检索结果-内蒙古大学图书馆

arXiv 2024年

作者： Zhang, Kejia Weng, Juanjuan Cai, Yuanzheng Luo, Zhiming Li, Shaozi Department of Artificial Intelligence Xiamen University Xiamen361005 China Fujian Provincial Key Laboratory of Information Processing and Intelligent Control College of Computer and Control Engineering Minjiang University Fuzhou350108 China College of Information Science and Technology Jinan University Guangzhou510632 China

Ensuring the robustness of deep neural networks against adversarial attacks remains a fundamental challenge in computer vision. While adversarial training (AT) has emerged as a promising defense strategy, our analysis reveals a critical limitation: AT-trained models exhibit a bias toward low-frequency features while neglecting high-frequency components. This bias is particularly concerning as each frequency component carries distinct and crucial information: low-frequency features encode fundamental structural patterns, while high-frequency features capture intricate details and textures. To address this limitation, we propose High-Frequency Feature Disentanglement and Recalibration (HFDR), a novel module that strategically separates and recalibrates frequency-specific features to capture latent semantic cues. We further introduce frequency attention regularization to harmonize feature extraction across the frequency spectrum and mitigate the inherent low-frequency bias of AT. Extensive experiments demonstrate our method’s superior performance against white-box attacks and transfer attacks, while exhibiting strong generalization capabilities across diverse scenarios. Copyright © 2024, 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

U-CCNet: Brain Tumor MRI Image Segmentation Model with Broader Global Context Semantic information Abstraction

U-CCNet: Brain Tumor MRI Image Segmentation Model with Broad...

引用

IEEE Advanced information technology, Electronic and Automation Control Conference (IAEAC)

作者： Xiaowei Han Jiwu Liu Jiayuan Zhao School of Computer and Information Engineering Harbin University of Commerce Harbin China Heilongjiang Provincial Key Laboratory of Electronic Commerce and Information Processing Harbin China Post-Doctoral Flow Station of Applied Economics Harbin China

Since brain tumor detection can help in the diagnosis of diseases such as Alzheimer’s disease, segmentation and classification of brain tumors are crucial for diagnosis and treatment. Although radiologists can find the tumor area using magnetic resonance imaging (MRI), it is more time-consuming. The high-precision segmentation results provided by deep learning models can assist radiologists in better-identifying brain tumors. In this paper, we propose a U-Net-based approach to detect brain tumors in MRI. In this paper, the cross-attention mechanism is added to the U-Net benchmark architecture and trained and tested on the LGG Segmentation dataset (TCGA-LGG) dataset. The test results show that the proposed model performs well in brain tumor MRI segmentation tasks. Finally, the proposed U-CCNet model is compared with other studies. Experimental results show that this study is superior to other models in the Dice similarity coefficient and IoU index. The proposed method is tested on the LGG brain MRI segmentation dataset. There is a significant improvement in segmentation accuracy in terms of dice score and mIoU, reducing the use of computationally expensive 3D networks.

关键词：

来源：评论

学校读者我要写书评

暂无评论

Deep Graph Convolutional Neural Network based on Adaptive Sampling and Aggregation Strategy

Deep Graph Convolutional Neural Network based on Adaptive Sa...

引用

Machine Vision, Image processing and Imaging technology (MVIPIT), International Conference on

作者： Xueli Wang Wei Zhang Gege Li Lin Zhou Hongkai Liu Zhonglin Ye School of Computer Qinghai Normal University Xining Qinghai Provincial Key Laboratory of Tibetan Information Processing and Machine Translation Xining Graduate School of Engineering Nagasaki Institute of Applied Science Nagasaki

ISBN: (数字)9798331543037

ISBN: (纸本)9798331543044

Graph neural networks learn node embeddings by recursively sampling and aggregating nodes in a graph, while existing methods have a fixed pattern of node sampling and aggregation, and usually only consider direct neighborhood. When sampling high-order neighborhood, it is easy to produce over-smoothing, which affects the model performance and accuracy. Therefore, this paper proposes an adaptive sampling and aggregation deep graph convolutional network (Adaptive Sampling and Aggregation Depth Graph Convolutional Network, ASA-GCN). A sampling strategy based on node importance is designed to adaptively sample high-order neighbors of nodes. Meanwhile, an aggregation strategy based on self-attention mechanism is proposed to adaptively aggregate node information within a given neighborhood. Finally, experimental results on multiple benchmark graph datasets show that the model has superior performance on the node classification task.

关键词： Representation learning Adaptation models Adaptive systems Accuracy Graph convolutional networks Aggregates Imaging Benchmark testing

来源：评论

学校读者我要写书评

暂无评论

HairDiffusion: vivid multi-colored hair editing via latent diffusion 24

HairDiffusion: vivid multi-colored hair editing via latent d...

