检索结果-内蒙古大学图书馆

IAENG International Journal of computer science 2023年第4期50卷 1347页

作者： Wang, Qiushi Zhang, Wenyu School of Computer Science and Software Engineering University of Science and Technology Liaoning Anshan114051 China

Session-based recommendation is a technique that leverages the user’s interaction sequence within a brief period to generate personalized recommendations. Currently, traditional transformer-based methods lack the capability to mitigate noise node interference in session sequences, while approaches based on heterogeneous graphs face challenges in capturing long-term sequential dependencies effectively. To tackle these challenges, a heterogeneous graph-enhanced Transformer session-based recommendation method is proposed in this paper. The proposed method constructs a heterogeneous session graph and utilizes a self-attention layer enhanced by the shortest path matrix. This enables the capturing of item embeddings that contains crucial graph structure information. Moreover, the method incorporates adversarial training to prevent excessive reliance on local graph information and achieve a more balanced integration of local and global information. After evaluating experiments on three public datasets, the results clearly demonstrate a remarkable enhancement compared to other robust baseline models, thus providing strong evidence of the method’s effectiveness. © 2023, International Association of Engineers. All rights reserved.

关键词： Graph theory

来源：评论

学校读者我要写书评

暂无评论

Object Detection for UAV Images Based on Improved YOLOv6

IAENG International Journal of Computer Science

引用

IAENG International Journal of computer science 2023年第2期50卷

作者： Li, Yingwei Zhang, Xiaoxia School of Computer Science and Software Engineering University of Science and Technology LiaoNing Anshan114051 China

This paper, we have improved the YOLOv6 object detection network to better detect high-density, small target objects in UAV aerial images. First of all, we designed a mixed data augmentation method which adds background samples to the training, so that the model can better identify falsely detected and falsely detected targets without changing the computational complexity. In addition, to prevent the loss of feature information caused by small target features in the upsampling stage, we have strengthened the Feature Pyramid Networks (FPN) with Feature Alignment Module (FAM) and Feature Selection Module (FSM). Finally, we have formed the Transformer Prediction Head (TPH) by using the Transformer Encoder Block to replace the prediction head of the original model. This ensures accurately locate targets in high-density scenes. The experiment used the Visdrone2019-DET drone aerial photography dataset. Compared with the original YOLOv6 network, the average precision (AP50) of the improved network on the verification set when the IoU is 0.5 was 59.73%, which is 6.02% higher than before the improvement. Comparison results with other mainstream object detection algorithms show that the performance of our proposed algorithm is better than other detectors. Therefore, the algorithm proposed in this paper can effectively detect objects with high density and small targets in UAV aerial photography scenes. © 2023, IAENG International Journal of computer science. All Rights Reserved.

关键词： Unmanned aerial vehicles (UAV)

来源：评论

学校读者我要写书评

暂无评论

A framework for Requirements specification of machine-learning systems 34

A framework for Requirements specification of machine-learni...

引用

34th International Conference on software engineering and Knowledge engineering, SEKE 2022

作者： Wang, Xi Miao, Weikai School of Computer Engineering and Science Shanghai University China Shanghai Key Laboratory of Computer Software Testing and Evaluating Shanghai201114 China Software Engineering Institute East China Normal University China

ISBN: (纸本)1891706543

The rapid development of machine learning (ML) systems has raised many concerns over their quality. Due to the inherent complexity and uncertainty, most of the traditional quality assurance techniques have been challenged, including requirements specification. Current strategies mainly focus on model extraction from existing neural networks to improve interpretability and facilitate system analysis, but failing to include user expectations on the system. To handle the problem, this paper proposes a specification framework for ML requirements where each ML system is regarded as a set of snapshot systems along the evolvement process. There are 3 layers in the framework and the hierarchy indicates that higher-level models need to be built based on lower-level ones. The bottom layer consists of meta snapshot model and meta data model serving as the meta models for snapshot systems and data requirements respectively. The middle layer is for snapshot models each describing a snapshot system through relations between its outputs produced with different inputs. The top layer is a learning model capturing the evolvement process by transitions among snapshot models. These transitions are activated by data models instantiated from meta data model. We adopt the specification of a self-driving system to illustrate the framework. © 2022 Knowledge Systems Institute Graduate school. All rights reserved.

