检索结果-内蒙古大学图书馆

27th International Conference on Pattern Recognition, ICPR 2024

作者： Jiang, Hao Fang, Zhijun Shao, Xuan Jiang, Xiaoyan Hwang, Jenq-Neng School of Computer Science and Technology Donghua University Shanghai China School of Electronic and Electrical Engineering Shanghai University of Engineering Science Shanghai China Department of Electrical and Computer Engineering University of Washington SeattleWA United States

ISBN: (纸本)9783031784439

Lightweight yet reliable depth estimation models that can deployed on edge devices are crucial for the practical application of fields such as autonomous driving, robot navigation, and augmented reality. However, previous research often fails to balance accuracy and complexity well. Existing lightweight models still have relatively high error rates in specific scenarios, which makes them unsuitable for industrial applications. Hence, we propose LiDUT-Depth, a lightweight architecture for self-supervised monocular depth estimation that is enhanced through dynamic upsampling and triplet loss optimization, which achieves competitive results with smaller model sizes and lower computational complexity. Specifically, an efficient dynamic upsampling module (EDU Module) is employed to preserve detailed local features, thereby obtaining more accurate depth estimation results. In addition, an improved semantic-aware triplet loss (SaTri Loss) is integrated into the training loss, significantly improving depth estimation accuracy in gradient-rich regions. Experiments show that our architecture achieves a considerably enhanced accuracy compared with previous works with a much lower model size. Our codes and models are available here. © The Author(s), under exclusive license to Springer Nature Switzerland AG 2025.

关键词： Augmented reality

来源：评论

学校读者我要写书评

暂无评论

A feature selection-aided deep learning based deepfake video detection method

引用

Multimedia Tools and Applications 2025年 1-24页

作者： Mohiuddin, Sk Roy, Ayush Pani, Saptarshi Malakar, Samir Sarkar, Ram Department of Computer Science Asutosh College Kolkata700026 India Department of Electrical Engineering Jadavpur University Kolkata700032 India Department of Computer Science UiT The Arctic University of Norway Tromsø9019 Norway Department of Computer Science and Engineering Jadavpur University Kolkata700032 India

The rapid advancements in deepfake technology are imposing significant challenges in detecting manipulated media contents. In this work, we have introduced a deepfake detection method that utilizes three pre-trained convolutional neural network models, namely DenseNet-121, EfficientNet-B0, and Xception to extract deep learning features from preprocessed extracted video frames. Next, these features are stacked to generate diverse feature representations, and thus end up with high-dimensional features. After that, a feature selection strategy is employed on the combined features to eliminate redundant and less effective features. The selected features are then fed into the K-nearest neighbors classifier to decide whether a questioned video is a fake video or not. Two widely used deepfake video datasets, namely FaceForensics++ and Celeb-DF (V2) are used to evaluate the performance of the proposed technique following two experimental setups: intra- and inter-datasets. The results show that the model achieves area under curve scores of 98.50% (85.70%) and 97.70% (79.00%) in the intra-dataset (inter-dataset) setups while trained on the FaceForensics++ and Celeb-DF (V2) datasets, respectively. The performance of the proposed model is comparable to that of state-of-the-art methods. Source code of the proposed method can be found at: here. © The Author(s), under exclusive licence to Springer science+Business Media, LLC, part of Springer Nature 2025.

关键词： Convolutional neural networks

来源：评论

学校读者我要写书评

暂无评论

Multi-Level and Energy-Efficient Partial Computation Offloading in Heterogeneous Edge Intelligence

引用

IEEE Internet of Things Journal 2025年第11期12卷 14993-15007页

作者： Xu, Baoyu Ruan, Yancheng Qiu, Chenghu He, Shuibing Shu, Feng Kang, Xiaoyang Zhang, Lihua Academy for Engineering and Technology Shanghai200433 China Shanghai University School of Computer Engineering and Science Shanghai200444 China Zhejiang University College of Computer Science and Technology Hangzhou310027 China University of Sydney School of Electrical and Information Engineering Australia Fudan University Laboratory for Neural Interface and Brain Computer Interface MOE Frontiers Center for Brain Science Academy for Engineering and Technology Shanghai200433 China Fudan University Academy for Engineering and Technology Shanghai200433 China

