检索结果-内蒙古大学图书馆

Deepfake Detection Using Spatio-Temporal-Structural Anomaly Learning and Fuzzy System-Based Decision Fusion

IEEE Access 2025年 13卷 82747-82758页

作者： Subburaj, Brindha Ragavendra, R. Vellore Institute of Technology School of Computer Science and Engineering Chennai600127 India

Deepfake techniques have been evolving rapidly in recent days and pose a severe security threat. Detecting such generated fake videos is a challenging task. Existing deep learning based deep fake detection model struggle in identifying fake videos when experimented on challenging dataset. In this paper, we propose a Fuzzy logic based Deepfake Detector system (Fuzzy-DFD) using a deep learning model ResNet-50 as encoder. The system workflow includes preprocessing, frame extraction, deep fake detection and decision fusion. In preprocessing, frames are extracted from video, facial region is cropped, and frame resizing is performed. Next, optical flow maps are generated using SEA-RAFT model and edge defined frames generated using Frei-Chen mask technique. These three frame variants present a rich representation of video for the deep learning model. 3 ResNet-50 models are employed as encoders to generate feature maps trained using above frame types, ensuring anomaly identification in spatial, temporal and structural domains. A novel three input and one output Fuzzy Inference System is designed for decision fusion combining the detection results of all three classifier models to make decision on video being a fake. The performance of the proposed Fuzzy-DFD system is evaluated by training the model on well-known deep fake classification datasets namely FakeForensics++ and Celeb-DF(v2). Further results are compiled and evaluated using performance indicators like accuracy, Precision, F1-Score, Recall and AUC. Fuzzy-DFD system attained accuracy of 99% and 93% for FaceForensics++ and Celeb-DF(v2) datasets respectively. Comparison study with other SOTA deep fake detection models is conducted, and Fuzzy-DFD significantly outperformed the other models taken for comparison. The anomaly learning through RGB, optical flow and structural characteristics with fuzzy system-based decision fusion technique contributes to a robust deep fake detection system. © 2013 IEEE.

关键词： Deep learning

来源：评论

学校读者我要写书评

暂无评论

Enhancing demand forecasting through combination of anomaly detection and continuous improvement

引用

Soft Computing 2025年第2期29卷 1243-1258页

作者： Jahani, Meysam Zojaji, Zahra Raji, Fatemeh Faculty of Computer Engineering University of Isfahan Isfahan Iran School of Computer Science and Informatics De Montfort University Leicester United Kingdom

Demand forecasting has emerged as a crucial element in supply chain management. It is essential to identify anomalous data and continuously improve the forecasting model with new data. However, existing literature fails to comprehensively cover both aspects of anomaly detection and continuous improvement in demand forecasting. This study proposes an enhanced model to improve accuracy in the demand forecasting. The proposed model introduces a novel data handling method that incorporates an anomaly detection autoencoder, improved with anomaly correction mechanisms. The data handling approach simultaneously detects data anomalies, distinguishes between expected and unexpected anomalies, and corrects anomalous data, ensuring cleaner input for demand forecasting. Then, the proposed model employs a long short-term memory architecture for demand forecasting, enhanced with a continuous improvement method. Thus, the model not only forecasts demand but also retrains the model when the anomaly data surpasses the predetermined threshold, thereby improving the accuracy of forecasting. The results show that the proposed model outperforms other models in detecting data anomalies, achieving an average precision-recall of 0.922, a receiver operating characteristic value of 0.739, and a significance level of less than 0.05. Finally, the model exhibits superior performance in demand forecasting, with average mean squared error, root mean squared error, and mean absolute error values of 33.167, 4.347, and 1.509, respectively, all with a significance level of less than 0.05. © The Author(s), under exclusive licence to Springer-Verlag GmbH Germany, part of Springer Nature 2025.

关键词： Demand response

来源：评论

学校读者我要写书评

暂无评论

A novel group tour trip recommender model for personalized travel systems

引用

PeerJ computer science 2025年 11卷 1-377页

作者： Alatiyyah, Mohammed Department of Computer Science College of Computer Engineering and Sciences Prince Sattam Bin Abdulaziz University Al-kharj Riyadh Saudi Arabia

Planning personalized travel itineraries for groups with diverse preferences is indeed challenging. This article proposes a novel group tour trip recommender model (GTTRM), which uses ant colony optimization (ACO) to optimize group satisfaction while minimizing conflicts between group members. Unlike existing models, the proposed GTTRM allows dynamic subgroup formation during the trip to handle conflicting preferences and provide tailored recommendations. Experimental results show that GTTRM significantly improves satisfaction levels for individual group members, outperforming state-of-the-art models in terms of both subgroup management and optimization efficiency. © 2025 Alatiyyah

关键词： Ant colony optimization

来源：评论

学校读者我要写书评

暂无评论

Editor-in-Chief's foreword: understanding internet of energy (IoE)

