检索结果-内蒙古大学图书馆

31st International Conference on Computational Linguistics, COLING 2025

作者： Wen, Ximing Tan, Wenjuan Weber, Rosina O. College of Computing and Informatics Drexel University Philadelphia United States Department of Computer Science Tsinghua University Beijing China

ISBN: (纸本)9798891761964

Pretrained transformer-based Language Models (LMs) are well-known for their ability to achieve significant improvement on text classification tasks with their powerful word embeddings, but their black-box nature, which leads to a lack of interpretability, has been a major concern. In this work, we introduce GAProtoNet, a novel white-box Multi-head Graph Attention-based Prototypical Network designed to explain the decisions of text classification models built with LM encoders. In our approach, the input vector and prototypes are regarded as nodes within a graph, and we utilize multi-head graph attention to selectively construct edges between the input node and prototype nodes to learn an interpretable prototypical representation. During inference, the model makes decisions based on a linear combination of activated prototypes weighted by the attention score assigned for each prototype, allowing its choices to be transparently explained by the attention weights and the prototypes. Experiments on multiple public datasets show our approach achieves superior results without sacrificing the accuracy of the original black-box LMs. We also compare with four alternative prototypical network variations and our approach achieves the best accuracy and F1 among all. Our case study and visualization of prototype clusters also demonstrate the efficiency in explaining the decisions of black-box models built with LMs. Our codes are available at https://***/ximingwen/GAProtoNet. © 2025 Association for Computational Linguistics.

关键词： Computational linguistics

来源：评论

学校读者我要写书评

暂无评论

Hybrid-Optimized Model for Deepfake Detection

引用

International Journal of Advanced computer science and Applications 2025年第4期16卷 148-160页

作者： Mancy, H. Elpeltagy, Marwa Eldahshan, Kamal Ismail, Aya Department of Computer Science College of Engineering and Computer Sciences Prince Sattam Bin Abdulaziz University Alkharj11942 Saudi Arabia Al-Azhar University Cairo Egypt Systems and Computers Department Al-Azhar University Egypt Mathematics Department Faculty of Science Al-Azhar University Egypt Mathematics Department Tanta University Egypt

The advancement of deep learning models has led to the creation of novel techniques for image and video synthesis. One such technique is the deepfake, which swaps faces among persons and then produces hyper-realistic videos of individuals saying or doing things that they never said or done. These deepfake videos pose a serious risk to everyone and countries if they are exploited for extortion, scamming, political disinformation, or identity theft. This work presents a new methodology based on a hybrid-optimized model for detecting deepfake videos. A Mask Region-based Convolutional Neural Network (Mask R-CNN) is employed to detect human faces from video frames. Then, the optimal bounding box representing the face region per frame is selected, which could help to discover many artifacts. An improved Xception-Network is proposed to extract informative and deep hierarchical representations of the produced face frames. The Bayesian optimization (BO) algorithm is employed to search for the optimal hyperparameters' values in the extreme gradient boosting (XGBoost) classifier model to properly discriminate the deepfake videos from the genuine ones. The proposed method is trained and validated on two different datasets;CelebDF-FaceForencics++ (c23) and FakeAVCeleb, and tested also on various datasets;CelebDF, DeepfakeTIMIT, and FakeAVCeleb. The experimental study proves the superiority of the proposed method over the state-of-the-art methods. The proposed method yielded %97.88 accuracy and %97.65 AUROC on the trained CelebDF-FaceForencics++ (c23) and tested CelebDF datasets. Additionally, it achieved %98.44 accuracy and %98.44 AUROC on the trained CelebDF-FaceForencics++ (c23) and tested DeepfakeTIMIT datasets. Moreover, it yielded %99.50 accuracy and %99.21 AUROC on the FakeAVCeleb visual dataset. © (2025), (science and Information Organization). All Rights Reserved.