引用

Proceedings of the 38th International Conference on Neural information processing Systems

作者： Yu Zeng Yang Zhang Jiachen Liu Linlin Shen Kaijun Deng Weizhao He Jinbao Wang Computer Vision Institute School of Computer Science & Software Engineering Shenzhen University Computer Vision Institute School of Computer Science & Software Engineering Shenzhen University and Shenzhen Institute of Artificial Intelligence and Robotics for Society and National Engineering Laboratory for Big Data System Computing Technology Shenzhen University National Engineering Laboratory for Big Data System Computing Technology Shenzhen University and Guangdong Provincial Key Laboratory of Intelligent Information Processing

ISBN: (纸本)9798331314385

Hair editing is a critical image synthesis task that aims to edit hair color and hairstyle using text descriptions or reference images, while preserving irrelevant attributes (e.g., identity, background, cloth). Many existing methods are based on StyleGAN to address this task. However, due to the limited spatial distribution of StyleGAN, it struggles with multiple hair color editing and facial preservation. Considering the advancements in diffusion models, we utilize Latent Diffusion Models (LDMs) for hairstyle editing. Our approach introduces Multi-stage Hairstyle Blend (MHB), effectively separating control of hair color and hairstyle in diffusion latent space. Additionally, we train a warping module to align the hair color with the target region. To further enhance multi-color hairstyle editing, we fine-tuned a CLIP model using a multi-color hairstyle dataset. Our method not only tackles the complexity of multi-color hairstyles but also addresses the challenge of preserving original colors during diffusion editing. Extensive experiments showcase the superiority of our method in editing multi-color hairstyles while preserving facial attributes given textual descriptions and reference images.

关键词：

来源：评论

学校读者我要写书评

暂无评论

A Sparsity Knowledge Transfer-Based Evolutionary Algorithm for Large-Scale Multitasking Multi-Objective Optimization

引用

IEEE Transactions on Evolutionary Computation 2024年

作者： Wu, Chengming Tian, Ye Zhang, Limiao Xiang, Xiaoshu Zhang, Xingyi Anhui University Key Laboratory of Intelligent Computing and Signal Processing of Ministry of Education School of Artificial Intelligence Hefei230601 China Anhui University Key Laboratory of Intelligent Computing and Signal Processing of Ministry of Education School of Computer Science and Technology Hefei230039 China Anhui University Information Materials and Intelligent Sensing Laboratory of Anhui Province Anhui Hefei230601 China Anhui University Key Laboratory of Intelligent Computing and Signal Processing of Ministry of Education Institutes of Physical Science and Information Technology Hefei230601 China

Multitasking multi-objective evolutionary algorithms (MMEAs) have been extensively studied in the past decade, which mainly concentrate on multitasking multi-objective optimization problems (MMOPs) with dozens of decision variables. Nevertheless, many real-world MMOPs have thousands of decision variables and are of sparse nature, which are regarded as large-scale MMOPs (LSMMOPs) in this study. To address LSMMOPs, a sparsity knowledge transfer-based evolutionary multi-objective algorithm, termed EMO-SKT, is proposed for efficiently finding high-quality sparse solutions of LSMMOPs. For each target optimization task, a sparsity knowledge transfer strategy extracts sparse distribution information from a source task and incorporates the information into the target task for two types of sparsity knowledge: variable importance and sparse degree. The variable importance is utilized to produce high-quality sparse solutions during the evolutionary search for EMO-SKT, while the sparse degree facilitates the reduction of search space and thus speeds up the convergence of the evolutionary search. Experimental results on eight benchmark problems and six practical LSMMOPs demonstrate the effectiveness of the sparsity knowledge transfer strategy. Furthermore, the proposed EMO-SKT is capable of efficiently finding high-quality sparse solutions on an LSMMOP with over 1000 decision variables. In comparison with five state-of-the-art multitasking or sparse optimization algorithms, the proposed EMO-SKT exhibits superior performance in terms of both solution quality and search efficiency. © 1997-2012 IEEE.

