检索结果-内蒙古大学图书馆

A 3d convolution-Incorporated dimension Preserved decomposition Model for Traffic data Prediction

IEEE TRANSACTIONS ON INTELLIGENT TRANSPORTATION SYSTEMS 2025年第1期26卷 673-690页

作者： Lin, Mingwei Liu, Jiaqi Chen, Hong Xu, Xiuqin Luo, Xin Xu, Zeshui Fujian Normal Univ Coll Comp & Cyber Secur Fuzhou 350117 Peoples R China Fujian Normal Univ Coll Math & Stat Fuzhou 350117 Peoples R China Southwest Univ Coll Comp & Informat Sci Chongqing 400715 Peoples R China Sichuan Univ Business Sch Chengdu 610064 Peoples R China

Traffic data prediction is a crucial component of Intelligent Transport Systems (ITS) as it contributes significantly to real-time navigation and congestion management. However, due to incomplete deployment of sensor and instability of data transmission, the traffic data that we collect is usually a high-dimensional and incomplete (HdI) matrix or tensor. Currently, there are two challenges with traffic data prediction as follows: a) Most of the existing models are designed for a full set of data, but traffic data are unavoidably missing. b) Most of the existing models often suffer from excessive complexity due to long sequences. To address these issues, we propose a novel 3d convolution-Incorporated dimension Preserved decomposition (3dCIdP) model for traffic data prediction with three main fold ideas: a) enhancing the low-rank property of traffic data to accurately capture its structure, b) learning the constraints of historical sequences to predict sequences through representation modelling and c) capturing spatio-temporal interaction information in traffic data through multidimensional interaction features. To evaluate the performance of the proposed 3dCIdP, we conduct extensive experiments using five publicly available datasets with three different missing rates. When the proposed 3dCIdP is compared to state-of-the-art models, experimental results show that the Root Mean Square Error (RMSE) is reduced by an average of 4.22% and the training time is reduced by an average of 89.76% on the large-scale traffic datasets.

关键词： Traffic tensor representation high-dimensional and incomplete data dimension preserved decomposition low rank approximation 3d convolution Traffic tensor representation high-dimensional and incomplete data dimension preserved decomposition low rank approximation 3d convolution

来源：评论

学校读者我要写书评

暂无评论

Mini-3dCvT: a lightweight lip-reading method based on 3d convolution visual transformer

引用

VISUAL COMPUTER 2025年第3期41卷 1957-1969页

作者： Wang, Huijuan Cui, Boyan Yuan, Quanbo Pu, Gangqiang Liu, Xueli Zhu, Jie North China Inst Aerosp Engn Sch Comp Langfang 065000 Peoples R China

Lip-reading has attracted more and more attention in recent years, and has wide application prospects and value in areas such as human-computer interaction, surveillance and security and audiovisual speech recognition. However, research on lip-reading has been slow due to the complexity of dealing with the fine spatial features of small-sized images of continuous video frames and the temporal features between images. In this paper, to address the challenges in extracting visual spatial features, temporal features and model light weighting, we propose a high-precision, highly robust and lightweight lip-reading method, Mini-3dCvT, which combines visual transforms and 3d convolution to extract spatiotemporal feature of continuous images, and makes full use of the properties of convolution and transforms to effectively extract local and global features of continuous images, use weight transformation and weight distillation in the convolution and transformer structures for model compression, and then send the extracted features to a bidirectional gated recurrent unit for sequence modeling. Experimental results on the large-scale public lip-reading datasets LRW and LRW-1000 show that this paper's method achieves 88.3% and 57.1% recognition accuracy on both datasets, and effectively reduces the model computation and number of parameters, improving the overall performance of the lip-reading model.

关键词： Lip-reading 3d convolution Visual transformer Spatial-temporal features Model light weighting

来源：评论

学校读者我要写书评

暂无评论

Multi-weighted graph 3d convolution network for traffic prediction

引用

NEURAL COMPUTING & APPLICATIONS 2023年第20期35卷 15221-15237页

作者： Liu, Yuqing Wang, Chen Xu, Sixuan Zhou, Wei Chen, Yuzhi Southeast Univ Sch Transportat Nanjing 211198 Jiangsu Peoples R China Purple Mt Labs Nanjing 211111 Jiangsu Peoples R China

