检索结果-内蒙古大学图书馆

Parameter-efficient convolutional neural networks using wavelet transforms

AIP conference Proceedings 2024年第1期2895卷

作者： Arnel L. Malubay Kurt Anthony C. de Los Santos Job A. Nable Ateneo de Manila University Quezon City Metro Manila Philippines

Convolutional Neural Networks (CNN's) are known to perform well on computer vision tasks such as image classification, image segmentation, and object detection. However, one major drawback of CNN's is the huge amount of computing and memory resources needed to train them. In this paper, we propose an architectural unit which we call Upsampling-Based wavelet Residual Block (UBWRB), that utilizes the 2D discrete wavelet transform coupled with upsampling operators and a residual connection to extract features from image data while having relatively fewer trainable parameters as compared to traditional convolutional layers. The discrete wavelet transform is a family of transforms that find extensive applications in signal processing and time-frequency analysis. For this paper, we use the filter-bank implementation of the discrete wavelet transform, allowing it to act in a similar fashion to a convolutional layer with fixed kernel weights. We demonstrate the performance and parameter-efficiency of CNN's with UBWRB's in the task of image classification by training them on the MNIST, Fashion-MNIST, and CIFAR-10 datasets. Our best-performing models achieve a test accuracy of 99.34% on the MNIST dataset while having less than 120,000 trainable parameters, and 92.90% and 84.27% on the Fashion-MNIST and CIFAR-10 datasets respectively, with both having less than 180,000 trainable parameters.

关键词：

来源：评论

学校读者我要写书评

暂无评论

Human-computer Interactive System Development for Cloud Removal from Single images 7

Human-computer Interactive System Development for Cloud Remo...

引用

conference on Optical Metrology and Inspection for Industrial applications Vii

作者： Wu, Feng Zhu, Xifang Xiang, Ruxi Xu, Qingquan Xu, Yang Sun, Qingcheng Changzhou Inst Technol Sch Elect & Informat Engn 666 Liaohe Rd Changzhou 213032 Jiangsu Peoples R China

ISBN: (数字)9781510639201

ISBN: (纸本)9781510639201

Remote sensing imaging is disturbed by clouds frequently and produces images with low contrast and poor resolution. Algorithms removing clouds from single remote sensing images are expected to improve their qualities and aroused worldwide interests. Compared with multi-image methods, cloud removal algorithms from single images reduce the demands for weather and improve the image collecting flexibility. In this paper, a human-computer interactive system for cloud removal from single images was developed according to the proposed cloud removal algorithm. Firstly, the distributions of clouds and sceneries in the contaminated remote sensing images were discussed. Features of the dual tree complex wavelet transform (DTCWT) to split image frequencies were investigated after the principle of DTCWT was analyzed. By combining the frequency features of DTCWT sub-bands and the frequency distributions of the remote sensing images, the cloud removal algorithm based on DTCWT was proposed, and its procedures were described completed. Then, a human-computer interactive system for cloud removal was developed in order to improve the cloud removal processing conveniences and achieve the best performance. The development schemes were introduced. The system structures were illustrated. The design methods of the camera driver and video signal output program were described. Communication streams between modules and human-computer interaction interfaces were designed further according to the cloud removal algorithm based on DTCWT. The system has a very good flexibility which permits the input of most parameters from users and is expect to achieve a best performance. Finally, the device of the system was completed and tested. image acquisition and cloud removal processing experiments were implemented by applying the human-computer interactive system. Cloud removal effects by adjusting parameters through the human-computer interaction interfaces were compared and evaluated. It proved that th

关键词： cloud removal dual tree complex wavelet transform human-computer interaction remote sensing image

来源：评论

学校读者我要写书评

暂无评论

Hand Movement Recognition Using Cross Spectrum image Analysis of EMG signals-A Deep Learning Approach

Hand Movement Recognition Using Cross Spectrum Image Analysi...

引用

2020 National conference on Emerging Trends on Sustainable Technology and Engineering applications, NCETSTEA 2020

作者： Roy, Sayanjit Singha Samanta, Kaniska Chatterjee, Soumya Dey, Sayantan Nandi, Arnab Bhowmik, Ronjoy Mondal, Shoumosree Techno India University Electrical Engineering Department Kolkata India Electrical Engineering Department St. Thomas College of Engineering and Technology Kolkata India

ISBN: (纸本)9781728143620

Accurate recognition of hand movements using electromyography (EMG) signals is important to develop robust human-machine interface system for upper limb prosthetic control. For accurate detection and classification of different movements, proper feature selection from EMG signals is necessary, failure of which may often lead to incorrect results. To this end, this article presents an image processing aided deep learning approach for detection and classification of hand movement EMG signals. In this study, EMG signals for both left hand and right hand movements were procured from an existing database. A reference signal for each category was chosen as reference and cross spectrum of the rest of the EMG signals was done with the reference signal. The resultant cross-wavelet spectrum images of both classes of EMG signals were fed to a pre-trained convolution neural network (CNN) for the purpose of detection and classification of hand movements. It has been observed that the proposed method returned an average classification accuracy of 97.6% in segregating different categories of EMG signals. Besides, the performance of the proposed method analyzed using different CNN architectures was also found to be robust. The proposed method can be implemented for real-time detection of hand movements. © 2020 IEEE.