关键词： Optimization algorithms

来源：评论

学校读者我要写书评

暂无评论

SIAVC: Semi-Supervised Framework for Industrial Accident Video Classification

arXiv

引用

arXiv 2024年

作者： Li, Zuoyong Lin, Qinghua Fan, Haoyi Zhao, Tiesong Zhang, David The Fujian Provincial Key Laboratory of Information Processing and Intelligent Control College of Computer and Control Engineering Minjiang University Fuzhou350121 China The School of Computer Science and Mathematics Fujian University of Technology Fuzhou350118 China The School of Computer and Artificial Intelligence Zhengzhou University Zhengzhou45000 China The Fujian Key Lab for Intelligent Processing and Wireless Transmission of Media Information Fuzhou University Fuzhou350108 China Shenzhen518172 China

Semi-supervised learning suffers from the imbalance of labeled and unlabeled training data in the video surveillance scenario. In this paper, we propose a new semi-supervised learning method called SIAVC for industrial accident video classification. Specifically, we design a video augmentation module called the Super Augmentation Block (SAB). SAB adds Gaussian noise and randomly masks video frames according to historical loss on the unlabeled data for model optimization. Then, we propose a Video Cross-set Augmentation Module (VCAM) to generate diverse pseudo-label samples from the high-confidence unlabeled samples, which alleviates the mismatch of sampling experience and provides high-quality training data. Additionally, we construct a new industrial accident surveillance video dataset with frame-level annotation, namely ECA9, to evaluate our proposed method. Compared with the state-of-the-art semi-supervised learning based methods, SIAVC demonstrates outstanding video classification performance, achieving 88.76% and 89.13% accuracy on ECA9 and Fire Detection datasets, respectively. The source code and the constructed dataset ECA9 will be released in https://***/AlchemyEmperor/SIAVC. Copyright © 2024, The Authors. All rights reserved.

关键词： Classification (of information)

来源：评论

学校读者我要写书评

暂无评论

NOMA Assisted Semi-Grant-Free Transmission for UAV Networks: A Multi-User Scheduling Approach

引用

IEEE Transactions on Communications 2025年

作者： Huang, Junlin Lu, Huabing Tang, Jie Zhao, Nan Shi, Zhaoyuan Wang, Xianbin Dalian University of Technology School of Information and Communication Engineering Dalian116024 China Nanchang University School of Information Engineering Jiangxi Provincial Key Laboratory of Advanced Signal Processing and Intelligent Communications Nanchang330031 China South China University of Technology School of Electronic and Information Engineering Guangzhou510640 China Anqing Normal University School of Computer and Information Anqing246052 China Western University Department of Electrical and Computer Engineering LondonN6A 5B9 Canada

Non-orthogonal multiple access (NOMA) assisted semi-grant-free (SGF) transmission, which enables grant-based users to share their spectrum with grant-free (GF) users, becomes an effective solution to the challenge of massive connections. However, this improvement is limited as most of the existing schemes can only admit one GF user. In this paper, we propose a novel NOMA-assisted SGF transmission scheme to support the access of multiple GF users in unmanned aerial vehicle networks. Moreover, in order to further enhance the advantages of the multi-user SGF scheme, two SGF schemes with power matching are devised to further improve the performance by employing the benefits of distributed contention. For ease of performance evaluation, the theoretical expressions of achievable sum rate and average age of information of the three SGF schemes are derived by applying order statistics. We also investigate the high signal-to- noise approximation expressions of sum rate for these schemes to give more insight. Finally, simulation results are provided to demonstrate the performance improvement of three proposed SGF schemes and validate the correctness of the theoretical analysis expressions. © 1972-2012 IEEE.

关键词： Unmanned aerial vehicles (UAV)

来源：评论

学校读者我要写书评

暂无评论

Orthogonal matrix of polarization combinations:concept and application to multichannel holographic recording

引用

Opto-Electronic Advances 2024年第10期7卷 4-13页

作者： Shujun Zheng Jiaren Tan Hongjie Liu Xiao Lin Yusuke Saita Takanori Nomura Xiaodi Tan Information Photonics Research Center College of Photonic and Electronic EngineeringFujian Normal UniversityFuzhou 350117China Department of Electrical and Computer Engineering Duke UniversityDurhamNC 27708USA College of Photonic and Electronic Engineering Key Laboratory of Opto-Electronic Science and for Medicine of Ministry of EducationFujian Provincial Key Laboratory of Photonics TechnologyFujian Provincial Engineering Technology Research Center of Photoelectric Sensing ApplicationFujian Normal UniversityFuzhou 350117China Faculty of Systems Engineering Wakayama University930 SakaedaniWakayama640-8510Japan.

Orthogonal matrices have become a vital means for coding and signal processing owing to their unique distributional *** orthogonal matrices based on amplitude or phase combinations have been extensively explored,the orthogonal matrix of polarization combinations(OMPC)is a novel,relatively unexplored ***,we propose a method for constructing OMPCs of any dimension encompassing 4n(where n is 1,2,4,8,…)mutually orthogonal 2ncomponent polarization *** the field of holography,the integration of polarization multiplexing techniques with polarization-sensitive materials is expected to emerge as a groundbreaking approach for multichannel hologram multiplexing,offering considerable enhancements in data storage capacity and security.A multidimensional OMPC enables the realization of multichannel multiplexing and dynamical modulation of information in polarization holographic *** consolidating all information into a single position within the material,we effectively avoided extraneous crosstalk during the reconstruction *** results show that achieving four distinct holographic images individually and simultaneously depends on the polarization combination represented by the incident *** discovery opens up a new avenue for achieving highly holographic information storage and dynamically displayed information,harnessing the potential of OMPC to expand the heretofore limited dimensionality of orthogonal polarization.