Predicting future traffic state (e.g., traffic speed, volume, travel time, etc.) accurately is highly desirable for traffic management and control. However, network-wide traffic flow has complicated spatial-temporal dependencies, making it challenging to predict. This study proposes a multi-weighted graph 3d convolution network (MWG3d) to predict future network-wide traffic speed, considering the spatial-temporal heterogeneous effects of multiple external factors (i.e., points of interests (POIs), roadway physical characteristics and incidents). The network is composed of a Graph-3d convolution (G3d) module and an incident impact module. In G3d module, a weighted graph convolution is developed first, which extracts complex spatial dependencies of traffic flow considering heterogeneous effects of POIs and roadway physical characteristics. These external factors have great influence on the periodicity of human daily activities, which in turn cyclically affect traffic flow. The weighted graph convolution is further connected with 3d convolutions to extract temporal dependencies of traffic flow, accounting for temporal heterogeneous effects of these external factors. An incident impact module is separately developed to account for spatial-temporal heterogeneous effects of incidents. These external factors could lead to abrupt and temporary changes in traffic flow. The proposed network is evaluated on two real-world datasets. The results show that MWG3d outperforms a selection of the state-of-the-art models. Furthermore, the spatial-temporal heterogeneous effects of external factors are crucial to prediction accuracy.

关键词： Traffic prediction Graph convolution 3d convolution Spatial-temporal dependencies Spatial-temporal heterogeneous effects

来源：评论

学校读者我要写书评

暂无评论

A corner preserving surface inference algorithm using 3d convolution

引用

PATTERN RECOGNITION LETTERS 2001年第3-4期22卷 259-269页

作者： Kim, JH Han, JH Pohang Univ Sci & Technol Dept Comp Sci & Engn Pohang 790784 South Korea

A corner preserving 3d convolution operator is proposed. This vector convolution operator is a different version of Guy and Medioni's diabolo field which is used for inferring surface normals from sparse and noisy 3d data. The new operator, named Quadratic field, is made by revolving a set of quadratic functions around the vertical axis. Other than the convolution operator, Guy and Medioni's method in surface normal inference is followed. Analysis and experimental results showed that the new operator performs better in recovering cornered surfaces from sparse 3d data. (C) 2001 Elsevier Science B.V. All rights reserved.

关键词： surface reconstruction surface inference corner preserving operation 3d convolution

来源：评论

学校读者我要写书评

暂无评论

Video Based Shuffling Step detection for Parkinsonian Patients Using 3d convolution

引用

IEEE TRANSACTIONS ON NEURAL SYSTEMS ANd REHABILITATION ENGINEERING 2021年 29卷 641-649页

作者： Cao, Xugang Xue, Youze Chen, Jiansheng Chen, Xiaohe Ma, Yu Hu, Chunhua Ma, Huimin Ma, Hongbing Changchun Univ Sci & Technol Dept Elect & Informat Engn Changchun 130022 Peoples R China Tsinghua Univ Dept Elect Engn Beijing 100084 Peoples R China Tsinghua Univ Beijing Natl Res Ctr Informat Sci & Technol Beijing 100084 Peoples R China Univ Sci & Technol Beijing Sch Comp & Commun Engn Beijing 100083 Peoples R China Chinese Acad Sci Suzhou Inst Biomed Engn & Technol Suzhou 215163 Peoples R China Tsinghua Univ Yuquan Hosp Beijing 100040 Peoples R China Tsinghua Univ Sch Aerosp Engn Natl Engn Lab Neuromodulat Beijing 100084 Peoples R China

Parkinson's disease (Pd) is a common neurodegenerative disease which impacts millions of people around the world. In clinical treatments, freezing of gait (FoG) is used as the typical symptom to assess Pd patients' condition. Currently, the assessment of FoG is usually performed through live observation or video analysis by doctors. Considering the aging societies, such a manual inspection based approach may cause serious burdens on the healthcare systems. In this study, we propose a pure video-based method to automatically detect the shuffling step, which is the most indistinguishable type of FoG. Firstly, the RGB silhouettes which only contain legs and feet are fed into the feature extraction module to obtain multi-level features. 3d convolutions are used to aggregate both temporal and spatial information. Then the multi-level features are aggregated by the feature fusion. Skip connections are implemented to reserve information of high resolution and period-wise horizontal pyramid pooling is utilized to fuse both global context and local features. To validate the efficacy of our method, a dataset containing 268 normal gait samples and 362 shuffling step samples is built, on which our method achieves an average detection accuracy of 90.8%. Besides shuffling step detection, we demonstrate that our method can also assess the severity of walking abnormity. Our proposal facilitates a more frequent assessment of FoG with less manpower and lower cost, leading to more accurate monitoring of the patients' condition.

