检索结果-内蒙古大学图书馆

International Conference on Acoustics, Speech, and Signal Processing (ICASSP)

作者： Xiang Zhang Qiang Zhu Tao Hu Qingsen Yan College of Information and Control Engineering Xi’an University of Architecture and Technology School of Computer Science Northwestern Polytechnical University

While recent progress in Multi-exposure HDR imaging is promising, the growing complexity of state-of-the-art (SOTA) methods poses challenges for their analysis and comparison. In this paper, we analyze the motivations and approaches behind previous SOTA works and introduce EiffHDR, an efficient Multi-exposure HDR imaging technique. In contrast to prior methods employing multiple branches spatial attention mechanisms, EiffHDR adopts a streamlined gating mechanism for information flow control at both spatial and channel levels, enabling implicit alignment. Subsequently, we process these features through proposed Efficient Merging Network, facilitating long-range correlations and multi-scale information perception, ultimately producing high-quality HDR images. Our experiments demonstrate that EiffHDR not only achieves outstanding performance but also significantly reduces computational complexity, making it a valuable contribution to the field.

关键词：

来源：评论

学校读者我要写书评

暂无评论

Real-Time Inference of Temporal Emotional and Behavioral Conditions of Learners in Problem Solving Within Mixed Reality Environments 1

引用

19th International Conference on Computing in Civil and Building Engineering, ICCCBE 2022

作者： Castets, Édouard Mutis, Ivan Agam, Gady Department of Mechanical Materials and Aerospace Engineering Illinois Institute of Technology ChicagoIL60616 United States Department of Civil Architecture and Environmental Engineering Illinois Institute of Technology ChicagoIL60616 United States Department of Computer Science Illinois Institute of Technology Chicago60616 United States

ISBN: (数字)9783031325151

ISBN: (纸本)9783031325144

Using new affordances of Machine Learning (ML) and Mixed Reality (MX) technology, the presented research looks at how learners’ decision-making in problem-solving can increase with the technology interaction. The approach empowers users for enhanced problem-solving outcomes by enabling them to perform adapted actions in an MX environment. For example, present an adapted user’s view of 3D design components in the immersive space for any action (locking and moving) in problem-solving (e.g., design interpretations). The aim is to incorporate ML to tailor the user’s ability to problem-solve complex 3D design components by capitalizing on the flexibility of MX immersive environments. The system uses a collection of physiological and pupillometry data to create real-time inferences of the user’s state (emotional and physiological conditions). This paper focuses on the first phase of developing the MX and ML system for the augmented technology-user interaction to allow inference of the data in real time. © 2023, The Author(s), under exclusive license to Springer Nature Switzerland AG.

关键词： Mixed reality

来源：评论

学校读者我要写书评

暂无评论

Future connected vehicles:Communications demands,privacy and cyber-security

引用

Communications in Transportation Research 2022年第1期2卷 39-41页

作者： Tomas Olovsson Tommy Svensson Jiaming Wu Department of Computer Science and Engineering Chalmers University of TechnologyGothenburgSE41296Sweden Department of Electrical Engineering Chalmers University of TechnologyGothenburgSE41296Sweden Department of Architecture and Civil Engineering Chalmers University of TechnologyGothenburgSE41296Sweden

*** With the advance of data collection,data processing,telecommunication and vehicular technologies,connected vehicles(CVs)have been emerging as a crucial branch of smart mobility(Olia et al.,2015;Li et al.,2021).Its basic idea is to realize real-time exchanges and processing of essential information,such as positions and destinations,among surrounding vehicles and infrastructures.

关键词： connected branch destination

来源：评论

学校读者我要写书评

暂无评论

A Design of Fetch Target Buffer Implemented on XiangShan Processor

A Design of Fetch Target Buffer Implemented on XiangShan Pro...

引用

2022 International Conference on Cloud Computing, Internet of Things, and computer Applications, CICA 2022

作者： Zou, Jiangrui Tang, Dan Cai, Ye Li, Zusong School of Computer and Software Shenzhen University Shenzhen China State Key Laboratory of Computer Architecture Institute of Computing Technology Chinese Academy of Sciences Beijing China Beijing Vcore Technology Co. Ltd. Beijing China

ISBN: (纸本)9781510656673

In the design of modern high-performance processors, branch prediction is a very important component, and good branch prediction can greatly improve the performance of the processor. And Branch Target Buffer (BTB) is an important part of branch prediction. BTB is used to store information about branch instructions. Different from storing each branch as a unit, Glenn et al. proposed a BTB that stores branch information in a unit of fetch block, which they call Fetch Target Buffer (FTB). This design can optimize the timing by limiting the branch prediction width. But they did not give specific implementation details. In this paper, we propose a specific FTB implementation, and it is used in the second version architecture of XiangShan high-performance processor. First, we give the fields in the FTB entry in this design and their corresponding functions. Most BTB entries need to store the following information: branch tag, branch type, target address. In addition to these fields in this paper, FTB will also store other information for managing entries and optimizing branch prediction performance. Secondly, we proposed the management strategy of FTB entries, when to add, modify, and replace. After all the instructions in a fetch block are committed, we need to decide whether this fetch block needs to be written to FTB, and which way to write, how to avoid the problem of multiple hits. Under the 14nm process, compared to the 1.8GHz of the first version of the architecture, the frequency of the new architecture using FTB can be increased to 2GHz. © 2022 SPIE.

