检索结果-内蒙古大学图书馆

Combination of temporal-channels correlation information and bilinear feature for action recognition

IET COMPUTER VISION 2020年第8期14卷 634-641页

作者： Cai, Jiahui Hu, Jianguo Li, Shiren Lin, Jialing Wang, Jun Sun Yat Sen Univ Sch Elect & Informat Technol Guangzhou Peoples R China Sun Yat Sen Univ Sch Microelect Sci & Technol Zhuhai Peoples R China Dev Res Inst Guangzhou Smart City Guangzhou Peoples R China

In this study, the authors focus on improving the spatio-temporal representation ability of three-dimensional (3D) convolutional neural networks (CNNs) in the video domain. They observe two unfavourable issues: (i) the convolutional filters only dedicate to learning local representation along input channels. Also they treat channel-wise features equally, without emphasising the important features;(ii) traditional global average pooling layer only captures first-order statistics, ignoring finer detail features useful for classification. To mitigate these problems, they proposed two modules to boost 3D CNNs' performance, which are temporal-channel correlation (TCC) and bilinear pooling module. The TCC module can capture the information of inter-channel correlations over the temporal domain. Moreover, the TCC module generates channel-wise dependencies, which can adaptively re-weight the channel-wise features. Therefore, the network can focus on learning important features. With regards to the bilinear pooling module, it can capture more complex second-order statistics in deep features and generate a second-order classification vector. We can get more accurate classification results by combining the first-order and second-order classification vector. Extensive experiments show that adding our proposed modules to 130 network could consistently improve the performance and outperform the state-of-the-art methods. The code and models are available at https://***/ caijh33/13D_TCC_Bilinear.

关键词： learning (artificial intelligence) vectors object recognition image representation image classification stereo image processing convolutional neural nets correlation methods higher order statistics video signal processing temporal‐channels correlation information bilinear feature three‐dimensional convolutional neural networks video domain 3D CNN convolutional filters channel‐wise features first‐order statistics temporal‐channel correlation bilinear pooling module TCC module channel‐wise dependencies second‐order statistics deep features second‐order classification vector first‐order classification vector spatio–temporal representation I3D network

来源：评论

学校读者我要写书评

暂无评论

Local signal adaptivity: provable feature learning in neural networks beyond kernels 21

Local signal adaptivity: provable feature learning in neural...

引用

Proceedings of the 35th International Conference on neural Information processing Systems

作者： Stefani Karp Ezra Winston Yuanzhi Li Aarti Singh Carnegie Mellon University and Google Research Carnegie Mellon University

ISBN: (纸本)9781713845393

neural networks have been shown to outperform kernel methods in practice (including neural tangent kernels). Most theoretical explanations of this performance gap focus on learning a complex hypothesis class; in some cases, it is unclear whether this hypothesis class captures realistic data. In this work, we propose a related, but alternative, explanation for this performance gap in the image classification setting, based on finding a sparse signal in the presence of noise. Specifically, we prove that, for a simple data distribution with sparse signal amidst high-variance noise, a simple convolutional neural network trained using stochastic gradient descent simultaneously learns to threshold out the noise and find the signal. On the other hand, the corresponding neural tangent kernel, with a fixed set of predetermined features, is unable to adapt to the signal in this manner. We supplement our theoretical results by demonstrating this phenomenon empirically: in CIFAR-10 and MNIST images with various backgrounds, as the background noise increases in intensity, a CNN's performance stays relatively robust, whereas its corresponding neural tangent kernel sees a notable drop in performance. We therefore propose the local signal adaptivity (LSA) phenomenon as one explanation for the superiority of neural networks over kernel methods.