关键词： Feature extraction Legged locomotion Three-dimensional displays Foot convolution Cameras Hospitals Parkinsonian shuffling step abnormal gait recognition 3d convolution severity assessment

来源：评论

学校读者我要写书评

暂无评论

Lightweight brain MR image super-resolution using 3d convolution

引用

MULTIMEdIA TOOLS ANd APPLICATIONS 2024年第3期83卷 8785-8795页

作者： Kim, Young Beom Van Le, The Lee, Jin Young Samsung Elect Seoul South Korea Sejong Univ Dept Intelligent Mechatron Engn Seoul South Korea

Magnetic resonance imaging (MRI) plays a very important role in a medical domain, such as image guided diagnostics and therapeutics. In particular, high resolution brain MRI has a great potential for preclinical and clinical procedures, because it is non-invasive imaging and shows a high level of anatomical detail. However, the high resolution MRI faces a number of challenges, such as long scan time, high magnetic field strength, and low signal to noise ratio. To solve these issues, deep learning based super-resolution networks, which provide high performance in various fields, can be employed in MRI. Since the super-resolution networks have been mainly developed to reconstruct high quality color images by using many parameters, they cannot be directly applied into MR scanners. Hence, this paper evaluates conventional networks with brain MR images, and then proposes a lightweight network employing 3d convolution, which consists of extraction, compression, and reconstruction parts. Experimental results show that the proposed network is very efficient, in terms of reconstruction quality and network complexity.

关键词： Brain MR image deep learning Magnetic resonance imaging (MRI) Super-resolution 3d convolution

来源：评论

学校读者我要写书评

暂无评论

Automatic Liver Tumor Segmentation on dynamic Contrast Enhanced MRI Using 4d Information: deep Learning Model Based on 3d convolution and convolutional LSTM

引用

IEEE TRANSACTIONS ON MEdICAL IMAGING 2022年第10期41卷 2965-2976页

作者： Zheng, Rencheng Wang, Qidong Lv, Shuangzhi Li, Cuiping Wang, Chengyan Chen, Weibo Wang, He Fudan Univ Key Lab Computat Neurosci & Brain Inspired Intell Minist Educ Shanghai 200437 Peoples R China Fudan Univ Inst Sci & Technol Brain Inspired Intelligence Shanghai 200437 Peoples R China Zhejiang Univ Affiliated Hosp 1 Sch Med Dept Radiol Hangzhou 310027 Zhejiang Peoples R China Shanghai Zhiyu Software Informat Co Ltd Shanghai 200023 Peoples R China Fudan Univ Human Phenome Inst Shanghai 200437 Peoples R China Philips Healthcare Shanghai 200233 Peoples R China

Objective: Accurate segmentation of liver tumors, which could help physicians make appropriate treatment decisions and assess the effectiveness of surgical treatment, is crucial for the clinical diagnosis of liver cancer. In this study, we propose a 4-dimensional (4d) deep learning model based on 3d convolution and convolutional long short-term memory (C-LSTM) for hepatocellular carcinoma (HCC) lesion segmentation. Methods: The proposed deep learning model utilizes 4d information on dynamic contrast enhanced (dCE) magnetic resonance imaging (MRI) images to assist liver tumor segmentation. Specifically, a shallow U-net based 3d CNN module was designed to extract 3d spatial domain features from each dCE phase, followed by a 4-layer C-LSTM network module for time domain information exploitation. The combined information of multi-phase dCE images and the manner by which tissue imaging features change on multi-contrast images allow the network to more effectively learn the characteristics of HCC, resulting in better segmentation performance. Results: The proposed model achieved a dice score of 0.825 +/- 0.077, a Hausdorff distance of 12.84 +/- 8.14 mm, and a volume similarity of 0.891 +/- 0.080 for liver tumor segmentation, which outperformed the 3d U-net model, RA-UNet model and other models in the ablation study in both internal and external test sets. Moreover, the performance of the proposed model is comparable to the nnU-Net model, which showed state-of-the-art performance in many segmentation tasks, with significantly reduced prediction time. Conclusion: The proposed 3d convolution and C-LSTM based model can achieve accurate segmentation of HCC lesions.