关键词： architecture

来源：评论

学校读者我要写书评

暂无评论

Source Camera Identification Algorithm Based on Multi-Scale Feature Fusion

引用

computers, Materials & Continua 2024年第8期80卷 3047-3065页

作者： Jianfeng Lu Caijin Li Xiangye Huang Chen Cui Mahmoud Emam School of Computer Science and Technology Hangzhou Dianzi UniversityHangzhou310018China Shangyu Institute of Science and Engineering Hangzhou Dianzi UniversityShaoxing312300China Key Laboratory of Public Security Information Application Based on Big-Data Architecture Ministry of Public SecurityZhejiang Police CollegeHangzhou310000China Faculty of Artificial Intelligence Menoufia UniversityShebin El-Koom32511Egypt

The widespread availability of digital multimedia data has led to a new challenge in digital *** source camera identification algorithms usually rely on various traces in the capturing ***,these traces have become increasingly difficult to extract due to wide availability of various image processing *** Neural Networks(CNN)-based algorithms have demonstrated good discriminative capabilities for different brands and even different models of camera ***,their performances is not ideal in case of distinguishing between individual devices of the same model,because cameras of the same model typically use the same optical lens,image sensor,and image processing algorithms,that result in minimal overall *** this paper,we propose a camera forensics algorithm based on multi-scale feature fusion to address these *** proposed algorithm extracts different local features from feature maps of different scales and then fuses them to obtain a comprehensive feature *** representation is then fed into a subsequent camera fingerprint classification *** upon the Swin-T network,we utilize Transformer Blocks and Graph Convolutional Network(GCN)modules to fuse multi-scale features from different stages of the backbone ***,we conduct experiments on established datasets to demonstrate the feasibility and effectiveness of the proposed approach.

关键词： Source camera identification camera forensics convolutional neural network feature fusion transformer block graph convolutional network

来源：评论

学校读者我要写书评

暂无评论

Efficient Content Reconstruction for High Dynamic Range Imaging

Efficient Content Reconstruction for High Dynamic Range Imag...

引用

International Conference on Acoustics, Speech, and Signal Processing (ICASSP)

作者： Xiang Zhang Tao Hu Jiashuang He Qingsen Yan College of Information and Control Engineering Xi’an University of Architecture and Technology School of Computer Science Northwestern Polytechnical University

High Dynamic Range (HDR) images can be reconstructed from multiple Low Dynamic Range (LDR) images using existing deep neural network (DNN) techniques. Despite notable advancements, DNN-based methods still exhibit ghosting artifacts when handling LDR images with saturation and significant motion. Recent Diffusion models (DMs) have been introduced in HDR imaging, showcasing promising performance, especially in achieving visually perceptible results. However, DMs typically require numerous inference iterations to recover the clean image from Gaussian noise, demanding substantial computational resources. Additionally, DM only learns a probability distribution of the added noise in each step but neglects image space constraints on HDR images, limiting distortion-based metrics. To tackle these challenges, we propose an efficient network that integrates DM modules into existing regression-based models, providing reliable content reconstruction for HDR while avoiding limitations in distortion-based metrics.

关键词：

来源：评论

学校读者我要写书评

暂无评论

Attributed Network Clustering through Mutual Affinity Propagated Non-negative Matrix Factorization

引用

SN computer Science 2025年第5期6卷 1-17页

作者： Pahari, Saikat Dey, Paramita Department of Computer Science Engineering OmDayal College of Engineering and Architecture West Bengal Howrah 711316 India Department of Information Technology Government College of Engineering and Ceramic Technology West Bengal Kolkata 700010 India

Attributed Network Clustering (ANC) has garnered significant attention in research for identifying communities within a complex network like a social, biological, or information network. Generally, such networks are represented as a graph where the nodes denote entities and edges are relationships between two adjacent nodes. In real-world networks, nodes are associated with attributes to express their properties. In this context, a community is defined as a group of nodes that interact more frequently with each other than with nodes outside their group. ANC involves creating a mapping function that converts nodes and associated attributes into lower-dimensional vector representations. Once transformed into these vectors, various machine learning algorithms are used to cluster the nodes and reveal the underlying communities. However, most ANC methods confront two major challenges. First, they assimilate node attribute information from the local neighborhood, as a result, they can’t explore the mutual affinity between node and attribute. Second, they do not consider the values of the community membership of the nodes at the time of clustering. This leads to poor performance, as the embedding method and clustering methods are independent. To address this, this article proposes a method named Mutual Affinity Propagation-based Nonnegative Matrix Factorization (MAP-NMF). The proposed method works in two phases. In the first phase, an attribute-augmented transition probability matrix is constructed to capture the mutual affinity between the node and attribute and obtain node representation using singular value decomposition of the same. The second phase produces community membership from the obtained latent representation through an NMF-based joint optimization. Moreover, a novel updating rule is proposed for the joint optimization method. Empirical study on both real-world and artificial networks proves the superiority of our proposed algorithm. © The Author(s), under exc

