen Kennesaw State University College of Computing and Software Engineering GA United States Sandia National Laboratories Department of Cognitive & Emerging Computing NM United States

ISBN: (纸本)9798350367942

Spiking Neural Networks (SNNs) have recently gained significant interest in on-chip learning in embedded devices and emerged as an energy-efficient alternative to conventional Artificial Neural Networks (ANNs). However, to extend SNNs to a Federated Learning (FL) setting involving collaborative model training, the communication between the local devices and the remote server remains the bottleneck, which is often restricted and costly. In this paper, we first explore the inherent robustness of SNNs under noisy communication in FL. Building upon this foundation, we propose a novel Federated Learning with Top-κ Sparsification (FLTS) algorithm to reduce the bandwidth usage for FL training. We discover that the proposed scheme with SNNs allows more bandwidth savings compared to ANNs without impacting the model's accuracy. Additionally, the number of parameters to be communicated can be reduced to as low as 6% of the size of the original model. We further improve the communication efficiency by enabling dynamic parameter compression during model training. Extensive experiment results demonstrate that our proposed algorithms significantly outperform the baselines in terms of communication cost and model accuracy and are promising for practical network-efficient FL with SNNs. © 2024 IEEE.

关键词： Federated learning

来源：评论

学校读者我要写书评

暂无评论

A novel fuzzy multimodal information fusion technology for human biometric traits identification

A novel fuzzy multimodal information fusion technology for h...

引用

10th IEEE International Conference on cognitive Informatics and cognitive computing, ICCI*CC 2011

作者： Monwar, Md. Maruf Gavrilova, Marina Wang, Yingxu Computer Science University of Calgary Canada Department of Computer Science University of Calgary Canada Department of Cognitive Computing and Software Engineering University of Calgary Canada

ISBN: (纸本)9781457716966

In recent years, biometric based security systems achieved more attention due to continuous terrorism threats around the world. However, a security system comprised of a single form of biometric information cannot fulfill users' expectations and may suffer from noisy sensor data, intra and inter class variations and continuous spoof attacks. To overcome some of these problems, multimodal biometric systems with multiple physiological, behavioral, and soft biometric information are becoming more popular due to increased recognition accuracy. In order to take full advantage of the multimodal approaches, one of the main issues is to implement the fusion mechanism for different biometric information. In this research, we utilize the physiological attributes (face, ear and iris) along with soft biometric information (gender, ethnicity and eye color). A fuzzy fusion mechanism for robust and reliable multimodal biometric based security systems is developed. The proposed fuzzy fusion scheme adopts rank, match score and soft biometrics information as the input and produces final identification decision via a fuzzy rule-based inference system. The experimental results show that the fuzzy fusion method can provide us faster, higher and more reliable recognition performance than conventional unimodal methods. The system can be effectively used at any security critical applications. © 2011 IEEE.

关键词： Information fusion

来源：评论

学校读者我要写书评

暂无评论

Landscape image composition analysis based on image processing and curve fitting

Landscape image composition analysis based on image processi...

引用

2013 5th International Conference on Computational and Information Sciences, ICCIS 2013

作者： Yang, Hengxiang Wang, Xiujin Bai, Lifang School of Computer Science and Technology Tianjin Key Laboratory of Cognitive Computing and Application Tianjin University Tianjin China Department of Software Engineering Tianjin University Tianjin China

ISBN: (纸本)9780769550046

The composition of the landscape image can convey the emotion of cameramen. Proposed in this article is one novel technique for landscape image composition analysis, which could lay the foundation for the further analysis of image emotion. This method is mainly aimed at analyzing the composition of landscape image with no foregrounds compos. Firstly, the image segmentation method based on edge detection and k-means clustering algorithm is used to extract the contour of each mountain. Secondly, curve fitting method is used to determine the curve type of each mountain in the landscape image. Finally, the normal distribution model is constructed to assign a weight to each pixel of the image from its location information, We can determine the composition of the whole image based on total weight comparison of each curve type which combined the contour information and location information. In this paper, we differentiated the curve from five types: Level-style, Inclined-style, Vertical-style, C-style and W-style. Corresponding, the composition of the whole image is divided into these five types. As a result, we can determine the composition type of landscape image. Experimental results demonstrate that this method is effective and meaningful in practice. © 2013 IEEE.