Due to the diversity of edge devices and applications, edge systems are heterogeneous and have been applied in artificial intelligence fields such as smart factories and intelligent transportation, which is called heterogeneous edge intelligence. Many studies employ computation offloading to transfer processing data from resource-scarce edge devices to resource-rich edge servers. These studies primarily focus on the overall resource consumption of homogeneous edge systems, neglecting the system heterogeneity and the details of resource consumption. In this article, we construct a system model from a parallel perspective for the heterogeneous edge system with different processors, memory, and applications, which perceives the cost of energy and delay from three levels: system, application, and component. A hybrid metaheuristic algorithm combined with a greedy rule, hybrid mutation, and whale optimization algorithm (GHMWOA) is proposed to realize partial computation offloading. A partial offloading architecture of heterogeneous edge intelligence is proposed to validate our model and algorithm with real-world hardware and software. Experiment results not only show GHMWOA outperforms multiple classical optimization algorithms in minimizing energy consumption, but also discover on which system component energy consumption depends, and how properties of application and system influence the cost of energy. © 2014 IEEE.

关键词： Heuristic algorithms

来源：评论

学校读者我要写书评

暂无评论

Optimized Temporary Frequency Support for Wind Power Plants Considering Expanded Operational Region of Wind Turbines

引用

CSEE Journal of Power and Energy Systems 2025年第1期11卷 51-64页

作者： Zhengyang Hu Bingtuan Gao Zhao Xu Sufan Jiang School of Electrical Engineering Southeast UniversityNanjing210096China Department of Electrical and Electronic Engineering The Hong Kong Polytechnic UniversityHong Kong SARChina Department of Electrical Engineering and Computer Science The University of TennesseeKnoxvilleTN 37996 USA

Wind power plants(WPPs)are increasingly mandated to provide temporary frequency support to power systems during contingencies involving significant power ***,the frequency support capabilities of WPPs under derated operations remain insufficiently investigated,highlighting the potential for further improvement of the frequency *** paper proposes a bi-level optimized temporary frequency support(OTFS)strategy for a *** implementation of the OTFS strategy is collaboratively accomplished by individual wind turbine(WT)controllers and the central WPP ***,to exploit the frequency support capability of WTs,the stable operational region of WTs is expanded by developing a novel dynamic power control approach in WT *** approach synergizes the WTs'temporary frequency support with the secondary frequency control of synchronous generators,enabling WTs to release more kinetic energy without causing a secondary frequency ***,a model predictive control strategy is developed for the WPP *** strategy ensures that multiple WTs operating within the expanded stable region are coordinated to minimize the magnitude of the frequency drop through efficient kinetic energy ***,comprehensive case studies are conducted on a real-time simulation platform to validate the effectiveness of the proposed strategy.

关键词： Active power control model predictive control optimized temporary frequency support wind power plant wind turbine

来源：评论

学校读者我要写书评

暂无评论

Measurement of Alkali Metal Vapor Number Density Based on Surface Plasmon Resonance

引用

IEEE Photonics Technology Letters 2025年第14期37卷 757-760页

作者： Liang, Jiehao Li, Yuhan Wang, Yangtao Li, Yang Liu, Xuejing University of Shanghai for Science and Technology School of Optica-Electrical and Computer Engineering Shanghai200093 China

Alkali metal vapor cells are essential components of atomic magnetometers for magnetic field measurements. In this letter, a novel measurement method based on the surface plasmon resonance (SPR) effect to determine the number density and temperature of alkali metal vapor in cells is proposed. Polarized alkali metal atoms with different number densities exhibit slight refractive index differences, which can be measured by SPR to calculate the vapor density. Temperature information is obtained by coating a polydimethylsiloxane (PDMS) film at a fixed location above the metallic layer in the vapor cell. A Kretschmann-type prism-coupled SPR structure with angular modulation is analyzed using the finite element method for its suitability in measuring the number density and temperature of rubidium (Rb) vapor cells. The results show that within the resonance angle range of 43.08 ° to 46 °, the rubidium vapor density ranges from 2.48×1012 cm−3 to 1.7×1016 cm−3, with a density sensitivity of 1.63×10-16 °/cm-3 and a temperature sensitivity of 6.5×10-2 °/K, covering most operating conditions of atomic magnetometers. The proposed sensing structure offers a novel, real-time, non-contact, and spatially distributed method for measuring alkali metal gas density and temperature, significantly advancing the development of atomic magnetometers.. © 1989-2012 IEEE.