引用

International Journal of computers and Applications 2025年第4期47卷 339-340页

作者： Pathan, Al-Sakib Khan Department of Computer Science and Engineering United International University Dhaka Bangladesh

来源：评论

学校读者我要写书评

暂无评论

Impact of machine learning-based imputation techniques on medical datasets- a comparative analysis

引用

Multimedia Tools and Applications 2025年第9期84卷 5905-5925页

作者： Tiwaskar, Shweta Rashid, Mamoon Gokhale, Prasad Department of Computer Engineering Faculty of Science and Technology Vishwakarma University Pune India

In the realm of medical datasets, particularly when considering diabetes, the occurrence of data incompleteness is a prevalent issue. Unveiling valuable patterns through medical data analysis is crucial for early and precise medical predictions. However, the quality of data and the proper handling of missing data hold significant significance. To address this challenge, imputation stands as a robust approach. The main goal of this paper aims to provide a comprehensive investigation into the effects brought about by Machine Learning (ML) based imputation techniques, specifically K Nearest Neighbor Imputation (KNNI), Multiple Imputation by Chained Equations (MICE), and MissForest. Results of all three techniques are compared with the complete dataset for five missing rates (10% to 50%), and evaluated using four categories of evaluation criteria i.e. (1) model performance, (2) imputation error rate (Mean Absolute Error (MAE), Root Mean Squared Error (RMSE), Coefficient of Determination (R^2) values), (3) Pearson correlation analysis and, (4) model selection basis (Bayesian information criterion (BIC), Akaike information criterion (AIC), values). Model performance includes accuracy, precision, recall, F1 score, and Matthews Correlation Coefficient (Mcoff) score of four ML classifiers viz. (a) Random Forest (RF), (b) Support vector machine (SVM), (c) AdaBoost, (d) XGBoost (XGB). For all missing rate cases, the MissForest technique is better than the KNNI and MICE in accuracy and Mcoff in 80% of cases, precision in 40% of cases, recall in 60% of cases, F1 score, MAE, RMSE, R^2 in 100% of cases, AIC in 80% of cases, and BIC values in 100% of cases. Also, the correlation analysis confirms that the MissForest imputation preserves association between the variables, like the complete dataset. Overall, our findings suggest that MissForest is a better machine learning-based imputation technique for handling missing data in diabetes research. © The Author(s), under exclusive lice

关键词： Adaptive boosting

来源：评论

学校读者我要写书评

暂无评论

Wide field of view large aperture meta-doublet eyepiece

引用

Light(science & Applications) 2025年第1期14卷 167-176页

作者： Anna Wirth-Singh Johannes E.Fröch Fan Yang Louis Martin Hanyu Zheng Hualiang Zhang Quentin T.Tanguy Zhihao Zhou Luocheng Huang Demis DJohn Biljana Stamenic Juejun Hu Tian Gu Arka Majumdar Department of Physics University of WashingtonSeattleWAUSA Department of Electrical and Computer Engineering University of WashingtonSeattleWAUSA Department of Materials Science and Engineering Massachusetts Institute of TechnologyCambridgeMAUSA Department of Electrical and Computer Engineering University of MassachusettsLowellMAUSA Department of Electrical and Computer Engineering University of CaliforniaSanta BarbaraCAUSA

Wide field of view and light weight optics are critical for advanced eyewear,with applications in augmented/virtual reality and night *** refractive lenses are often stacked to correct aberrations at a wide field of view,leading to limited performance and increased size and *** particular,simultaneously achieving a wide field of view and large aperture for light collection is desirable but challenging to realize in a compact ***,we demonstrate a wide field of view(greater than 60°)meta-optic doublet eyepiece with an entrance aperture of 2.1 *** the design wavelength of 633 nm,the meta-optic doublet achieves comparable performance to a refractive lens-based eyepiece *** meta-doublet eyepiece illustrates the potential for meta-optics to play an important role in the development of high-quality monochrome near-eye displays and night vision systems.

关键词： optics refractive illustrate

来源：评论

学校读者我要写书评

暂无评论

A high-capacity reversible data hiding with contrast enhancement and brightness preservation for medical images

引用

Multimedia Tools and Applications 2025年第8期84卷 5239-5264页

作者： Gandhi, Sonal Kumar, Rajeev Department of Computer Science and Engineering Delhi Technological University Delhi110042 India

The healthcare industry has witnessed an increase in the use of cloud storage, resulting in a significant demand for safeguarding medical records from potential attackers. In response to this challenge, reversible data hiding (RDH) has emerged as a lifesaver. The RDH ensures the concealment of private and confidential data with minimal loss to the cover image while offering opportunities for image enhancement. This paper introduces a new RDH method with contrast enhancement for medical images. The method aims to provide high embedding capacity (EC) while preserving brightness. To achieve this, the proposed method initially segments cover images into regions of interest (ROI) and non-regions of interest (NROI) and employs different embedding strategies based on the characteristics of each region, thereby enhancing embedding performance. Furthermore, a novel pre-processing technique is introduced for processing ROI pixels. This technique re-organizes and creates empty bins to provide enlarged EC with less distortion by capitalizing on the unique properties of medical images. Additionally, the proposed method embeds secret data in the NROI to further increase the EC. Experimental results show that the proposed method effectively preserves brightness (with 0.99 Relative Mean Brightness Error) while providing higher EC, exceeding the best-known methods by 20%. Furthermore, the method enhances contrast with minimal distortion in the output images. © The Author(s), under exclusive licence to Springer science+Business Media, LLC, part of Springer Nature 2024.