关键词： Convolutional neural networks

来源：评论

学校读者我要写书评

暂无评论

An Automated Framework of Superpixels-Saliency Map and Gated Recurrent Unit Deep Convolutional Neural Network for Land Cover and Crops Disease Classification

引用

IEEE Access 2025年 13卷 76370-76387页

作者： Haider, Irfan Khan, Muhammad Attique Nazir, Muhammad Masood, Saleha Kraiem, Naoufel Abdulaziz Alhammadi, Dina HITEC University Department of Computer Science Taxila47080 Pakistan Prince Mohammad Bin Fahd University College of Computer Engineering and Science Department of Artificial Intelligence Dhahran34754 Saudi Arabia King Fahd University of Petroleum and Minerals IRC for Finance and Digital Economy Dhahran31261 Saudi Arabia King Khalid University College of Computer Science Abha61413 Saudi Arabia Princess Nourah bint Abdulrahman University College of Computer and Information Sciences Department of Information Systems P.O.Box 84428 Riyadh11671 Saudi Arabia

In this work, we proposed an automated deep learning and saliency map architecture for the segmentation of crops, leaf disease segmentation, and land cover classification. The proposed framework is based on two embedded steps. In the first step, crop leaf disease segmentation was performed using superpixel clustering-based saliency maps and Bayesian optimization. The contrast enhancement technique is designed in the first segmentation phase and is passed to the saliency technique for the disease segmentation. In the second phase, EfficientNet-b0 architecture is fine-tuned with hyperparameters optimized via Bayesian Optimization. Also, the fine-tuned model is embedded with a single self-attention residual block fused with an efficient average pool layer. Training has been performed on segmented and contrast-enhanced images that were later fused using a serial-embedded approach. The extracted features in the testing phase are further optimized using the modified moth flame-controlled bisection (MFcB) technique. Finally, the extracted features are classified using machine learning classifiers for the final classification. Experiments are performed on the publically available cucumber leaf dataset and Remote sensing dataset with an improved accuracy of 97.6% and 92.90%, respectively. A comparison with state-of-the-art techniques shows that the proposed architecture has improved performance. © 2013 IEEE.

关键词： Superpixels

来源：评论

学校读者我要写书评

暂无评论

A Robust Security Detection Strategy for Next Generation IoT Networks

引用

computers, Materials & Continua 2025年第1期82卷 443-466页

作者： Hafida Assmi Azidine Guezzaz Said Benkirane Mourade Azrour Said Jabbour Nisreen Innab Abdulatif Alabdulatif Technology Higher School Essaouira Cadi Ayyad UniversityEssaouira44000Morocco IMIA Laboratory MSIA TeamFaculty of Sciences and TechniquesMoulay Ismail University of MeknesErrachidia50050Morocco CRIL-CNRS Artois UniversityLens62300France Department of Computer Science and Information Systems College of Applied SciencesAlMaarefa UniversityDiriyahRiyadh13713Saudi Arabia Department of Computer Science College of ComputerQassim UniversityBuraydah52571Saudi Arabia

Internet of Things(IoT)refers to the infrastructures that connect smart devices to the Internet,operating *** connectivitymakes it possible to harvest vast quantities of data,creating new opportunities for the emergence of unprecedented *** ensure IoT securit,various approaches have been implemented,such as authentication,encoding,as well as devices to guarantee data integrity and *** these approaches,Intrusion Detection systems(IDS)is an actual security solution,whose performance can be enhanced by integrating various algorithms,including Machine Learning(ML)and Deep Learning(DL),enabling proactive and accurate detection of *** study proposes to optimize the performance of network IDS using an ensemble learning method based on a voting classification *** combining the strengths of three powerful algorithms,Random Forest(RF),K-Nearest Neighbors(KNN),and Support Vector Machine(SVM)to detect both normal behavior and different categories of *** analysis focuses primarily on the NSL-KDD dataset,while also integrating the recent Edge-IIoT dataset,tailored to industrial IoT *** results show significant enhancements on the Edge-IIoT and NSL-KDD datasets,reaching accuracy levels between 72%to 99%,with precision between 87%and 99%,while recall values and F1-scores are also between 72%and 99%,for both normal and attack *** the promising results of this study,it suffers from certain limitations,notably the use of specific datasets and the lack of evaluations in a variety of *** work could include applying this model to various datasets and evaluating more advanced ensemble strategies,with the aim of further enhancing the effectiveness of IDS.