关键词： Normal distribution

来源：评论

学校读者我要写书评

暂无评论

The third competition on knowledge engineering for planning and scheduling

引用

AI Magazine 2010年第1期31卷 95-98页

作者： Barták, Roman Fratini, Simone McCluskey, Lee Department of Computer Science Faculty of Mathematics and Physics Charles University Prague Czech Republic Institute of Cognitive Science and Technology Italian National Research Council Italy Software Technology University of Huddersfield United Kingdom University's School of Computing and Engineering United Kingdom

We report on the staging of the third competition on knowledge engineering for AI planning and scheduling systems, held during ICAPS-09 at Thessaloniki, Greece, in September 2009. We give an overview of how the competition has developed since its first run in 2005 and its relationship with the AI planning field. This run of the competition focused on translators that, when input with some formal description in an application-area-specific language, output solver-ready domain models. Despite a fairly narrow focus within knowledge engineering, seven teams took part in what turned out to be a very interesting and successful competition. Copyright © 2010, Association for the Advancement of Artificial Intelligence. All rights reserved.

关键词： Scheduling

来源：评论

学校读者我要写书评

暂无评论

The Robustness of Spiking Neural Networks in Communication and its Application towards Network Efficiency in Federated Learning

The Robustness of Spiking Neural Networks in Communication a...

引用

IEEE International Conference on Performance, computing and Communications (IPCCC)

作者： Manh V. Nguyen Liang Zhao Bobin Deng William Severa Honghui Xu Shaoen Wu College of Computing and Software Engineering Kennesaw State University Georgia USA Department of Cognitive & Emerging Computing Sandia National Laboratories New Mexico USA

ISBN: (数字)9798350367942

ISBN: (纸本)9798350367959

Spiking Neural Networks (SNNs) have recently gained significant interest in on-chip learning in embedded devices and emerged as an energy-efficient alternative to conventional Artificial Neural Networks (ANNs). However, to extend SNNs to a Federated Learning (FL) setting involving collaborative model training, the communication between the local devices and the remote server remains the bottleneck, which is often restricted and costly. In this paper, we first explore the inherent robustness of SNNs under noisy communication in FL. Building upon this foundation, we propose a novel Federated Learning with Top-κ Sparsification (FLTS) algorithm to reduce the bandwidth usage for FL training. We discover that the proposed scheme with SNNs allows more bandwidth savings compared to ANNs without impacting the model’s accuracy. Additionally, the number of parameters to be communicated can be reduced to as low as 6% of the size of the original model. We further improve the communication efficiency by enabling dynamic parameter compression during model training. Extensive experiment results demonstrate that our proposed algorithms significantly outperform the baselines in terms of communication cost and model accuracy and are promising for practical network-efficient FL with SNNs.

关键词： Training Accuracy Costs Federated learning Noise Spiking neural networks Bandwidth Robustness System-on-chip Noise measurement

来源：评论

学校读者我要写书评

暂无评论

The Robustness of Spiking Neural Networks in Communication and its Application towards Network Efficiency in Federated Learning

arXiv

引用

arXiv 2024年

作者： Nguyen, Manh V. Zhao, Liang Deng, Bobin Severa, William Xu, Honghui Wu, Shaoen College of Computing and Software Engineering Kennesaw State University Georgia United States Department of Cognitive & Emerging Computing Sandia National Laboratories NM United States