关键词： Metadata

来源：评论

学校读者我要写书评

暂无评论

Graph-based Multi-view Clustering for Web services 21

Graph-based Multi-view Clustering for Web services

引用

21st IEEE International Symposium on Parallel and Distributed Processing with Applications, 13th IEEE International Conference on Big Data and Cloud Computing, 16th IEEE International Conference on Social Computing and Networking and 13th International Conference on Sustainable Computing and Communications, ISPA/BDCloud/SocialCom/SustainCom 2023

作者： Yang, Wang Zhenzhen, Yuan Guosheng, Kang Buqing, Cao Jianxun, Liu Yong, Xiao Hnust Xiangtan China Hunan University of Science and Technology Hunan Provincial Key Lab. for Services Computing and Novel Software Technology Xiangtan China School of Computer Science and Engineering China

ISBN: (纸本)9798350329223

The number of Web services on the Internet has been steadily increasing in recent years due to their growing popularity. Under the big data environment, how to effectively manage Web services is of significance for service discovery, service recommendation, etc. It is widely studied that Web services clustering is an effective way for service management. However, most of the current Web service clustering only extracts the information of Web services for clustering from one view, such as Web service content descriptions, networks in which Web services participate, and so on. Extracting information from Web services only unilaterally will not be able to provide a three-dimensional and comprehensive description of Web services, which may diminish the effect of Web service clustering. In addition, some Web service resources will be wasted if other information of Web services is not used at the same time. We find that multi-view clustering can simultaneously consider multiple information of a data at the same time, and multiple information can complement and enhance each other according to the characteristics of multiview clustering. Therefore, in this paper, we apply Web services to graph-based multi-view clustering in multi-view clustering to improve the performance of Web service clustering by simultaneously considering multiple feature information about Web services and distributing different weights to different information in the clustering process. © 2023 IEEE.

关键词： Web services

来源：评论

学校读者我要写书评

暂无评论

ATMLP: Attention and Time Series MLP for Fall Detection

引用

IAENG International Journal of Applied Mathematics 2023年第1期53卷 1页

作者： Zhang, Huan Cui, Wenhua Shi, Tianwei Tao, Ye Zhang, Jianfeng School of Computer Science and Software Engineering University of Science and Technology Liaoning Anshan114051 China

Addressing the problem that the human body is difficult to be correctly identified in the process of falling, this study proposes an Attention Mechanism and Time Series MLP (AT-MLP) model. To realize the capture of body features during the state and process of the fall of the human body, the time series is combined in the submodule. Thus, the subspace Angle Time Series algorithm (ATS) and the subspace Angle Difference-value Time Series algorithm (ADTS) are designed. The AT-MLP stage composed of Convolutional Block Attention Module (CBAM) and MLP blocks is constructed in the model to strengthen the relation between the time and space dimensions. Comparing with other methods on the Le2i Fall detection Dataset (LFDD) to verify the effectiveness of the AT-MLP model, the experimental results show that the AT-MLP model effectively improves the detection efficiency of human falls and has strong generalization ability. © 2023,IAENG International Journal of Applied Mathematics. All Rights Reserved.

关键词： Time series

来源：评论

学校读者我要写书评

暂无评论

An attention and denoising enhanced network for semi-supervised left atrial segmentation 4

An attention and denoising enhanced network for semi-supervi...