关键词： IoT security intrusion detection RF KNN SVM EL NSL-KDD Edge-IIoT

来源：评论

学校读者我要写书评

暂无评论

Driver Drowsiness Detection Using Swin Transformer and Diffusion Models for Robust Image Denoising

引用

IEEE Access 2025年 13卷 71880-71907页

作者： El-Nabi, Samy Abd Ibrahim, Ahmed F. El-Rabaie, El-Sayed M. Hassan, Osama F. Soliman, Naglaa F. Ramadan, Khalil F. El-Shafai, Walid Faculty of Computer Science and Engineering Department of Artificial Intelligence Engineering South Sinai 46511 Egypt Menoufia University Faculty of Electronic Engineering Department of Electronics and Electrical Communications Engineering Menouf32952 Egypt Suez Canal University Faculty of Computers and Informatics Information Systems Department Ismailia41522 Egypt Princess Nourah bint Abdulrahman University College of Computer and Information Sciences Department of Information Technology P.O. Box 84428 Riyadh11671 Saudi Arabia Computer Science Department Riyadh11586 Saudi Arabia

With the rapid development of intelligent transportation systems and growing emphasis on driver safety, real-time detection of driver drowsiness has become a critical area of research. This study presents a robust and scalable driver drowsiness detection framework that integrates a Swin Transformer-based deep learning model with a diffusion model for image denoising. While conventional convolutional neural networks (CNNs) are effective in standard vision tasks, they often suffer performance degradation in real-world driving scenarios due to noise, poor lighting, motion blur, and adversarial attacks. To address these challenges, the proposed model focuses on eye-state detection, specifically, prolonged eye closure, as a primary indicator of driver disengagement and fatigue. Our system introduces a novel preprocessing stage using a denoising diffusion model built on a U-Net encoder-decoder architecture, effectively mitigating the impact of Gaussian noise and adversarial perturbations. Additionally, we incorporate adversarial training with Fast Gradient Sign Method (FGSM) and Projected Gradient Descent (PGD) attacks, demonstrating significant improvements in classification accuracy and resilience. Evaluations are conducted on two benchmark datasets, Eye-Blink and Closed Eyes in the Wild (CEW), under both clean and noisy conditions. Comparative experiments show that the proposed system outperforms several state-of-the-art models, including ViT, ResNet50V2, InceptionV3, MobileNet, DenseNet169, and VGG19, in terms of accuracy (up to 99.82%), PSNR (up to 41.61 dB), and SSIM (up to 0.984), while maintaining competitive inference times suitable for practical deployment. Moreover, a detailed sensitivity analysis of data augmentation strategies reveals that techniques such as rotation and horizontal flip substantially enhance the model’s generalization across variable visual inputs. The system also demonstrates improved robustness under real-world black-box scenarios and adver

关键词： Sensitivity analysis

来源：评论

学校读者我要写书评

暂无评论

A novel hybrid deep learning framework for pedestrian detection based on thermal infrared and visible spectrum images

引用

Multimedia Tools and Applications 2025年 1-27页

作者： Defaoui, Mahassine Koutti, Lahcen El Ansari, Mohamed Lahmyed, Redouan Masmoudi, Lhoussaine LabSIV Department of Computer Science Faculty of Science Ibn Zohr University Agadir80000 Morocco Informatics and Applications Laboratory Faculty of Science My Ismail University Meknes Morocco Conception and Systems Laboratory Faculty of Sciences Mohamed V University Rabat Morocco

For many intelligent systems, designing accurate pedestrian detection approaches is a fundamental task. This paper describes a novel hybrid system to detect pedestrians using both visible and thermal infrared sensors. The proposed method is achieved in two primary stages. The first stage, hypotheses generation (HG), aims to extract the regions that represent suspected pedestrians from the image. It is performed using both thermal and visible images of the same scene. Here, an improved deep learning-based solution with a convolutional neural network named PedTherm-U-net is proposed. It segments the thermal images and maps the segmentation results into the corresponding visible image to extract the ROIs (candidate pedestrians). The second stage, hypotheses verification (HV), classifies the ROIs extracted from the first stage into pedestrians and non-pedestrians. In the HV step, a novel PedVis-ResNet framework to classify the extracted ROIs is introduced. It is a modified version of ResNet model dedicated to classify the visible pedestrian samples. The proposed method has been tested on the publicly accessible OSU Color-Thermal dataset, and the obtained results demonstrate the effectiveness of the proposed method when compared to the recent state of the art pedestrian detection approaches. © The Author(s), under exclusive licence to Springer science+Business Media, LLC, part of Springer Nature 2025.

