检索结果-内蒙古大学图书馆

arXiv 2024年

作者： Wan, Yao Wan, Guanghua Zhang, Shijie Zhang, Hongyu Zhou, Pan Jin, Hai Sun, Lichao National Engineering Research Center for Big Data Technology and System Services Computing Technology and System Lab Cluster and Grid Computing Lab School of Computer Science and Technology Huazhong University of Science and Technology Wuhan China Huazhong University of Science and Technology China University of Leigh United States

Recent years have witnessed significant progress in developing deep learning-based models for automated code completion. Examples of such models include CodeGPT and StarCoder. These models are typically trained from a large amount of source code collected from open-source communities such as GitHub. Although using source code in GitHub has been a common practice for training deep-learning-based models for code completion, it may induce some legal and ethical issues such as copyright infringement. In this paper, we investigate the legal and ethical issues of current neural code completion models by answering the following question: Is my code used to train your neural code completion model? To this end, we tailor a membership inference approach (termed CodeMI) that was originally crafted for classification tasks to a more challenging task of code completion. In particular, since the target code completion models perform as opaque black boxes, preventing access to their training data and parameters, we opt to train multiple shadow models to mimic their behavior. The acquired posteriors from these shadow models are subsequently employed to train a membership classifier. Subsequently, the membership classifier can be effectively employed to deduce the membership status of a given code sample based on the output of a target code completion model. We comprehensively evaluate the effectiveness of this adapted approach across a diverse array of neural code completion models, (i.e., LSTM-based, CodeGPT, CodeGen, and StarCoder). Experimental results reveal that the LSTM-based and CodeGPT models suffer the membership leakage issue, which can be easily detected by our proposed membership inference approach with an accuracy of 0.842, and 0.730, respectively. Interestingly, our experiments also show that the data membership of current large language models of code, e.g., CodeGen and StarCoder, is difficult to detect, leaving amper space for further improvement. Finally, we also t

关键词： Deep learning

来源：评论

学校读者我要写书评

暂无评论

Metamorphic Testing for Traffic Sign Detection and Recognition

Metamorphic Testing for Traffic Sign Detection and Recogniti...

引用

IEEE International Conference on Software Quality, Reliability and Security Companion (QRS-C)

作者： Yuanyuan Long Yong Fan Ya Pan School of Computer Science and Technology of Southwest University of Science and Technology Mianyang China Sichuan Big Data and Intelligent System Engineering Technology Research Center Mianyang China

Traffic sign detection and recognition (TSDR) is a crucial technology to realize vehicle autonomous driving and maintain road safety. Misdetection or misclassification of traffic signs can lead to serious traffic accidents. Therefore, it is essential to ensure the correctness of traffic sign detection and recognition systems. However, most existing methods use labeled data, but it isn't easy to obtain labeled data. The metamorphic testing can reduce the dependence on the labeled data, but the current metamorphic relations do not consider the domain characteristics of traffic signs, so it is difficult to effectively detect traffic sign detection and recognition systems faults. To solve the above problems, this paper introduces metamorphic testing to verify traffic sign detection and recognition systems and designs and implements a tool, MTSDR, which provides a series of metamorphic relations in the traffic domain. It can generate many follow-up test images based on these metamorphic relations and automatically performs tests for evaluation, focusing on practical failure situations. The experimental evaluation was carried out on four advanced traffic sign detection and recognition models on open datasets. The results show that MTSDR can generate natural images and find thousands of traffic sign detection and recognition faults. Compared with the previous methods, our method has a higher fault detection capability, and the highest fault detection rate reaches more than 40%. In addition, it finds that retraining the traffic sign detection and recognition models using the images that found the fault can improve the model's performance, reduce its fault, and reduce the fault rate by 11.77% on average.

关键词：

来源：评论

学校读者我要写书评

暂无评论

Real-Time Violent Action Recognition Using Key Frames Extraction and Deep Learning

引用

computers, Materials & Continua 2021年第11期69卷 2217-2230页

作者： Muzamil Ahmed Muhammad Ramzan Hikmat Ullah Khan Saqib Iqbal Muhammad Attique Khan Jung-In Choi Yunyoung Nam Seifedine Kadry Deparment of Computer Science and Information Technology The University of LahoreSargodha CampusSargodha40100Pakistan Department of Computer Science COMSATS University IslamabadWah CampusWah Cantt47040Pakistan School of System and Technology University of Management and TechnologyLahore54782Pakistan Department of Computer Science and Information Technology University of SargodhaSargodha40100Pakistan College of Engineering Al Ain UniversityAl AinUnited Arab Emirates Department of Computer Science HITEC University TaxilaTaxilaPakistan Applied Artificial Intelligence Ajou UniversitySuwonKorea Department of Computer Science and Engineering Soonchunhyang UniversityAsanKorea Department of Mathematics and Computer Science Faculty of ScienceBeirut Arab UniversityLebanon