关键词： Magnetic field measurement

来源：评论

学校读者我要写书评

暂无评论

Effect of Low-Energy Proton Radiation on the Degradation of Silicon Carbide Schottky Diodes

引用

Journal of Electronic Materials 2025年 1-9页

作者： Baba, Tamana Hasbullah, Nurul Fadzlin Javed, Yasir Khan, Zafar Iqbal Sulaiman, Nurul Nabiilah Department of Electrical and Computer Engineering Kulliyyah of Engineering International Islamic University Malaysia Selangor Kuala Lumpur Malaysia College of Computer Science and Information Sciences Prince Sultan University P.O. Box 66833 Riyadh11586 Saudi Arabia

Silicon carbide (SiC) Schottky diodes are gaining traction for their potential in high-voltage, high-temperature, and radiation-resistant applications. While superior to traditional silicon (Si) diodes, their radiation hardness requires further exploration, particularly at lower energies. Although high-energy radiation undoubtedly degrades devices, understanding the impact of lower-energy radiation is crucial for real-world applications. This study employed 8 MeV proton radiation, considered relatively low energy in radiation physics, to investigate the response of SiC Schottky diodes. Interestingly, the irradiated devices displayed a slight decrease in reverse leakage current, indicating improved blocking ability. Encouragingly, both forward current and capacitance remained relatively unaffected, with a minor change in the ideality factor. Microscopy confirmed these findings by revealing no surface-level defects or leakage current hotspots, solidifying the exceptional resilience of SiC Schottky diodes to even lower-energy radiation. © The Minerals, Metals & Materials Society 2025.

关键词： Silicon carbide

来源：评论

学校读者我要写书评

暂无评论

An enhanced QoS approach for multi-objective optimization using social spider optimization 5G enable IoMT network

引用

International Journal of Information Technology (Singapore) 2025年第3期17卷 1851-1864页

作者： Jena, Rasmita Barik, Ram Chandra Yadav, Devendra Kumar Pradhan, Saroj Computer Science and Engineering C.V. Raman Global University Odisha Bhubaneswar India School of Computer Science and Engineering XIM University Odisha Bhubaneswar India Electrical Engineering Parala Maharaja Engineering College Odisha Berhampur India

Social Spider Optimization (SSO) is a swarm algorithm, based on the mating and cooperating behaviour of social spiders. This approach is basically used for global optimum search during allocation of resources and data compression in Internet of Things (IoTs) networks. Some of the Swaram Intelligence (SI) algorithm such as Particle Swarm Optimization (PSO) algorithm, Ant Colony Optimization (ACO), Social Spider Optimization (SSO), Parallel Social Spider Optimization (PSSO) are numerically simulating to decrease computation time for resource allocation, to improve the energy efficiency and data compression rate in IoT networks. However, still some deficiencies exist in the system. To overcome these deficiencies, an Adaptive Social Spider Optimization (ASSO) algorithm is suggested for global search optimum. This proposed model will update the both the female and male individuals’ positions, and each individual position can be calculated during the search process. This will help to strengthen the ability of the search performance. ASSO has been compared with the existing algorithm to obtain experimental results that shows the suggested ASSO conquered better result with an allocation rate of 96.90% and consumed energy 0.6mj that is comparatively better than PSSO. © Bharati Vidyapeeth's Institute of computer Applications and Management 2024.

关键词： Internet of Things (IoT) Resource Allocation Sensor Social Spider Optimization (SSO)

来源：评论

学校读者我要写书评

暂无评论

A 0.75mm2407µW Real-Time Speech Audio Denoiser with Quantized Cascaded Redundant Convolutional Encoder-Decoder for Wearable IoT Devices 38

A 0.75mm2407µW Real-Time Speech Audio Denoiser with Quantiz...

