检索结果-内蒙古大学图书馆

AI and Ethics 2025年 1-21页

作者： Sebastian, Rinu Ann Ehinger, Kris Miller, Tim School of Computing and Information Systems The University of Melbourne Carlton Australia School of Electrical Engineering and Computer Science The University of Queensland St Lucia Australia

In this paper, we critically examine the relevant ethical concerns of using computer vision-based surveillance in workplaces and propose an intent—and priority-based ethical framework for such systems. With the growing capabilities of computer vision technologies, its application in monitoring workplaces brings forth significant concerns. Organisations increasingly leverage computer vision for workplace surveillance to improve productivity, safety, and security. Unlike electronic surveillance techniques that monitor workers’ wire or electronic communication, computer vision-based workplace surveillance (CVWS) captures highly detailed visual and personal information about workers, including body language, emotional state, and actions. This makes CVWS potentially more intrusive than traditional electronic surveillance, raising a more comprehensive range of ethical considerations. However, this topic has received minimal attention in the current literature. Our proposed framework combines the intention for deploying surveillance with the moral notions of privacy, data security, fairness, transparency, explainability, autonomy, beneficence, nonmaleficence, dignity, and reliability to morally scrutinise CVWS systems. The paper proposes a second framework that aims to establish accountability among key stakeholders of the CVWS system. Further, we discuss two critical questions to consider when evaluating the necessity for CVWS systems in a work environment. In practice, this work will serve as a groundwork for stakeholders such as technical developers, employers, and regulatory and advocacy teams to make ethical design decisions during the developmental, operational, and maintenance stages of CVWS systems and devise proactive strategies to minimise potential harm.

关键词：

来源：评论

学校读者我要写书评

暂无评论

Advancing Air Quality Monitoring: TinyML-Based Real-Time Ozone Prediction with Cost-Effective Edge Devices

arXiv

引用

arXiv 2025年

作者： Ken, Huam Ming Behjati, Mehran Rafsanjani, Ahmad Sahban Aslam, Saad Meng, Yap Kian Abdul Majeed, Anwar P.P. Zheng, Yufan Department of Computing and Information Systems School of Engineering and Technology Sunway University 47500 Malaysia Department of Computing and Information Systems School of Engineering and Technology Sunway University 47500 Malaysia Taicang Suzhou215400 China

The escalation of urban air pollution necessitates innovative solutions for real-time air quality monitoring and prediction. This paper introduces a novel TinyML-based system designed to predict ozone concentration in real-time. The system employs an Arduino Nano 33 BLE Sense microcontroller equipped with an MQ7 sensor for carbon monoxide (CO) detection and built-in sensors for temperature and pressure measurements. The data, sourced from a Kaggle dataset on air quality parameters from India, underwent thorough cleaning and preprocessing. Model training and evaluation were performed using Edge Impulse, considering various combinations of input parameters (CO, temperature, and pressure). The optimal model, incorporating all three variables, achieved a mean squared error (MSE) of 0.03 and an R-squared value of 0.95, indicating high predictive accuracy. The regression model was deployed on the microcontroller via the Arduino IDE, showcasing robust real-time performance. Sensitivity analysis identified CO levels as the most critical predictor of ozone concentration, followed by pressure and temperature. The system's low-cost and low-power design makes it suitable for widespread implementation, particularly in resource-constrained settings. This TinyML approach provides precise real-time predictions of ozone levels, enabling prompt responses to pollution events and enhancing public health protection. Copyright © 2025, The Authors. All rights reserved.

关键词： Air quality

来源：评论

学校读者我要写书评

暂无评论

AER-LLM: Ambiguity-aware Emotion Recognition Leveraging Large Language Models

AER-LLM: Ambiguity-aware Emotion Recognition Leveraging Larg...

引用

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

作者： Hong, Xin Gong, Yuan Sethu, Vidhyasaharan Dang, Ting School of Computing and Information Systems University of Melbourne Melbourne Australia CSAIL SLS Massachusetts Institute of Technology Boston United States School of Electrical Engineering and Telecommunications University of New South Wales Sydney Australia

ISBN: (纸本)9798350368741

Recent advancements in Large Language Models (LLMs) have demonstrated great success in many Natural Language Processing (NLP) tasks. In addition to their cognitive intelligence, exploring their capabilities in emotional intelligence is also crucial, as it enables more natural and empathetic conversational AI. Recent studies have shown LLMs' capability in recognizing emotions, but they often focus on single emotion labels and overlook the complex and ambiguous nature of human emotions. This study is the first to address this gap by exploring the potential of LLMs in recognizing ambiguous emotions, leveraging their strong generalization capabilities and in-context learning. We design zero-shot and few-shot prompting and incorporate past dialogue as context information for ambiguous emotion recognition. Experiments conducted using three datasets indicate significant potential for LLMs in recognizing ambiguous emotions, and highlight the substantial benefits of including context information. Furthermore, our findings indicate that LLMs demonstrate a high degree of effectiveness in recognizing less ambiguous emotions and exhibit potential for identifying more ambiguous emotions, paralleling human perceptual capabilities. © 2025 IEEE.

