检索结果-内蒙古大学图书馆

Design, Automation and Test in Europe Conference and Exhibition (DATE)

作者： Hanif, Muhammad Abdullah Hafiz, Rechan Shafique, Muhammad Vienna Univ Technol Inst Comp Engn Vienna Austria Informat Technol Univ Dept Elect Engn Lahore Pakistan

ISBN: (纸本)9783981926309

Approximate computing is an emerging paradigm for error resilient applications as it leverages accuracy loss for improving power, energy, area, and/or performance of an application. The spectrum of error resilient applications includes the domains of image and video processing, Artificial intelligence (AI) and Machine Learning (ML), data analytics, and other Recognition, Mining, and Synthesis (RMS) applications. In this work, we address one of the most challenging question, i.e., how to systematically employ approximate computing in Convolution Neural Networks (CNNs), which are one of the most compute-intensive and the pivotal part of AI. Towards this, we propose a methodology to systematically analyze error resilience of deep CNNs and identify parameters that can be exploited for improving performance/efficiency of these networks for inference purposes. We also present a case study for significance-driven classification of filters for different convolutional layers, and propose to prune those baring the least significance, and thereby enabling accuracy vs. efficiency tradeoffs by exploiting their resilience characteristics in a systematic way.

关键词： Resilience Hardware Convolutional neural networks Memory management Artificial intelligence Quantization (signal) computational efficiency

来源：评论

学校读者我要写书评

暂无评论

Face Recognition Based on Multi-source Superpixel Fusion

Face Recognition Based on Multi-source Superpixel Fusion

引用

symposium of image, signal processing, and Artificial Vision (STSIVA)

作者： Xiaoxiang Xu Fanzhang Li Li Zhang School of Computer Science and Technology Soochow University Suzhou Jiangsu china Soochow University Suzhou Jiangsu china

ISBN: (数字)9781728170862

ISBN: (纸本)9781728170879

Face recognition is an important field of artificial intelligence. Single face recognition is one of the indispensable and challenging directions. There have been many existing single face recognition algorithms, such as PCA, 2DPCA, LDA, SVD, HMM, maximum likelihood linear regression, neural network and so on. These algorithms mainly solve the problem of single face recognition from three perspectives: expanding samples, expanding feature space and creating a learning general framework. The general performance of the algorithm for learning the general framework and extending the feature space is better than that of the method for expanding samples. In this paper,a face recognition learning algorithm based on multi-source superpixel fusion is proposed, which combines extended feature space and learning general framework. This algorithm mainly combines the proposed multi-source fusion representation model with super-pixel face features, and uses the theory of multi-source fusion model to solve the shortcomings of super-pixel face features. In particular, LBP algorithm is introduced into super-pixel segmentation to reduce the impact of illumination, rotation and occlusion on superpixel segmentation. Finally, the feasibility and validity of the super-pixel face recognition learning algorithm based on multi-source fusion model are demonstrated by simulation experiments.

关键词： Face recognition computational modeling Neural networks Maximum likelihood linear regression Hidden Markov models Lighting Principal component analysis

来源：评论

学校读者我要写书评

暂无评论

DiCE-M: Distributed Code Generation and Execution for Marine Applications - An Edge-Cloud Approach

DiCE-M: Distributed Code Generation and Execution for Marine...

引用

ieee/ACM symposium on Edge Computing (SEC)

作者： Giuseppe Coviello Kunal Rao Gennaro Mellone Ciro Giuseppe De Vita Srimat Chakradhar NEC Laboratories America Inc. Princeton NJ

