检索结果-内蒙古大学图书馆

Deep superpixel cut for unsupervised image segmentation

学校读者我要写书评

暂无评论

arXiv 2021年

作者： Lin, Qinghong Zhong, Weichan Lu, Jianglin College of Computer Science and Software Engineering Shenzhen University Shenzhen518060 China Institute of Computer Mathematics and Information Technologies Kazan Federal University Kazan420008 Russia

Image segmentation, one of the most critical vision tasks, has been studied for many years. Most of the early algorithms are unsupervised methods, which use hand-crafted features to divide the image into many regions. Recently, owing to the great success of deep learning technology, CNNs based methods show superior performance in image segmentation. However, these methods rely on a large number of human annotations, which are expensive to collect. In this paper, we propose a deep unsupervised method for image segmentation, which contains the following two stages. First, a Superpixelwise Autoencoder (SuperAE) is designed to learn the deep embedding and reconstruct a smoothed image, then the smoothed image is passed to generate superpixels. Second, we present a novel clustering algorithm called Deep Superpixel Cut (DSC), which measures the deep similarity between superpixels and formulates image segmentation as a soft partitioning problem. Via backpropagation, DSC adaptively partitions the superpixels into perceptual regions. Experimental results on the BSDS500 dataset demonstrate the effectiveness of the proposed method. © 2021, CC BY.

关键词： Image segmentation

Towards High-resolution 3D Anomaly Detection via Group-Level Feature Contrastive Learning

学校读者我要写书评

暂无评论

arXiv 2024年

作者： Zhu, Hongze Xie, Guoyang Hou, Chengbin Dai, Tao Gao, Can Wang, Jinbao Shen, Linlin National Engineering Laboratory for Big Data System Computing Technology Shenzhen University Shenzhen China Department of Computer Science City University of Hong Kong Hong Kong Department of Intelligent Manufacturing CATL Ningde China Fuzhou Fuyao Institute for Advanced Study Fuyao University of Science and Technology Fuzhou China College of Computer Science and Software Engineering Shenzhen University Shenzhen China Guangdong Provincial Key Laboratory of Intelligent Information Processing Shenzhen China Shenzhen University Shenzhen China Shenzhen Institute of Artificial Intelligence and Robotics for Society Shenzhen China

High-resolution point clouds (HRPCD) anomaly detection (AD) plays a critical role in precision machining and high-end equipment manufacturing. Despite considerable 3D-AD methods that have been proposed recently, they still cannot meet the requirements of the HRPCD-AD task. There are several challenges: i) It is difficult to directly capture HRPCD information due to large amounts of points at the sample level;ii) The advanced transformer-based methods usually obtain anisotropic features, leading to degradation of the representation;iii) The proportion of abnormal areas is very small, which makes it difficult to characterize. To address these challenges, we propose a novel group-level feature-based network, called Group3AD, which has a significantly efficient representation ability. First, we design an Intercluster Uniformity Network (IUN) to present the mapping of different groups in the feature space as several clusters, and obtain a more uniform distribution between clusters representing different parts of the point clouds in the feature space. Then, an Intracluster Alignment Network (IAN) is designed to encourage groups within the cluster to be distributed tightly in the feature space. In addition, we propose an Adaptive Group-Center Selection (AGCS) based on geometric information to improve the pixel density of potential anomalous regions during inference. The experimental results verify the effectiveness of our proposed Group3AD, which surpasses Reg3D-AD by the margin of 5% in terms of object-level AUROC on Real3DAD. We provide the code and supplementary information on our website: https://***/M-3LAB/Group3AD. Copyright © 2024, The Authors. All rights reserved.

关键词： Anomaly detection

A comparative study of neural network techniques for automatic software vulnerability detection

