检索结果-内蒙古大学图书馆

第十六届国际体视学与图像分析学术会议

作者： Hairong SUN Li KANG Jianjun HUANG Guangdong Key Laboratory of Intelligent Information Processing College of Electronics and Information Engineering Shenzhen University

Infrared small target detection has widespread applications across various domains,including military,aviation,and ***,improving the detection rate of Infrared small targets in complex backgrounds remains *** order to detect Infrared small targets,we propose the variable-structure U-shaped network known as CAFUNet.A central differential convolution-based encoder,ASPP,an Attention Fusion module,and a decoder module are the critical components of *** encoder module based on central difference convolution effectively extracts shallow detail information from infrared images,complemented by rich contextual information obtained from the deep features in the decoder ***,the direct fusion of the shallow detail features with semantic features may lead to feature ***,to address this,we incorporate an Attention Fusion(AF) module to enhance the network's performance *** performed ablation studies on each module to evaluate its *** results show that our proposed algorithm outperforms the state-of-the-art methods on publicly available datasets.

关键词： infrared small target detection central difference convolution ASPP AF

来源：评论

学校读者我要写书评

暂无评论

A NETWORK LIGHTWEIGHTING METHOD FOR DIFFICULT SEGMENTATION TASK IN 3D MEDICAL IMAGES

A NETWORK LIGHTWEIGHTING METHOD FOR DIFFICULT SEGMENTATION T...

引用

第十六届国际体视学与图像分析学术会议

作者： Li KANG Zhixin GONG Jianjun HUANG Ziqi ZHOU College of Electronics and Information Engineering Shenzhen University Guangdong Key Laboratory of Intelligent Information Processing

Currently,deep learning is widely used in medical image segmentation and has achieved good ***,some 3 D medical image segmentation tasks with the characteristics of diverse lesion representations,blurred edges,and unstable positions require networks with high complexity,large number of parameters and calculations,and high requirements on operating equipment,making the network unable to be deployed in *** this work,we propose a method to make the network lightweight and applied the method to a 3 D CNN based *** experimented with the network on the COVID-19 lesion segmentation ***,we use three cascaded one-dimensional convolutions to replace 3 D convolutions,and integrate instance normalization with the previous layer of one-dimensional convolutions to improve network inference *** addition,we else simplify test-time augmentation and deep supervision of the *** show that our method has a certain *** with the original network,the lightweight network has reduced the prediction time of a single sample and the memory usage by half and the number of parameters has been reduced by 60%.The training time of one round has also been reduced by half and the segmentation accuracy dropped within the acceptable range.

关键词： 3D medical image segmentation 3D U-net lightweight network COVID-19 lesion segmentation

来源：评论

学校读者我要写书评

暂无评论

IMPROVING THE LEARNING PERFORMANCE OF CLIENT'S LOCAL DISTRIBUTION IN CYCLIC FEDERATED LEARNING FOR CLASSIFICATION OF ALZHEIMER'S DISEASE

IMPROVING THE LEARNING PERFORMANCE OF CLIENT'S LOCAL DISTRIB...

引用

第十六届国际体视学与图像分析学术会议

作者： Li KANG Bin LUO Jianjun HUANG Guangdong Key Laboratory of Intelligent Information Processing College of Electronics and Information Engineering Shenzhen University

Cyclic federated learning based on distribution information sharing and knowledge distillation（CFLSD） aims to address the challenges of non-iid data distribution and reduce communication ***,when client data is extremely heterogeneous and scarce,it becomes challenging for clients to fully learn the distribution of local data using GANs,thereby affecting the overall model *** overcome this limitation,we propose a transfer learning approach where clients first pretrain their generators on a source domain and then fine-tune them on their local *** results on the classification of Alzheimer's disease demonstrate that this method effectively improves client distribution learning performance and enhances the overall model performance.

关键词： federated learning transfer learning medical image processing

来源：评论

学校读者我要写书评

暂无评论

CHRONIC STROKE LESIONS SEGMENTATION BASED ON COMPOUND LOSS FUNCTION AND TRANSFORMER

CHRONIC STROKE LESIONS SEGMENTATION BASED ON COMPOUND LOSS F...

引用

第十六届国际体视学与图像分析学术会议

作者： Zhiyong GAO Li KANG Jianjun HUANG College of Electronics and Information Engineering Shenzhen University Guangdong Key Laboratory of Intelligent Information Processing

Ischemic stroke is a cerebrovascular disease with high disability and mortality rates *** and accurate segmentation of stroke lesions from medical images is crucial for early diagnosis and ***,stroke lesion segmentation faces many challenges,such as different locations of stroke occurrences,large shape variations,unclear lesion boundaries,and strong data *** accurately segment stroke lesions,our paper proposes an end-to-end deep learning method based on a compound loss function and Transformer to automatically segment lesion regions from stroke MRI *** our proposed method,we use the unified segmentation framework MMSegmentation,with a UNet network as the main architecture,and introduce a Transformer-based attention mechanism to explore more detailed local *** addition,to alleviate the problem of data heterogeneity,we propose a weighted compound loss function and hard example mining strategy,which enables our network parameters to be updated in a faster and smoother *** evaluate the proposed method on the ATLAS dataset and compare it with several state-of-the-art *** the experimental results,the proposed method achieves better stroke segmentation performance.