关键词： Deep learning

来源：评论

学校读者我要写书评

暂无评论

Color image Segmentation Via wavelet Frames 4

Color Image Segmentation Via Wavelet Frames

引用

4th IEEE International conference on signal and image processing (ICSIP)

作者： Yang, Jianbin Zhang, Enlin Wang, Cong Hohai Univ Dept Math Coll Sci Nanjing Jiangsu Peoples R China Xidian Univ Sch Electromech Engn Xian Shaanxi Peoples R China

ISBN: (纸本)9781728136608

image segmentation aims at partitioning an image into its constituent parts, which plays a crucial role in practical applications. In this paper, we present a wavelet frame-based model for color images segmentation, which can be regarded as a discretization to the classical Chan-Vese (C-V) model. The advantage of the wavelet frame-based approach is that it has fast algorithm and is able to extract important features of the input images. We then apply the alternating direction method of multipliers (ADAM) algorithm to solve the model. The experiments on some color image segmentation tasks indicate that our algorithm performs favorably against several existing methods.

关键词： wavelet frame image segmentation ADMM algorithm color images

来源：评论

学校读者我要写书评

暂无评论

Real-time PCG diagnosis using FPGA 18th

Real-time PCG diagnosis using FPGA

引用

Joint conferences on 18th International conference on Intelligent Systems Design and applications, ISDA 2018 and 10th World Congress on Nature and Biologically Inspired Computing , NaBIC 2018

作者： Fadhel, Mohammed Abdulraheem Al-Shamma, Omran Oleiwi, Sameer Razzaq Taher, Bahaa Hussein Alzubaidi, Laith University of Information Technology and Communications Baghdad Iraq Muthanna University Muthanna Iraq University of Sumer Dhi Qar Iraq Faculty of Science and Engineering Queensland University of Technology Brisbane Australia

ISBN: (纸本)9783030166564

Phono-Cardio-Gram (PCG) is an effective technique of detecting various heart abnormalities and malfunctions. Several PCG segmentation algorithms, each with its advantages and disadvantages, have been developed and tested. Unfortunately, most of these algorithms fail to diagnose heart conditions in real time. This paper introduces a real-time methodology by utilizing the field programmable gate array (FPGA) hardware to speed up the processing time very successfully. The type of FPGA hardware is Altera DE2 Cyclone ii board with NIOS ii soft processor. The algorithm of discrete wavelet transforms (DWT) with Shannon energy and spectra signal are used to extract the PCG features. The algorithm is also applied to identify the normal heart condition and nine abnormal cases. The results show that the methodology worked extremely efficient and suitable for real-time heart diagnosis purposes. © Springer Nature Switzerland AG 2020.

关键词： Discrete wavelet transforms

来源：评论

学校读者我要写书评

暂无评论

Towards the Definition of a Low-Cost Toolbox for Qualitative Inspection of Painted Historical Vaults by Means of Modified DSLR Cameras, Open Source Programs and signal processing Techniques 20th

Towards the Definition of a Low-Cost Toolbox for Qualitative...

引用

20th International conference on Computational Science and Its applications (ICCSA)

作者： Piroddi, Luca Calcina, Sergio Vincenzo Trogu, Antonio Vignoli, Giulio UniCA DICAAR Dept Civil Engn Environm Engn & Architecture Cagliari Italy Geol Survey Denmark & Greenland GEUS Aarhus Denmark

ISBN: (纸本)9783030588205;9783030588199

Historical architecture is a primary element containing the identity values of a society. The wide diffusion of many ancient buildings gathering part of these values on painting walls over territories often characterized by poor technological or economic resources brings to consider the development of low-cost protocols to inspect valued surfaces and to give the authorities in charge of preservation and restoration adequate technical information. Here we present the preliminary results of a recent application of remote sensing micro-geophysical techniques to typical architectural targets such as vaults. A modified commercial Digital Single-Lens Reflex (DSLR) camera was used to acquire multispectral datasets on portions of a painted vault. Multispectral datasets were used raw or after the application of a pre-processing step with a Multi images Stacking (MIS) algorithm. Multispectral images were then processed with spatial wavelet decomposition, histogram enhancing, thresholds application, image fusion, false colors compositing and Principal Component Analysis (PCA) techniques. Software used have been GNU image Manipulation Program (GIMP) and Mathworks MATLAB (which can be substituted for the processing steps proposed by the built-in functions of GNU OCTAVE open-source software). Processed images were able to highlight features on vault paintings revealing details of the surface or its very shallow layers which were impossible or very difficult to distinguish in raw data. In fact, they emphasized low-visible details, differences in apparently similar finishes or pigments, cracks and probably details of surface preparation.

