检索结果-内蒙古大学图书馆

arXiv 2025年

作者： Zhang, Wei Ma, Mengting Jiang, Yizhen Lian, Rongrong Wu, Zhenkai Cui, Kangning Ma, Xiaowen School of Software Technology Zhejiang University Hangzhou310027 China Innovation Center of Yangtze River Delta Zhejiang University Zhejiang Jiaxing314103 China School of Computer Science and Technology Zhejiang University Hangzhou310027 China Department of Mathematics City University of Hong Kong Hong Kong Noah’s Ark Lab Huawei Shanghai201206 China

Compared with natural images, remote sensing images (RSIs) have the unique characteristic. i.e., larger intraclass variance, which makes semantic segmentation for remote sensing images more challenging. Moreover, existing semantic segmentation models for remote sensing images usually employ a vanilla softmax classifier, which has three drawbacks: (1) non-direct supervision for the pixel representations during training;(2) inadequate modeling ability of parametric softmax classifiers under large intraclass variance;and (3) opaque process of classification decision. In this paper, we propose a novel classifier (called CenterSeg) customized for RSI semantic segmentation, which solves the abovementioned problems with multiple prototypes, direct supervision under Grassmann manifold, and interpretability strategy. Specifically, for each class, our CenterSeg obtains local class centers by aggregating corresponding pixel features based on ground-truth masks, and generates multiple prototypes through hard attention assignment and momentum updating. In addition, we introduce the Grassmann manifold and constrain the joint embedding space of pixel features and prototypes based on two additional regularization terms. Especially, during the inference, CenterSeg can further provide interpretability to the model by restricting the prototype as a sample of the training set. Experimental results on three remote sensing segmentation datasets validate the effectiveness of the model. Besides the superior performance, CenterSeg has the advantages of simplicity, lightweight, compatibility, and interpretability. Code is available at https://***/xwmaxwma/rssegmentation. © 2025, CC BY.

关键词： Semantic Segmentation

来源：评论

学校读者我要写书评

暂无评论

SYNTACTIC ERROR REPAIR USING REPAIR PATTERNS

引用

INFORMATION PROCESSING LETTERS 1993年第4期47卷 189-196页

作者： YUN, IS CHOE, KM HAN, T Programming Languages Laboratory Department of Computer Science Korea Advanced Institute of Science and Technology 373-1 Kusong-Dong Yusung-Cu Taejon 305-701 South Korea

A syntactic error repair model is proposed, and is defined as a partial function from strings to sentences. The replacement of a substring of a string with a substring of a sentence is described by a repair pattern, which is roughly a pair of strings of grammar symbols. The model can be efficient with some restriction on repair patterns. An LR-based implementation of the model is discussed.

关键词： COMPILERS SYNTACTIC ERROR RECOVERY SYNTACTIC ERROR REPAIR MODEL REPAIR PATTERN

来源：评论

学校读者我要写书评

暂无评论

Enhanced Multi-Scale Cross-Attention for Person Image Generation

arXiv

引用

arXiv 2025年

作者： Tang, Hao Shao, Ling Sebe, Nicu Van Gool, Luc The School of Computer Science Peking University Beijing100871 China Terminus AI Lab University of Chinese Academy of Sciences Beijing100049 China University of Trento Trento38123 Italy The Department of Information Technology and Electrical Engineering ETH Zurich Switzerland The Department of Electrical Engineering KU Leuven Belgium INSAIT Sofia Un. Bulgaria