关键词： chronic stroke medical image segmentation transformer

来源：评论

学校读者我要写书评

暂无评论

MMR-HAD: Multi-scale Mamba Reconstruction Network for Hyperspectral Anomaly Detection

引用

IEEE Transactions on Geoscience and Remote Sensing 2025年

作者： Fu, Xiyou Zhang, Ting Cheng, Juan Jia, Sen Shenzhen University College of Computer Science and Software Engineering Shenzhen518060 China Shaanxi Key Laboratory of Optical Remote Sensing and Intelligent Information Processing Xi'an710119 China

Hyperspectral anomaly detection (HAD) aims to identify anomalous objects from hyperspectral images (HSIs) whose spectral features significantly deviate from their surroundings. Existing HAD methods still reconstruct some anomalies during the background reconstruction process, which can seriously affect the detection accuracy. Consequently, inspired by Mamba's ability to effectively model long sequences, we propose a Multi-scale Mamba Reconstruction Network for Hyperspectral Anomaly Detection (MMR-HAD) by enhancing the representation of the background and inhibit anomalies from being reconstructed. MMR-HAD first removes most of the anomalous pixels in hyperspectral images using the Random Mask (RM) strategy, which reduces the interference of anomalous pixels on the background reconstruction and makes the background features more prominent. To further filter out the small amount of residual anomalous pixels, we propose the Multi-scale Dilated Attention Background Enhancement (MDABE) mechanism, which enhances the background representation. Finally, we apply the Multi-scale Dynamic Feature Fusion (MDFF) strategy to reconstruct the background image, further extracting and strengthening the background information, thus obtaining a pure background image. MMR-HAD, which focuses on generating a pure background, has been experimentally validated on seven real hyperspectral datasets. The results demonstrate that it excels in background enhancement and anomaly suppression, significantly improving detection accuracy. Introducing Mamba offers a promising solution for HAD, with substantial potential for practical application. © 1980-2012 IEEE.

关键词： Image reconstruction

来源：评论

学校读者我要写书评

暂无评论

Positioning Error Compensation by Channel Knowledge Map in UAV Communication Missions

arXiv

引用

arXiv 2024年

作者： Zhang, Chiya Wang, Ting He, Chunlong School of Electronic and Information Engineering Harbin Institute of Technology Shenzhen518055 China Guangdong Key Laboratory of Intelligent Information Processing Shenzhen University Shenzhen518060 China

When Unmanned Aerial Vehicles (UAVs) perform high-precision communication tasks, such as searching for users and providing emergency coverage, positioning errors between base stations and users make it challenging to deploy trajectory planning algorithms. To address these challenges caused by position errors, a framework was proposed to compensate it by Channel Knowledge Map (CKM), which stores channel state information (CSI). By taking the positions with errors as input, the generated CKM could give a prediction of signal attenuation which is close to true positions. Based on that, the predictions are utilized to calculate the received power and a PPO-based algorithm is applied to optimize the compensation. After training, the framework is able to find a strategy that minimize the flight time under communication constraints and positioning error. Besides, the confidence interval is calculated to assist the allocation of power and the update of CKM is studied to adapt to the dynamic environment. Simulation results show the robustness of CKM to positioning error and environmental changes, and the superiority of CKM-assisted UAV communication design. © 2024, CC BY.

关键词： Unmanned aerial vehicles (UAV)

来源：评论

学校读者我要写书评

暂无评论

Simple and Efficient Knowledge Graph Attention Network for Recommendation

Simple and Efficient Knowledge Graph Attention Network for R...

引用

2023 International Conference on Cyber-Physical Social Intelligence, ICCSI 2023

作者： Li, Yangding Fu, Shaobin Feng, Hao Zeng, Yangyang Wang, Jinghao Jiang, Zhihao Zhang, Lvyun Hunan Provincial Laboratory of Intelligent Computing and Language Information Processing Hunan Normal University School of Information Science and Engineering Changsha China Shenzhen Lazada Software Technology Company Limited Shenzhen China Hechi University School of Big Data and Computer Science Guangxi China

ISBN: (纸本)9798350312492

Existing methods for modeling recommendation systems based on knowledge graphs include embedding-based, pathbased, and propagation-based methods. The embedding-based approach is flexible but more suitable for intra-graph applications, the path-based approach can model complex relationships but has a high computational cost, and the propagation-based approach considers global information but may introduce noise. This study proposed a simple and efficient model, called SEKGAT, which comprehensive the ideology of path-based and propagation approach to personalized recommendation by aggregating the user preferences through graph attention mechanism and fusing multiple feature representations on the knowledge graph into item features through pooling aggregators. Experimental results for the CTR prediction and Top-K recommendation tasks on three datasets of real-world scenarios show that our model approach is competitive. © 2023 IEEE.

