检索结果-内蒙古大学图书馆

STTG-net:a Spatio-temporal network for human motion prediction based on transformer and graph convolution network

Visual Computing for Industry,Biomedicine,and Art 2022年第1期5卷 224-238页

作者： Lujing Chen Rui Liu Xin Yang Dongsheng Zhou Qiang Zhang Xiaopeng Wei tional and Local Joint Engineering Laboratory of Computer Aided Design School of Software EngineeringDalian UniversityDalian 116622China School of Computer Science and Technology Dalian University of TechnologyDalian 116024China

In recent years,human motion prediction has become an active research topic in computer ***,owing to the complexity and stochastic nature of human motion,it remains a challenging *** previous works,human motion prediction has always been treated as a typical inter-sequence problem,and most works have aimed to capture the temporal dependence between successive ***,although these approaches focused on the effects of the temporal dimension,they rarely considered the correlation between different joints in ***,the spatio-temporal coupling of human joints is considered,to propose a novel spatio-temporal network based on a transformer and a gragh convolutional network(GCN)(STTG-Net).The temporal transformer is used to capture the global temporal dependencies,and the spatial GCN module is used to establish local spatial correlations between the joints for each *** overcome the problems of error accumulation and discontinuity in the motion prediction,a revision method based on fusion strategy is also proposed,in which the current prediction frame is fused with the previous *** experimental results show that the proposed prediction method has less prediction error and the prediction motion is smoother than previous prediction *** effectiveness of the proposed method is also demonstrated comparing it with the state-of-the-art method on the Human3.6 M dataset.

关键词： Human motion prediction Transformer Gragh convolutional network

来源：评论

学校读者我要写书评

暂无评论

ASFNet: Adaptive multiscale segmentation fusion network for real-time semantic segmentation

ASFNet: Adaptive multiscale segmentation fusion network for ...

引用

作者： Zha, Hengfeng Liu, Rui Yang, Xin Zhou, Dongsheng Zhang, Qiang Wei, Xiaopeng National and Local Joint Engineering Laboratory of Computer Aided Design School of Software Engineering Dalian University Dalian China School of Computer Science and Technology Dalian University of Technology Dalian China

Recently, the development of deep learning has facilitated continuous progress in the field of computer vision. Pixel-level semantic segmentation serves as a fundamental task in computer vision. It achieves significant results by connecting wider and deeper backbone networks and building fine-grained segmentation heads. However, applications such as self-driving cars are more critical to the computational speed of the algorithms. The trade-off between accuracy and real-time performance of existing algorithms is still a challenging task. To address this challenge, this article proposes an adaptive multiscale segmentation fusion network to fuse multiscale contextual, which designs an adaptive multiscale segmentation fusion module based on an attention mechanism. Using segmentation fusion instead of feature fusion, the multiscale segmentation results are aggregated to obtain more precise segmentation results. The final results achieved 70.9% mIoU of accuracy in the Cityspace test set, processing images at 61 FPS when the input is 1024 × 2048. In addition, when adjusting the input size to 512 × 1024, the images are processed at 185 FPS. © 2021 John Wiley & Sons, Ltd.

关键词： Semantic Segmentation

来源：评论

学校读者我要写书评

暂无评论

PSA-CLNet: Pedestrian Search Method Based on Polarized Self-Attention and COIM Loss 9

PSA-CLNet: Pedestrian Search Method Based on Polarized Self-...

引用

9th International Conference on Virtual Reality, ICVR 2023

作者： Fu, Junjie Liu, Rui Yi, Pengfei Dong, Jing Zhou, Dongsheng Wei, Xiaopeng Dalian University National and Local Joint Engineering Laboratory of Computer Aided Design School of Software Engineering Dalian China Dalian University of Technology School of Computer Science and Technology Dalian China

