检索结果-内蒙古大学图书馆

Multilevel Attention Based U-Shape Graph Neural Network for Point Clouds Learning

IEEE TRANSACTIONS ON INDUSTRIAL INFORMATICS 2022年第1期18卷 448-456页

作者： Zou, Xiaofeng Li, Kenli Chen, Cen Hunan Univ Coll Informat Sci & Engn Changsha 410082 Peoples R China Natl Supercomp Ctr Changsha Changsha 410082 Peoples R China Infocomm Res Inst I2R Singapore 126777 Singapore

With the popularity of 3-D sensors in industrial Internet of Things (IIoT), point clouds learning is increasingly important. In this article, we propose a novel multilevel attention based U-shape graph neural network (MAUGNN) for point clouds learning, which can effectively learn the features from low-level to high-level and fuse multiple-level features based on the graph neural networks and attention mechanism. There are three parts in MAUGNN: encoder, decoder, and connections. In the encoder and decoder, we design an attention-based graph convolution to explore the structural information for point clouds. During the encoder, a structure-aware attention pooling is proposed to support down-sampling on point cloud data. To adaptively fuse coarse-grained features from the encoder and fine-grained features from the decoder together, we also propose a structure-aware attention skip connection mechanism. Extensive experiments on popular point cloud datasets demonstrate the superior performance of our MAUGNN over state-of-the-art baselines.

关键词： Three-dimensional displays Decoding Fuses Task analysis Feature extraction Encoding Semantics Attention mechanism encoder-decoder graph neural networks (GNNs) point clouds

来源：评论

学校读者我要写书评

暂无评论

A review of sign language recognition research

引用

JOURNAL OF INTELLIGENT & FUZZY SYSTEMS 2022年第4期43卷 3879-3898页

作者： Yu, Ming Jia, Jingli Xue, Cuihong Yan, Gang Guo, Yingchun Liu, Yuehao Hebei Univ Technol Sch Artificial Intelligence Tianjin Peoples R China Tianjin Univ Technol Tech Coll Deaf Tianjin Peoples R China

Sign language is the primary way of communication between hard-of-hearing and hearing people. Sign language recognition helps promote the better integration of deaf and hard-of-hearing people into society. We reviewed 95 types of research on sign language recognition technology from 1993 to 2021, analyzing and comparing algorithms from three aspects of gesture, isolated word, and continuous sentence recognition, elaborating the evolution of sign language acquisition equipment and we summarized the datasets of sign language recognition research and evaluation criteria. Finally, the main technology trends are discussed, and future challenges are analyzed.

关键词： Sign language recognition convolutional neural network encoder-decoder dataset

来源：评论

学校读者我要写书评

暂无评论

FRDD-Net: Automated Carotid Plaque Ultrasound Images Segmentation Using Feature Remapping and Dense Decoding

引用

SENSORS 2022年第3期22卷 887页

作者： Li, Yanhan Zou, Lian Xiong, Li Yu, Fen Jiang, Hao Fan, Cien Cheng, Mofan Li, Qi Wuhan Univ Elect Informat Sch Wuhan 430072 Peoples R China Wuhan Univ Zhongnan Hosp Cardiovasc Ultrasound Dept Wuhan 430071 Peoples R China Huazhong Univ Sci & Technol Cent Hosp Wuhan Tongji Med Coll Dept Ultrasound Wuhan 430014 Peoples R China

Automated segmentation and evaluation of carotid plaques ultrasound images is of great significance for the diagnosis and early intervention of high-risk groups of cardiovascular and cerebrovascular diseases. However, it remains challenging to develop such solutions due to the relatively low quality of ultrasound images and heterogenous characteristics of carotid plaques. To address those problems, in this paper, we propose a novel deep convolutional neural network, FRDD-Net, with an encoder-decoder architecture to automatically segment carotid plaques. We propose the feature remapping modules (FRMs) and incorporate them into the encoding and decoding blocks to ameliorate the reliability of acquired features. We also propose a new dense decoding mechanism as part of the decoder, thus promoting the utilization efficiency of encoded features. Additionally, we construct a compound loss function to train our network to further enhance its robustness in the face of numerous cases. We train and test our network in multiple carotid plaque ultrasound datasets and our method yields the best performance compared to other state-of-the-art methods. Further ablation studies consistently show the advancement of our proposed architecture.