Violence recognition is crucial because of its applications in activities related to security and law *** semi-automated systems have issues such as tedious manual surveillances,which causes human errors and makes these systems less *** approaches have been proposed using trajectory-based,non-object-centric,and deep-learning-based *** studies have shown that deep learning techniques attain higher accuracy and lower error rates than those of other ***,the their performance must be *** study explores the state-of-the-art deep learning architecture of convolutional neural networks(CNNs)and inception V4 to detect and recognize violence using video *** the proposed framework,the keyframe extraction technique eliminates duplicate consecutive *** keyframing phase reduces the training data size and hence decreases the computational cost by avoiding duplicate *** feature selection and classification tasks,the applied sequential CNN uses one kernel size,whereas the inception v4 CNN uses multiple kernels for different layers of the *** empirical analysis,four widely used standard datasets are used with diverse *** results confirm that the proposed approach attains 98%accuracy,reduces the computational cost,and outperforms the existing techniques of violence detection and recognition.

关键词： Violence detection violence recognition deep learning convolutional neural network inception v4 keyframe extraction

来源：评论

学校读者我要写书评

暂无评论

Dual Encoders for Diffusion-based Image Inpainting

Dual Encoders for Diffusion-based Image Inpainting

引用

International Conference on Acoustics, Speech, and Signal Processing (ICASSP)

作者： Dezhi Zheng Kaijun Deng Jinbao Wang Linlin Shen Computer Vision Institute School of Computer Science and Software Engineering Shenzhen University Shenzhen China National Engineering Laboratory for Big Data System Computing Technology Shenzhen University Shenzhen China Guangdong Provincial Key Laboratory of Intelligent Information Processing Shenzhen China

ISBN: (数字)9798350368741

ISBN: (纸本)9798350368758

Current diffusion-based inpainting models struggle to preserve unmasked regions or generate highly coherent content. Additionally, it is hard for them to generate meaningful content for 3D inpainting. To tackle these challenges, we design a plug-and-play branch that runs through the entire generation process to enhance existing models. Specifically, we utilize dual encoders - a Convolutional Neural Network (CNN) encoder and the pre-trained Variational AutoEncoder (VAE) encoder, to encode masked images. The latent code and the feature map from the dual encoders are fed to diffusion models simultaneously. In addition, we apply Zero-padded initialization to solve the problem of mode collapse caused by this branch. Experiments on BrushBench and EditBench demonstrate that models with our plug-and-play branch can improve the coherence of inpainting, and our model achieves new state-of-the-art results.

关键词： Visualization Three-dimensional displays Codes Autoencoders Coherence Signal processing Diffusion models Decoding Convolutional neural networks Speech processing

来源：评论

学校读者我要写书评

暂无评论

CMRM: Collaborative Multi-agent Reinforcement Learning for Multi-objective Traffic Signal Control

引用

IEEE Transactions on Consumer Electronics 2025年

作者： Nie, Lei Qi, Dandan Liu, Bingyi Li, Peng Bao, Haizhou He, Heng Wuhan University of Science and Technology School of Computer Science and Technology Wuhan430065 China Wuhan University of Science and Technology Hubei Province Key Laboratory of Intelligent Information Processing and Real-Time Industrial System Wuhan430065 China Wuhan University of Technology School of Computer Science and Artificial Intelligence Wuhan430070 China Jilin University Key Laboratory of Symbolic Computation and Knowledge Engineering of Ministry of Education Changchun130012 China

Efficient traffic signal control is a cost-effective way to ease urban traffic congestion. Multi-agent reinforcement learning (MARL) has become a widely adopted method for optimizing traffic signal control (TSC). However, existing MARL-based methods often focus on a single optimization objective, lacking a comprehensive consideration of traffic efficiency, environmental pollution, and traffic safety. Simultaneously, these methods often fail to effectively capture the dynamic and complex interactions among agents in multi-intersection scenarios, which negatively impacts traffic efficiency. In this article, we propose a collaborative MARL-based method for multi-objective TSC, called CMRM. First, we introduce a multi-objective reward mechanism that integrates traffic efficiency, environmental impact, and safety to guide agents toward more comprehensive optimization. Second, we design a cooperation enhancement module (CEM) based on the graph attention mechanism to dynamically capture neighboring agents’ state information. This mitigates the partial observability problem in independent proximal policy optimization (IPPO) and enhances the model’s ability to capture dynamic and complex interactions among agents. Finally, we assess the performance of the proposed CMRM method using SUMO on two real traffic networks. Experimental results demonstrate that our method significantly improves traffic efficiency while reducing environmental pollution and enhancing traffic safety, compared to the best performing baseline, our method reduces CO2 emission by approximately 17.53% and 9.57%, and lowers vehicle collision risks by 44.39% and 42.85% in two different traffic networks. © 1975-2011 IEEE.

