检索结果-内蒙古大学图书馆

arXiv 2022年

作者： Wang, Rui Wu, Xiao-Jun Chen, Ziheng Xu, Tianyang Kittler, Josef School of Artificial Intelligence and Computer Science Jiangnan University Wuxi214122 China Jiangsu Provincial Engineering Laboratory of Pattern Recognition and Computational Intelligence Jiangnan University China Centre for Vision Speech and Signal Processing University of Surrey GuildfordGU2 7XH United Kingdom School of Artificial Intelligence and Computer Science Jiangnan University China

Image set-based visual classification methods have achieved remarkable performance, via characterising the image set in terms of a non-singular covariance matrix on a symmetric positive definite (SPD) manifold. To adapt to complicated visual scenarios better, several Riemannian networks (RiemNets) for SPD matrix nonlinear processing have recently been studied. However, it is pertinent to ask, whether greater accuracy gains can be achieved by simply increasing the depth of RiemNets. The answer appears to be negative, as deeper RiemNets tend to lose generalization ability. To explore a possible solution to this issue, we propose a new architecture for SPD matrix learning. Specifically, to enrich the deep representations, we adopt SPDNet [1] as the backbone, with a stacked Riemannian autoencoder (SRAE) built on the tail. The associated reconstruction error term can make the embedding functions of both SRAE and of each RAE an approximate identity mapping, which helps to prevent the degradation of statistical information. We then insert several residual-like blocks with shortcut connections to augment the representational capacity of SRAE, and to simplify the training of a deeper network. The experimental evidence demonstrates that our DreamNet can achieve improved accuracy with increased depth of the network. Copyright © 2022, The Authors. All rights reserved.

关键词： Covariance matrix

来源：评论

学校读者我要写书评

暂无评论

Estimating Difficulty Score of Visual Search in Images for Semi-supervised Object Detection 1

引用

16th Pacific Rim Knowledge Acquisition Workshop, PKAW 2019 held in conjunction with the 16th Pacific Rim International Conference on Artificial intelligence, PRICAI 2019

作者： Ma, Dongliang Zhang, Haipeng Wu, Hao Zhang, Tao Sun, Jun JiangSu Provincial Engineering Laboratory of Pattern Recognition and Computational Intelligence JiangNan University Wuxi China

ISBN: (数字)9783030306397

ISBN: (纸本)9783030306380

Humans can intuitively understand the content of images, and often reach a consensus that some images are more difficult to visual search tasks than others. However, this is quite challenging for computers as it is a subjective task which may be influenced by human emotional factors. Instead of focusing on how the models make reactions on datasets, our method has a capability of assigning scores to samples respectively within a dataset that estimating the difficulty of visual search tasks. Our model shows better performance for predicting visual search difficulty scores of samples produced by human annotators in PASCAL VOC2012. Eventually, we demostrate with experiments that our method has an ability of selecting suitable samples to improve the performance of detectors in a semi-supervised task. © 2019, Springer Nature Switzerland AG.

关键词： Object detection

来源：评论

学校读者我要写书评

暂无评论

Spectral Efficiency for Multi-pair Massive MIMO Two-Way Relay Networks with Hybrid Processing 8th

Spectral Efficiency for Multi-pair Massive MIMO Two-Way Rela...

引用

8th International Conference on Communications, Signal Processing, and Systems, CSPS 2019

作者： Wang, Hongyan Li, Zhengquan Xue, Xiaomei Li, Baolong Liu, Yang Wu, Guilu Wu, Qiong Graduate School of Engineering Nagasaki University Nagasaki852-8521 Japan Jiangsu Provincial Engineering Laboratory of Pattern Recognition and Computational Intelligence Jiangnan University Wuxi214122 China National Mobile Communication Research Laboratory Southeast University Nanjing210096 China