关键词：

来源：评论

学校读者我要写书评

暂无评论

Semi-automated and interactive segmentation of contrast-enhancing masses on breast DCE-MRI using spatial fuzzy clustering

引用

BIOMEDICAL signal processing AND CONTROL 2022年第PartA期71卷 103113-103113页

作者： Militello, Carmelo Rundo, Leonardo Dimarco, Mariangela Orlando, Alessia Conti, Vincenzo Woitek, Ramona D'Angelo, Ildebrando Bartolotta, Tommaso Vincenzo Russo, Giorgio Italian Natl Res Council IBFM CNR Inst Mol Bioimaging & Physiol I-90015 Cefalu PA Italy Univ Cambridge Dept Radiol Cambridge CB2 0QQ England Canc Res UK Cambridge Ctr Cambridge CB2 0RE England Univ Hosp Paolo Giaccone Dept Biomed Neurosci & Adv Diagnost BiND Sect Radiol I-90127 Palermo Italy Univ Enna KORE Fac Engn & Architecture I-94100 Enna Italy Med Univ Vienna Dept Biomed Imaging & Image Guided Therapy A-1090 Vienna Austria Fdn Ist G Giglio Dept Radiol I-90015 Cefalu PA Italy

Multiparametric Magnetic Resonance Imaging (MRI) is the most sensitive imaging modality for breast cancer detection and is increasingly playing a key role in lesion characterization. In this context, accurate and reliable quantification of the shape and extent of breast cancer is crucial in clinical research environments. Since conventional lesion delineation procedures are still mostly manual, automated segmentation approaches can improve this time-consuming and operator-dependent task by annotating the regions of interest in a reproducible manner. In this work, a semi-automated and interactive approach based on the spatial Fuzzy C-Means (sFCM) algorithm is proposed, used to segment masses on dynamic contrast-enhanced (DCE) MRI of the breast. Our method was compared against existing approaches based on classic image processing, namely (i) Otsu's method for thresholding-based segmentation, and (ii) the traditional FCM algorithm. A further comparison was performed against state-of-the-art Convolutional neural Networks (CNNs) for medical image segmentation, namely SegNet and U-Net, in a 5-fold cross-validation scheme. The results showed the validity of the proposed approach, by significantly outperforming the competing methods in terms of the Dice similarity coefficient (84.47 +/- 4.75). Moreover, a Pearson's coefficient of rho = 0.993 showed a high correlation between segmented volume and the gold standard provided by clinicians. Overall, the proposed method was confirmed to outperform the competing literature methods. The proposed computer-assisted approach could be deployed into clinical research environments by providing a reliable tool for volumetric and radiomics analyses.

关键词： Semi-automated segmentation Breast cancer Unsupervised fuzzy clustering Spatial information Computer-assisted lesion detection Magnetic Resonance Imaging

来源：评论

学校读者我要写书评

暂无评论

Uni-COAL: A Unified Framework for Cross-Modality Synthesis and Super-Resolution of MR images

arXiv

引用

arXiv 2023年

作者： Song, Zhiyun Qi, Zengxin Wang, Xin Zhao, Xiangyu Shen, Zhenrong Wang, Sheng Fei, Manman Wang, Zhe Zang, Di Chen, Dongdong Yao, Linlin Wang, Qian Wu, Xuehai Zhang, Lichi The School of Biomedical Engineering Shanghai Jiao Tong University Shanghai200030 China Department of Neurosurgery Huashan Hospital Shanghai Medical College Fudan University National Center for Neurological Disorders Shanghai Key Laboratory of Brain Function and Restoration and Neural Regeneration State Key Laboratory of Medical Neurobiology and MOE Frontiers Center for Brain Science School of Basic Medical Sciences Institutes of Brain Science Fudan University China School of Biomedical Engineering ShanghaiTech University Shanghai201210 China Shanghai Clinical Research Trial Center Shanghai201210 China

