检索结果-内蒙古大学图书馆

Conference on Computer Vision and Pattern Recognition (CVPR)

作者： Tiancheng Lin Zhimiao Yu Hongyu Hu Yi Xu Chang Wen Chen Shanghai Key Lab of Digital Media Processing and Transmission Shanghai Jiao Tong University MoE Key Lab of Artificial Intelligence AI Institute Shanghai Jiao Tong University The Hong Kong Polytechnic University Hong Kong China

Multi-instance learning (MIL) is an effective paradigm for whole-slide pathological images (WSIs) classification to handle the gigapixel resolution and slide-level label. Prevailing MIL methods primarily focus on improving the feature extractor and aggregator. However, one deficiency of these methods is that the bag contextual prior may trick the model into capturing spurious correlations between bags and labels. This deficiency is a confounder that limits the performance of existing MIL methods. In this paper, we propose a novel scheme, Interventional Bag Multi-Instance Learning (IBMIL), to achieve deconfounded bag-level prediction. Unlike traditional likelihood-based strategies, the proposed scheme is based on the backdoor adjustment to achieve the interventional training, thus is capable of suppressing the bias caused by the bag contextual prior. Note that the principle of IBMIL is orthogonal to existing bag MIL methods. Therefore, IBMIL is able to bring consistent performance boosting to existing schemes, achieving new state-of-the-art performance. Code is available at https://***/HHHedo/IBMIL.

关键词：

来源：评论

学校读者我要写书评

暂无评论

Feature-Centric Video transmission and Analytics in Large-Scale Internet of Video Things

引用

CAAI Artificial Intelligence Research 2024年第1期3卷 171-177页

作者： Hongan Wei Yuxiang Liu Kejian Hu Liqun Lin Youjia Chen Tiesong Zhao Wanjian Feng Fujian Key Lab for Intelligent Processing and Wireless Transmission of Media Information College of Physics and Information EngineeringFuzhou UniversityFuzhou 350116China Fujian Science&Technology Innovation Laboratory for Optoelectronic Information of China Fuzhou 350108China Yealink Inc. Xiamen 361009China

The interconnection of large-scale visual sensors is called the Internet of Video Things(IoVT),which brings a qualitative leap to the interaction of urban ***,communication delay and resource allocation have brought challenges to the development of *** this paper,we propose a novel city surveillance IoVT architecture to improve *** paradigm consists of front-end target region capture,edge computing and cloud-end feature matching,which can adapt the channel and computing resource allocation ratio flexibly,avoiding communication link congestion caused by unnecessary video *** results show that the proposed scheme is feasible,and can realize efficient data transmission and analysis in an IoVTbased smart city.

关键词： Internet of Video Things(IoVT) edge computing video transmission resource allocation

来源：评论

学校读者我要写书评

暂无评论

MSFEN-AM: A Non-intrusive Load Identification Method for Power Saving

MSFEN-AM: A Non-intrusive Load Identification Method for Pow...

引用

International Conference on Information, Communication and Networks (ICICN)

作者： Yiwen Xu Dengfeng Liu Zhiquan Lin Tiesong Zhao Nian He Fujian Key Lab for Intelligent Processing and Wireless Transmission of Media Information Fuzhou University Fuzhou China Zhicheng College Fuzhou University Fuzhou China Peng Cheng Laboratory Shenzhen China

In order to conserve power resources, researchers have been increasing their focus on load monitoring and identification technologies. However, current image classification methods for load identification face difficulties in extracting overall image contour features and complementing them based on image detail textures. To address this problem, we propose the MSFEN-AM network model, which consists of five parts: input part, downsampling module, bottlenick, upsampling module, and output part. We design a multi-scale feature dense extraction module in the network for feature reuse to improve information utilization. Additionally, we incorporated an attention mechanism to focus on extracting texture and edge information from the image. Furthermore, we introduced an attention gate mechanism to strengthen network selective features and suppress invalid information. Finally, we conducted experiments on the PLAID dataset and performed ablation tests on the multi-scale feature extraction modules and attention gating mechanism. The results show the superiority of our method.

关键词：

来源：评论

学校读者我要写书评

暂无评论

Multi-source Data Fusion Base on Block-Term Decomposition in Federated Learning

Multi-source Data Fusion Base on Block-Term Decomposition in...