In this paper, we propose a novel cross-attention-based generative adversarial network (GAN) for the challenging person image generation task. Cross-attention is a novel and intuitive multi-modal fusion method in which an attention/correlation matrix is calculated between two feature maps of different modalities. Specifically, we propose the novel XingGAN (or CrossingGAN), which consists of two generation branches that capture the person’s appearance and shape, respectively. Moreover, we propose two novel cross-attention blocks to effectively transfer and update the person’s shape and appearance embeddings for mutual improvement. This has not been considered by any other existing GAN-based image generation work. To further learn the long-range correlations between different person poses at different scales and sub-regions, we propose two novel multi-scale cross-attention blocks. To tackle the issue of independent correlation computations within the cross-attention mechanism leading to noisy and ambiguous attention weights, which hinder performance improvements, we propose a module called enhanced attention (EA). Lastly, we introduce a novel densely connected co-attention module to fuse appearance and shape features at different stages effectively. Extensive experiments on two public datasets demonstrate that the proposed method outperforms current GAN-based methods and performs on par with diffusion-based methods. However, our method is significantly faster than diffusion-based methods in both training and inference. Copyright © 2025, The Authors. All rights reserved.

关键词： Generative adversarial networks

来源：评论

学校读者我要写书评

暂无评论

Middleware for Wireless Sensor Networks:A Survey

引用

Journal of computer science & technology 2008年第3期23卷 305-326页

作者：王苗苗曹建农李京 Sajal K.Dasi Department of Computer Science The University of Science and Technology of China Internets and Mobile Computing Lab Department of ComputingThe Hong Kong Polytechnic University Department of Computer Science and Engineering The University of Texas at Arlington

Wireless Sensor Networks （WSNs） have found more and more applications in a variety of pervasive computing environments. However, how to support the development, maintenance, deployment and execution of applications over WSNs remains to be a nontrivial and challenging task, mainly because of the gap between the high level requirements from pervasive computing applications and the underlying operation of WSNs. Middleware for WSN can help bridge the gap and remove impediments. In recent years, research has been carried out on WSN middleware from different aspects and for different purposes. In this paper, we provide a comprehensive review of the existing work on WSN middleware, seeking for a better understanding of the current issues and future directions in this field. We propose a reference framework to analyze the functionalities of WSN middleware in terms of the system abstractions and the services provided. We review the approaches and techniques for implementing the services. On the basis of the analysis and by using a feature tree, we provide taxonomy of the features of WSN middleware and their relationships, and use the taxonomy to classify and evaluate existing work. We also discuss open problems in this important area of research.

关键词： middleware wireless sensor network pervasive computing

来源：评论

学校读者我要写书评

暂无评论

YET ANOTHER EFFICIENT BACKWARD EXECUTION ALGORITHM IN THE AND OR PROCESS MODEL

引用

INFORMATION PROCESSING LETTERS 1991年第4期40卷 201-211页

作者： KIM, DH CHOE, KM Programming Languages Laboratory Department of Computer Science Korea Advanced Institute of Science and Technology 371-1 Kusong-Dong Yusang-Gu Taejon 305-701 South Korea

An efficient backward execution algorithm in the AND/OR Process Model for parallel evaluation of logic programs is proposed. The efficiency of the algorithm is achieved by means of information acquired during execution of clauses. The algorithm is considered to be efficient in the sense that it issues fewer number of cancel messages and avoids unnecessary resetting operations. Furthermore, it performs independent redoing and resetting more concurrently than other related works.

关键词： APPLICATIVE (LOGIC) programming PARALLEL EVALUATION OF LOGIC PROGRAMS AND OR PROCESS MODEL AND PARALLELISM BACKWARD EXECUTION ALGORITHM MULTIPLE FAILURE SELECTIVE RESETTING

来源：评论

学校读者我要写书评

暂无评论

Multiple-Instance Learning with Instance Selection via Constructive Covering Algorithm

引用

Tsinghua science and technology 2014年第3期19卷 285-292页

作者： Yanping Zhang Heng Zhang Huazhen Wei Jie Tang Shu Zhao the Department of Computer Science and Technology and Key Lab of Intelligent Computing and Signal Processing Anhui University the Department of Computer Science and Technology and Key Lab of Intelligent Computing and Signal Processing Anhui University1Hefei 230601 China the Department of Computer Science and Technology Tsinghua University