ISBN: (纸本)9798350345810

Pedestrian search is a combined task of pedestrian detection and pedestrian re-identification, i.e., how to find a specific pedestrian in a panoramic picture. However, due to the impact of different imaging conditions and environments, the accuracy of pedestrian search is still low at present, which is difficult to meet the needs of practical applications. To improve the accuracy of pedestrian search, this paper proposes PSA-CLNet pedestrian search method, which integrates polarized self-attention mechanism (PSA) and cosine online instance matching (COIM) loss. Firstly, the channel attention is added to enhance the pedestrian feature information and improve the expression ability of the feature information. Secondly, spatial attention is added to enhance the areas that have more contribution to the pedestrian search task to increase the correlation between the pictures and improve the recognition ability. Meanwhile, this paper proposes a COIM loss based on the online instance matching (OIM) loss. The loss function can enlarge the class distance between pedestrians and background, and reduce the class distance between pedestrians. This could improve the accuracy of pedestrian search algorithm effectively. Finally, this paper uses ResNet as the main backbone network to verify the proposed method. The pedestrian search method proposed in this paper could achieve 96.60% of mAP index and 96.83% of top-1 index in CUHK-SYSU dataset, and achieve 48.78% of mAP index and 88.38% of top-1 index in PRW dataset. © 2023 IEEE.

关键词： COIM loss pedestrian search PSA

来源：评论

学校读者我要写书评

暂无评论

ER-Net: Efficient Recalibration Network for Multi-view Multi-person 3D Pose Estimation 8

ER-Net: Efficient Recalibration Network for Multi-view Multi...

引用

8th International Conference on Virtual Reality, ICVR 2022

作者： Zhou, Mi Liu, Rui Yi, Pengfei Zhou, Dongsheng Zhang, Qiang Wei, Xiaopeng School of Software Engineering Dalian University National and Local Joint Engineering Laboratory of Computer Aided Design Dalian China School of Computer Science and Technology Dalian University of Technology Dalian China

ISBN: (数字)9781665479110

ISBN: (纸本)9781665479110

Multi-view multi-person 3D human pose estimation is a hot topic in the field of human pose estimation due to its wide range of application scenarios. With the introduction of end-To-end direct regression methods, the field has reached a new stage of development. However, the regression results are still not accurate enough even for the optimal method for the joints that are more heavily influenced by external factors. In this paper, we propose an effective feature recalibration module based on the channel attention mechanism and an optimal calibration strategy, which is applied to a multi-view multi-person 3D human pose estimation task backbone network to achieve improved detection accuracy for joints that are more severely affected by external factors. Specifically, we achieve optimal weight adjustment of joint feature information through a recalibration module based on a channel attention mechanism and an optimal calibration strategy, which enables the model to learn the dependencies between joints and the dependencies between people and their corresponding joints. We call this method the Efficient Recalibration Network (ER-Net). Finally, experiments were conducted on two benchmark datasets for this task, Campus, and Shelf, PCP reached 97.3 percent and 98.3 percent, respectively. © 2022 IEEE.

关键词： computer vision

来源：评论

学校读者我要写书评

暂无评论

DETG: Dual encoding document level relation extraction based on tree graph

DETG: Dual encoding document level relation extraction based...

引用

computer Information Science and Application Technology (CISAT), International Conference on

作者： Jiaxing Li Zhaoqian Zhong National and Local Joint Engineering Laboratory of Computer Aided Design Dalian University Dalian China

ISBN: (数字)9798350375107

ISBN: (纸本)9798350375114

In recent years, graph-based and sequence-based document-level relation extraction models have attracted much attention. However, when performing document-level relation extraction tasks, traditional graph-based models struggle to address the overfitting issue of graph convolutional networks when analyzing document information, and sequence-based models may not effectively capture all document features. To address the above problems, we propose a document level relation extraction model based on dendrogram with dual encoders. The model uses dual encoders to extract document features of various granularities to construct a dendrogram structure. Experimental results on the public dataset DocRED show that the performance of the proposed model is significantly improved compared to other models that solely rely on graph structure or sequence structure.

