检索结果-内蒙古大学图书馆

International Conference on Knowledge and Smart technology (KST)

作者： Daichi Kato Yu Ding Chao Zhang Shogo Tokai Chunzhi Gu Graduate School of Engineering University of Fukui Fukui-city Fukui Japan Graduate School of Science and Engineering University of Toyama Toyama-city Toyama Japan Faculty of Engineering University of Toyama Toyama-city Toyama Japan Faculty of Engineering University of Fukui Fukui-city Fukui Japan Department of Computer Science and Engineering Toyohashi University of Technology Toyohashi-city Aichi Japan

ISBN: (数字)9798331520403

ISBN: (纸本)9798331520410

Fish freshness estimation is a key procedure to ensure the quality of products in the marine industry. Recently, this field has progressed towards using computer vision-based techniques to realize fast and convenient classification using fisheye images considering the high freshness sensitivity. However, prior approaches can only handle the fish images taken in ideal shooting environments, which limits the applications. In this paper, we propose to address the task of fish freshness estimation under low-light environments using dark fish-eye images. Specifically, we first learn to recover the bright image from the dark one, and then segment the fish-eye area to produce a score mask. Eventually, the mask image is regarded as a guidance to improve the estimation accuracy. Our method jointly classifies the freshness into three pre-determined and realizes image recovery. Experiments on a publicly available fish eye dataset demonstrate the effectiveness of our method to the fisheye images captured in low-light shooting environments.

关键词： Industries Degradation Image segmentation Accuracy Sensitivity Estimation computer architecture Fish Feature extraction Image restoration

来源：评论

学校读者我要写书评

暂无评论

An end-to-end model for time series classification in the presence of missing values

引用

Expert Systems with Applications 2025年 284卷

作者： Liu, Mengna Yao, Pengshuai Cheng, Xu Chen, Shengyong The School of Computer Science and Engineering Tianjin University of Technology No. 391 Bin Shui Xi Dao Road Xiqing District Tianjin300384 China

The presence of missing data frequently impairs the performance of time series classification in practical applications, leading to incomplete time series classification (ITSC) challenges. One-stage methods, which integrate label information into the imputation process, often overlook the potential for imputation errors that can distort temporal patterns and subsequently degrade feature learning. To address this limitation, we propose a novel joint learning framework comprising a temporal imputation module (TIM) and a noise-robust multi-scale feature learning module (MSFL). The TIM employs a Gated Recurrent Unit to impute missing values, introducing potential imputation error. To mitigate this, the MSFL leverages dilated convolutional neural networks to extract robust features at multiple scales, enabling the model to capture both fine-grained local details and broader, longer-range dependencies while simultaneously reducing the impact of imputation error. Extensive evaluations on 68 univariate UCR datasets demonstrate a substantial improvement in classification accuracy compared to existing methods across missing data ratios ranging from 20 % to 80 %. The superiority of our model increase as the missing ratio increases, at 80 % missingness, average accuracy improves by 20.83 %. Furthermore, validation on multivariate PAM and real-world datasets confirms the model's superior performance, underscoring the efficacy of jointly optimizing imputation and multi-scale feature learning for robust ITSC. © 2025 Elsevier Ltd

关键词： Convolutional neural networks

来源：评论

学校读者我要写书评

暂无评论

Superior Flexible Tactile Sensor with AI-based Unit Microstructure Design for Human and Robot Parameters Monitoring Application

引用

IEEE Sensors Journal 2025年

作者： Yao, Pei Liu, Zhuo Xiao, Xuan Duo, Yongchao Dai, Guang Wang, Liang Xu, Hongchen Tiangong University Tianjin Key Laboratory of Autonomous Intelligence Technology and Systems School of Computer Science and Technology Tianjin300387 China Tiangong University School of Textiles Science and Engineering Tianjin300387 China Tianjin University of Technology School of Science Key Laboratory of Quantum Optics and Intelligent Photonics School of Science Tianjin300384 China Shanxi Xi'an710049 China Xi'an Jiaotong University School of Instrument Science and Technology Shanxi Xi'an710049 China

