检索结果-内蒙古大学图书馆

IEEE Advanced Information Technology, Electronic and Automation Control Conference (IAEAC)

作者： Zhongyu Wu Hua Huang Qishen Li Penghui Chen School of Information Engineering Nanchang Hangkong University Nanchang Jiangxi China Key Laboratory of Jiangxi Province for Image Processing and Pattern Recognition Nanchang Jiangxi China School of Software Nanchang Hangkong University Nanchang Jiangxi China

Monocular depth estimation is a fundamental task in computer vision and has drawn increasing attention. Recently, attention-based models and encoder-decoder architectures have led to great improvements in monocular depth estimation. Typically, most of the previous methods used repeated simple up-sampling operations during decoding, which may not make full use of the potential properties of the features extracted by the encoder, and there are problems of inaccurate prediction of the edge and depth maximum region. We propose an attention-based feature fusion module for encoder and decoder. We treat the monocular depth estimation as a pixel-level optimization problem, where the coarsest encoder feature is used to initialize the pixel-level optimization, which is then refined to higher resolution by the proposed attentional feature fusion (AFF). We formulate the prediction problem as ordinal regression over the bin centers that discretize the continuous depth range. It predicts a correspondingly different distribution of bins based on different pictures and we predict bins at the coarsest level using global pooling and MLP layers. In the NYUV2 dataset, the proposed architecture improving original model by 2.5.% and 1.1%, in terms of Log10 and Absolute relative error, respectively.

关键词：

来源：评论

学校读者我要写书评

暂无评论

Research on Multi-level Attention-based Human Pose Estimation

Research on Multi-level Attention-based Human Pose Estimatio...

引用

Algorithms, High Performance Computing and Artificial Intelligence (AHPCAI), International Conference on

作者： Jun Gao Hua Huang QiShen Li ShaoQi Xia School of Information Engineering Nanchang Hangkong University Nanchang Jiangxi China Key Laboratory of Jiangxi Province for Image Processing and Pattern Recognition Nanchang Jiangxi China School of Software Nanchang Hangkong University Nanchang Jiangxi China

Detecting human key points from a single image is very challenging due to occlusion, blurring, illumination and scale changes. In this paper, this problem is addressed by designing an effective network structure. Since global and local information plays an important role in reasoning about human body structure and invisible keypoints, Multi-level Attention Network (MAN) is proposed. First, compared with traditional multi-resolution networks, it enables multi-resolution feature maps with greater information variance by generating them directly from the highest resolution feature map, which in turn increases the abundance of feature information after final fusion. Secondly, it effectively integrates global and local information in different resolution feature maps through the Feature Alignment Attention Block(FAAB), and intensifies them in a targeted manner. On the COCO dataset, with HRNet (Sun K. et al [1]) as the baseline network, HRNet of inserted MAN improves 1.1-2.3 AP points over the baseline network.

关键词： Computational modeling High performance computing Pose estimation Lighting Feature extraction Cognition Sun

来源：评论

学校读者我要写书评

暂无评论

Estimating the Connection Probability for 2-D Mesh Networks by Monte Carlo Methods

Estimating the Connection Probability for 2-D Mesh Networks ...

引用

International Conference on Computer and Management (CAMAN)

作者： Meilian Liang Binmei Liang Zehui Shao Xiaodong Xu School of Mathematics and Information Science Guangxi University Nanning China Key Laboratory of Pattern Recognition and Intelligent Information Processing School of Information Science & Technology University of Chengdu Chengdu China Guangxi Academy of Sciences Nanning China

Fault tolerance is a central issue in the design and implementation of interconnection networks for large parallel systems. Connection probability of a network is a good network fault tolerance measure. For a mesh of given size and node failure probability, the gap between the known upper and lower bounds on the connection probability is often very large. In this paper we design algorithms to estimate the connection probability for 2-D meshes by Monte Carlo methods. The experiment is carefully designed and performed, and the simulation results give good estimates of the connection probability for 2-D meshes and can be used to evaluate the known lower and upper bounds on connection probability for 2-D meshes.

关键词： Mesh networks Fault tolerance Fault tolerant systems Upper bound Monte Carlo methods Computers Simulation

来源：评论

学校读者我要写书评

暂无评论

Research on attention-based multiscale information fusion with the real-time instance segmentation method

Research on attention-based multiscale information fusion wi...

引用

Algorithms, High Performance Computing and Artificial Intelligence (AHPCAI), International Conference on

作者： Yifan He Hua Huang Qishen Li Guiwen Zhang School of Information Engineering Nanchang Hangkong University Nanchang Jiangxi China Key Laboratory of Jiangxi Province for Image Processing and Pattern Recognition Nanchang Jiangxi China School of Software Nanchang Hangkong University Nanchang Jiangxi China

Instance segmentation is a comprehensive computer vision task that involves a wide range of other tasks. Recently, the study of real-time instance segmentation methods has received more attention for the development of autonomous driving. Although existing real-time instance segmentation methods are fast, their accuracy does not meet practical needs. Most methods go for segmentation based on object detection, and their effectiveness is overly dependent on the effectiveness of detection. This paper proposes a new attention-based multiscale information fusion method based on Cheng, T. et al. [1]. Firstly, the PPM module of the baseline network is replaced with the module Multiscale Context Attention (MSCA) designed in this paper based on the baseline network, which uses atrous convolution with different ratios to obtain information of four scales, and then uses non-local attention to enhance the information of features. It can effectively suppress the interference of redundant information on the instance segmentation results. Secondly, a new feature fusion approach is designed, which no longer uses bilinear interpolation, but sub-pixel up sampling combined with attention. We did experiments related to this module on the coco dataset and demonstrated its effectiveness, with a 0.5% improvement over the baseline network.