引用

4th International Conference on Algorithms, Microchips, and Network Applications, AMNA 2025

作者： Zhang, Wenhui Zhu, Jinbo Lin, Jiayi School of Computer Science and Information Security Guilin University of Electronic Technology Guangxi Guilin 541004 China College of Computer Information and Software Engineering Guangxi University Guangxi Nanning 530004 China

ISBN: (纸本)9781510690608

Accurate left atrial (LA) segmentation is vital for cardiovascular disease management. However, current meth- ods struggle with complex LA structures, noise, and especially limited annotated data. This paper introduces ADeNet, a novel semi-supervised framework (Attention and Denoising Enhancement Network) that effectively leverages unlabeled data to improve segmentation accuracy and robustness. ADeNet integrates an enhanced multi-scale self-attention mechanism, an adaptive denoising module, and a multi-decoder consistency regular- ization strategy. By incorporating multi-scale self-attention into an encoder-decoder architecture with diverse upsampling decoders, ADeNet better adapts to complex LA shapes and captures both global and local contex- tual information. The adaptive denoising module estimates uncertainty from decoder discrepancies and dynam- ically adjusts loss weights, improving noise robustness. Crucially, the consistency regularization scheme enforces consistent predictions across multiple decoders under diverse perturbations, including data augmentation and strategically applied feature-level noise. This maximizes the information extracted from unlabeled data. Exper- iments on a public LA dataset show ADeNet significantly outperforms state-of-the-art methods in accuracy and robustness, offering an effective approach for high-precision LA segmentation. © 2025 SPIE.

关键词： Attention Mechanism Consistency Regularization Denoising Left Atrial Segmentation Semi-Supervised Learning

来源：评论

学校读者我要写书评

暂无评论

DDVC: Deep Distributed Video Coding Using Quality Enhancement Network

引用

IEEE Transactions on Circuits and Systems for Video Technology 2025年

作者： Zhou, Junwei Ye, Zhuang Yi, Xiangbo Zhang, Weijian Lin, Qiuzhen Xiang, Jianwen Wuhan University of Technology School of Computer Science and Artificial Intelligence Wuhan430070 China Shenzhen University College of Computer Science and Software Engineering Shenzhen518061 China

Distributed video coding (DVC) transfers the complex process of the encoder to the decoder, which is suitable for video applications with limited encoding resources. Deep learning has shown impressive performance in video coding tasks in learning nonlinear compact representations of input frames and reconstructing video frame details. It is worth exploring whether deep learning implementation of the DVC paradigm is feasible and whether performance gains can be obtained. This paper proposes a deep DVC scheme (DDVC) using a quality enhancement network (QEN), which maps pixels to a more compressible latent space via an autoencoder resulting in a compact representation of Wyner-Ziv (WZ) frames. Moreover, considering the spatio-temporal correlation between the WZ frame and the Key frame, the QEN on the decoder side, using CNN and LSTM iteratively extracts common information between the WZ frame and the Key frame, which could further finetune the WZ frame reconstruction. We evaluated DDVC in limited encoding resources application scenarios with 19 related video sequences. Results on the video sequences with different motion intensity levels show that DDVC significantly outperforms existing schemes in reconstruction quality with the same compression ratio. We open-sourced the implementation at GitHub1. © 1991-2012 IEEE.

关键词： Video streaming

来源：评论

学校读者我要写书评

暂无评论

High-Performance GPU based Planetary-Scale Terrain Visualization 2022

High-Performance GPU based Planetary-Scale Terrain Visualiza...

引用

5th International Conference on Image and Graphics Processing, ICIGP 2022

作者： Masood, Zafar Zheng, Jiangbin Ahmad, Idrees Irfan, Muhammad Saeed, Amir School of Software Northwestern Polytechnical University Xi’an China Department of Computer Science and Software Engineering International Islamic University Islamabad Pakistan School of Automation and Electrical Engineering Beihang University China