Due to the development of wearable monitoring systems and the gradually expanding application demand, research interest has turned to free-form microstructures design capable of realizing superior and stable piezoresistive responses. The deep learning methods can accelerate and improve microstructure design with data-driven precision and increased efficiency. In this study, the concept of programmable microstructure is introduced to digitalize microstructure unit, convolutional neural network (CNN) model and Tabu-Search algorithm are respectively used to predict and search iteratively novel stress-voltage (S-V) responses from the finite element modeling (FEM) training set. Therefore, efficient exploration of parameter space and faster generating of novel microstructure designs can be achieved by the proposed algorithm framework, and the optimized microstructure has also been experimentally validated. Moreover, the working mechanism of the sensing medium layer is explicated by the resistance evolution of Ti2CTx-MXene atomic layers under the compressive strain via density functional theory calculations. The proposed AI-based tactile sensor exhibits high sensitivity, large linearity, fast response, and excellent cycling stability, it leverages strong capabilities to detect human being's subtle activities and terrain perception of snake robot. This study could achieve a higher degree in human physiological monitoring and robot tactile construction, and gain a new insight for digital microstructure in sensor design. © 2001-2012 IEEE.

关键词： Programmable robots

来源：评论

学校读者我要写书评

暂无评论

Negative Binomial Regression Model for Predicting the Number of Tourists in National Parks of Thailand 10

Negative Binomial Regression Model for Predicting the Number...

引用

10th International Conference on Digital Arts, Media and technology and 8th ECTI Northern Section Conference on Electrical, Electronics, computer and Telecommunications engineering, ECTI DAMT and NCON 2025

作者： Thepsumritporn, Sthaporn Jitpattanakul, Anuchit Mekruksavanich, Sakorn Phaphan, Wikanda King Mongkut's University of Technology North Bangkok Faculty of Applied Science Department of Applied Statistics Bangkok10800 Thailand King Mongkut's University of Technology North Bangkok Faculty of Applied Science Department of Mathematics Bangkok10800 Thailand Intelligent and Nonlinear Dynamic Innovations Research Center Thailand University of Phayao School of Information and Communication Department of Computer Engineering Phayao56000 Thailand Research Group in Statistical Learning and Inference Thailand

ISBN: (纸本)9798331543273

This article examines the use of count data models to predict the number of tourists visiting Thailand's national parks. The dataset, encompassing observations from 2016 to 2022 across 146 national parks, exhibits characteristics such as discreteness, over dispersion, and non-negative integer values. Two models - Poisson (POI), and Negative Binomial (NB) - were compared to identify the most effective approach for modeling and prediction. Model evaluation metrics, including the Akaike Information Criterion (AIC) and log-likelihood, were employed. Results demonstrate that the Negative Binomial regression model best captures the data's overdispersion compared to the POI model. The NB model achieved the lowest AIC and highest log-likelihood, affirming its superior performance for this application. Our findings underscore the NB model's utility in predicting tourism-related count data and suggest its broader applicability to other domains characterized by overdispersed count data. © 2025 IEEE.

关键词： Logistic regression

来源：评论

学校读者我要写书评

暂无评论

Predictive Analytics for Ear Health Management Using IoT and Ensemble Learning

Predictive Analytics for Ear Health Management Using IoT and...

引用

2025 International Conference on Intelligent Control, Computing and Communications, IC3 2025

作者： Saravanan, S.K. Parvathy, Palaniraj Rajidurai Meenakshi, R. Prabu, K. Ganeshbabu, T.R. Meenakshi, B. Saveetha School of Engineering Saveetha Institute of Medical and Technical Sciences Saveetha University Department of Big Data and Network Security Tamil Nadu Chennai India Mphasis Corporation ChandlerAZ United States Chennai Institute of Technology Department of Computer Science and Engineering Tamil Nadu Chennai India Gojan School of Business and Technology Department of Electronics and Communication Engineering Redhills Tamil Nadu Chennai India Department of Electronics and Communication Tamil Nadu Namakkal India Sri Sairam Engineering College Department of Electrical and Electronics Engineering Tamil Nadu Chennai India

ISBN: (纸本)9798331527495

Proactive health monitoring solutions are becoming more important due to a growing population of older people. To better manage the ear health of the elderly, this research investigates ensemble learning methods within the domain of predictive analytics using the Internet of Things (IoT). It provides an ensemble learning approach that uses IoT capabilities to predict potential ear health problems in this population. It evaluates the efficacy of many ensembles learning approaches, including bagging, boosting, and stacking, using real-world sensor data acquired from IoT devices. Our results show that ensemble learning methods considerably improve prediction accuracy compared to standalone models. Ensemble learning overcomes the limitations of individual algorithms and maximizes their combined prediction potential by efficiently combining varied models. Our findings have important implications for the management of healthcare for the elderly, providing new opportunities to enhance early diagnosis and intervention techniques for ear health concerns via the use of predictive analytics provided by the IoT. It highlights the potential of ensemble learning approaches to improve predictive analytics in the senior healthcare sector, especially with IoT-driven solutions. © 2025 IEEE.