引用

International Conference on Computer and Communications (ICCC)

作者： Junjun Nie Xinxin Feng Lixia Xu Haifeng Zheng Fujian Key Lab for Intelligent Processing and Wireless Transmission of Media Information College of Physics and Information Engineering Fuzhou University Fuzhou China

With the development of the Internet of Things technology and the popularization of artificial intelligence, the number of edge devices in the network is growing exponentially. These devices are generating a large amount of data all the time. However, the traditional methods of centralized processing of these data usually require huge storage space and cause high computation burden. In addition, massive data is generally diversified, making the unified processing of these heterogeneous data more complicated, and the utilization of data is low. In this paper, we propose a Block-Term Decomposition (BTD) based algorithm in federated learning to make a fully utilization of multi-source data on edges. For the fusion problem of heteroge-neous data without interaction in Federated Learning (FL), the proposed algorithm introduces BTD method to capture the multi-dimensional characteristics of heterogeneous data by constructing high-order tensors. Experimental results on IEMOCAP dataset show that the proposed algorithm improve calculation accuracy in a single node and under FL framework.

关键词： Tensors Federated learning Computational modeling Data integration Data models Internet of Things Artificial intelligence

来源：评论

学校读者我要写书评

暂无评论

Interventional Bag Multi-Instance Learning On Whole-Slide Pathological Images

arXiv

引用

arXiv 2023年

作者： Lin, Tiancheng Yu, Zhimiao Hu, Hongyu Xu, Yi Chen, Chang Wen Shanghai Key Lab of Digital Media Processing and Transmission Shanghai Jiao Tong University China MoE Key Lab of Artificial Intelligence AI Institute Shanghai Jiao Tong University China The Hong Kong Polytechnic University Hong Kong

关键词： Machine learning

来源：评论

学校读者我要写书评

暂无评论

Robust Spatial-Temporal Graph-Tensor Recovery for Network Latency Estimation

Robust Spatial-Temporal Graph-Tensor Recovery for Network La...

引用

GLOBECOM 2022 - 2022 IEEE Global Communications Conference

作者： Huiyu Lin Lei Deng Lingzhen Wang Haifeng Zheng Xinxin Feng Fujian Key Lab for Intelligent Processing and Wireless Transmission of Media Information College of Physics and Information Engineering Fuzhou University Fuzhou China

ISBN: (纸本)9781665435413

Network latency is an important metric for network performance evaluation. However, device faults inevitably occur during the data collection process, resulting in abnormal data or even missing data. It is desirable to accurately estimate the network latencies and detect the abnormal data. Existing approaches can not provide reliable performance due to the limitation in exploiting the spatial-temporal correlations of latency data. In this paper, we propose a novel Robust Spatial-Temporal Graph-Tensor Recovery (RSTGTR) algorithm which simultaneously recovers the missing data and detects the anomalies in network latencies. Firstly, we model the network latency data as a novel graph-tensor for exploring the topological structure of networks. Secondly, we develop spatial-temporal constraints and propose a graph-tensor recovery algorithm (RSTGTR). Finally, we conduct extensive experiments to evaluate the performance of the proposed algorithm by using a real-world latency dataset. Experimental results show the proposed algorithm outperforms existing methods in terms of latency estimation and anomaly detection.

关键词： Performance evaluation Correlation Estimation Data collection Data models Reliability Global communication

来源：评论

学校读者我要写书评

暂无评论

DesnowFormer: an effective transformer-based image desnowing network

DesnowFormer: an effective transformer-based image desnowing...

引用

IEEE visual Communications and Image processing (VCIP)

作者： Ting Zhang Nanfeng Jiang Junhong Lin Jielian Lin Tiesong Zhao Fujian Key Lab for Intelligent Processing and Wireless Transmission of Media Information College of Physics and Information Engineering Fuzhou University Fuzhou China

ISBN: (纸本)9781665475938

Single image desnowing is an important and challenge task for lots of computer vision applications, such as visual tracking and video surveillance. Although existing deep learning-based methods have achieved promising results, most of them rely on the local deep features and neglect global relationship information between the local regions. Therefore, inevitably leading to over-smooth or detail loss results. To solve this issue, we design a UNet-based end-to-end architecture for image desnowing. Specially, to better characterize global information and preserve image detail, we combine Window-based Self-Attention (WSA) transformer block with Residue Spatial Attention (RSA) to build basic unit of our network. Besides, to protect the structure of the image effectively, we also introduce a Residue Channel (RC) loss to guide high-quality image restoration. Extensive experimental results on both synthetic and real-world datasets demonstrate that the proposed model achieves new state-of-the-art results.

关键词： Training Learning systems visualization Computer vision visual communication Semantics Transformers

来源：评论

学校读者我要写书评

暂无评论

Incremental Unsupervised Adversarial Domain Adaptation for Federated Learning in IoT Networks

Incremental Unsupervised Adversarial Domain Adaptation for F...