ISBN: (纸本)9789811394089

In millimeter wave (mm-wave) communication systems, large antenna array can be employed for higher data rates. Hybrid digital and analogue beamforming design is adopted to reduce the digital signal processing (DSP) power consumption and the circuitry complexity. In this paper, we investigate the multi-pair massive MIMO two-way relay networks with hybrid processing architecture. Furthermore, the power scaling scheme is also considered. We got some conclusions that when the number of antennas at relay station M goes huge, the asymptotic spectral efficiency can be obtained. We also compared three cases of power scaling schemes and obtained the optimal case specially when M tends to considerable large. We can draw the conclusion that when, the Case 2 (only the transmit power of users scaled by M in power scaling schemes) is the optimal case compared with the other two cases. Finally, the Monte-Carlo is employed to demonstrate the validity of analytical results. © 2020, Springer Nature Singapore Pte Ltd.

关键词： Spectrum efficiency

来源：评论

学校读者我要写书评

暂无评论

Editorial

引用

International Journal of Bio-Inspired Computation 2020年第2期16卷 67-67页

作者： Fang, Wei Wu, Xiaojun Jiangsu Provincial Engineering Laboratory of Pattern Recognition and Computational Intelligence Department of Computer Science and Technology Jiangnan University 214122 China

来源：评论

学校读者我要写书评

暂无评论

MoNET: No-reference image quality assessment based on a multi-depth output network

引用

Journal of Electronic Imaging 2021年第4期30卷 043007-043007页

作者： Sang, Qingbing Su, Chenfei Zhu, Lingying Liu, Lixiong Wu, Xiaojun Bovik, Alan C. Jiangnan University School of Artificial Intelligence and Computer Science Wuxi China Jiangsu Provincial Engineering Laboratory of Pattern Recognition and Computational Intelligence Wuxi China Beijing Institute of Technology School of Computer Science and Technology Beijing China University of Texas at Austin Laboratory for Image and Video Engineering AustinTX United States

When deep convolutional neural networks perform feature extraction, the features computed at each layer express different abstractions of visual information. The earlier layers extract highly compact low-level features such as bandpass and directional primitives, whereas deeper layers extract structural features of increasing abstraction, similar to contours, shapes, and edges, becoming less effable as the depth increases. We propose a different kind of end-to-end no-reference (NR) image quality assessment (IQA) model, which is defined as a multi-depth output convolutional neural network (MoNET). It accomplishes this by mapping both shallow and deep features to perceived quality. MoNET delivers three outputs that express shallow (lower-level) and deep (high-level) features, and maps them to subjective quality scores. The multiple outputs are combined into a single, final quality score. MoNET does this by combining the responses of three learning machines, so it may be viewed as a form of ensemble learning. The experimental results on three public image quality databases show that our proposed model achieves better performance than other state-of-the-art NR IQA algorithms. © 2021 SPIE and IS and T.

关键词： Convolutional neural networks

来源：评论

学校读者我要写书评

暂无评论

Low-Complexity Channel Estimation Method Based on ISSOR-PCG for Massive MIMO Systems 8th

Low-Complexity Channel Estimation Method Based on ISSOR-PCG ...

引用

8th International Conference on Communications, Signal Processing, and Systems, CSPS 2019

作者： Zhou, Cheng Li, Zhengquan Wu, Qiong Liu, Yang Li, Baolong Wu, Guilu Zhao, Xiaoqing Jiangsu Provincial Engineering Laboratory of Pattern Recognition and Computational Intelligence Jiangnan University Wuxi214122 China National Mobile Communication Research Laboratory Southeast University Nanjing210096 China Department of Electronic Engineering Tsinghua University Beijing100084 China The National Key Laboratory of millimeter wave Southeast University Nanjing210096 China