关键词： ultrasound segmentation deep convolutional neural networks carotid plaques encoder-decoder

来源：评论

学校读者我要写书评

暂无评论

Combine unlabeled with labeled MR images to measure acute ischemic stroke lesion by stepwise learning

引用

IET IMAGE PROCESSING 2022年第14期16卷 3965-3976页

作者： Zhao, Bin Liu, Zhiyang Liu, Guohua Wu, Mengran Cao, Chen Jin, Song Wu, Hong Ding, Shuxue Nankai Univ Tianjin Key Lab Optoelect Sensor & Sensing Networ Coll Elect Informat & Opt Engn Tianjin 300350 Peoples R China Guilin Univ Elect Technol Sch Artificial Intelligence Guilin Guangxi Peoples R China Tianjin Huanhu Hosp Key Lab Cerebral Artery & Neural Degenerat Tianji Dept Med Imaging Tianjin Peoples R China

Acute ischemic stroke is a common threat to human health and may obtain timely treatment by fast localizing and quantitatively evaluating the lesions. Most CNN-based methods try to segment and measure the lesions, however, they require a training on a large number of labeled subjects that are labor-intensive and time-consuming to obtain. In this paper, a method is proposed that can combine limited labeled subjects with abundant unlabeled subjects to alleviate the problem. The proposed method consists of two stages: stepwise learning process and segmentation process. Stepwise learning is used to obtain the pretrained encoder. The pretrained encoder and the proposed decoder are connected into a new end-to-end segmentation network, which is retrained on the labeled subjects in the segmentation process. By using 5 labeled subjects and 79 unlabeled subjects, the proposed method achieves a mean dice coefficient of 0.663 +/-$\,{\pm}\,$0.205, a mean average symmetric surface distance (ASSD) of 2.17 mm and a mean 95 percentile Hausdorff distance (HD) of 18.38 mm on a clinical MR dataset with 179 subjects. More importantly, it achieves lesion-wise F-1 score of 0.857 and a subject-wise detection rate of 0.966.

关键词： end-to-end segmentation network medical image processing convolutional neural nets image segmentation labeled MR image unlabeled MR image CNN acute ischemic stroke lesion biomedical MRI stepwise learning process learning (artificial intelligence) encoder-decoder diseases segmentation process

来源：评论

学校读者我要写书评

暂无评论

Using Quadcopter for Vision-based Tracking of High-Rise Vertical Structure 20

Using Quadcopter for Vision-based Tracking of High-Rise Vert...

引用

20th IEEE/ASME International Conference on Mechatronic and Embedded Systems and Applications (MESA)

作者： Gupta, Ayush Shukla, Amit Sivarathri, Ashok Kumar Pandya, Naisarg Indian Inst Technol Mandi Sch Mech & Mat Engn Mandi Himachal Prades India Indian Inst Technol Mandi Ctr Artificial Intelligence & Robot Mandi Himachal Prades India

ISBN: (纸本)9798331516246;9798331516239

This research paper focuses on developing vision-based autonomous navigation and motion control algorithms for quadcopters to track high-rise industrial vertical structures. A camera, fixed on the drone will capture the Red-Green-Blue (RGB) channel image and will feed to the image processing node, which is a convolution neural network (CNN) based encoder-decoder model for image segmentation. The segmented image will go to the classical computer vision algorithm to get the desired features for navigation and the motion control algorithm. In this paper, we have implemented a nonlinear sliding mode control algorithm for motion control and discuss the results of positioning and motion control of the drone for the motion in circular and spiral paths. The entire experimentation is performed on the DJI M100 drone with the Nvidia Jetson Xavier NX as an onboard computer.