关键词： Image segmentation Liver Tumors Three-dimensional displays Magnetic resonance imaging deep learning Solid modeling 4d information deep learning 3d convolution convolutional LSTM tumor segmentation

来源：评论

学校读者我要写书评

暂无评论

Learning Modal and Spatial Features With Lightweight 3d convolution for RGB Guided depth Completion

引用

IEEE TRANSACTIONS ON CONSUMER ELECTRONICS 2021年第3期67卷 195-201页

作者： Zhao Tao Pan Shuguo Gao Wang Sun Yingchun Southeast Univ Sch Instrument Sci & Engn Nanjing 210096 Peoples R China

RGB guided depth completion aims to recover a complete depth map from a sparse set of depth measurements and one corresponding RGB image, which is efficient for 3d applications to generate high-quality depth maps. Most prevailing approaches feed the sparse depth data and RGB image collected by consumer devices into a 2d convolutional network performed only at the spatial level. We argue that there is a correlation in registered multi-modal data between different modalities, which is ignored in 2d convolutional operations, resulting in loss of accuracy. In order to acquire the extra modal information between different modalities, we adopt 3d convolution for the depth completion task. Meanwhile, to decrease the significantly increased parameter size 3d convolutions, we propose a simple and effective method to reduce this increase while retaining their modal information. We verified the effectiveness of our proposed method for modal and spatial features learning on NYUv2 and KITTI depth completion datasets. Our lightweight 3d convolution achieved approximate accuracy as the standard 3d convolution, but with the same parameter size of 2d convolution. Our proposed modal features and lightweight 3d convolution are helpful to inform the development of depth sensors for consumer devices.

关键词： convolution Three-dimensional displays Feature extraction Kernel Cameras Laser radar Standards depth completion 3d convolution lightweight model modal features

来源：评论

学校读者我要写书评

暂无评论

Constant Velocity 3d convolution

引用

IEEE ACCESS 2018年 6卷 76490-76501页

作者： Sekikawa, Yusuke Ishikawa, Kohta Saito, Hideo Denso IT Lab Inc Tokyo 1500002 Japan Keio Univ Yokohama Kanagawa 2238522 Japan

We propose a novel 3-d convolution method, cv3dconv, for extracting spatiotemporal features from videos. It reduces the number of sum-of-products operations in 3-d convolution by thousands of times by assuming the constant moving velocity of the features. We observed that a specific class of video sequences, such as video captured by an in-vehicle camera, can be well approximated with piece-wise linear movements of 2-d features in a temporal dimension. Our principal finding is that a 3-d kernel, represented by constant velocity, can be decomposed into a convolution of a 2-d-shaped kernel and a 3-d-velocity kernel, which is parameterized using only two parameters. We derived an efficient recursive algorithm for this class of 3-d convolution, which is exceptionally suited for sparse spatiotemporal data, and this parameterized decomposed representation imposes a structured regularization along a temporal direction. We experimentally verified the validity of our approximation using a controlled dataset, and we also showed the effectiveness of the cv3dconv by adopting it for deep neural networks (dNNs) in visual odometry estimation task using publicly available event-based camera dataset captured in urban road scene. Our dNN architecture improves the estimation accuracy for about 30% compared with the existing states-of-the-arts architecture designed for event data.

关键词： convolutional neural network 3d convolution spatiotemporal convolution event-based camera

来源：评论

学校读者我要写书评

暂无评论

Effective polarization-based image dehazing through 3d convolution network

引用

SIGNAL IMAGE ANd VIdEO PROCESSING 2024年第SUPPL 1期18卷 463-474页

作者： Wang, Xin Fu, Wei Yu, Haichao Zhang, Yong Hefei Univ Technol Sch Comp & Informat Hefei 230601 Anhui Peoples R China Intelligent Interconnected Syst Lab Anhui Prov Hefei 230601 Anhui Peoples R China

The presence of numerous microscopic particles within haze contributes to the scattering of atmospheric light, leading to a notable degradation in the quality of captured images. Polarization, an intrinsic attribute of the light field, encapsulates a wealth of information concerning both targets and their surrounding environment and has demonstrated robust performance in dehazing real-world hazy images. In this study, we introduce an end-to-end dehazing network tailored for processing polarization-based hazy images. Leveraging the capabilities of 3d convolution, our novel POL-3d Encoder adeptly harnesses the correlation information situated within the polarization angle dimension, derived from fused polarization images. Furthermore, we integrate a Spatial Redundancy-Reducing module, effectively mitigating redundancy within feature maps across the spatial dimension. To assess the efficacy of our network and ensure its robustness, we curate a dataset rooted in the principles of polarization-based image formation. Experimental results conclusively demonstrate that our method achieves state-of-the-art performance on both synthetic data and real-world hazy images, thereby highlighting its potential for advancing image dehazing techniques.

关键词： Image dehazing Polarimetric imaging 3d convolution End-to-end

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：