关键词： Interpolation Computer vision Convolution High performance computing Interference Object detection Real-time systems

来源：评论

学校读者我要写书评

暂无评论

Nonconvex penalties with analytical solutions for one-bit compressive sensing

arXiv

引用

arXiv 2017年

作者： Huang, Xiaolin Yan, Ming Institute of Image Processing and Pattern Recognition Shanghai Jiao Tong University MOE Key Laboratory of System Control and Information Processing Shanghai200240 China Department of Computational Mathematics Science and Engineering Department of Mathematics Michigan State University East LansingMI48824 United States

One-bit measurements widely exist in the real world and can be used to recover sparse signals. This task is known as one-bit compressive sensing (1bit-CS). In this paper, we propose novel algorithms based on both convex and nonconvex sparsity-inducing penalties for robust 1bit-CS. We consider the dual problem, which has only one variable and provides a sufficient condition to verify whether a solution is globally optimal or not. For positive homogeneous penalties, a globally optimal solution can be obtained in two steps: a proximal operator and a normalization step. For other penalties, we solve the dual problem, and it needs to evaluate the proximal operators for many times. Then we provide fast algorithms for finding analytical solutions for three penalties: minimax concave penalty (MCP), 0norm, and sorted 1penalty. Specifically, our algorithm is more than 200 times faster than the existing algorithm for MCP. Its efficiency is comparable to the algorithm for the 1penalty in time, while its performance is much better than 1. Among these penalties, sorted 1is most robust to noise in different settings. Copyright © 2017, The Authors. All rights reserved.

关键词： Compressed sensing

来源：评论

学校读者我要写书评

暂无评论

An efficient algorithm and implementation for residue to binary number conversion

An efficient algorithm and implementation for residue to bin...

引用

International Conference on Communications, Circuits and Systems (ICCCAS)

作者： Chengyi Xiong Zhirong Gao Jinwen Tian Key Laboratory of Education Commission for Image Processing and Intelligent Control Institute of Pattern Recognition & Artificial Intelligence Huazhong University of Science and Technology Wuhan China Department of Computer Science Wuhan University of Science & Engineering Wuhan China Key Laboratory of Education Commission for Image Processing Huazhong University of Science & Technology China

The residue number system (RNS) has computational advantages in addition and multiplication compared with weighted number systems, such as the binary number system (BNS), since operations on residue digits are performed independently and these processes can be performed in parallel. Thus they are widely used in digital signal processing etc. Since residue to binary conversion is critical and difficult for the practicality of RNS, in this paper, a novel residue to binary (R/B) conversion algorithm for the restricted moduli set (2/sup n/ -1, 2/sup n/, 2n+1), based on exploring the periodicity of modulo (2/sup n/ /spl plusmn/ 1) operations is presented. A new 2n-bit adder based R/B converter is also proposed. The performance comparison results demonstrate that the new converter is faster and requires less area compared with the others reported in the previous literature.

关键词： Signal processing algorithms Image converters Educational institutions Computer science Concurrent computing Digital signal processing Arithmetic Laboratories Image processing Intelligent control

来源：评论

学校读者我要写书评

暂无评论

Adaptive dynamic loading and unloading mechanism applied to development environment for image processing algorithm

Adaptive dynamic loading and unloading mechanism applied to ...

引用

2011 International Conference on Computer Science and Service System, CSSS 2011

作者： Lu, Haifeng Zhang, Tianxu Yan, Wei Cheng, Zhao State Key Laboratory for Multispectral Information Processing Technologies Institute for Pattern Recognition and Artificial Intelligence Huazhong University of Science and Technology Wuhan 430074 China Department of Technology Hubei Daily Media Group Wuhan 430077 China

ISBN: (纸本)9781424497638

An advanced development environment contributes to research process greatly. This paper presents a novel adaptive dynamic loading and unloading mechanism applied to development environment for image processing algorithm and two implementation methods of the mechanism. Based on the inherent characteristic of variable arguments of functions in programming language and explicit linking technique for dynamic link library (DLL), the mechanism is to integrate the image processing algorithm into the DLL file which is independent from the development environment, and to make the development environment identify the normative algorithm DLL and load functions of image processing algorithm successfully when needed. The proposed mechanism has been implemented in a development environment written by C/C++ language. The results show that the development environment based on the novel mechanism can improve work efficiency by making researchers only focus on the development of image processing algorithm rather than the other inefficient and repetitive work. At the same time the standardability, reusability and confidentiality of image processing algorithm can be promoted greatly. This mechanism is also applicable to other similar systems. © 2011 IEEE.