学校读者我要写书评

暂无评论

arXiv 2021年

作者： Tang, Gaigai Meng, Lianxiao Wang, Huiqiang Ren, Shuangyin Wang, Qiang Yang, Lin Cao, Weipeng School of computer science and technology Harbin Engineering University Harbin China National Key Laboratory of Science and Technology on Information System Security Institute of System Engineering Chinese Academy of Military Science Beijing China College of Computer Science and Software Engineering Shenzhen University Shenzhen China

software vulnerabilities are usually caused by design flaws or implementation errors, which could be exploited to cause damage to the security of the system. At present, the most commonly used method for detecting software vulnerabilities is static analysis. Most of the related technologies work based on rules or code similarity (source code level) and rely on manually defined vulnerability features. However, these rules and vulnerability features are difficult to be defined and designed accurately, which makes static analysis face many challenges in practical applications. To alleviate this problem, some researchers have proposed to use neural networks that have the ability of automatic feature extraction to improve the intelligence of detection. However, there are many types of neural networks, and different data preprocessing methods will have a significant impact on model performance. It is a great challenge for engineers and researchers to choose a proper neural network and data preprocessing method for a given problem. To solve this problem, we have conducted extensive experiments to test the performance of the two most typical neural networks (i.e., Bi-LSTM and RVFL) with the two most classical data preprocessing methods (i.e., the vector representation and the program symbolization methods) on software vulnerability detection problems and obtained a series of interesting research conclusions, which can provide valuable guidelines for researchers and engineers. Specifically, we found that 1) the training speed of RVFL is always faster than Bi-LSTM, but the prediction accuracy of Bi-LSTM model is higher than RVFL;2) using doc2vec for vector representation can make the model have faster training speed and generalization ability than using word2vec;and 3) multi-level symbolization is helpful to improve the precision of neural network models. © 2021, CC BY.

关键词： Long short-term memory

Wavelength Allocation and Dispersion Penalty Analysis of 50G PON Coexisting with EPON

学校读者我要写书评

暂无评论

Wavelength Allocation and Dispersion Penalty Analysis of 50G...

Asia Communications and Photonics Conference and Exhibition (ACP)

作者： Yang Zou Chenghao Chay Zhenfeng Xue Suyi Wang Qiongxia Shen Rui Li Ming Jiang Dezhi Zhang Xiaoxiao Dai Zhijun Yan Qi Yang Deming Liu National Engineering Laboratory for Next Generation Internet Access System School of Optical and Electronic Information Huazhong University of Science and Technology Wuhan China Fiberhome Telecommunication Technologies Co. Ltd. Wuhan China State Key Laboratory of Optical Fiber and Cable Manufacture Technology China Telecom Research Institute Beijing China Hubei Optical Fundamental Research Center Wuhan China Jinyinhu Laboratory Wuhan China

To address the wavelength collision between 50G PON and EPON, this paper discusses the power budget penalty when changing the downstream wavelength of 50G PON with varying EML chirp factors and DSP combinations.

ISBN: (数字)9798350379266

ISBN: (纸本)9798350379273

关键词： Chirp Simulation Asia IEEE 802.3 Standard EPON Optical fiber communication Complexity theory Resource management Dispersion Photonics

PICO: Accelerating All k-Core Paradigms on GPU

学校读者我要写书评

暂无评论

arXiv 2024年

作者： Zhao, Chen Yu, Ting Zheng, Zhigao Jin, Song Jiang, Jiawei Du, Bo Tao, Dacheng School of Computer Science Wuhan University Wuhan430072 China National Engineering Research Center for Multimedia Software Wuhan University Wuhan430072 China Institute of Artificial Intelligence Wuhan University Wuhan430072 China Hubei Key Laboratory of Multimedia and Network Communication Engineering Wuhan University Wuhan430072 China Hubei Luojia Laboratory Wuhan430072 China Zhejiang Lab Hangzhou China UBTECH Sydney Artificial Intelligence Center University of Sydney Australia School of Information Technologies University of Sydney Australia

Core decomposition is a well-established graph mining problem with various applications that involves partitioning the graph into hierarchical subgraphs. Solutions to this problem have been developed using both bottom-up and top-down approaches from the perspective of vertex convergence dependency. However, existing algorithms have not effectively harnessed GPU performance to expedite core decomposition, despite the growing need for enhanced performance. Moreover, approaching performance limitations of core decomposition from two different directions within a parallel synchronization structure has not been thoroughly explored. This paper introduces an efficient GPU acceleration framework, PICO, for the Peel and Index2core paradigms of k-core decomposition. We propose PeelOne, a Peel-based algorithm designed to simplify the parallel logic and minimize atomic operations by eliminating vertices that are'under-core'. We also propose an Index2core-based algorithm, named HistoCore, which addresses the issue of extensive redundant computations across both vertices and edges. Extensive experiments on NVIDIA RTX 3090 GPU show that PeelOne outperforms all other Peel-based algorithms, and HistoCore outperforms all other Index2core-based algorithms. Furthermore, HistoCore even outperforms PeelOne by 1.1× ∼ 3.2× speedup on six datasets, which breaks the stereotype that the Index2core paradigm performs much worse than the Peel in a shared memory parallel setting. Copyright © 2024, The Authors. All rights reserved.