关键词： Traffic congestion

来源：评论

学校读者我要写书评

暂无评论

A 16-channel 14-Bit, 1.2-Msps/channel, Three-stage Hybrid ADC with SAR/SS/TDC Architecture for Dark Matter Detection

A 16-channel 14-Bit, 1.2-Msps/channel, Three-stage Hybrid AD...

引用

IEEE Symposium on Nuclear science (NSS/MIC)

作者： B. Yang C. Yu H. Dong M. Xu Y. Qian W. Gao School of Computer Science Interdisciplinary Research Center of Bio-inspired System-on-chip Northwestern Polytechnical University Xi’an China Chinese Academy of Science Institute of Modern Physics Lanzhou China Ministry of Education Engineering Research Center of Embedded System Integrated Xi’an China

ISBN: (数字)9798350388152

ISBN: (纸本)9798350388169

Energy-efficient front-end electronics plays an important role to construct space-bone γ-ray radiation detector systems for the next-generation dark matter detection mission of China. A multichannel analog-to-digital converter (ADC) application-specific integrated circuit (ASIC) is required to digitize the analog signals represented charge quantity from a PMT detector with the plastic scintillator. The challenges of the ADC lie in multiple channels, a resolution of 14 bits, a sampling rate of large than 1 Msps with a low power condition. Two hybrid-architecture ADC prototype chips that combines successive approximate register (SAR), single-slope (SS) ramp, and a flash time-to-digital converter (TDC) have been designed by our group in the last three years. Both typical topologies and fabricated process have been verified. However, the developed ADCs cannot meet the engineering application of front-end electronics. In this paper, we propose an improved version of ADC to enhance the channel numbers and to improve the overall performances. A 16-channel ADC prototype chip has been designed and fabricated in 0.18-μm CMOS technology with a 1.8/3.3V power supply. The die size is 2.85 mm × 3.35 mm. With a clock frequency of 50 MHz, the simulated ENOB and the sampling rate are 12.87 bits and 1.27 Msps, respectively. The power dissipation is 3.9 mW/channel. The 16-channel ADC chip is currently under testing, and preliminary results have been obtained. For future work, the ADC circuits will be integrated into a multichannel front-end readout ASIC together for all-digital data output.

关键词： Scintillators Semiconductor detectors Power supplies Dark matter Radiation detectors Prototypes Topology

来源：评论

学校读者我要写书评

暂无评论

DSDFormer: An Innovative Transformer-Mamba Framework for Robust High-Precision Driver Distraction Identification

arXiv

引用

arXiv 2024年

作者： Chen, Junzhou Zhang, Zirui Yu, Jing Huang, Heqiang Zhang, Ronghui Xu, Xuemiao Sheng, Bin Yan, Hong Guangdong Provincial Key Laboratory of Intelligent Transportation System School of Intelligent Systems Engineering Sun Yat-Sen University Guangzhou510275 China The School of Computer Science and Engineering South China University of Technology China The Department of Computer Science and Engineering Shanghai Jiao Tong University China The Department of Electrical Engineering City University of Hong Kong Kowloon Hong Kong

Driver distraction remains a leading cause of traffic accidents, posing a critical threat to road safety globally. As intelligent transportation systems evolve, accurate and real-time identification of driver distraction has become essential. However, existing methods struggle to capture both global contextual and fine-grained local features while contending with noisy labels in training datasets. To address these challenges, we propose DSDFormer, a novel framework that integrates the strengths of Transformer and Mamba architectures through a Dual State Domain Attention (DSDA) mechanism, enabling a balance between long-range dependencies and detailed feature extraction for robust driver behavior recognition. Additionally, we introduce Temporal Reasoning Confident Learning (TRCL), an unsupervised approach that refines noisy labels by leveraging spatiotemporal correlations in video sequences. Our model achieves state-of-the-art performance on the AUC-V1, AUC-V2, and 100-Driver datasets and demonstrates real-time processing efficiency on the NVIDIA Jetson AGX Orin platform. Extensive experimental results confirm that DSDFormer and TRCL significantly improve both the accuracy and robustness of driver distraction detection, offering a scalable solution to enhance road safety. Copyright © 2024, The Authors. All rights reserved.