Cross-modality synthesis (CMS), super-resolution (SR), and their combination (CMSR) have been extensively studied for magnetic resonance imaging (MRI). Their primary goals are to enhance the imaging quality by synthesizing the desired modality and reducing the slice thickness. Despite the promising synthetic results, these techniques are often tailored to specific tasks, thereby limiting their adaptability to complex clinical scenarios. Therefore, it is crucial to build a unified network that can handle various image synthesis tasks with arbitrary requirements of modality and resolution settings, so that the resources for training and deploying the models can be greatly reduced. However, none of the previous works is capable of performing CMS, SR, and CMSR using a unified network. Moreover, these MRI reconstruction methods often treat alias frequencies improperly, resulting in suboptimal detail restoration. In this paper, we propose a Unified Co-Modulated Alias-free framework (Uni-COAL) to accomplish the aforementioned tasks with a single network. The co-modulation design of the image-conditioned and stochastic attribute representations ensures the consistency between CMS and SR, while simultaneously accommodating arbitrary combinations of input/output modalities and thickness. The generator of Uni-COAL is also designed to be alias-free based on the Shannon-Nyquist signal processing framework, ensuring effective suppression of alias frequencies. Additionally, we leverage the semantic prior of Segment Anything Model (SAM) to guide Uni-COAL, ensuring a more authentic preservation of anatomical structures during synthesis. Experiments on three datasets demonstrate that Uni-COAL outperforms the alternatives in CMS, SR, and CMSR tasks for MR images, which highlights its generalizability to wide-range applications. Copyright © 2023, The Authors. All rights reserved.

关键词： Magnetic resonance imaging

来源：评论

学校读者我要写书评

暂无评论

Digital Halftone Classification using Simplified CNN and stochastic Statistics

Digital Halftone Classification using Simplified CNN and Sto...

引用

Asia-Pacific signal and Information processing Association Annual Summit and Conference (APSIPA)

作者： Jing-Ming Guo Sankarasrinivasan S National Taiwan University of Science and Technology

ISBN: (纸本)9781665441629

Digital halftone technique is widely utilized in various printing applications to generate high quality printing output. The halftone classification is a fundamental requirement for the optimal reconstruction of binary printed images during inverse half toning and image authentication. In this study, a new classification scheme based on convolutional neural networks (CCN) is proposed. For effective training, a new approach exploiting the properties of stochastic halftone parameters is proposed to obtain the favorable halftone patch. The CNN network has been optimized to perform classification with minimal layers, and the model is trained using cross entropy, and make use of softmax classification layer. During testing, the sub-halftone patches are classified individually and then majority voting strategy is performed to identify the actual halftone class. For performance evaluation, a deep learning halftone database has been developed for the 24 halftone varieties which also includes the latest multitone images. From experimental result, it has been validated that the proposed architecture achieves average correct classification rate (ACCR) of 100% for several halftone class, and showed superior performance in comparison with state-of-the-art methods.

关键词： Printing Training Performance evaluation Databases stochastic processes Information processing Quality assessment

来源：评论

学校读者我要写书评

暂无评论

BP-DIP: A Backprojection based Deep image Prior

BP-DIP: A Backprojection based Deep Image Prior

引用

European signal processing Conference (EUSIPCO)

作者： Jenny Zukerman Tom Tirer Raja Giryes Tel Aviv University Tel Aviv Israel School of Electrical Engineering Tel Aviv University Tel Aviv Israel

ISBN: (数字)9789082797053

ISBN: (纸本)9781728150017

Deep neural networks are a very powerful tool for many computer vision tasks, including image restoration, exhibiting state-of-the-art results. However, the performance of deep learning methods tends to drop once the observation model used in training mismatches the one in test time. In addition, most deep learning methods require vast amounts of training data, which are not accessible in many applications. To mitigate these disadvantages, we propose to combine two image restoration approaches: (i) Deep image Prior (DIP), which trains a convolutional neural network (CNN) from scratch in test time using the given degraded image. It does not require any training data and builds on the implicit prior imposed by the CNN architecture; and (ii) a backprojection (BP) fidelity term, which is an alternative to the standard least squares loss that is usually used in previous DIP works. We demonstrate the performance of the proposed method, termed BP-DIP, on the deblurring task and show its advantages over the plain DIP, with both higher PSNR values and better inference run-time.