引用

38th International Conference on VLSI Design, VLSID 2025

作者： Kochar, Dimple Vijay Ashok, Maitreyi Chandrakasan, Anantha P. Massachusetts Institute of Technology Electrical Engineering and Computer Science CambridgeMA United States

ISBN: (纸本)9798331522445

Audio denoising is crucial for delivering high-quality sound in applications ranging from communication devices to entertainment systems. On-device denoising is critical for en-suring consistent performance across various host platforms. Machine learning (ML) models exhibit strong audio processing performance in the frequency domain but require efficient hardware design. This paper focuses on enhancing audio quality using convolutional encoder-decoder ML models with low power consumption while meeting real-time processing constraints. We achieve this by developing a quantized network that optimally reduces computational costs without compromising enhancement quality. Furthermore, our hardware quantization scheme reduces memory usage by up to 75% while maintaining accuracy. Next, we design a complementary processing element activation routing scheme tailored to our algorithm, significantly reducing the on-chip memory accesses by 5-9×. Fabricated in 28nm CMOS process, our chip demonstrates real-time audio denoising, processing each frame within 8ms while consuming only 407µ or 3.24 µJ/frame at 0.65V, 18.5 MHz, making it ideal for battery-powered IoT devices. In terms of performance, our chip also achieves the highest evaluation score for audio quality (PESQ), outperforming previous works. © 2025 IEEE.

关键词： CMOS integrated circuits

来源：评论

学校读者我要写书评

暂无评论

Integrated Global-Local Gaussian Attention for Image Compression

Integrated Global-Local Gaussian Attention for Image Compres...

引用

2025 IEEE International Conference on Acoustics, Speech, and Signal Processing, ICASSP 2025

作者： Khoshkhahtinat, Atefeh Mehta, Piyush M. Department of Computer Science and Electrical Engineering West Virginia University WV United States

ISBN: (纸本)9798350368741

Transformers have demonstrated outstanding performance in learned image compression (LIC) due to their high capacity for modeling complex dependencies. However, existing methods employ window-based attention mechanisms to reduce overhead, which limits the long-range dependency modeling capabilities of transformers. In this study, we present a novel transformer block designed to efficiently leverage global spatial information while keeping computational costs low. Additionally, our proposed transformer incorporates a local attention module that utilizes Gaussian positional encoding to enhance spatial awareness. These improvements contribute to achieving less redundant latent representations and improving compression efficiency. In addition to enhance the decorrelating capability of the transformation component, we propose a novel channel-wise entropy model that effectively leverages channel dependencies, thereby providing a more accurate estimation of the latent space distribution. Experimental results indicate that our framework surpasses both conventional and neural codecs. © 2025 IEEE.

关键词： Entropy model Guassian positional encoding Learned Image Compression Transformer

来源：评论

学校读者我要写书评

暂无评论

Adaptive Structure-Aware Connectivity-Preserving Loss for Improved Road Segmentation in Remote Sensing Images

Adaptive Structure-Aware Connectivity-Preserving Loss for Im...

引用

2025 IEEE/CVF Winter Conference on Applications of computer Vision Workshops, WACVW 2025

作者： Shojaei, Sara Bohl, Trevor Palaniappan, Kannappan Bunyak, Filiz University of Missouri-Columbia Department of Electrical Engineering and Computer Science MO United States

ISBN: (纸本)9798331536626

Accurate road extraction from remote sensing images is a critical step for many applications, including navigation, environmental monitoring, and disaster response. In remote sensing images, roads appear as thin curvilinear structures and are very prone to fragmentation due to challenges such as low-contrast materials and occlusions by surrounding vegetation or buildings. Although deep learning has greatly advanced image segmentation techniques, pixel-level loss functions are insufficient for preserving connectivity and topology of thin curvilinear structures such as roads. Small pixel-level errors can disrupt the extraction of accurate road graphs, complicating subsequent characterizations and analysis tasks. In this paper, we propose a novel, adaptive, structure-aware, connectivity-preserving loss function SAC-Loss. This loss function combines global and local processing to improve structure awareness and uses a proximity-based weighting scheme with asymmetric penalties to fill gaps in the road structures while limiting spurious detections elsewhere. Experimental results demonstrate improved road segmentation in terms of visual quality and quantitative performance. © 2025 IEEE.

关键词： Image segmentation

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：