关键词： Highway accidents

来源：评论

学校读者我要写书评

暂无评论

ACDR-CRAFF Net: A Multi-Scale Network Based on Adaptive Channel and Coordinate Relational Attention Network for Remote Sensing Scene Classification

引用

IET Image Processing 2025年第1期19卷

作者： Dai, Wei Xu, Haixia Shi, Furong Yuan, Liming Wang, Xinyu Wen, Xianbin School of Computer Science and Engineering Tianjin University of Technology Tianjin China Key Laboratory of Computer Vision and System Ministry of Education Tianjin University of Technology Tianjin China Tianjin Key Laboratory of Intelligence Computing and Novel Software Technology Tianjin University of Technology Tianjin China

Accurate classification of remote sensing scene images is crucial for diverse applications, from environmental monitoring to urban planning. While convolutional neural networks (CNNs) have dramatically improved classification accuracy, challenges remain due to the complex distribution of small objects, varied spatial configurations, and intra-class multimodality in remote sensing images. In this work, we make three key contributions to address these challenges. (1) We propose the adaptive channel and coordinate relational attention network (ACDR-CRAFF), a novel multi-scale feature fusion framework designed to enhance feature representation across scales. (2) We introduce two innovative modules: the adaptive channel dimensionality reduction (ACDR) module, which dynamically adjusts channel representations to retain essential low-dimensional features, and the coordinate relational attention multi-scale feature fusion (CRAFF) module, which effectively captures and transfers spatial information between feature levels. (3) By integrating ACDR and CRAFF, our model achieves a progressive fusion of local to global features, ensuring robust feature expressiveness at multiple scales. Experimental results on four widely used benchmark datasets demonstrate that ACDR-CRAFF consistently outperforms several state-of-the-art methods, achieving significant improvements in classification accuracy and setting a new benchmark for complex remote sensing scene classification tasks. These results underscore the effectiveness of our approach in addressing the limitations of existing methods and advancing the state of the art in remote sensing image analysis. © 2025 The Author(s). IET Image Processing published by John Wiley & Sons Ltd on behalf of The Institution of engineering and Technology.

关键词： Photointerpretation

来源：评论

学校读者我要写书评

暂无评论

Locality-Aware Spatial-Temporal Attention Neural Network with Contrast Learning for Traffic Flow Prediction 3

Locality-Aware Spatial-Temporal Attention Neural Network wit...

引用

3rd International Conference on Internet of Things and Smart City, IoTSC 2023

作者： Shi, Zheng Wang, Jingping Zhang, Yingjun Yin, Hui Huang, Hua Zhong, Yanglong School of Computer and Information Technology Beijing Jiaotong University Beijing100044 China Frontiers Science Center for Smart High-speed Railway System Beijing Jiaotong University Beijing100044 China Information Technology Center Beijing Jiaotong University Beijing100044 China School of Civil Engineering Beijing Jiaotong University Beijing100044 China

ISBN: (纸本)9781510666375

Traffic flow prediction is a crucial component of establishing the intelligent transportation system. Accurate and real-time traffic flow prediction is of great significance for urban traffic management. With the recent development of artificial intelligence, deep learning-based methods have been effective tools for traffic flow prediction. However, locality-unawareness and data scarcity are still two open issues to be considered in traffic flow prediction tasks. To address these problems, we propose a novel locality-aware spatial-temporal attention neural network, named LASTANN, in this paper. Specifically, we propose two elaborate attention modules to perceive local information in spatial and temporal dimensions. We also propose an auxiliary module based on contrast learning to strengthen the representation ability of model. We verify the effectiveness of the proposed model on two real-world traffic flow datasets. The experimental results demonstrate that the proposed LASTANN outperforms other baselines consistently and each component enhances the prediction performances significantly. © 2023 SPIE.

关键词： Deep learning

来源：评论

学校读者我要写书评

暂无评论

Exploring the usefulness of light field super-resolution for object detection

引用

The Journal of China Universities of Posts and Telecommunications 2021年第5期28卷 68-81页

作者： Zhang Wenzhe Shi Fan Zhao Meng Chen Shengyong Engineering Research Center of Learning-Based Intelligent System of Ministry of Education Tianjin University of TechnologyTianjin 300384China The Key Laboratory of Computer Vision and System of Ministry of Education Tianjin University of TechnologyTianjin 300384China School of Computer Science and Engineering Tianjin University of TechnologyTianjin 300384China The MOE Key Laboratory of Weak-Light Nonlinear Photonics Nankai UniversityTianjin 300457China

In order to solve the impact of image degradation on object detection, an object detection method based on light field super-resolution(LFSR) is proposed. This method takes LFSR as an image enhancement step to provide high-quality images for object detection without using expensive imaging equipment. To evaluate this method, three types of objects: person, bicycle, and car, are chosen and the results are compared from 5 parts: detected object quantity, mean confidence score, detection results in different scenes, error detection, and detection results from different images sizes and detection speed. Experimental results based on the common object in context(COCO) dataset show that the method incorporated LFSR improves performance of object detection models.

关键词： light-field(LF) super-resolution(SR) object detection RetinaNet YOLOv3 TinyYOLOv3

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：