关键词： ambiguous emotion emotion recognition large language models multimodal prompt design

来源：评论

学校读者我要写书评

暂无评论

MuLAD: Multimodal Aggression Detection from Social Media Memes Exploiting Visual and Textual Features 27th

MuLAD: Multimodal Aggression Detection from Social Media Me...

引用

27th International Conference on Pattern Recognition, ICPR 2024

作者： Hasan, Md. Maruf Ahsan, Shawly Hoque, Mohammed Moshiul Dewan, M. Ali Akber Chittagong University of Engineering and Technology Chittagong4349 Bangladesh School of Computing and Information Systems Faculty of Science and Technology Athabasca University AthabascaABT9S 3A3 Canada

ISBN: (纸本)9783031781186

Aggression detection from memes is challenging due to their region-specific interpretation and multimodal nature. Detecting or classifying aggressive memes is complicated in low-resource languages (including Bengali) because benchmark datasets and primary language processing software are needed. This paper proposes an innovative meme classification technique that harnesses deep learning (DL) approaches to leverage memes’ visual and textual features in Bengali. Various DL frameworks, such as VGG16, VGG19, ResNet50, CNN, BiLSTM, and BiLSTM+CNN, extract visual and textual features from memes. A novel corpus named the Bengali Meme Dataset (AMemD) is also introduced, comprising a substantial amount of multimodal data, including text and image components. Experimental results on AMemD demonstrate the effectiveness of the proposed approach. The CNN combined with VGG16 obtained the highest f1-score of 0.738 among all multimodal techniques tested. This pioneering research offers valuable insights into the complex task of aggression detection from memes in Bengali and provides a foundation for future studies in this area. The dataset is available at https://***/Maruf089/Multimodal-Aggression-Detection. © The Author(s), under exclusive license to Springer Nature Switzerland AG 2025.

关键词： Deep learning

来源：评论

学校读者我要写书评

暂无评论

AttentionPoolMobileNeXt: An automated construction damage detection model based on a new convolutional neural network and deep feature engineering models

引用

Multimedia Tools and Applications 2025年第4期84卷 1821-1843页

作者： Aydin, Mehmet Barua, Prabal Datta Chadalavada, Sreenivasulu Dogan, Sengul Tuncer, Turker Chakraborty, Subrata Acharya, Rajendra U. Department of Digital Forensics Engineering Technology Faculty Firat University Elazig Turkey University of Southern Queensland Toowoomba Australia School of Engineering University of Southern Queensland Springfield LakesQLD4300 Australia School of Science and Technology Faculty of Science Agriculture Business and Law University of New England ArmidaleNSW2351 Australia Center for Advanced Modelling and Geospatial Information Systems Faculty of Engineering and IT University of Technology Sydney SydneyNSW2007 Australia School of Mathematics Physics and Computing University of Southern Queensland Toowoomba Australia

In 2023, Turkiye faced a series of devastating earthquakes and these earthquakes affected millions of people due to damaged constructions. These earthquakes demonstrated the urgent need for advanced automated damage detection models to help people. This study introduces a novel solution to address this challenge through the AttentionPoolMobileNeXt model, derived from a modified MobileNetV2 architecture. To rigorously evaluate the effectiveness of the model, we meticulously curated a dataset comprising instances of construction damage classified into five distinct classes. Upon applying this dataset to the AttentionPoolMobileNeXt model, we obtained an accuracy of 97%. In this work, we have created a dataset consisting of five distinct damage classes, and achieved 97% test accuracy using our proposed AttentionPoolMobileNeXt model. Additionally, the study extends its impact by introducing the AttentionPoolMobileNeXt-based Deep Feature engineering (DFE) model, further enhancing the classification performance and interpretability of the system. The presented DFE significantly increased the test classification accuracy from 90.17% to 97%, yielding improvement over the baseline model. AttentionPoolMobileNeXt and its DFE counterpart collectively contribute to advancing the state-of-the-art in automated damage detection, offering valuable insights for disaster response and recovery efforts. © The Author(s) 2024.