关键词： Intelligent systems

来源：评论

学校读者我要写书评

暂无评论

Smart Sanitation Management: Integrating IoT, AI-Driven Analytics, and Automated Sensing for Efficient Public Toilet Maintenance 2

Smart Sanitation Management: Integrating IoT, AI-Driven Anal...

引用

2nd International Conference on Machine Learning and Autonomous systems, ICMLAS 2025

作者： Haripriya, K. Preethi, C. Hammed, S.Shahul Pavalarajan, S. Sharon, S. Akhil Suryakumar, S. PSNA College of Engineering and Technology Department of Computer Science and Business Systems Dindigul India

ISBN: (纸本)9798331505745

An IoT-based self-sustained public toilet maintenance system aims to improve hygiene, efficiency, and resource management. It integrates smart sensors, microcontrollers, and cloud connectivity to monitor parameters such as water levels, air quality, and toilet usage in real-time. AI-driven predictive maintenance optimizes cleaning schedules, minimizes downtime, and ensures efficient resource use. Automated water control prevents wastage, while air quality sensors detect harmful gases and activate ventilation. A smart notification system alerts municipal authorities about supply shortages, enabling timely replenishment of water, soap, and tissues. By reducing manual intervention, the system lowers operational costs and improves service quality. Cloud connectivity allows seamless data processing, offering insights for better sanitation management. The integration of IoT and AI ensures proactive maintenance responses. A comparative analysis highlights the novelty of this system in automating and optimizing maintenance. © 2025 IEEE.

关键词： Replenishment (water resources)

来源：评论

学校读者我要写书评

暂无评论

IoT in urban development: insight into smart city applications, case studies, challenges, and future prospects

引用

PeerJ computer science 2025年 11卷 1-42页

作者： Salih, Sayeed Abdelmaboud, Abdelzahir Husain, Omayma Motwakel, Abdelwahed Elshaﬁe, Hashim Sharif, Mahir Hamdan, Mosab Department of Management Information Systems College of Business Administration in Hawtat Bani Tamim Prince Sattam bin Abdulaziz University Al-Kharj Saudi Arabia Humanities Research Center Sultan Qaboos University Muscat Oman Department of Information Technology Ajman University Ajman United Arab Emirates Department of Computer Science University of Khartoum Khartoum Sudan College of Computer Science King Khalid University Abha Saudi Arabia Computer Science Department Faculty of Computer Science Information Technology Omdurman Islamic University Omdurman Sudan Computer and Self Development Prince Sattam bin Abdelaziz University Al-Kharj Saudi Arabia School of Computing National College of Ireland Dublin Ireland

With the integration of Internet of Things (IoT) technology, smart cities possess the capability to advance their public transportation modalities, address prevalent trafﬁc congestion challenges, reﬁne infrastructure, and optimize communication frameworks, thereby augmenting their progression towards heightened urbanization. Through the integration of sensors, cell phones, artiﬁcial intelligence (AI), data analytics, and cloud computing, smart cities worldwide are evolving to be more efﬁcient, productive, and responsive to their residents’ needs. While the promise of smart cities has been marked over the past decade, notable challenges, especially in the realm of security, threaten their optimal realization. This research provides a comprehensive survey on IoT in smart cities. It focuses on the IoT-based smart city components. Moreover, it provides explanation for integrating different technologies with IoT for smart cities such as AI, sensing technologies, and networking technologies. Additionally, this study provides several case studies for smart cities. In addition, this study investigates the challenges of adopting IoT in smart cities and provides prevention methods for each challenge. Moreover, this study provides future directions for the upcoming researchers. It serves as a foundational guide for stakeholders and emphasizes the pressing need for a balanced integration of innovation and safety in the smart city landscape. © 2024 Salih et al. Distributed under Creative Commons CC-BY 4.0

关键词： Smart city

来源：评论

学校读者我要写书评

暂无评论

Extracting Facial Features to Detect Deepfake Videos Using Machine Learning

引用

International Journal of Advanced computer science and Applications 2025年第4期16卷 834-842页