关键词： mHealth

来源：评论

学校读者我要写书评

暂无评论

Metaverse meets distributed machine learning: A contemporary review on the development with privacy-preserving concerns

引用

ICT Express 2025年

作者： Gardner, Samuel Harry Hoang, Trong-Minh Na, Woongsoo Dao, Nhu-Ngoc Cho, Sungrae Department of Computer Science and Engineering Sejong University Seoul 05006 South Korea Intelligently Connected Networks Lab Posts and Telecommunications Institute of Technology Hanoi 100000 Viet Nam Division of Computer Science and Engineering Kongju National University Cheonan 31080 South Korea School of Computer Science and Engineering Chung-Ang University Seoul 05006 South Korea

Distributed machine learning utilization in the metaverse exposes many potential benefits. However, the combination of these advanced technologies raises significant privacy concerns due to the potential exploitation of sensitive user and system data. This paper provides a systematic investigation of over 100 recent studies across key academic databases obtained by initial keyword-filter screening followed by a thorough full-text review. Particularly, metaverse evolution and enabling infrastructure technologies are briefly summarized. Subsequently, the distributed learning architectures and their features are analyzed as well as possibly associated vulnerability discussions. Then, envisioned metaverse applications and future research challenges are highlighted before concluding remarks. © 2025 The Authors

关键词： Distributed machine learning Mobile metaverse Privacy preservation Virtual reality

来源：评论

学校读者我要写书评

暂无评论

AVENet: Disentangling Features by Approximating Average Features for Voice Conversion

arXiv

引用

arXiv 2025年

作者： Wang, Wenyu Zhou, Yiquan Zhu, Jihua Ding, Hongwu Xu, Jiacheng Li, Shihao School of Software Engineering Xi’an Jiaotong University Xi’an China School of Computer Science and Technology Anhui University Hefei China School of Software Engineering East China Normal University Shanghai China Division of Music and Audio Union Wheatland Culture and Media Ltd Chengdu China

Voice conversion (VC) has made progress in feature disentanglement, but it is still difficult to balance timbre and content information. This paper evaluates the pre-trained model features commonly used in voice conversion, and proposes an innovative method for disentangling speech feature representations. Specifically, we first propose an ideal content feature, referred to as the average feature, which is calculated by averaging the features within frame-level aligned parallel speech (FAPS) data. For generating FAPS data, we utilize a technique that involves freezing the duration predictor in a Text-to-Speech system and manipulating speaker embedding. To fit the average feature on traditional VC datasets, we then design the AVENet to take features as input and generate closely matching average features. Experiments are conducted on the performance of AVENet-extracted features within a VC system. The experimental results demonstrate its superiority over multiple current speech feature disentangling methods. These findings affirm the effectiveness of our disentanglement approach. Copyright © 2025, The Authors. All rights reserved.

关键词： Feature extraction

来源：评论

学校读者我要写书评

暂无评论

Multi-label feature selection with feature reconstruction and label correlations

引用

Expert Systems with Applications 2025年 285卷

作者： Zhang, Ning Wang, Ao Lu, Pengwei Feng, Tao Xu, Ying Du, Guodong School of Information Science and Engineering Yanshan University Qinhuangdao066004 China Hebei Key Laboratory of Computer Virtual Technology and System Integration Yanshan University Qinhuangdao066004 China Department of Computer Science and Technology Tsinghua University Beijing100084 China Shenzhen Research Institute of Yanshan University Shenzhen518063 China