Multiple-Instance Learning （MIL） is used to predict the unlabeled bags＇ label by learning the labeled positive training bags and negative training *** bag is made up of several unlabeled instances.A bag is labeled positive if at least one of its instances is positive,otherwise *** multiple-instance learning methods with instance selection ignore the representative degree of the selected *** example,if an instance has many similar instances with the same label around it,the instance should be more representative than *** on this idea,in this paper,a multiple-instance learning with instance selection via constructive covering algorithm （MilCa） is *** MilCa,we firstly use maximal Hausdorff to select some initial positive instances from positive bags,then use a Constructive Covering Algorithm （CCA） to restructure the structure of the original instances of negative *** an inverse testing process is employed to exclude the false positive instances from positive bags and to select the high representative degree instances ordered by the number of covered instances from training ***,a similarity measure function is used to convert the training bag into a single sample and CCA is again used to classification for the converted *** results on synthetic data and standard benchmark datasets demonstrate that MilCa can decrease the number of the selected instances and it is competitive with the state-of-the-art MIL algorithms.

关键词： multiple-instance learning instance selection constructive covering algorithm maximal Hausdorff

来源：评论

学校读者我要写书评

暂无评论

Underwater Perception and Autonomous Operation for Underwater Vehicle-Manipulator Systems: Methodology and Experiment

引用

IEEE/ASME Transactions on Mechatronics 2025年

作者： Jiang, Chaicheng Xiong, Xinyang Xiang, Xianbo Zeng, Zhigang Zhang, Fumin Huazhong University of Science and Technology School of Artificial Intelligence and Automation Wuhan 430074 China Huazhong University of Science and Technology School of Naval Architecture and Ocean Engineering Wuhan 430074 China Huazhong University of Science and Technology School of Naval Architecture and Ocean Engineering School of Artificial Intelligence and Automation State Key Lab of Intelligent Manufacturing Equipment and Technology Wuhan 430074 China Huazhong University of Science and Technology School of Artificial Intelligence and Automation Key Laboratory of Image Processing and Intelligent Control The Education Ministry of China Wuhan 430074 China Hong Kong University of Science and Technology Department of Electronic and Computer Engineering Clear Water Bay Hong Kong Hong Kong

Camera-equipped underwater vehicle-manipulator systems (UVMS) can achieve autonomous underwater operations through visual servoing. However, current methods excessively rely on auxiliary means for target's visual information and do not sufficiently consider the target's pose when designing visual servoing schemes. As a result, UVMS can only perform dynamic positioning or grasp simple-shaped targets like sea urchins. In this article, we are devoted to making the UVMS autonomously complete more complex underwater operations, such as valve switching and pipe handling. First, an underwater target perception system is proposed to address the issue of discontinuous target position estimation caused by various interferences. Second, a deep neural network-based grasping mechanism controller is proposed to enable the manipulator to grasp targets in various initial poses using the target image. Third, a decoupled visual servoing scheme is proposed for the motion control of the UVMS, and in collaboration with a designed task supervisor to ensure that the UVMS can complete the underwater operation autonomously. To validate the effectiveness of the proposed methods, underwater experiments are conducted based on a UVMS prototype. The UVMS succeeds to autonomously switch different types of valves and handle the pipe. Experiments conducted under dynamic currents validated the reliability and robustness of the proposed methods. © 1996-2012 IEEE.