ISBN: (纸本)9789811394089

With the boost of the quantity of antennas at the base station (BS) in massive multiple-input multiple-output (MIMO) systems, the channel capacity and spectral efficiency are also increased. Conventional channel estimation method, such as the classical minimum mean square error (MMSE), which involves the matrix inversion in large size with enormous computational complexity, especially in massive MIMO systems due to large antenna arrays. To degrade the complexity caused by the inversion of the matrix, a low-complexity channel estimation scheme is proposed based on the improved symmetric successive over relaxation preconditioned conjugate gradient (ISSOR-PCG) method to avoid computing the matrix inversion directly. A simple way is also introduced to address the optimal relaxation parameter for the proposed scheme, by utilizing the channel asymptotic orthogonality in massive MIMO systems. Analysis shows that the proposed channel estimator is able to degrade the complexity effectively compared with MMSE channel estimator. Simulation results illustrate that the proposed scheme can obtain near-optimal performance to the classical MMSE estimation method and outperforms other baseline schemes with increased number of iterations. © 2020, Springer Nature Singapore Pte Ltd.

关键词： Mean square error

来源：评论

学校读者我要写书评

暂无评论

PPT Fusion: Pyramid Patch Transformer for a Case Study in Image Fusion

arXiv

引用

arXiv 2021年

作者： Fu, Yu Xu, Tianyang Wu, Xiao-Jun Kittler, Josef Jiangsu Provincial Engineering Laboratory of Pattern Recognition and Computational Intelligence School of Artificial Intelligence and Computer Science Jiangnan University Wuxi 214122 China Centre for Vision Speech and Signal Processing University of Surrey GuildfordGU2 7XH United Kingdom

The Transformer architecture has witnessed a rapid development in recent years, outperforming the CNN architectures in many computer vision tasks, as exemplified by the Vision Transformers (ViT) for image classification. However, existing visual transformer models aim to extract semantic information for high-level tasks, such as classification and detection. These methods ignore the importance of the spatial resolution of the input image, thus sacrificing the local correlation information of neighboring pixels. In this paper, we propose a Patch Pyramid Transformer(PPT) to effectively address the above issues. Specifically, we first design a Patch Transformer to transform the image into a sequence of patches, where transformer encoding is performed for each patch to extract local representations. In addition, we construct a Pyramid Transformer to effectively extract the non-local information from the entire image. After obtaining a set of multi-scale, multi-dimensional, and multi-angle features of the original image, we design the image reconstruction network to ensure that the features can be reconstructed into the original input. To validate the effectiveness, we apply the proposed Patch Pyramid Transformer to image fusion tasks. The experimental results demonstrate its superior performance, compared to the state-of-the-art fusion approaches, achieving the best results on several evaluation indicators. Thanks to the underlying representational capacity of the PPT network, it can directly be applied to different image fusion tasks without redesigning or retraining the network. Copyright © 2021, The Authors. All rights reserved.

关键词： Deep learning

来源：评论

学校读者我要写书评

暂无评论

RFN-Nest: An end-to-end residual fusion network for infrared and visible images

arXiv

引用

arXiv 2021年

作者： Li, Hui Wu, Xiao-Jun Kittler, Josef Jiangsu Provincial Engineering Laboratory of Pattern Recognition and Computational Intelligence School of Artificial Intelligence and Computer Science Jiangnan University Wuxi214122 China The Center for Vision Speech and Signal Processing University of Surrey GuildfordGU2 7XH United Kingdom

In the image fusion field, the design of deep learning-based fusion methods is far from routine. It is invariably fusion-task specific and requires a careful consideration. The most difficult part of the design is to choose an appropriate strategy to generate the fused image for a specific task in hand. Thus, devising learnable fusion strategy is a very challenging problem in the community of image fusion. To address this problem, a novel end-to-end fusion network architecture (RFN-Nest) is developed for infrared and visible image fusion. We propose a residual fusion network (RFN) which is based on a residual architecture to replace the traditional fusion approach. A novel detail-preserving loss function, and a feature enhancing loss function are proposed to train RFN. The fusion model learning is accomplished by a novel two-stage training strategy. In the first stage, we train an auto-encoder based on an innovative nest connection (Nest) concept. Next, the RFN is trained using the proposed loss functions. The experimental results on public domain data sets show that, compared with the existing methods, our end-to-end fusion network delivers a better performance than the state-of-the-art methods in both subjective and objective evaluation. The code of our fusion method is available at https://***/hli1221/imagefusion-rfn-nest. © 2021, CC BY-NC-ND.