关键词： Image classification

来源：评论

学校读者我要写书评

暂无评论

YOLOv8n-CGW: A novel approach to multi-oriented vehicle detection in intelligent transportation systems

引用

Multimedia Tools and Applications 2025年第7期84卷 3809-3840页

作者： Berwo, Michael Abebe Fang, Yong Sarwar, Nadeem Mahmood, Jabar Aljohani, Mansourah Elhosseini, Mostafa School of Information and Engineering Chang’an University Shaanxi Xi’an710064 China Department of Computer Science Bahria University Lahore Campus Punjab Lahore54600 Pakistan Department of Information Technology Faculty of Computing and Information Technology University of Sialkot Punjab Sialkot51040 Pakistan College of Computer Science and Engineering Taibah University Yanbu46421 Saudi Arabia Computers and Control Systems Engineering Department Faculty of Engineering Mansoura University Mansoura35516 Egypt

In the context of Intelligent Transportation systems (ITS), the role of vehicle detection and classification is indispensable for streamlining transportation management, refining traffic control, and conducting in-depth accident analyses. However, the intricate task of accurately detecting multi-oriented vehicles in diverse scenarios remains a challenge, even with the advancements in ITS. Factors such as vehicle morphology, design, road conditions, structural details, and climatic variables provide a significant part in the detection. To address this, our study introduces an advanced YOLOv8n model, specifically the YOLOv8n-CGW, which is built upon pre-trained deep learning frameworks. This model is meticulously tailored to enhance the detection capabilities for multi-oriented vehicles. Comprehensive evaluations on esteemed datasets like MtV, MERGED, Custom dataset, DAWN, WEDGE, and UA-DETRAC underscore the superiority of our approach over existing models. Notably, our YOLOv8n-CGW model achieves a remarkable mAP of 83.3% and boasts a swift inference time of 0.0019s, setting a new benchmark in the domain. © The Author(s), under exclusive licence to Springer Science+Business Media, LLC, part of Springer Nature 2024.

关键词： Deep neural networks

来源：评论

学校读者我要写书评

暂无评论

Congruent Feature Selection Method to Improve the Efficacy of Machine Learning-Based Classification in Medical Image Processing

引用

Computer Modeling in engineering & Sciences 2025年第1期142卷 357-384页

作者： Mohd Anjum Naoufel Kraiem Hong Min Ashit Kumar Dutta Yousef Ibrahim Daradkeh Department of Computer Engineering Aligarh Muslim UniversityAligarh202002India College of Computer Science King Khalid UniversityAbha61413Saudi Arabia School of Computing Gachon UniversitySeongnam13120Republic of Korea Department of Computer Science and Information Systems College of Applied SciencesAlMaarefa UniversityRiyadh13713Saudi Arabia Department of Computer Engineering and Information College of Engineering in Wadi AlddawasirPrince Sattam bin Abdulaziz UniversityAl-Kharj16273Saudi Arabia

Machine learning(ML)is increasingly applied for medical image processing with appropriate learning *** applications include analyzing images of various organs,such as the brain,lung,eye,etc.,to identify specific flaws/diseases for *** primary concern of ML applications is the precise selection of flexible image features for pattern detection and region *** of the extracted image features are irrelevant and lead to an increase in computation ***,this article uses an analytical learning paradigm to design a Congruent Feature Selection Method to select the most relevant image *** process trains the learning paradigm using similarity and correlation-based features over different textural intensities and pixel *** similarity between the pixels over the various distribution patterns with high indexes is recommended for disease ***,the correlation based on intensity and distribution is analyzed to improve the feature selection ***,the more congruent pixels are sorted in the descending order of the selection,which identifies better regions than the ***,the learning paradigm is trained using intensity and region-based similarity to maximize the chances of ***,the probability of feature selection,regardless of the textures and medical image patterns,is *** process enhances the performance of ML applications for different medical image *** proposed method improves the accuracy,precision,and training rate by 13.19%,10.69%,and 11.06%,respectively,compared to other models for the selected *** mean error and selection time is also reduced by 12.56%and 13.56%,respectively,compared to the same models and dataset.

关键词： Computer vision feature selection machine learning region detection texture analysis image classification medical images

来源：评论

学校读者我要写书评

暂无评论

Seeing through Dense Fog with an Intelligent Adaptive Dehazing System

引用

IEEE Sensors Journal 2025年第10期25卷 17696-17705页

作者： Chen, Pengyun Cao, Ning Xu, Ziqin Cui, Shuang Jin, Shaohui Liu, Hao Xu, Mingliang Zhengzhou University School of Computer Science and Artificial Intelligence Zhengzhou450001 China National University of Defense Technology College of Advanced Interdisciplinary Studies Changsha410073 China Ministry of Education Engineering Research Center of Intelligent Swarm Systems Zhengzhou450001 China National Super-computing Center Zhengzhou450001 China