关键词： Autonomous underwater operations deep neural network underwater vehicle-manipulator systems (UVMS) visual perception visual servoing

来源：评论

学校读者我要写书评

暂无评论

Priority-Based Routing Resource Assignment Considering Crosstalk

引用

Journal of computer science & technology 2006年第6期21卷 913-921页

作者：蔡懿慈刘斌熊焰周强洪先龙 Department of Computer Science University of Science and Technology of China Hefei 230027 P.R. China EDA Lab Department of Computer Science and Technology Tsinghua University Beijing 100084 P.R. China

Crosstalk has become one of the most critical concerns in very deep sub-micron era. This paper deals with the problem of crosstalk mitigation at both methodological and algorithmic levels. Noting that intermediate operations between global routing and detailed routing are very effective in crosstalk estimation and reduction, the authors propose to incorporate several intermediate steps that are separated in traditional design flow into an integrated routing resource assignment stage, so that the operations could easily cooperate to fully exert their power on crosstalk reduction. An efficient priority-based heuristic algorithm is developed, which works slice by slice. Crosstalk avoidance, and ,nany other aspects that are critical in routing practice including congestion, vias, layer preference, etc., are taken into account. A track reservation strategy is adopted in the algorithm framework to compensate the undesired effects caused by sequential routing. Experimental results on a series of ISPD98 and industrial benchmarks show that the proposed approach is able to reduce capacitive crosstalk by about 70% on average without compromising completion ratio compared with a previously reported graph based algorithm, demonstrating the advantages of the approach.

关键词： crosstalk resource assignment routing track reservation VLSI

来源：评论

学校读者我要写书评

暂无评论

SalienTR: A closer look at multi-modal transformer for RGB-T salient object detection

引用

Expert Systems with Applications 2025年 286卷

作者： Guo, Ruohao Yue, Wenzhen Qu, Liao Qi, Yanyu Niu, Dantong Ying, Xianghua School of Intelligence Science and Technology State Key Laboratory of General Artificial Intelligence Peking University Beijing100871 China Electrical and Computer Engineering department Carnegie Mellon University PA15213 United States College of Information and Electrical Engineering China Agricultural University Beijing100091 China BAIR Lab University of California BerkeleyCA94709 United States

Recently, transformer-based methods show impressive performance on challenging dense prediction tasks, yet far from perfect. In this paper, we present SalienTR, a novel and effective transformer-based model for high-quality RGB-T salient object detection. First, it builds on two parallel Swin Transformer encoders to extract hierarchical appearance representations of color maps and spatial information of thermal infrared maps. Next, a well-designed cross-modal fusion transformer, namely ComFormer, is devised to mine and fuse two-modality features. Specifically, the proposed ComFormer consists of three consecutive components: (1) A local cross-modal multi-head self-attention (LoC-MSA) to capture local correlations along space-modal dimensions;(2) A global cross-modal multi-head self-attention (GLoC-MSA) to build long-range dependencies between different modalities;(3) A uni-modal convolution (Uni-Conv) that exploits fused information and further learn latent semantics. Besides, we develop a dual-stream decoder that integrates multi-level features to generate both the final salient map and its corresponding edge map. Extensive experiments show that our SalienTR performs well on widely used RGB-T benchmarks and surpasses other state-of-the-art approaches. In addition, SalienTR exhibits high robustness under various challenging scenarios and transferable generalization on RGB-D benchmark datasets. © 2025 Elsevier Ltd

关键词： Object detection

来源：评论

学校读者我要写书评

暂无评论

AmbientOriented programming in AmbientTalk: Combining mobile hardware with simplicity and expressiveness

AmbientOriented Programming in AmbientTalk: Combining mobile...

引用

OOPSLA'05 - 20th Annual ACM Conference on Object-Oriented programming, Systems, Languages, and Applications

作者： Dedecker, Jessie Cutsem, Tom Van Mostinckx, Stijn De Meuter, Wolfgang D Hond, Theo Programming Technology Laboratory Department of Computer Science Vrije Universiteit Brussel Belgium

A new field in distributed computing, called Ambient Intelligence, has emerged as a consequence of the increasing availability of wireless devices and the mobile networks they induce. Developing software for such mobile networks is extremely hard in conventional programming languages because the network is dynamically defined. We demonstrate a new distributed programming language named AmbientTalk, a research vehicle to experiment with expressive high-level language constructs for Ambient-Oriented programming. © 2005 ACM.

关键词： High level languages

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：