关键词： Graphics processing unit

Interfacial spintronic THz emission

学校读者我要写书评

暂无评论

arXiv 2023年

作者： Agarwal, Piyush Medwal, Rohit Dongol, Keynesh Mohan, John Rex Yang, Yingshu Asada, Hironori Fukuma, Yasuhiro Singh, Ranjan Division of Physics and Applied Physics School of Physical and Mathematical Sciences Nanyang Technological University 21 Nanyang Link Singapore637371 Singapore Center for Disruptive Photonic Technologies The Photonics Institute Nanyang Technological University Singapore639798 Singapore Department of Physics Indian Institute of Technology Uttar Pradesh Kanpur208016 India Department of Electronic Devices and Engineering Graduate School of Science and Engineering Yamaguchi University Ube755-8611 Japan Department of Physics and Information Technology Faculty of Computer Science and System Engineering Kyushu Institute of Technology Iizuka820-8502 Japan

The broken inversion symmetry at the ferromagnet (FM)/heavy-metal (HM) interface leads to spin-dependent degeneracy of the energy band, forming spin-polarized surface states. As a result, the interface serves as an effective medium for converting spin accumulation into two-dimensional charge current through the inverse Rashba-Edelstein effect. Exploring and assessing this spin-to-charge conversion (SCC) phenomenon at the FM/HM interface could offer a promising avenue to surpass the presumed limits of SCC in bulk HM layers. We utilize spintronic heterostructures as a platform to measure the spin-to-charge conversion (SCC) experienced by photoexcited spin currents. These heterostructures emit terahertz electric field when illuminated by femtosecond laser pulses, enabling us to quantitatively assess the ultrafast SCC process. Our results demonstrate a robust interfacial spin-to-charge conversion (iSCC) within a synthetic antiferromagnetic heterostructure, specifically for the NiFe/Ru/NiFe configuration, by isolating the SCC contribution originating from the interface itself, separate from the bulk heavy-metal (HM) region. Moreover, the iSCC at the NiFe/Ru interface is discovered to be approximately 27% of the strength observed in the highest spin-Hall conducting heavy-metal, Pt. Our results thus highlight the significance of interfacial engineering as a promising pathway for achieving efficient ultrafast spintronic devices. Copyright © 2023, The Authors. All rights reserved.

关键词： Electric fields

An IoT-Based Air Pollution Monitoring system for Smart City

学校读者我要写书评

暂无评论

An IoT-Based Air Pollution Monitoring System for Smart City

Computer systems (ICCS), IEEE International Conference on

作者： Alibek Aubakirov Kanat Makhanov Aliya Issainova Natalya Glazyrina Bibigul Razkhova Dina Satybaldina School of Computer Science Hangzhou Dianzi University Hangzhou China Department of Radio Engineering Electronics and Telecommunications L.N Gumilyov Eurasian National University Astana Kazakhstan Information Security Department L.N. Gumilyov Eurasian National University Astana Kazakhstan Computer and Software Engineering Department L.N Gumilyov Eurasian National University Astana Kazakhstan Department of Artificial Intelligence Technologies) L.N. Gumilyov Eurasian National University Astana Kazakhstan Information Security and Cryptology Research Institute L.N. Gumilyov Eurasian National University Astana Kazakhstan

ISBN: (数字)9798350366341

ISBN: (纸本)9798350366358

In this paper, we present the results of research on the design and development of a gas analytical unit, an electronic nose, with an autonomous power supply, as well as the function of wireless data transmission via the Long-Range Wide Area Networks protocol. We built the device on a circuit board featuring eight gas sensors (MQ series), an STM32F103C8T6 microcontroller with 12-bit analog-to-digital conversion, a RAK3172 wireless data transmission module, voltage dividers, and a USB serial adapter. It also has an HDC2080DMBR temperature and humidity sensor. We evaluated the electronic nose performance towards standard mixtures of carbon dioxide, carbon monoxide, nitrogen dioxide, ammonia, and hydrogen sulfide, which are important analytes in outdoor and indoor air quality monitoring. Multidimensional linear regression methods were employed to build calibration curves for electronic nose sensors, which allowed for converting the measured sensor responses into quantitative values of the concentrations of the analyzed substances. The estimated coefficients of determination, R2, approach 1 indicating a strong correlation between the approximating model and the original data. The multivariate chemiresistive responses were successfully clustered using the k-means algorithm, revealing clear distinctions between sensor data associated with different gas concentrations. The k-NN classifier, applied to the clustered data, achieved 100% accuracy in identifying gas concentrations, underscoring the potential of the proposed methodology for gas sensing applications. We proposed and implemented the architecture of a network of mobile gas analyzers as the Internet of Things devices for environmental monitoring using the VK Cloud Platform.