关键词： Multispectral analysis Digital image processing PCA Cultural heritage Historical architecture Painted walls inspection Low cost diagnostics

来源：评论

学校读者我要写书评

暂无评论

Combining Scatter Transform and Deep Neural Networks for Multilabel Electrocardiogram signal Classification

Combining Scatter Transform and Deep Neural Networks for Mul...

引用

Computing in Cardiology conference (CinC)

作者： Oppelt, Maximilian P. Riehl, Maximilian Kemeth, Felix P. Steffan, Jan Fraunhofer IIS Dept Image Proc & Med Engn Wolfsmantel 33 D-91058 Erlangen Germany Johns Hopkins Univ Dept Chem & Biomol Engn Whiting Sch Engn Baltimore MD 21218 USA

ISBN: (纸本)9781728173825

An essential part for the accurate classification of electrocardiogram (ECG) signals is the extraction of informative yet general features, which are able to discriminate diseases. Cardiovascular abnormalities manifest themselves in features on different time scales: small scale morphological features, such as missing P-waves, as well as rhythmical features apparent on heart rate scales. For this reason we incorporate a variant of the complex wavelet transform, called a scatter transform, in a deep residual neural network (ResNet). The former has the advantage of being derived from theory, making it well behaved under certain transformations of the input. The latter has proven useful in ECG classification, allowing feature extraction and classification to be learned in an end-to-end manner. Through the incorporation of trainable layers in between scatter transforms, the model gains the ability to combine information from different channels, yielding more informative features for the classification task and adapting them to the specific domain. For evaluation, we submitted our model in the official phase in the PhysioNet/Computing in Cardiology Challenge 2020. Our (Team Triage) approach achieved a challenge validation score of 0.640, and full test score of 0.485, placing us 4th out of 41 in the official ranking.

关键词： Classification (of information)

来源：评论

学校读者我要写书评

暂无评论

Single image Ear Recognition Using wavelet-Based Multi-Band PCA 27

Single Image Ear Recognition Using Wavelet-Based Multi-Band ...

引用

27th European signal processing conference (EUSIPCO)

作者： Zarachoff, Matthew Sheikh-Akbari, Akbar Monekosso, Dorothy Leeds Beckett Univ Sch Comp Creat Technol & Engn Leeds W Yorkshire England

ISBN: (纸本)9789082797039

Principal Component Analysis (PCA) has been successfully used for many applications, including ear recognition. This paper presents a 2D wavelet based Multi-Band PCA (2D-WMBPCA) method, inspired by PCA based techniques for multispectral and hyperspectral images, which have shown a significantly higher performance to that of standard PCA. The proposed method performs 2D non-decimated wavelet transform on the input image dividing the image into its subbands. It then splits each resulting subband into a number of bands evenly based on the coefficient values. Standard PCA is then applied on each resulting set of bands to extract the subbands eigenvectors, which are used as features for matching. Experimental results on images of two benchmark ear image datasets show that the proposed 2D-WMBPCA significantly outperforms both the standard PCA method and the eigenfaces method.

关键词： Ear recognition principal component analysis multi-band image creation non-decimated wavelet transform

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

Systolic Murmurs Diagnosis Improvement by Feature Fusion and Decision Fusion

Systolic Murmurs Diagnosis Improvement by Feature Fusion and...

引用

IEEE International conference on signal and image processing applications (IEEE ICSIPA)

作者： Akbari, Somayeh Ghassemian, Hassan Akbari, Zahra Tarbiat Modares Univ Fac Elect & Comp Engn Image Proc & Informat Anal Lab Tehran Iran

ISBN: (纸本)9781728133775

Acoustic sound generated by the heart mechanical activity, can provide useful information about the condition of heart valves. The heart sound auscultation is the fundamental tool in the evaluation of the cardiovascular system. The advantage of this method is fast, inexpensive and noninvasive. Due to human auscultatory limitation and non-stationary characteristics of phonocardiogram signals (PCG), diagnosis based on sounds that are heard via a stethoscope is difficult skill, therefor it requires a lot of practice. This study has proposed a biomedical automatic system for classification of PCG signals, which, recorded by a digital stethoscope. In order to extract various characteristics of PCG signals, the power spectrum estimation, wavelet transform (WT) and Mel frequency Cepstrum coefficients (MFCC) have been used in feature extraction step. Features are given to four classifiers: support vector machine (SVM), k-nearest neighbor (k-NN), multilayer perceptron (MLP) and maximum likelihood (ML). The majority voting combination rule is utilized for fusion of different classifiers. The proposed method has been examined on dataset of 90 PCG records containing healthy and three types of cardiac valve diseases (pulmonary stenosis (PS), Atrial Septal Defect (ASD) and Ventricular Septal Defect (VSD)). The experimental results demonstrate that the classifier fusion rule significantly increases the diagnostic accuracy of abnormal PCG. Our proposed method can be used for online classification of PCG in intelligent diagnosis systems.

关键词： PCG signal MFCC algorithm power spectrum estimation wavelet transform feature fusion decision fusion

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：