ISBN: (数字)9798350378283

ISBN: (纸本)9798350378290

Edge computing has emerged as a transformative technology that reduces application latency, improves cost efficiency, enhances security, and enables large-scale deployment of applications across various domains. In environmental monitoring, systems such as MegaSense[49], use low-cost sensors to gather and process real-time air quality data through edge-cloud collaboration, highlighting the critical role of edge computing in enabling scalable, efficient solutions. Similarly, marine science increasingly requires real-time processing and analysis of marine data from remote, resource-constrained environments. In this paper, we extend the power of edge computing by integrating it with Generative Artificial intelligence(GenAI),specifically large language models (LLMs), to address challenges in marine science applications. We propose DiCE-M (Distributed Code generation and Execution for Marine applications), a robust system that uses LLM to generate distributed code for marine applications and then utilizes a runtime to efficiently execute it on an edge+cloud computing infrastructure. Specifically, DiCE-M leverages edge computing to execute lightweight AI models locally on unmanned surface vehicles(USVs)while offloading complex tasks to the cloud, thus balancing computational load and enabling realtime monitoring in marine environments. We use marine litter identification as an example application to demonstrate the utility of DiCE-M. Our results show that DiCE-M reduces latency by more than 2X when marine litter is not detected and cuts cloud computing costs by more than half compared to traditional cloud-based approaches. By selectively cropping and transmitting relevant image portions, DiCE-M further improves bandwidth efficiency, making it a reliable and cost-effective solution for deploying AI-drivenapplications on resource-constrained USVs in dynamic marine environments.

关键词： Cloud computing Costs Codes Spectral efficiency image edge detection computational modeling Real-time systems Oceanography Vehicle dynamics Edge computing

来源：评论

学校读者我要写书评

暂无评论

Hidden Challenge in Deep-Learning Real-Time Object Detection on Edge Devices

Hidden Challenge in Deep-Learning Real-Time Object Detection...

引用

Midwest symposium on Circuits and Systems (MWSCAS)

作者： Marcus F. Nicolas Dalila B. Megherbi Department of Electrical and Computer Engineering UMASS Lowell CMINDS Research Center Lowell USA

ISBN: (数字)9798350387179

ISBN: (纸本)9798350387186

As the need arises for edge intelligence, which combines edge-computing with artificial intelligence, efficient and accurate algorithms are increasingly needed. Edge computing allows AI applications to be executed in real-time, closest to hardware devices/systems, thus reducing latency that could be introduced if traversing the cloud or sent to some storage unit. The aim is the needed real-time information processing, among other benefits. Notwithstanding the advantages of Edge-AI, a challenge that is not generally addressed in many such real-time detection and recognition DL models is the effect of an object rotation on real-time object detection accuracy. In particular, object label uncertainty can be introduced in rotated object detection because the detected shape is not foreseen. The focus of this paper is to highlight how real-time object detection can be affected by rotation, a problem of importance in the field. We selected a particular architecture, Efficientdet, and implemented it on a Raspberry Pi 4 to study this rotational object detection handicap. This paper also aims to increase awareness and tender possible solutions in hopes that future researchers and engineers can design DL models on edge computing devices that will significantly diminish the impacts of an object rotation on its real-time object detection accuracy.

关键词： Accuracy Uncertainty image edge detection computational modeling Object detection Real-time systems Hardware

来源：评论

学校读者我要写书评

暂无评论

A phantom study on temporal and subband Minimum Variance adaptive beamforming

A phantom study on temporal and subband Minimum Variance ada...

引用

ieee International Ultrasonics symposium

作者： Konstantinos Diamantis Iben H. Voxen Alan H. Greenaway Tom Anderson Jorgen Arendt Jensen Vassilis Sboros School of Engineering & Physical Sciences Heriot-Watt University Edinburgh UK Greenwood Engineering Brendby Denmark Medical Physics The University of Edinburgh Edinburgh UK Dept. of Electrical Engineering Technical University of Denmark Lyngby Denmark

ISBN: (纸本)9781479970506

This paper compares experimentally temporal and subband implementations of the Minimum Variance (MV) adaptive beamformer for medical ultrasound imaging. The performance of the two approaches is tested by comparing wire phantom measurements, obtained by the research ultrasound scanner SARUS. A 7 MHz BK8804 linear transducer was used to scan a wire phantom in which wires are separated by 10 mm. Performance is then evaluated by the lateral Full-Width-Half-Maximum (FWHM), the Peak Sidelobe Level (PSL), and the computational load. Beamformed single emission responses are also compared with those from conventional Delay-and-Sum (DAS) beamformer. FWHM measured at the depth of 46.6 mm, is 0.02 mm (0.09λ) for both adaptive methods while the corresponding values for Hanning and Boxcar weights are 0.64 and 0.44 mm respectively. Between the MV beamformers a -2 dB difference in PSL is noticed in favor of the subband approach (-31 and -33 dB), whereas values from conventional are not lower than -29 dB. This slight improvement in the case of the subband implementation comes at the expense of increased computational cost;3.7 TFLOPs per image are required in contrast to 130 GFLOPs of the temporal one, when only 0.5 GFLOPs are needed in DAS beamforming.