关键词： Analytical models Information science Tree graphs Graph convolutional networks Semantics Knowledge graphs Feature extraction Encoding Cognition Data mining

来源：评论

学校读者我要写书评

暂无评论

joint Convolutional and Self-Attention Network for Occluded Person Re-Identification

Joint Convolutional and Self-Attention Network for Occluded ...

引用

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

作者： Chuxia Yang Wanshu Fan Dongsheng Zhou Qiang Zhang National and Local Joint Engineering Laboratory of Computer Aided Design School of Software Engineering Dalian University Dalian China

Occluded person Re-Identification (Re-ID) is built on cross views, which aims to retrieve a target person in occlusion scenes. Under the condition that occlusion leads to the interference of other objects and the loss of personal information, the efficient extraction of personal feature representation is crucial to the recognition accuracy of the system. Most of the existing methods solve this problem by designing various deep networks, which are called convolutional neural networks (CNN)-based methods. Although these methods have the powerful ability to mine local features, they may fail to capture features containing global information due to the limitation of the gaussian distribution property of convolution operation. Recently, methods based on Vision Transformer (ViT) have been successfully employed to person Re-ID task and achieved good performance. However, since ViT-based methods lack the capability of extracting local information from person images, the generated results may severely lose local details. To address these deficiencies, we design a convolution and self-attention aggregation network (CSNet) by combining the advantages of both CNN and ViT. The proposed CSNet consists of three parts. First, to better capture personal information, we adopt Dual-Branch Encoder (DBE) to encode person images. Then, we also embed a local Information Aggregation Module (LIAM) in the feature map, which effectively leverages the useful information in the local feature map. Finally, a Multi-Head Global-to-local Attention (MHGLA) module is designed to transmit global information to local features. Experimental results demonstrate the superiority of the proposed method compared with the state-of-the-art (SOTA) methods on both the occluded person Re-ID datasets and the holistic person Re-ID datasets.

关键词： Convolution Interference Gaussian distribution Feature extraction Transformers Sensors Data mining

来源：评论

学校读者我要写书评

暂无评论

Semantic Image Synthesis via Location Aware Generative Adversarial Network

Semantic Image Synthesis via Location Aware Generative Adver...

引用

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

作者： Jiawei Xu Rui Liu Jing Dong Pengfei Yi Wanshu Fan Dongsheng Zhou National and Local Joint Engineering Laboratory of Computer Aided Design School of Software Engineering Dalian University Dalian China

Semantic image synthesis aims to synthesize photo-realistic images through the given semantic segmentation masks. Most existing models use conditional batch normalization (CBN) to regulate normalization activation by spatially varying modulation parameters. It can prevent semantic information from being eliminated during normalization. But the modulation parameters in CBN lack location constraint, resulting in the lack of structural information in the synthetic image. And CBN is highly dependent on the batch size. To address these limitations, we propose location aware conditional group normalization (LACGN) and construct a location aware generative adversarial network (LAGAN) based on this method. LACGN can learn spatial location aware information in a weakly supervised manner that relies on the current image synthesis process to guide transformations spatially. It allows the synthetic image to have more structural information and detailed features. At the same time, group normalization(GN) replace the traditional BN to eliminate the dependence on batch size. Extensive experiments show that LAGAN is better than other methods.

关键词： Location awareness Image synthesis Semantic segmentation Semantics Modulation Generative adversarial networks Sensors

来源：评论

学校读者我要写书评

暂无评论

MCFNet: Multi-scale Cross Fusion Network for 3D Human Pose Estimation

MCFNet: Multi-scale Cross Fusion Network for 3D Human Pose E...