关键词： Deep learning Training Training data Tools image restoration Task analysis Electronics packaging

来源：评论

学校读者我要写书评

暂无评论

Classification of Hyperspectral image by CNN Based on Shadow Area Enhancement Through Dynamic stochastic Resonance

引用

IEEE ACCESS 2019年 7卷 134862-134870页

作者： Liu, Xuefeng Wang, Hao Meng, Yue Fu, Min Qingdao Univ Sci & Technol Coll Automat & Elect Engn Qingdao Peoples R China Ocean Univ China Coll Informat Sci & Engn Qingdao Peoples R China

Target detection and classification in the shadow area of hyperspectral images (HSIs) has always been an important problem in the field of HSI data processing. However, there are few methods to detect or classify targets in the shadow area effectively because of occlusion of objects or oblique solar radiation. Dynamic stochastic resonance (DSR) theory shows that under the synergistic action of weak input signal, noise and non-linear system, the energy of noise can be transferred into signal partially. It breaks the idea that signal can be enhanced only by eliminating noise and has been proved to be effective in many fields. In this paper, DSR is introduced into the shadow area enhancement of HSI from both spatial and spectral dimensions. Then, the enhanced HSI data can be obtained by fusing the processed shadow with the original HSI data. Finally, 2D convolutional neural networks (2D-CNN) is used to classify the enhanced HSI. To evaluate the proposed method, a real-world HSI has been processed in the experiment and the results show that DSR can improve the contrast and the spectral intensity of HSI shadow area. Compared to other state-of-the-art methods, the classification accuracy is better at containing different targets with small samples, especially in the spectral dimension.

关键词： Roads Two dimensional displays Hyperspectral imaging stochastic resonance Feature extraction Convolutional neural networks Spatial databases Convolutional neural network remote sensing images shadow region stochastic resonance DSR parameter optimization

来源：评论

学校读者我要写书评

暂无评论

Breast cancer detection using an ensemble deep learning method

引用

BIOMEDICAL signal processing AND CONTROL 2021年第0期70卷 103009-103009页

作者： Das, Abhishek Mohanty, Mihir Narayan Mallick, Pradeep Kumar Tiwari, Prayag Muhammad, Khan Zhu, Hongyin Siksha O Anusandhan ITER Bhubaneswar 751030 Odisha India KIIT Sch Comp Engn Bhubaneswar 751024 Odisha India Aalto Univ Dept Comp Sci Espoo 02150 Finland Sungkyunkwan Univ Sch Convergence Coll Comp & Informat Visual Analyt Knowledge Lab VIS2KNOW Lab Seoul 03063 South Korea Tsinghua Univ Dept Comp Sci & Technol Beijing Peoples R China

In this work, the effectiveness of the deep learning model is applied for one-dimensional data when converted to images. This work is based on the effective conversion of one-dimensional data to images and designing a stacked ensemble deep learning model that can increase the performance of classification accuracy in comparison to single models. Breast cancer detection from gene expression dataset and breast histopathology images is considered using the proposed ensemble model. The gene expression data is one-dimensional. Using the tDistributed stochastic Neighbor embedding technique and Convex Hull algorithm the one-dimensional data is converted to an image. Existing methods are using the datasets directly for training and classification, whereas the proposed method uses the dataset as well as the decomposed forms of the same for improving the performance. It involves two-stage classification. The first stage consists of three Convolutional neural Networks as the base classifiers. Empirical Wavelet Transform and Variational Mode Decomposition are the two methods used to decompose the dataset so that the models can be trained at the molecular level, making our model robust in comparison to state-of-the-art methods. The first stage classification outcomes are used to train the second stage classifier "Multilayer Perceptron". The gene expression dataset collected from Mendeley and is used for the generation of two-dimensional synthetic datasets. The synthetic datasets and breast histopathology image datasets are used for the training and validation of the proposedmodel. The improved results obtained in this work show the effectiveness of our method

关键词： Convolutional neural networks Multilayer perceptron Empirical wavelet transform Variational mode decomposition Stacking ensemble Breast cancer Invasive ductal carcinoma

来源：评论

学校读者我要写书评

暂无评论

Analysis of deep learning architectures for turbulence mitigation in long-range imagery 2

Analysis of deep learning architectures for turbulence mitig...