Most existing image dehazing methods are mainly suitable for synthetic datasets and often perform poorly in real-world, complex hazy scenarios. To address this issue, this paper proposes an intelligent adaptive dehazing system that integrates range-gated imaging with deep learning, combining image acquisition and restoration for enhanced dehazing performance. The range-gated imaging system reduces scattered light interference. However, our approach primarily focuses on enhancing image quality through advanced dehazing methods. Specifically, we introduce the MSCENAFormer dehazing network, which achieves high-quality reconstruction of target scenes in dense fog by effectively removing fog and improving visibility. The core modules of MSCENAFormer include the multi-scale enhanced neighborhood attention module (MSENA) and the comprehensive attention refinement module (CARM). MSENA is designed to capture rich local information in harsh environment, improving the dehazing effect and enhancing image details. CARM integrates the local and global information to optimize the visual effect further. In addition, the adaptive feature mixing module (AFM) is used to fuse multi-scale features for better performance. To validate the performance of our method, we utilize our lab-collected non-homogeneous haze real dataset, O-ITDF, along with the public datasets NH-HAZE, NTIRE2021 and NTIRE2023. Experimental results demonstrate that our proposed MSCENAFormer outperforms many methods. We share our code at https://***/NingCao-zzu/MSCENAFormer. © 2001-2012 IEEE.

关键词： Light interference

来源：评论

学校读者我要写书评

暂无评论

Directed Brain Network Transformer for Psychiatric Diagnosis 27th

Directed Brain Network Transformer for Psychiatric Diagnosis

引用

27th International Conference on Pattern Recognition, ICPR 2024

作者： Zhu, Xu Qi, Zhiwei Yue, Kun Su, Yunshan Duan, Liang School of Information Science and Engineering Yunnan University Kunming China Yunnan Key Laboratory of Intelligent Systems and Computing Yunnan University Kunming China School of Education Yunnan University Kunming China Department of Radiology Affiliated Hospital of Yunnan University Kunming China

ISBN: (纸本)9783031781940

Human brain is a complex organ that consists of billions of neurons and trillions of connections among the neurons. To describe the correlations among the time series of the brain regions, we model the brain as a functional brain network to diagnose psychosis. However, traditional methods of functional brain network construction are usually noisy and do not consider the causal relationships among brain regions. To obtain the causal relationships and improve diagnosis interpretability, we propose a directed brain network Transformer (DBNT) for psychiatric diagnosis. First, the causal relationships in the blood-oxygen-level-dependent time series of brain regions are extracted to generate a directed brain network. Then, the feature encoding method is proposed to obtain local and global features of the brain networks by using the DBNT. Experimental results demonstrate that the accuracy of DBNT increases by 8.1% and 6.4% compared to state-of-the-art methods on two large-scale brain network datasets. DBNT also highlights the brain regions associated with psychosis and provides interpretation for diagnosis. © The Author(s), under exclusive license to Springer Nature Switzerland AG 2025.

关键词： Network coding

来源：评论

学校读者我要写书评

暂无评论

DFedADMM: Dual Constraint Controlled Model Inconsistency for Decentralize Federated Learning

引用

IEEE Transactions on Pattern Analysis and Machine Intelligence 2025年第6期47卷 4803-4815页

作者： Li, Qinglun Shen, Li Li, Guanghao Yin, Quanjun Tao, Dacheng National University of Defense Technology College of Systems Engineering China Shenzhen Campus of Sun Yat-Sen University School of Cyber Science and Technology Shenzhen518107 China JD Explore Academy Beijing100000 China Nanyang Technological University College of Computing & Data Science 50 Nanyang Avenue 639798 Singapore

To address the communication burden issues associated with Federated Learning (FL), Decentralized Federated Learning (DFL) discards the central server and establishes a decentralized communication network, where each client communicates only with neighboring clients. However, existing DFL methods still suffer from two major challenges: local inconsistency and local heterogeneous overfitting, which existing DFL methods have not fundamentally addressed. To tackle these issues, we propose novel DFL algorithms, DFedADMM and its enhanced version DFedADMM-SAM, to improve the performance for DFL. The DFedADMM algorithm employs primal-dual optimization (ADMM) by utilizing dual variables to control the model inconsistency raised from the decentralized heterogeneous data distributions. The DFedADMM-SAM algorithm further improves on DFedADMM by employing a Sharpness-Aware Minimization (SAM) optimizer, which uses gradient perturbations to generate locally flat models and searches for models with uniformly low loss values to mitigate local heterogeneous overfitting. Theoretically, we derive convergence rates of (Formula Presented) and (Formula Presented) in the non-convex setting for DFedADMM and DFedADMM-SAM, respectively, where 1−ψ represents the spectral gap of the gossip matrix. Empirically, extensive experiments on MNIST, CIFAR10, and CIFAR100 datasets demonstrate that our algorithms exhibit superior performance in terms of generalization, convergence speed, and communication overhead compared to existing state-of-the-art (SOTA) optimizers in DFL. © 2025 IEEE. All rights reserved.

关键词： Federated learning

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：