Spiking Neural Networks (SNNs) have recently gained significant interest in on-chip learning in embedded devices and emerged as an energy-efficient alternative to conventional Artificial Neural Networks (ANNs). However, to extend SNNs to a Federated Learning (FL) setting involving collaborative model training, the communication between the local devices and the remote server remains the bottleneck, which is often restricted and costly. In this paper, we first explore the inherent robustness of SNNs under noisy communication in FL. Building upon this foundation, we propose a novel Federated Learning with Top-κ Sparsification (FLTS) algorithm to reduce the bandwidth usage for FL training. We discover that the proposed scheme with SNNs allows more bandwidth savings compared to ANNs without impacting the model’s accuracy. Additionally, the number of parameters to be communicated can be reduced to as low as 6% of the size of the original model. We further improve the communication efficiency by enabling dynamic parameter compression during model training. Extensive experiment results demonstrate that our proposed algorithms significantly outperform the baselines in terms of communication cost and model accuracy and are promising for practical network-efficient FL with SNNs. Copyright © 2024, The Authors. All rights reserved.

关键词： Federated learning

来源：评论

学校读者我要写书评

暂无评论

cognitive models of causal inferences using causation networks

Cognitive models of causal inferences using causation networ...

引用

IEEE International Conference on cognitive Informatics

作者： Yingxu Wang International Institute of Cognitive Informatics and Cognitive Computing IICICC Theoretical and Empirical Software Engineering Research Centre TESERC Department of Electrical and Computer Engineering Schulich School of Engineering University of Calgary Calgary AB Canada

Causal inference is one of the central capabilities of the natural intelligence that plays a crucial role in thought, perception, reasoning, and problem solving. This paper presents a set of cognitive models for causation analyses and causal inferences. The taxonomy and mathematical models of causations are created. The framework and properties of causal inferences are elaborated. Methodologies for uncertain causal inferences are discussed. The formalization of causal inference methodologies enables machines to mimic complex human reasoning mechanisms in cognitive informatics, cognitive computing, and computational intelligence.

关键词： Uncertainty Cognition Computational intelligence cognitive informatics Humans Analytical models Context

来源：评论

学校读者我要写书评

暂无评论

cognitive computing and World Wide Wisdom (WWW+)

Cognitive computing and World Wide Wisdom (WWW+)

引用

IEEE International Conference on cognitive Informatics

It is recognized that the key theoretical and technical problems toward the next generation internet are not only a speed issue, but also a more fundamental issue of the increasingly demands for the sharing of computational intelligent capabilities. According to cognitive informatics, the cognitive information that humans acquire, process, retain, and share can be classified into four profound forms known as knowledge, experience, skill, and wisdom. Among them, wisdom is the most advanced cognitive objects, which is a form of natural intelligence of humans that transfers a query or instruction into an action or behavior based on a well developed reasoning and judgment. However, the current Internet is still remains as an information network. Towards the development of next generation Internet as a wisdom network, the World Wide Wisdom (WWW+) network infrastructures and technologies are yet to be sough t on the basis of cognitive informatics and cognitive computers. In WWW+, each node is a cognitive computer (CC), which is a form of autonomous and intelligent computers that think, perceive, and learn. CC enables the simulation of machinable thought such as computational inferences, reasoning, and causality analyses by autonomous inferences and perceptions mimicking the mechanisms of the brain [3, 15]. The cognitive learning engine of a CC is an autonomous learning system that enables machines learn in natural languages and symbolic notations. The cognitive search engine of a CC is a machine-learning-based search system that results in cognitive knowledge acquisitions and manipulations. On the basis of the development of CCs, the next generation internet, WWW+, will be developed as a world-wide intelligent network for knowledge processing, autonomous learning, and machine-supported problem solving. The theoretical foundations for WWW+ and cognitive computing are cognitive informatics, with underpinning contemporary denotational mathematics, such as concept algebra, sy

关键词： cognitive informatics Algebra Computational intelligence World Wide Web Computers software engineering software

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：