关键词： Array signal processing Ultrasonic imaging Wires Covariance matrices Biomedical imaging Transducers

来源：评论

学校读者我要写书评

暂无评论

Towards Wearable Ultrafast Ultrasound: A FPGA Solution

Towards Wearable Ultrafast Ultrasound: A FPGA Solution

引用

ieee International symposium on Applications of Ferroelectrics (ISAF)

作者： Zhengchang Kou Michael L. Oelze Beckman Institute University of Illinois Urbana-Champaign Urbana IL USA Electrical and Computer Engineering Department University of Illinois Urbana-Champaign Urbana IL USA Carle Illinois College of Medicine University of Illinois Urbana-Champaign Urbana IL USA

ISBN: (数字)9798350371901

ISBN: (纸本)9798350371918

Wearable ultrasound has been extensively studied in the past several years. Multiple emerging applications including wearable cardiac and brain imaging have been demonstrated with wearable ultrasound. However, current wearable ultrasound still suffers from several key limitations. High image quality and fully integrated wearable ultrasound imaging hardware have not been achieved in combination. To date what has been constructed is either a limited channel count on a fully integrated wearable ultrasound system that has limited image quality and frame rate or a larger channel count wearable transducer connected to a cart-based ultrasound research platform. As a result, fully integrated wearable ultrafast ultrasound, which can be used as the foundation of wearable microvessel imaging, functional ultrasound or even brain computer interface, has not been demonstrated. The major roadblocks to enabling high image quality and frame rate on a fully integrated wearable ultrasound system are the computational cost and requirement for low power. To solve this problem, we propose a low power ultrafast beamformer solution using an FPGA with a highly efficient parallel IQ beamforming algorithm. A post-compounding beamforming rate of over 1000 frames per second was been achieved utilizing only 0.4 Watts power consumption. This result paves the way for fully integrated wearable ultrafast ultrasound imaging systems.

关键词： image quality Neuroimaging Ultrasonic imaging Transducers Power demand Array signal processing Imaging Hardware Wearable devices Field programmable gate arrays

来源：评论

学校读者我要写书评

暂无评论

QUANTOF: multi-GPU-based list-mode fully quantitative TOF PET image reconstruction

QUANTOF: multi-GPU-based list-mode fully quantitative TOF PE...

引用

ieee symposium on Nuclear Science (NSS/MIC)

作者： Z. Yuan F. Zhan H. Lu C. Li Y. Hou H. Wang J. Hua R. Liao J. Jiang Key Laboratory of Beam Technology of Ministry of Education College of Nuclear Science and Technology Beijing Normal University Beijing China Department of Nuclear Medicine The First Affiliated Hospital of USTC Division of Life Sciences and Medicine University of Science and Technology of China Anhui China School of Artificial Intelligence Beijing Normal University Beijing China

ISBN: (数字)9798350388152

ISBN: (纸本)9798350388169

Positron Emission Tomography (PET) plays a pivotal role in the early detection and precise diagnosis of various diseases, including cancer, neurological disorders, and heart conditions. Ordered subset expectation maximization (OSEM) algorithm is the most commonly used image reconstruction algorithm. To address the high computational costs associated with this algorithm, numerous studies have explored the effectiveness of GPU acceleration. However, these studies have predominantly focused on the forward and backward projection components of iterative algorithms, neglecting correction techniques which are essential for PET image reconstruction, such as scatter correction. This work introduces a multi-GPU-based list-mode fully quantitative Time-of-Flight (TOF) PET image reconstruction technique, named QUANTOF, that integrates all corrections during PET image iterative reconstruction on GPUs. Additionally, optimizations have been applied to the backprojection algorithm, resulting in a two-fold increase in speed compared to the pre-optimized version. Validation using experimental datasets from a Siemens Biograph Vision PET/CT scanner demonstrates that this technique significantly reduces reconstruction time costs while achieving exceptional quantitative accuracy.

关键词： Neurological diseases Heart Microwave integrated circuits Semiconductor detectors Graphics processing units Iterative methods Positron emission tomography

来源：评论

学校读者我要写书评

暂无评论

image Semantic Segmentation Based on the GAN Auxiliary Network

Image Semantic Segmentation Based on the GAN Auxiliary Netwo...