关键词： Image segmentation

来源：评论

学校读者我要写书评

暂无评论

On the Difficulty to Beat the First Linear Programming Bound for Binary Codes

引用

IEEE Transactions on Information Theory 2025年第4期71卷 2383-2388页

作者： Samorodnitsky, Alex The Hebrew University of Jerusalem School of Engineering and Computer Science Jerusalem91904 Israel

The first linear programming bound is the best known asymptotic upper bound for binary codes, for a certain subrange of distances. Starting from the work of Friedman and Tillich (2005), there are, by now, some arguably easier and more direct arguments for this bound. We show that this more recent line of argument runs into certain difficulties if one tries to go beyond this bound [say, towards the second linear programming bound of McEliece et al. (1977)]. Stated more constructively, we show that certain necessary requirements have to be met in order to produce a feasible solution to the dual linear program of Delsarte (1973), which improves on the first linear programming bound, following this line of argument. © 1963-2012 IEEE.

关键词： Integer linear programming

来源：评论

学校读者我要写书评

暂无评论

A verifiable variable threshold visual image secret sharing scheme

引用

Multimedia Systems 2025年第3期31卷 1-18页

作者： Yang, Yu-Guang Cheng, Wen Xu, Guang-Bao Jiang, Dong-Huan Zhou, Yi-Hua Shi, Wei-Min Jiang, Dong-Hua College of Computer Science and Technology Beijing University of Technology Beijing100124 China College of Computer Science and Engineering Shandong University of Science and Technology Qingdao266590 China College of Mathematics and Systems Science Shandong University of Science and Technology Qingdao266590 China School of Computer Science and Engineering Sun Yat-Sen University Guangzhou511400 China

This paper proposes a visual image encryption algorithm based on 2D compressive sensing and variable threshold secret sharing, aiming to achieve efficient, secure, and flexible image encryption and decryption. Firstly, the plain image is compressed and encrypted using the 2D compressive sensing technique, while the measurement matrix and disordered sequence are generated through an improved chaotic system. Secondly, the shadow image is produced using the variable threshold secret sharing scheme, and a corresponding authentication signature is generated to verify the integrity and authenticity of the shadow image. Finally, the shadow image and its signature are embedded into a carrier image to generate a cipher image with visual steganography. Experimental results demonstrate that the proposed algorithm achieves excellent performance in terms of operational efficiency, visual encryption quality, and decryption performance. The encryption execution time is 0.694 s, the PSNR of decrypted images is approximately 32 dB, and the PSNR of cipher images is around 42 dB, ensuring high visual concealment and reconstruction quality while maintaining security. © The Author(s), under exclusive licence to Springer-Verlag GmbH Germany, part of Springer Nature 2025.

关键词： Image compression

来源：评论

学校读者我要写书评

暂无评论

LoRaDIP: Low-Rank Adaptation With Deep Image Prior for Generative Low-Light Image Enhancement

IEEE Transactions on Artificial Intelligence

引用

IEEE Transactions on Artificial Intelligence 2025年第4期6卷 909-920页

作者： Zhao, Zunjin Shi, Daming Shenzhen University College of Computer Science and Software Engineering Shenzhen518060 China

This article presents LoRaDIP, a novel low-light image enhancement (LLIE) model based on deep image priors (DIPs). While DIP-based enhancement models are known for their zero-shot learning, their expensive computational cost remains a challenge. In addressing this issue, our proposed LoRaDIP introduces a low-rank adaptation technique, significantly reducing computational expenses without compromising performance. The contributions of this work are threefold. First, we eliminate the need for estimating initial illumination and reflectance, opting instead to directly estimate the illumination map from the observed image in a generative fashion. The illumination is parameterized by a DIP network. Second, considering the overparameterization of DIP networks, we introduce a low-rank adaptation technique to decrease the number of trainable parameters, thereby reducing computational demands. Third, differing from the existing DIP-based models that rely on a preset fixed number of iterations to halt the optimization process of Retinex decomposition, we propose an automatic stopping criterion based on stable rank, preventing unnecessary iterations. LoRaDIP not only inherits the advantage of requiring only the single input image but also exhibits reduced computational costs while maintaining or even surpassing the performance of state-of-the-art models. © 2020 IEEE.

关键词： Zero-shot 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：