关键词： Attribute augmented graph Attributed network embedding Joint optimization Non-negative matrix factorization

来源：评论

学校读者我要写书评

暂无评论

An alert-situation text data augmentation method based on MLM

引用

High technology Letters 2024年第4期30卷 389-396页

作者： DING Weijie MAO Tingyun CHEN Lili ZHOU Mingwei YUAN Ying HU Wentao Department of Computer and Information Security Zhejiang Police CollegeHangzhou 310053P.R.China Key Laboratory of Public Security Information Application Based on Big-Data Architecture Ministry of Public SecurityHangzhou 310053P.R.China Zhejiang Dahua Technology Co. LtdHangzhou 310053P.R.China

The performance of deep learning models is heavily reliant on the quality and quantity of train-ing *** training data will lead to ***,in the task of alert-situation text classification,it is usually difficult to obtain a large amount of training *** paper proposes a text data augmentation method based on masked language model(MLM),aiming to enhance the generalization capability of deep learning models by expanding the training *** method em-ploys a Mask strategy to randomly conceal words in the text,effectively leveraging contextual infor-mation to predict and replace masked words based on MLM,thereby generating new training *** Mask strategies of character level,word level and N-gram are designed,and the performance of each Mask strategy under different Mask ratios is analyzed and *** experimental results show that the performance of the word-level Mask strategy is better than the traditional data augmen-tation method.

关键词： deep learning text data augmentation masked language model(MLM) alert-sit-uation text classification

来源：评论

学校读者我要写书评

暂无评论

Transferable Physical Adversarial Patch Attack for Remote Sensing Object Detection

Transferable Physical Adversarial Patch Attack for Remote Se...

引用

IEEE International Symposium on Geoscience and Remote Sensing (IGARSS)

作者： Di Wang Wenxuan Zhu Ke Li Xiao Gao Pengfei Yang School of Computer Science and Technology Xidian University Xi’an China College of Information and Control Engineering Xi’an University of Architecture and Technology Xi’an China

ISBN: (数字)9798350360325

ISBN: (纸本)9798350360332

Deep neural networks (DNNs) have been widely used in remote sensing but demonstrated to be vulnerable with adversarial examples. By adding elaborately designed perturbations on the clean images, DNNs may output wrong prediction. Research on adversarial attack contributes to the study of model robustness. However, previous methods mainly focus on white-box scenario or digital domain for classification tasks, while the vulnerability of remote sensing detectors has not been fully explored. Aiming at attacking black-box remote sensing detectors in physical domain, we propose to generate a transferable physical adversarial patch (TPAP) as the perturbations. Specifically, the initial patch is optimized by a U-Net and modified by the plane mask and position mask before applied to the clean image. By attacking a surrogate model, TPAP can be transferred to the target model. Abundant experimental results validate the attack ability of TPAP and evaluate the robustness of current one-stage detectors.

关键词： YOLO Perturbation methods Closed box Detectors Robustness Sensors Task analysis

来源：评论

学校读者我要写书评

暂无评论

GraphTensor: Comprehensive GNN-Acceleration Framework for Efficient Parallel Processing of Massive Datasets

GraphTensor: Comprehensive GNN-Acceleration Framework for Ef...

引用

International Symposium on Parallel and Distributed Processing (IPDPS)

作者： Junhyeok Jang Miryeong Kwon Donghyun Gouk Hanyeoreum Bae Myoungsoo Jung Computer Architecture and Memory Systems Laboratory Korea Advanced Institute of Science and Technology (KAIST)

We present GraphTensor, a comprehensive open-source framework that supports efficient parallel neural network processing on large graphs. GraphTensor offers a set of easy-to-use programming primitives that appreciate both graph and neural network execution behaviors from the beginning (graph sampling) to the end (dense data processing). Our framework runs diverse graph neural network (GNN) models in a destination-centric, feature-wise manner, which can significantly shorten training execution times in a GPU. In addition, GraphTensor rearranges multiple GNN kernels based on their system hyperparameters in a self-governing manner, thereby reducing the processing dimensionality and the latencies further. From the end-to-end execution viewpoint, GraphTensor significantly shortens the service-level GNN latency by applying pipeline parallelism for efficient graph dataset preprocessing. Our evaluation shows that GraphTensor exhibits 1.4× better training performance than emerging GNN frameworks under the execution of large-scale, real-world graph workloads. For the end-to-end services, GraphTensor reduces training latencies of an advanced version of the GNN frameworks (optimized for multi-threaded graph sampling) by 2.4×, on average.

关键词：

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：