引用

symposium of image, signal processing, and Artificial Vision (STSIVA)

作者： Jinshuo Zhang Zhicheng Wang Songyan Zhang Gang Wei Zhiqiang Xiong Meng Yang CAD Research Center Tongji University Shanghai China China Ship Development and Design Center Wuhan China

ISBN: (纸本)9781665426565

To enhance the performance of existing deep neural networks in semantic segmentation while preserving efficiency at the same time, a semantic segmentation network with the help of GAN (Generative Adversarial Networks) is proposed. The method consists of a generator and a discriminator. The segmentation results obtained from the generator are encoded and then fed into the discriminator to obtain the pixel-wise uncertainty values. Such uncertainty values are taken as weights for the calculation of CEGU (Cross-Entropy with GAN Uncertainty) to help the optimization of the generator. The discriminator is removed after training. Experiment results show that the mean IoU (Intersection over Union) scores of the segmentation results grow by 4.7% and 3.2% respectively on ResNet-50 and ResNet18, after using the GAN auxiliary method along with the CEGU. It shows that such a GAN auxiliary network can significantly improve the performance of basic end-to-end methods with various backbones on the semantic segmentation task, without introducing extra computation cost in the test phase.

关键词： Training image segmentation Uncertainty Costs computational modeling Semantics Generative adversarial networks

来源：评论

学校读者我要写书评

暂无评论

EEG channel selection-based binary particle swarm optimization with recurrent convolutional autoencoder for emotion recognition

引用

BIOMEDICAL signal processing AND CONTROL 2023年第1期84卷

作者： Kouka, Najwa Fourati, Rahma Fdhila, Raja Siarry, Patrick Alimi, Adel M. Natl Engn Sch Sfax ENIS Univ Sfax Res Grp Intelligent Machines BP 1173 1173 Sfax 3038 Tunisia Univ Paris Est Lab Image Signaux & Syst Intelligents LISSI F-94400 Creteil France Univ Johannesburg Fac Engn & Built Environm Dept Elect & Elect Engn Sci Johannesburg South Africa

Electroencephalography (EEG) signals can demonstrate the activities of the human brain and recognize different emotional states. Emotion recognition based on full EEG channels leads to the use of redundant data and increases the hardware complexity. This paper presents a channel selection method based on a new Binary Many-Objective Particle Swarm Optimization with Cooperative Agents (BMaOPSO-CA). More specifically, we perform unsupervised feature learning with recurrent convolutional layers based on autoencoder architecture directly from clean EEG signals. Extensive validation on three public effective benchmarks, i.e. DASPS, DEAP, and SEED which are different in channel number used stimuli, and participant ratings, was carried out with subject independent scheme. As result, the experimental study highlights the optimal electrode locations related to emotions, leading to analyze the relationship between specific brain regions and emotions.

关键词： Unsupervised EEG feature learning Convolutional LSTM Channel selection Binary particle swarm optimization Local learning strategy

来源：评论

学校读者我要写书评

暂无评论

Single Dendritic Neural Classification with Functional Weight-enhanced Differential Evolution

Single Dendritic Neural Classification with Functional Weigh...

引用

symposium of image, signal processing, and Artificial Vision (STSIVA)

作者： Ziqian Wang Kaiyu Wang Jiaru Yang Zheng Tang Faculty of Engineering University of Toyama Toyama Japan

ISBN: (纸本)9781665426565

As current mainstream deep learning models based on neural networks have been long criticized because of their complex structures, attempts in formulating a single neural model have raised much attention. Owing to the nonlinear information processing ability, dendritic neuron model (DNM) has shown its great potential in classification problems over the past decades. However, designing an effective learning algorithm for training DNM is still an open question due to the issues of local optima trapping and overfiting caused by traditional back-propagation (BP) algorithm. In this study, a novel functional weight-enhanced differential evolutionary algorithm (termed FWDE) is proposed to solve the aforementioned problems. By introducing Gaussian distribution function into weight generation of fitness-distance balance selection strategy, FWDE obtains significantly better classification accuracy with faster convergence speed compared with other representative non-BP and BP algorithms. The experimental results verify the great performance of FWDE, indicating that DNM with an powerful learning algorithm is considerably more effective.

关键词： Deep learning Training Proteins computational modeling Neurons Text categorization Gaussian distribution

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：