引用

IEEE International Conference on Signal and Image Processing (ICSIP)

作者： Dazhong Wang Rui Liu Pengfei Yi Jing Dong Dongsheng Zhou National and Local Joint Engineering Laboratory of Computer Aided Design School of Software Engineering Dalian University Dalian China

ISBN: (数字)9798350350920

ISBN: (纸本)9798350350937

3D human pose estimation has attracted increasing attention due to it frequently used in fields such as human-computer interaction. Due to the structure of the human skeleton is a kind of undirected topology, Graph Convolutional Network(GCN) is a suitable method to study 3D human pose estimation. However, traditional GCNs overlook the distinctions between joints and the implicit connections among nonadjacent joints. Additionally, the topological structure of the human body leads to significant error accumulation in pose estimation, which make the joints have higher error levels in highly flexible regions. We present Multi-scale Cross Fusion Network(MCFNet), a network that sequentially learns the human skeleton structure from joint, limb, and body semantic levels by segmenting original features. The MCFNet can combine the High-order MGCN(Modulated Graph Convolutional Network) module for joint scale capture and the Conv-SE module for limb scale feature fusion, thus achieving excellent joint feature extraction and reduction of the error accumulation. Experimental results on datasets like Human3.6M demonstrate that our proposed method surpasses the performance of state-of-the-art networks and achieves the highest accuracy.

关键词： Human computer interaction Three-dimensional displays Accuracy Network topology Graph convolutional networks Pose estimation Semantics

来源：评论

学校读者我要写书评

暂无评论

3D Human Pose Estimation via Graph Extended Spatio-Temporal Convolutional Network

3D Human Pose Estimation via Graph Extended Spatio-Temporal ...

引用

Virtual Reality (ICVR), IEEE International Conference on

作者： Yanhui Jia Wanshu Fan Dongsheng Zhou Qiang Zhang National and Local Joint Engineering Laboratory of Computer Aided Design School of Software Engineering Dalian University Dalian China

3D human pose estimation is an important premise for human behavior analysis and understanding, which has a wide range of applications in intelligent transportation, human-computer interaction, and animation production. Most existing works focus on extracting the feature relationship between frames by combining spatio-temporal information to reduce the error of attitude reconstruction. However, the majority of them often suffer from insufficient joint correlation characteristics. To address this problem, we propose a Graph Expand Spatiotemporal Convolutional Network, named GESC-Net, to improve the limitation of extracting human spatial structure features. To better enrich the feature of extracting local information, we develop a learnable symmetric connection (LSC) block in the spatial structure. Moreover, a CbAttantion block is also designed to obtain a larger view of the acquisition of global structure and get more effective features. We evaluate our approach on two standard benchmark datasets: Human3.6M and HumanEva-I. The quantitative and qualitative evaluation results demonstrate that the GESC-Net can achieve better 3D human posture estimation than existing state-of-the-art methods.

关键词：

来源：评论

学校读者我要写书评

暂无评论

MHFDN:Multi-Branch Hybrid Frequency Domain Networks for 3D Human Motion Prediction

MHFDN:Multi-Branch Hybrid Frequency Domain Networks for 3D H...

引用

IEEE International Conference on Signal and Image Processing (ICSIP)

作者： Xin Zhou Rui Liu Jing Dong Pengfei Yi Ling Wang National and Local Joint Engineering Laboratory of Computer Aided Design School of Software Engineering Dalian University Dalian China

ISBN: (数字)9798350350920

ISBN: (纸本)9798350350937

Recent studies have shown great potential in 3D human motion prediction using Graph Convolutional Networks (GCN)-based modeling. However, these methods have two key problems: First, there is a lack of explicit extraction of the spatial features of joint data. Second, the relevant high-frequency components are often ignored in the feature extraction process. In order to solve these two problems, we propose a novel and universal hybrid frequency domain networks based on multi-branch structure, which contains multiple mixed frequency domain blocks to give more focus on high frequencies information during training. Through the multi-branch structure, we can learn both local and global movements of the human body. Since the bottom layer is mainly responsible for capturing high-frequency details, the top layer is more concerned with the modeling of low-frequency global information, we further introduce the structure of frequency component modification module and perform information trade-offs based on different networks layers, enabling each layer to effectively extract appropriate high-frequency and low-frequency human features. Experiments show that our method has high prediction performance and higher robustness on the Human3.6M and CMU Mocap datasets.

关键词： Training Solid modeling Three-dimensional displays Convolution Frequency-domain analysis Predictive models Feature extraction

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：