关键词： Autonomous tracking Computer Vision encoder-decoder and sliding mode control

来源：评论

学校读者我要写书评

暂无评论

Attention-based dual residual network-based for multi-spectral pan-sharpening

引用

INTERNATIONAL JOURNAL OF REMOTE SENSING 2022年第13期43卷 4911-4950页

作者： Zhang, Xinyu Li, Jinjiang Hua, Zhen Shandong Technol & Business Univ Sch Informat & Elect Engn Yantai Peoples R China Shandong Technol & Business Univ Sch Comp Sci & Technol Yantai 264005 Peoples R China Coinnovat Ctr Shandong Coll & Univ Future Intelligent Comp Yantai Peoples R China

Pansharpening belongs to an important part in the field of remote sensing image fusion, which refers to the fusion of low spatial resolution multispectral (MS) images with high spatial resolution (PAN) images to finally obtain high resolution multispectral (HRMS) images. The existing deep learning-based pan-sharpening methods have achieved better results compared with the traditional methods, but there are still two problems: spectral distortion and loss of spatial detail information. We propose an end-to-end attention-based dual-residual multi-stage remote sensing image fusion network (ADRPN) for MS image and PAN image pan-sharpening based on an in-depth study of spectral information of MS images and spatial information fusion of PAN images, which consists of three main stages, each resembling an encoder-decoder, and cross-stage fusion to achieve channel domain feature stitching. The first two stages are the feature extraction stage, where the features of MS images and PAN images are extracted using the residual module, and the different features learned are used to guide the training of individual networks using CRFB (Cross residual feature block). In the third stage (image reconstruction), we use the CA (Coordinate attention) and SE (Squeeze-and-Excitation) attention mechanism to enable the network to more precisely locate the region of interest by the precise location information obtained, which allows the features extracted in the first two stages to be better fused with the original image, thus reducing the occurrence of spectral distortion and loss of spatial detail information. Qualitative and quantitative analyses of real and simulated data from the benchmark datasets QuickBird (QB), GF-2, and WorldView-2 (WV2) show that the method can better preserve the spectral and spatial detail information and obtain high-quality HRMS images.

关键词： Pansharpening multispectral (MS) images panchromatic (PAN) images residual block encoder-decoder attention CRFB

来源：评论

学校读者我要写书评

暂无评论

Global Stock Market Prediction Using Transformer-Based Deep Learning Techniques 15

Global Stock Market Prediction Using Transformer-Based Deep ...

引用

15th International Conference on Computing Communication and Networking Technologies, ICCCNT 2024

作者： Gothai, E. Thamilselvan, R. Natesan, P. Vignesh, K. Kongu Engineering College Department of Computer Technology Tamilnadu Perundurai India Kongu Engineering College Department of Cse Tamilnadu Perundurai India Kongu Engineering College Department of It Tamilnadu Perundurai India Kongu Engineering College Software Systems Department of Ct Tamilnadu Perundurai India

ISBN: (纸本)9798350370249

Deep learning approaches' use in financial market forecasting has recently drawn a lot of attention from both investors and scholars. The Transformer framework, initially created for natural language processing, is used in this study to propose a revolutionary approach for forecasting the stock market indices of seven different economies: China, Canada, India, Japan, Russia, the United Kingdom, and the United States. The well-known Transformer architecture, which excels at capturing complex non-linear patterns, is modified to study the unique dynamics of each nation's stock market. The model successfully determines the basic principles driving market behavior by using an encoder-decoder structure and a multi-head attention mechanism. The aforementioned indices are used in back-testing trials, which cover a variety of economic environments. The outcomes demonstrate that the Transformer performs better than conventional techniques, with Mean Squared Error (MSE) acting as a crucial evaluation criterion. Notably, the model shows promise for investors in these widely separated markets to receive excess profits. This study offers a thorough analysis of stock market forecasting, expanding its relevance to a wider global context. comprehension of how this strategy might adjust to various financial ecosystems is improved by the addition of seven unique economies. Future research may go more deeply into applications that are industry-specific and investigate potential extensions to additional international markets. © 2024 IEEE.

关键词： Deep Learning encoder-decoder Stock Prediction Transformer

来源：评论

学校读者我要写书评

暂无评论

A Multiple-Integration encoder for Multi-Turn Text-to-SQL Semantic Parsing

引用

IEEE-ACM TRANSACTIONS ON AUDIO SPEECH AND LANGUAGE PROCESSING 2021年 29卷 1503-1513页

作者： Wang, Run-Ze Ling, Zhen-Hua Zhou, Jing-Bo Hu, Yu Univ Sci & Technol China Natl Engn Lab Speech & Language Informat Hefei 230027 Peoples R China Baidu Res Business Intelligence Lab Beijing 100084 Peoples R China Univ Sci & Technol China Hefei 230027 Peoples R China iFLYTEK Res Hefei 230088 Peoples R China