作者： Aslam, Ayesha Mir, Jamaluddin Zaman, Gohar Rahman, Atta Salam, Asiya Abdus Ali, Farhan Alhiyafi, Jamal Bakry, Aghiad Gul, Mustafa Jamal Gollapalli, Mohammed Mahmud, Maqsood Department of Computer Science Abbottabad University of Science and Technology Havelian Pakistan Department of Computer Science College of Computer Science and Information Technology Imam Abdulrahman Bin Faisal University P.O. Box 1982 Dammam31441 Saudi Arabia Department of Computer Information Systems College of Computer Science and Information Technology Imam Abdulrahman Bin Faisal University P.O. Box 1982 Dammam31441 Saudi Arabia College of Electronics and Information Engineering Shenzhen University Shenzhen China Department of Computer Science Kettering University FlintMI United States Department of Business Administration University of York York YO10 5DD Heslington United Kingdom Department of Information Technology & Engineering Sydney Met SydneyNSW2000 Australia School of Computing Ulster University Northern Ireland Belfast United Kingdom

Generative adversarial networks (GANs) have gained popularity for their ability to synthesize images from random inputs in deep learning models. One of the notable applications of this technology is the creation of realistic videos known as deepfakes, which have been misused on social media platforms. The difficulty lies in distinguishing these fake videos from real ones with the naked eye, leading to significant concerns. This study proposes a supervised machine learning approach to effectively differentiate between real and counterfeit videos by detecting visual artifacts. To achieve this, two facial features are extracted: eye blinking and nose position, utilizing landmark detection techniques. Both features were trained on supervised machine learning classifiers and evaluated using the publicly available UADFV and Celeb-DF deepfake datasets. The experiments successfully demonstrate that the proposed method achieves a promising and superior performance, with an area under the curve (AUC) of 97% for deepfake detection in contrast to state-of-the-art methods investigating the same datasets. © (2025), (science and Information Organization). All rights reserved.

关键词： Support vector machines

来源：评论

学校读者我要写书评

暂无评论

MediGuide AI using Machine Learning 2

MediGuide AI using Machine Learning

引用

2nd International Conference on Machine Learning and Autonomous systems, ICMLAS 2025

作者： Pooja, R. Elugubanti, Nikitha Mary Shiba, C. Kalai priya, V. Sasirekha, K. Sabapathy, S Rathana R.M.K Engineering College Computer Science and Business Systems Tiruvallur District Kavaraipettai India R.M.K Engineering College Department Of Computer Science and Business Systems Kavaraipettai India R.M.K College of Engineering and Technology Department Of Computer Science and Engineering Tiruvallur District Puduvoyal India R.M.D Engineering College Department Of Computer Science and Business Systems Kavaraipettai India Dhaanish Ahmed College of Engineering Department Of Computer Science and Engineering Vanchuvanchery Chennai India

ISBN: (纸本)9798331505745

Providing diverse patient groups with their specific medication requirements, using a wide range of drugs, poses meaningful challenges for prescribing. Standardized approaches often lead to large waste, many side effects and suboptimal outcomes. Therefore, we need improved, more personalized healthcare. MediGuide AI's drug recommendation system is powered by machine learning and it offers a thorough solution. Optimized data pre-processing and analysis, leading to the desired outcome, is achieved using advanced machine learning algorithms like Naive Bayes, TF-IDF and the Natural Language Toolkit. Highly personalized prescriptions are generated using thorough patient data, including demographics and meaningful clinical factors to achieve substantially better accuracy. Important enhancements include dual recommendation outputs, which provide many drug recommendations alongside a large number of alternative Ayurvedic treatments presented side by side for effortless comparison. An advanced personalized dosage calculation module precisely determines drug dosages using advanced regression models based on age, weight and metabolism. This system helps patients track and record their medications and it also suggests several nearby Ayurvedic shops based on their location. MediGuide AI employs exceptionally powerful algorithms and its advanced tools deliver remarkably precise and exceptionally user-friendly healthcare solutions. Current systems make use of rule-based methods that cannot learn from new data, whereas MediGuide AI leverages a hybrid Naïve Bayes and PassiveAggressiveClassifier model for higher accuracy and real-time adaptability. It also includes Ayurvedic options to ensure personalized healthcare experience. © 2025 IEEE.

关键词： Drug dosage

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：