关键词： Reusability

来源：评论

学校读者我要写书评

暂无评论

VIRTUAL WORK SPACE FOR ASSEMBLY

引用

SYSTEMS AND COMPUTERS IN JAPAN 1994年第7期25卷 92-100页

作者： HIRATA, Y MIZUGUCHI, T SATO, M KAWARADA, H Precision and Intelligence Laboratory Tokyo Institute of Technology Yokohama Japan 227 Members Hiroshi Kawarda is currently a Professor in the Precision and Intelligence Laboratory Tokyo Institute of Technology. His research interests include digital processing and pattern recognition. He received a B.E. in Electrical Engineering in 1962 and a doctorate in 1967 from the Tokyo Institute of Technology.

To implement a human interface for efficiently modeling 3-D shapes on a computer, a virtual works space enabling direct manipulation of a shape model on a computer similar to manipulating a real 3-D object must be constructed. In this virtual work space, the tactile and force sensations arising while the object is manipulated must be fed back to the person. We previously proposed the space interface device SPIDAR capable of force feedback to the person's fingertips. However, only one finger could be used in SPIDAR, so operations where the object is held were not possible. In this paper, we extend SPIDAR and propose the space interface device SPIDAR II to implement a virtual work space where operations involving holding and moving objects are possible. This device measures the position of the minimum of the two fingers needed to manipulate objects and can provide tactile feedback to the fingertips while an object is being held. A prototype device is built and a virtual work space is constructed where holding and moving objects is possible. Further, an experiment is conducted to investigate the effect of tactile feedback to work operability in this virtual work space, and its effectiveness is confirmed.

关键词： VIRTUAL REALITY VIRTUAL WORK SPACE ASSEMBLY SHAPE MODELING HUMAN INTERFACE

来源：评论

学校读者我要写书评

暂无评论

New design for 9/7-tap wavelet filter and its VLSI implementation

引用

Jisuanji Xuebao/Chinese Journal of Computers 2008年第3期31卷 411-418页

作者： Tian, Xin Tan, Yi-Hua Tian, Jin-Wen State Key Laboratory for Multi-Spectral Information Processing Technologies Institute for Pattern Recognition and Artificial Intelligence Wuhan 430074 China

This paper designs a new kind of lifting 9-7-tap wavelet filter with binary characteristics and presents a high speed VLSI structure for the wavelet filter. The coefficients of the lifting filters are binary numbers, so it can simplify the VLSI design. The test results for compression performances have shown that the new filter is good almost as CDF97 under PSNR. when the data has a finite accuracy, it may have a better performance than CDF97. The filter can be implemented with using adders and shifters. Therefore the hardware resources can be saved and the critical path can be shortened. According to using the folding technology and the retiming technology, the architecture of the design can be transformed into a kind of embedded folding architecture which leads to the parallel computation of addition operations. So the critical path can be shortened to nearly the time of a addition operation and the utilization of the hardware resources also has a good performance. Simulation results show that the max working frequency could almost reach 250 MHz which is the four times by the CDF97;The occupied logic cell is reduced by 66.7 compared with the CDF97+4 stages pipeline. So it is especially suitable for the high speed VLSI design.

关键词： VLSI circuits

来源：评论

学校读者我要写书评

暂无评论

A Semantic Segmentation Method of Buildings in Remote Sensing Image Based on Improved UNet 2

A Semantic Segmentation Method of Buildings in Remote Sensin...

引用

2nd International Conference on Signal Image processing and Communication, ICSIPC 2022

作者： Li, Zhongyu Liu, Yang Kuang, Yin Wang, Huajun Liu, Cheng College of Computer Science Chengdu Normal University Chengdu611130 China College of Geophysics Chengdu University of Technology Chengdu610059 China Key Laboratory of Pattern Recognition and Intelligent Information Processing of Sichuan Chengdu University Chengdu610106 China Artificial Intelligence Key Laboratory of Sichuan Province Zigong643000 China Key Laboratory of interior Layout optimization and Security Institutions of Higher Education of Sichuan Province Chengdu Normal University Sichuan Chengdu611130 China College of Movie and Media Sichuan Normal University Chengdu610066 China

ISBN: (纸本)9781510657694

Aiming at the problem of model instability and overfitting of deep neural networks with the deepening of the number of network layers, the current mainstream method is to use batch normalization (BN) to alleviate them. However, since the BN method is more sensitive to batch size when the batch size is small, the model performance will be poor. For a relatively large model, due to the limitation of video memory, the batch size cannot take a large value, limiting the model's performance. Because of the dependence of BN on batch size, this paper adopts group normalization (GN) instead of batch normalization (BN) in the UNet network to alleviate the impact of the model on batch size. Then experiments are carried out on the WHUBuilding dataset. The experimental results show that the improved model (UNet-GN) improves the mean intersection over union (MIoU) and mean pixel accuracy (MPA) by 10.66% and 1.65% respectively compared with the original model (UNet-BN). © 2022 SPIE.

关键词： Deep neural networks

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：