引用

Conference on Artificial Intelligence and Machine Learning in Defense Applications ii

作者： Vint, David Di Caterina, Gaetano Soraghan, John J. Lamb, Robert A. Humphreys, David Univ Strathclyde Ctr Signal & Image Proc Glasgow Lanark Scotland Leonardo MW Ltd Airborne & Space Syst Div Edinburgh Midlothian Scotland

ISBN: (纸本)9781510639003

In long range imagery, the atmosphere along the line of sight can result in unwanted visual effects. Random variations in the refractive index of the air causes light to shift and distort. When captured by a camera, this randomly induced variation results in blurred and spatially distorted images. The removal of such effects is greatly desired. Many traditional methods are able to reduce the effects of turbulence within images, however they require complex optimisation procedures or have large computational complexity. The use of deep learning for image processing has now become commonplace, with neural networks being able to outperform traditional methods in many fields. This paper presents an evaluation of various deep learning architectures on the task of turbulence mitigation. The core disadvantage of deep learning is the dependence on a large quantity of relevant data. For the task of turbulence mitigation, real life data is difficult to obtain, as a clean undistorted image is not always obtainable. Turbulent images were therefore generated with the use of a turbulence simulator. This was able to accurately represent atmospheric conditions and apply the resulting spatial distortions onto clean images. This paper provides a comparison between current state of the art image reconstruction convolutional neural networks. Each network is trained on simulated turbulence data. They are then assessed on a series of test images. It is shown that the networks are unable to provide high quality output images. However, they are shown to be able to reduce the effects of spatial warping within the test images. This paper provides critical analysis into the effectiveness of the application of deep learning. It is shown that deep learning has potential in this field, and can be used to make further improvements in the future.

关键词： Turbulence Mitigation Deep Learning CNN Turbulence Simulation

来源：评论

学校读者我要写书评

暂无评论

Detection of Knocking Combustion Using the Continuous Wavelet Transformation and a Convolutional neural Network

引用

ENERGIES 2021年第2期14卷 439-439页

作者： Kefalas, Achilles Ofner, Andreas B. Pirker, Gerhard Posch, Stefan Geiger, Bernhard C. Wimmer, Andreas Graz Univ Technol Inst Internal Combust Engines & Thermodynam A-8010 Graz Austria Know Ctr GmbH Res Ctr Data Driven Business & Big Data Analyt A-8010 Graz Austria LEC GmbH Large Engine Competence Ctr A-8010 Graz Austria

The phenomenon of knock is an abnormal combustion occurring in spark-ignition (SI) engines and forms a barrier that prevents an increase in thermal efficiency while simultaneously reducing CO2 emissions. Since knocking combustion is highly stochastic, a cyclic analysis of in-cylinder pressure is necessary. In this study we propose an approach for efficient and robust detection and identification of knocking combustion in three different internal combustion engines. The proposed methodology includes a signal processing technique, called continuous wavelet transformation (CWT), which provides a simultaneous analysis of the in-cylinder pressure traces in the time and frequency domains with coefficients. These coefficients serve as input for a convolutional neural network (CNN) which extracts distinctive features and performs an image recognition task in order to distinguish between non-knock and knock. The results revealed the following: (i) The CWT delivered a stable and effective feature space with the coefficients that represents the unique time-frequency pattern of each individual in-cylinder pressure cycle;(ii) the proposed approach was superior to the state-of-the-art threshold value exceeded (TVE) method with a maximum amplitude pressure oscillation (MAPO) criterion improving the overall accuracy by 6.15 percentage points (up to 92.62%);and (iii) The CWT + CNN method does not require calibrating threshold values for different engines or operating conditions as long as enough and diverse data is used to train the neural network.

关键词： knocking combustion SI engines continuous wavelet transformation convolutional neural networks pressure trace time series

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：