With the arrival of the big data era, multi-label data have frequently appeared in various fields. However, multi-label data often contain rich feature information, leading to a high-dimensional feature space. Thus, multi-label feature selection plays a crucial role in addressing this challenge as a preprocessing step. Multi-label data assign a set of labels to each instance simultaneously;therefore, capturing the correlations among labels is an effective means of enhancing the learning performance. However, current multi-label feature selection methods seldom consider label correlations. Moreover, redundant features can cause overfitting, increase the complexity, and reduce the interpretability. To address these issues, we propose a multi-label feature selection method that incorporates feature reconstruction and label correlations. Specifically, we construct a low-dimensional embedding from the original feature space using induction, which preserves the manifold structure of the feature space. Next, we explore the global and local label correlations using a manifold regularization term designed by the generated low-dimensional space. We then achieve feature reconstruction by minimising the reconstruction error and sparse regularization terms to learn the relationships between the features. Finally, we construct an optimal objective function with 2,1-norm regularization to facilitate multi-label feature selection and ensure convergence. Comparative experiments on 11 datasets show that the proposed method exhibits notable performance improvements over existing methods(e.g., achieves 5.6 % higher Average precision than RFS on Scene and 3.6 % lower Coverage than CRASG on science), validating the effectiveness and superiority of the proposed method. © 2025 Elsevier Ltd

关键词： Feature extraction

来源：评论

学校读者我要写书评

暂无评论

Motion Direction Awareness: A Biomimetic Dynamic Capture Mechanism for Video Prediction

引用

IEEE Transactions on Multimedia 2025年

作者： Gan, Lianqiang Lai, Junyu Zhu, Junhong Liu, Huashuo Gao, Lianli University of Electronic Science and Technology of China School of Aeronautics and Astronautics Chengdu611731 China University of Electronic Science and Technology of China School of Aeronautics and Astronautics China Aircraft Swarm Intelligent Sensing and Cooperative Control Key Laboratory of Sichuan Province Chengdu611731 China University of Electronic Science and Technology of China School of Computer Science and Engineering Chengdu611731 China

Video prediction is an important yet challenging task that generates future frames based on previous observations. Despite recent progress, existing methods still suffer from motion blur, due to weak motion perception capabilities leading to uncertainty in motion direction. To address this, we propose a Motion Direction Awareness (MDA) mechanism inspired by the direction-selective mechanism in animal visual systems. Specifically, MDA can decompose complex motions into horizontal and vertical components, allowing dimension reduction and independent processing, thereby effectively enhancing motion perception and reducing uncertainty in predicted motion directions. Based on MDA, we design a multi-scale feature fusion network named MDANet for video prediction, which incorporates different scales of spatially encoded features in conjunction with MDA mechanism to extract the temporal evolution information of global and local spatial features. Extensive experiments on representative datasets demonstrate that MDANet can alleviate motion blurring, improving prediction accuracy and temporal consistency over state-of-the-art models. Furthermore, we validate the generalizability and effectiveness of our MDA mechanism by integrating it into other advanced models. The code is available at supplementary. © 1999-2012 IEEE.

关键词： Deep learning

来源：评论

学校读者我要写书评

暂无评论

Leveraging Artificial Intelligence for Tomato Leaf Disease Diagnosis: A Performance Comparison of DL Models

Leveraging Artificial Intelligence for Tomato Leaf Disease D...

引用

Industrial Electronics: Developments & Applications (ICIDeA), IEEE International Conference on-

作者： Ankita Sharma Sheifali Gupta Nagendar Yamsani Saif Obbayed Chitkara University Institute of Engineering and Technology Chitkara University Punjab India School of Computer Science and Artificial Intelligence SR University Warangal Telangana India Department of Computers Techniques Engineering College of Technical Engineering The Islamic University Najaf Iraq

ISBN: (数字)9798331533205

ISBN: (纸本)9798331533212

Early prediction of the tomato leaf disease could benefit farmers in the agriculture field. It can reduce yield loss by detecting the disease promptly. This paper employed three distinct models of Deep Learning for the early prediction of the leaf disease, such models are Convolutional Neural Network, EfficientNetB3, and VGG16. This research article works on the classification of the various kinds of tomato leaf disorders, 10 different leaf disorders are used with a total of 10,000 images, and each disorder has 1000 images. Among the three models, the EfficientNetB3 model performed better with a testing accuracy of 98%, with a minimum testing loss of 0.041 loss. This model performed better because of its depth-wise, and widthwise scaling, and optimized compound scaling of the networks.

关键词： Deep learning Technological innovation Accuracy Predictive models Agriculture Convolutional neural networks Medical diagnosis Compounds Artificial intelligence Testing

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：