关键词： Recommender systems

来源：评论

学校读者我要写书评

暂无评论

Adaptive Neighborhoods Locality Preserving Projections

Adaptive Neighborhoods Locality Preserving Projections

引用

IEEE International Conference on information Communication and Signal processing (ICICSP)

作者： Jiayi Xu Wendong Huang Yang Zhao Jihong Pei Xuan Yang Guangdong Key Laboratory of Intelligent Information Processing Shenzhen University College of Electronics and Information Engineering Shenzhen China Shenzhen University College of Electronics and Information Engineering Shenzhen China Shenzhen University College of Computer Science and Software Engineering Shenzhen China

ISBN: (数字)9798350355895

ISBN: (纸本)9798350355901

In contemporary society, the analysis and utilisation of big data is of paramount importance. However, traditional feature extraction methods frequently encounter the curse of dimensionality when dealing with high-dimensional data, which is no longer able to adapt to the current needs of practical applications. The manifold learning method, which is capable of effectively extracting features of high-dimensional data, ne-cessitates the resolution of the problem of how to choose an appropriate local neighbourhood. Consequently, this paper proposes an Adaptive neighborhood Locality Preserving Projection (LAPP) based on the Locality Preserving Projection (LPP). TThis method combines two neighborhood selection method. To achieve an adaptive radius selection for local neighborhoods of sample points based on the distance characteristics between sample points. The LAPP proposed in this paper markedly reduces the dimensionality reduction and enhances the computational efficiency while maintaining a certain degree of accuracy.

关键词： Manifolds Accuracy High dimensional data Signal processing Big Data Feature extraction Computational efficiency Manifold learning Data mining

来源：评论

学校读者我要写书评

暂无评论

Identity and Modality Attributes Driven Multimodal Fusion Networks for Emotion Recognition in Conversations

引用

IEEE Transactions on Multimedia 2025年

作者： Shi, Wuzhen Chen, Xuping Yao, Biyun Wen, Yang Sheng, Bin Shenzhen University Guangdong Province Engineering Laboratory for Digital Creative Technology Guangdong Provincial Key Laboratory of Intelligent Information Processing College of Electronics and Information Engineering Shenzhen518060 China Shanghai Jiao Tong University Department of Computer Science and Engineering Shanghai China

Emotion recognition in conversations (ERC) is a crucial aspect of human-computer interaction and plays an important role in various domains, including healthcare, entertainment, and education. Since the conversation data in the form of multimodal sequences is well suited to be constructed into graphs, the methods based on graph convolutional network (GCN) show incomparable advantages. However, existing methods attempt to model the highly uncertain emotional relationships between different speakers, which is not an easy task and may even introduce interference information. Therefore, we propose an identity and modality attributes driven multimodality fusion network (dubbed IMDNet) for emotion recognition in conversations. Specifically, we construct a speaker-centric graph that only connects nodes of the same speaker within modalities to each other, reducing the interference between the emotions of different speakers. We also introduce the attribute embedding mechanism, which facilitates the correct calculation of correlations between nodes for better multimodal feature fusion. Considering that the emotional correlation between utterances will decrease over time, we present an utterance distance attention to make the fusion network pay more attention to the adjacent utterances. Furthermore, we explore the solution to the data imbalance problem suitable for conversation scenarios. Given the presence of possible anomalous samples in the dataset, we opt for the BoundaryFocalLoss. Experiments on the IEMOCAP and MELD datasets show that our IMDNet outperforms the state-of-the-art methods. © 1999-2012 IEEE.

关键词： Convolution

来源：评论

学校读者我要写书评

暂无评论

Distributed intelligent Reflecting Surfaces-Aided Device-to-Device Communications System

引用

Journal of Communications and information Networks 2021年第3期6卷 197-207页

作者： Chiya Zhang Wenyu Chen Qian Chen Chunlong He School of Electronic and Information Engineering Harbin Institute of TechnologyShenzhen 518055China Peng Cheng Laboratory(PCL) Shenzhen 5180523China Guangdong Key Laboratory of Intelligent Information Processing Shenzhen UniversityShenzhen 518060China National Mobile Communications Research Laboratory Southeast UniversityNanjing 211189China

intelligent reflecting surface(IRS)is a revolutionizing and promising technology to improve the high transmission rate of the wireless communication ***,an IRS consists of a great amount of low-cost passive reflecting elements,which reflect the incident signals to the desired user by collaboratively using passive *** paper introduces IRSs into a device-to-device(D2D)underlying cellular system to enhance transmission rate performance of the D2D *** formulate an optimization problem of maximizing the transmission rate of the D2D pairs while satisfying the minimum required rate of the cellular *** address this problem by jointly optimizing the reuse indicator,received beamforming,power allocation,and phase shift *** coordinate descent(BCD)algorithm is adopted to decouple the original problem into four *** form solutions are obtained by solving the sub-problems of optimizing the received beamforming and power ***,Kuhn-Munkres(KM)algorithm and minimization-majorization(MM)algorithm are adopted to solve the sub-problems of optimizing the reuse indicator and phase shift matrices,*** results demonstrate that IRSs can effectively improve the transmission rate of the D2D pairs and our proposed distributed IRSs scheme outperforms the other benchmark schemes.

关键词： intelligent reflecting surface device-to-device communications block coordinate descent

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：