引用

International Conference on Mobile Ad-hoc and Sensor Networks, MSN

作者： Yan Huang Mengxuan Du Haifeng Zheng Xinxin Feng Fujian Key Lab for Intelligent Processing and Wireless Transmission of Media Information College of Physics and Information Engineering Fuzhou University Fuzhou China

Federated learning, as an effective machine learning paradigm, can collaboratively training an efficient global model by exchanging the network parameters between edge nodes and the cloud server without sacrificing data privacy. Unfortunately, the obtained global model cannot generalize to newly collected unlabeled data since the unlabeled data collected by different edge devices are diverse. Furthermore, the distributions of collected labeled data and unlabeled data are also different for edge devices. In this paper, we propose a method named Incremental Unsupervised Adversarial Domain Adaptation (IUADA) for federated learning, which aims to reduce the domain shift between the labeled data and unlabeled data in the edge nodes and enhance the performance of the personalized target domain models based on the local unlabeled data. Finally, we evaluate the performance of the proposed method by using three real-world datasets. Extensive experimental results demonstrate that the proposed method is efficient to solve the problem of domain shift and achieves a better performance for unlabeled data for federated learning.

关键词： Training Performance evaluation Knowledge engineering Adaptation models Data privacy Federated learning Data models

来源：评论

学校读者我要写书评

暂无评论

Spatial-Temporal Constrained Pseudo-labeling for Unsupervised Person Re-identification via GCN Inference 18th

Spatial-Temporal Constrained Pseudo-labeling for Unsupervis...

引用

18th International Forum of Digital Multimedia Communication, IFTC 2021

作者： Ling, Sen Yang, Hua Liu, Chuang Chen, Lin Zhao, Hongtian The Institute of Image Communication and Network Engineering Department of Electronic Engineering Shanghai Jiao Tong University Shanghai China Shanghai Key Laboratory of Digital Media Processing and Transmission Shanghai Jiao Tong University Shanghai China MoE Key Lab of Artificial Intelligence AI Institute Shanghai Jiao Tong University Shanghai China

ISBN: (纸本)9789811922657

Most existing unsupervised person re-identification (Re-ID) methods primarily depend on the cluster distance, and merely exploit the available source labeled data to assign pseudo labels for the unannotated data. Whereas, the cluster distance usually fails to adapt to different datasets due to the domain gap. Besides, learning exclusively from the source data can not generate accurate pseudo labels for the lack of the target data information. To address this problem, we propose to exploit the spatial-temporal constraints to facilitate the pseudo label generation process. Specifically, graphs for the labeled source data are constructed and the graph convolution network (GCN) is used to learn graph embeddings. Based on these graph embeddings, the likelihood of linkages between graph nodes is estimated and utilized to assign pseudo labels for the unlabeled data. Then, with the pseudo labels, a smoothed spatial-temporal probability distribution model is generated to amend the likelihood of linkages between graph nodes as well as correct the visual similarity scores for person Re-ID. Finally, we optimize the pseudo label assignment, feature extraction networks, and spatial-temporal model alternatively and iteratively to improve the person Re-ID performance. Comprehensive experiments demonstrate that the proposed method outperforms state-of-the-art methods. © 2022, Springer Nature Singapore Pte Ltd.

关键词： Graph theory

来源：评论

学校读者我要写书评

暂无评论

Traffic Flow Prediction Base on Feature Fusion of Multi-modal Data

Traffic Flow Prediction Base on Feature Fusion of Multi-moda...

引用

International Conference on Computer and Communications (ICCC)

作者： Lixia Xu Xinxin Feng Qiang Zheng Haifeng Zheng Fujian Key Lab for Intelligent Processing and Wireless Transmission of Media Information College of Physics and Information Engineering Fuzhou University Fuzhou China

Accurate traffic flow prediction is important for the application of intelligent transportation systems. The transportation networks have complicated spatial structures and have short-term/long-term dependence. In addition, transportation networks also face various external factors, including the instability of hardware equipments in the data collection process and the network delay in the data transmission process. These factors might lead to result abnormal and missing traffic data, and reduce the accuracy. However, most prediction methods are based on a single model and a single type of traffic data, and the prediction performance is limited, especially when the data set is affected by noise and missing. We propose a traffic flow prediction method based on feature fusion of multi-modal data. We first mine the temporal and spatial characteristics of traffic data by designing multiple sub-networks, and then make full use of the correlation of traffic data. The extracted multi-modal data (such as flow data, occupancy data, speed data) features are fused through three different fusion models, and finally, the traffic flow data is predicted. We use real highway data sets for the experiments. The experimental results show that the traffic flow prediction model based on feature fusion has higher robustness and prediction accuracy than the unfused ones.

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：