关键词： Wireless communication Wireless sensor networks Atmospheric measurements Smart cities Clustering algorithms Electronic noses Pollution measurement Classification algorithms Calibration Data communication

Analytical Models for QoT Estimations in Super-C and Super-C + L Bands Optical Transmission systems

学校读者我要写书评

暂无评论

Analytical Models for QoT Estimations in Super-C and Super-C...

Conference on Lasers and Electro-Optics (CLEO)

作者： Baoluo Yan Yijun Sun Ziwen Zhou Hong Liu Hongya Wang Chen Huan Zhiyong Zhao Lipeng Feng Anxu Zhang Zhenhua Feng Hu Shi WDM System Department of Wireline Product R&D Institute ZTE Corp. Beijing China State Key Laboratory of Mobile Network and Mobile Multimedia Technology China School of Software Engineering Chongqing University of Posts and Telecommunications Chongqing China Optics Valley Laboratory School of Optical and Electronic Information and Wuhan National Laboratory for Optoelectronics Huazhong University of Science and Technology Wuhan China China Telecom Research Institute Beijing China

We propose a novel physical model-based QoT estimation with achieved OSNR estimation error ≤ 1.5 dB in primary paths and OSNR prediction error ≤ 1.76 dB in unestablished lightpaths for both super-C bands and super-C... 详细信息

ISBN: (纸本)9781957171258

关键词： Hole burning Optical amplifiers Optical networks Optical signal to noise ratio Stimulated scattering Variable optical attenuators

Depth-Aware Multi-Modal Fusion for Generalized Zero-Shot Learning

学校读者我要写书评

暂无评论

Depth-Aware Multi-Modal Fusion for Generalized Zero-Shot Lea...

IEEE International Conference on Industrial Informatics (INDIN)

作者： Weipeng Cao Xuyang Yao Zhiwu Xu Yinghui Pan Yixuan Sun Dachuan Li Bohua Qiu Muheng Wei Guangdong Laboratory of Artificial Intelligence and Digital Economy (Shenzhen) Shenzhen China National Engineering Laboratory for Big Data System Computing Technology Shenzhen University Shenzhen China College of Computer Science and Software Engineering Shenzhen University Shenzhen China Stony Brook University New York United States Research Institute of Trustworthy Autonomous Systems Southern University of Science and Technology Shenzhen China Department of Computer Science and Engineering Southern University of Science and Technology Shenzhen China ZhenDui Industry Artificial Intelligence Co. Ltd Shenzhen China Department of Automation Shanghai Jiao Tong University Shanghai China Key Laboratory of System Control and Information Processing Ministry of Education of China Shanghai China

ISBN: (数字)9798331527471

ISBN: (纸本)9798331527488

Realizing Generalized Zero-Shot Learning (GZSL) based on large models is emerging as a prevailing trend. However, most existing methods merely regard large models as black boxes, solely leveraging the features output by the final layer while disregarding potential performance enhancements from other layers. Indeed, numerous researchers have visually depicted variations in the features learned across different layers of neural networks. Motivated by this observation, we propose a Vision Transformer (ViT)-based GZSL method named Depth-Aware Multi-Modal ViT (DAM2ViT), which exploits multi-level features of ViT. DAM2ViT incorporates a multi-modal interaction block to align semantic information of categories across multiple layers, thereby augmenting the model's capacity to learn associations between visual and semantic spaces. Extensive experiments conducted on three benchmark datasets (i.e., CUB, SUN, AWA2) have showcased that DAM2ViT achieves competitive results compared to state-of-the-art methods.

关键词： Visualization Adaptation models Semantics Zero shot learning Neural networks Termination of employment Transformers Market research Sun Optimization