关键词： Image fusion

来源：评论

学校读者我要写书评

暂无评论

The optimal graph regularized sparse coding with application to image representation 1

引用

2nd Chinese Conference on pattern recognition and Computer Vision, PRCV 2019

作者： Shu, Zhenqiu Wu, Xiaojun Liu, Zhen You, Congzhe Fan, Honghui School of Computer Engineering Jiangsu University of Technology Changzhou231001 China Jiangsu Provincial Engineering Laboratory of Pattern Recognition and Computational Intelligence Jiangnan University Wuxi214122 China

ISBN: (数字)9783030317232

ISBN: (纸本)9783030317225

Sparse representation has shown its superiority and effectiveness in many real applications in recent years. However, it is still an open problem to effectively preserve the intrinsic geometric structure of data in new representation space. In this paper, we propose a novel method, called the Optimal Graph regularized Sparse Coding (OGSC), to deal with the high dimensional data. Specifically, we impose a rank constraint on the Laplacian matrix of the graph model, and thus can learn the optimal graph to preserve the manifold structure of data in each iteration. Additionally, the optimization scheme for our proposed method is also provided in this paper. The experimental results on three benchmark datasets have shown that our proposed OGSC method outperforms other stat-of-the-art methods. © Springer Nature Switzerland AG 2019.

关键词： Laplace transforms

来源：评论

学校读者我要写书评

暂无评论

Differentiable neural architecture learning for efficient neural network design

arXiv

引用

arXiv 2021年

作者： Guo, Qingbei Wu, Xiao-Jun Kittler, Josef Feng, Zhiquan Jiangsu Provincial Engineering Laboratory of Pattern Recognition and Computational Intelligence Jiangnan University Wuxi214122 China Shandong Provincial Key Laboratory of Network based Intelligent Computing University of Jinan Jinan250022 China Centre for Vision Speech and Signal Processing University of Surrey GuildfordGU2 7XH United Kingdom

Automated neural network design has received ever-increasing attention with the evolution of deep convolutional neural networks (CNNs), especially involving their deployment on embedded and mobile platforms. One of the biggest problems that neural architecture search (NAS) confronts is that a large number of candidate neural architectures are required to train, using, for instance, reinforcement learning and evolutionary optimisation algorithms, at a vast computation cost. Even recent differentiable neural architecture search (DNAS) samples a small number of candidate neural architectures based on the probability distribution of learned architecture parameters to select the final neural architecture. To address this computational complexity issue, we introduce a novel architecture parameterisation based on scaled sigmoid function, and propose a general Differentiable Neural Architecture Learning (DNAL) method to optimize the neural architecture without the need to evaluate candidate neural networks. Specifically, for stochastic supernets as well as conventional CNNs, we build a new channel-wise module layer with the architecture components controlled by a scaled sigmoid function. We train these neural network models from scratch. The network optimization is decoupled into the weight optimization and the architecture optimization, which avoids the interaction between the two types of parameters and alleviates the vanishing gradient problem. We address the non-convex optimization problem of neural architecture by the continuous scaled sigmoid method with convergence guarantees. Extensive experiments demonstrate our DNAL method delivers superior performance in terms of neural architecture search cost, and adapts to conventional CNNs (e.g., VGG16 and ResNet50), lightweight CNNs (e.g., MobileNetV2) and stochastic supernets (e.g., ProxylessNAS). The optimal networks learned by DNAL surpass those produced by the state-of-the-art methods on the benchmark CIFAR-10 and ImageN

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：