This paper studies multi-turn text-to-SQL generation, which is a new but important task in semantic parsing. In order to deal with its two challenges, i.e., multi-turn interaction and cross-domain evaluation, this paper proposes a multiple-integration encoder, which derives the vector representations of user utterances and database schemas using three custom-designed modules for information integration. First, an utterance representation enhancing module is built to integrate the information of history utterances into the representation of each token in current utterance by attentive selection. Second, a schema discrepancy enhancing module is designed to integrate previous predicted SQL query into the representation of schema items. Third, a latent schema linking module is employed to integrate schema information into utterance representations for better dealing with unseen database schemas. These three modules are all implemented based on a lightweight multi-head attention mechanism, which reduces the number of parameters in conventional multi-head attention. Experimental results on the SParC dataset show that our method achieved better accuracy of multi-turn text-to-SQL generation than the most advanced benchmarks. Further ablations studies and analysis also demonstrate the effectiveness of the three modules designed for information integration in the encoder.

关键词： Task analysis Structured Query Language Decoding Databases Semantics History Bit error rate Text-to-SQL cross-domain multi-turn encoder-decoder lightweight multi-head attention

来源：评论

学校读者我要写书评

暂无评论

VIG-UNET: VISION GRAPH NEURAL NETWORKS FOR MEDICAL IMAGE SEGMENTATION 20

VIG-UNET: VISION GRAPH NEURAL NETWORKS FOR MEDICAL IMAGE SEG...

引用

20th IEEE International Symposium on Biomedical Imaging (ISBI)

作者： Jiang, Juntao Chen, Xiyu Tian, Guanzhong Liu, Yong Zhejiang Univ Coll Control Sci & Engn Hangzhou Peoples R China Zhejiang Univ Polytech Inst Hangzhou Peoples R China Zhejiang Univ Ningbo Innovat Ctr Ningbo Peoples R China

ISBN: (纸本)9781665473583

Deep neural networks have been widely used in medical image analysis and medical image segmentation is one of the most important tasks. U-shaped neural networks with encoder-decoder are prevailing and have succeeded greatly in various segmentation tasks. While CNNs treat an image as a grid of pixels in Euclidean space and Transformers recognize an image as a sequence of patches, graph-based representation is more generalized and can construct connections for each part of an image. In this paper, we propose a novel ViG-UNet, a graph neural network-based U-shaped architecture with the encoder, the decoder, the bottleneck, and skip connections. The downsampling and upsampling modules are also carefully designed. The experimental results on ISIC 2016, ISIC 2017 and Kvasir-SEG datasets demonstrate that our proposed architecture outperforms most existing classic and state-of-the-art U-shaped networks.

关键词： Medical image segmentation ViG-UNet Graph neural networks encoder-decoder

来源：评论

学校读者我要写书评

暂无评论

Generating Textual Description Using Modified Beam Search 29th

Generating Textual Description Using Modified Beam Search

引用

29th International Conference on Neural Information Processing

作者： Rai, Divyansh Agarwal, Arpit Kumar, Bagesh Vyas, O. P. Khan, Suhaib Shourya, S. Indian Inst Informat Technol Allahabad Prayagraj India

ISBN: (纸本)9789819916412;9789819916429

Generating textual descriptions of images by describing them in words is a fundamental problem that connects computer vision and natural language processing. A single image may include several entities, their orientations, appearance, and position in a scene as well as their complex spatial interactions, thus leading to a lot of possible captions for an image. Search algorithm of Beam Search has been employed for the task of sentence for the last couple of decades, although it returns around the similar captions with minor changes of wordings. We came across another search strategy, Diverse M-Best which uses M (M denotes the number of independent, diverse beam searches) beam searches from diverse starting statements and keeps the best output from each beam search, and removes the rest of (B-1) captions. This method would mostly lead us to many possible diverse generated sequences, but running Beam Search M several times would be computationally expensive. With the above stated works in vision, we have devised and implemented a novel algorithm, Modified Beam Search (MBS), for generation of Diverse and better captions, with an increase in the computational complexity as compared to the Beam Search. We obtained improvements on BLEU-3 and BLEU-4 scores by 1-3% over the top-2 predicted captions from the original beam search captions.

关键词： CNN encoder-decoder LSTM Modified Beam Search Search 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：