关键词： orthogonal matrix of polarization combinations polarization-modulated multiplexing multichannel recording polarization holography dynamical information modulation

来源：评论

学校读者我要写书评

暂无评论

DAMPER: A Dual-Stage Medical Report Generation Framework with Coarse-Grained MeSH Alignment and Fine-Grained Hypergraph Matching 39

DAMPER: A Dual-Stage Medical Report Generation Framework wit...

引用

39th Annual AAAI Conference on Artificial Intelligence, AAAI 2025

作者： Huang, Xiaofei Chen, Wenting Liu, Jie Lu, Qisheng Luo, Xiaoling Shen, Linlin Computer Vision Institute College of Computer Science and Software Engineering Shenzhen University Shenzhen China Department of Electrical Engineering City University of Hong Kong Kowloon Hong Kong National Engineering Laboratory for Big Data System Computing Technology Shenzhen University China Guangdong Provincial Key Laboratory of Intelligent Information Processing China

ISBN: (纸本)157735897X

Medical report generation is crucial for clinical diagnosis and patient management, summarizing diagnoses and recommendations based on medical imaging. However, existing work often overlook the clinical pipeline involved in report writing, where physicians typically conduct an initial quick review followed by a detailed examination. Moreover, current alignment methods may lead to misaligned relationships. To address these issues, we propose DAMPER, a dual-stage framework for medical report generation that mimics the clinical pipeline of report writing in two stages. In the first stage, a MeSH-Guided Coarse-Grained Alignment (MCG) stage that aligns chest X-ray (CXR) image features with medical subject headings (MeSH) features to generate a rough keyphrase representation of the overall impression. In the second stage, a Hypergraph-Enhanced Fine-Grained Alignment (HFG) stage that constructs hypergraphs for image patches and report annotations, modeling high-order relationships within each modality and performing hypergraph matching to capture semantic correlations between image regions and textual phrases. Finally,the coarse-grained visual features, generated MeSH representations, and visual hypergraph features are fed into a report decoder to produce the final medical report. Extensive experiments on public datasets demonstrate the effectiveness of DAMPER in generating comprehensive and accurate medical reports, outperforming state-of-the-art methods across various evaluation metrics. Copyright © 2025, Association for the Advancement of Artificial Intelligence (***). All rights reserved.

关键词： Diagnosis

来源：评论

学校读者我要写书评

暂无评论

A Multi-Stage Goal-Driven Network for Pedestrian Trajectory Prediction

A Multi-Stage Goal-Driven Network for Pedestrian Trajectory ...

引用

computer Vision, Image and Deep Learning (CVIDL), International Conference on

作者： Xiuen Wu Tao Wang Yuanzheng Cai Lingyu Liang George Papageorgiou Fujian Provincial Key Laboratory of Information Processing and Intelligent Control Minjiang University Fuzhou China College of Computer and Data Science Fuzhou University Fuzhou China School of Electronic and Information Engineering South China University of Technology Guangzhou China Department of Computer Science and Engineering European University Cyprus Nicosia Cyprus

ISBN: (数字)9798350373820

ISBN: (纸本)9798350373837

Pedestrian trajectory prediction plays a pivotal role in ensuring the safety and efficiency of various applications, including autonomous vehicles and traffic management systems. This paper proposes a novel method for pedestrian trajectory prediction, called multi-stage goal-driven network (MGNet). Diverging from prior approaches relying on stepwise recursive prediction and the singular forecasting of a long-term goal, MGNet directs trajectory generation by forecasting intermediate stage goals, thereby reducing prediction errors. The network comprises three main components: a conditional variational autoencoder (CVAE), an attention module, and a multi-stage goal evaluator. Trajectories are encoded using conditional variational autoencoders to acquire knowledge about the approximate distribution of pedestrians’ future trajectories, and combined with an attention mechanism to capture the temporal dependency between trajectory sequences. The pivotal module is the multi-stage goal evaluator, which utilizes the encoded feature vectors to predict intermediate goals, effectively minimizing cumulative errors in the recursive inference process. The effectiveness of MGNet is demonstrated through comprehensive experiments on the JAAD and PIE datasets. Comparative evaluations against state-of-the-art algorithms reveal significant performance improvements achieved by our proposed method.

关键词： Deep learning Pedestrians Computational modeling Predictive models Prediction algorithms Vectors Inference algorithms

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：