ISBN: (纸本)9781450395465

High-performance visualization of planetary-scale multi-resolution terrain for high-fidelity display systems is a challenging task. Real-world terrain visualization requires real-time modeling and rendering of the large-scale multi-resolution model. This work proposes a novel method for high-performance multi-resolution terrain modeling and rendering using hardware tessellation. The proposed method submits patch primitives using geometry instancing for high-performance. Patch Level-of-detail (LOD) is simplified using a view-dependent simplification algorithm with error control. The proposed simplification algorithm simplifies the model using viewer information and local height variation for optimum triangulation. Patch simplified LOD is tessellated utilizing a fixed-function tessellator. Height grid-aligned post-tessellated vertices are displaced in 3D to construct accurate high-frequency surface details. Finally, multi-resolution satellite imagery and normal maps are utilized for real-world features display. We evaluated the proposed method for high display resolutions and achieved high rendering rate. The proposed method results are compared with clipmap-based and hardware tessellation based methods. © 2022 ACM.

关键词： Visualization

来源：评论

学校读者我要写书评

暂无评论

Vehicle And Pedestrian Detection Algorithm Based on Improved YOLOv5

IAENG International Journal of Computer Science

引用

IAENG International Journal of computer science 2023年第4期50卷 1401页

作者： Sun, Jiuhan Wang, Zhifeng School of Computer Science and Software Engineering University of Science and Technology Liaoning Anshan114051 China

As urbanization progresses, urban road congestion has intensified, highlighting the need for effective vehicle and pedestrian detection as a cornerstone of public safety transportation. This area holds significant relevance in video surveillance and public safety domains. Despite its importance, achieving precise vehicle and pedestrian detection in complex road environments remains challenging. This paper presents a vehicle-pedestrian detection algorithm based on the improved YOLOv5. Key modifications include the integration of a small target detection layer and alterations to the feature pyramid using the feature fusion technique inherent to the weighted Bidirectional Feature Pyramid Network (BIFPN). This ensures efficient multi-scale feature fusion. A coordinated attention mechanism is introduced to preserve accurate target location data. Furthermore, the paper incorporates the SIOU metric to refine the localization loss function, bolstering both speed and edge regression accuracy. Experimental outcomes indicate that our improved YOLOv5 algorithm augments detection accuracy by 1.9% and achieves a detection speed of 67 FPS, which surpasses many competing target detection algorithms. © 2023, International Association of Engineers. All rights reserved.

关键词： Vehicles

来源：评论

学校读者我要写书评

暂无评论

Lightweight Vehicle Detection Based on Improved Yolox-nano

IAENG International Journal of Computer Science

引用

IAENG International Journal of computer science 2023年第1期50卷 152页

作者： Jing, XiuPing Tian, Ying School of Computer Science and Software Engineering University of Science and Technology Liaoning Liaoning 114051 China

Vehicle detection based on deep learning plays a vital role in various fields, such as autopilot and intelligent transportation. Moreover, it presents a major development direction to computer vision in recent years. Lightweight vehicle detection aims to reduce the burden of computer storage and computing, including exploration of network structure and computing efficiency. But it is difficult to cope with complex and changeable traffic scenes. The urgent problem is to enhance lightweight network performance while maintaining inference speed. In this context, this study proposes an improved Yolox-nano to alleviate the above-mentioned problems by splitting the vanilla convolution unit into two parts and utilizing fewer convolution kernels to generate several feature maps. Subsequently, several linear transformations are further applied to generate cheap feature maps efficiently. The cheap convolutional unit Ghost Module can control feature map ratio of vanilla convolution and linear transformation. The activation function in Ghost Module is improved by SiLU, which has a more powerful nonlinear ability. In addition, the convolution and the batch normalization in Ghost Module can be fused in the pre-processing stage to speed up inference time. Experimental results show that the improved network provides 4.9% and 3.1% mAP on Pascal VOC and MS COCO vehicle datasets. The improved network enhances the performance and maintains an efficient inference speed, and detection results become further competitive using Yolox-nano © 2023, IAENG International Journal of computer *** Rights Reserved.

关键词： Convolution

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：