检索结果-内蒙古大学图书馆

Seeing Through the Lies: A Vision Transformer-Based Solution 3rd

Seeing Through the Lies: A Vision Transformer-Based Solution

3rd International Conference on Artificial-Business Analytics, Quantum and Machine Learning: Trends, Perspectives, and Prospects, Com-IT-Con 2023

作者： Rasool, Aale Katarya, Rahul Big Data Analytics and Web Intelligence Laboratory Department of Computer Science Delhi Technological University New Delhi India

ISBN: (纸本)9789819725076

Deepfake technology has become increasingly sophisticated and poses a growing threat to society, as it can be used to create convincing fake videos for malicious purposes. Therefore, detecting deepfakes has become crucial to ensure the authenticity of visual content. This paper proposes a novel deepfake detection approach that utilizes InceptionResNetV2 for feature extraction and Vision Transformer (ViT) with Nyström Attention mechanism for classification. Our model achieves high accuracy, AUC, precision, and recall on the CelebDFv2 dataset and outperforms state-of-the-art techniques like EfficientNet, XceptionNet, and ResNet. We also perform evaluations on the CelebDFv1 and DFDC datasets. Furthermore, we conduct a comprehensive literature survey to highlight the existing research in deepfake detection. We provide a roadmap for future work in this field, highlighting potential research directions that can further improve the effectiveness and applicability of deepfake detection models. The proposed model shows great potential for real-world applications, and the suggested future work can help to address some of the remaining challenges in this area. © The Author(s), under exclusive license to Springer Nature Singapore Pte Ltd. 2024.

关键词： Face recognition

来源：评论

学校读者我要写书评

暂无评论

Semantic-Aware Human Object Interaction Image Generation 41

Semantic-Aware Human Object Interaction Image Generation

引用

41st International Conference on Machine Learning, ICML 2024

作者： Xu, Zhu Chen, Qingchao Peng, Yuxin Liu, Yang Wangxuan Institute of Computer Technology Peking University China National Institute of Health Data Science Peking University China

Recent text-to-image generative models have demonstrated remarkable abilities in generating realistic images. Despite their great success, these models struggle to generate high-fidelity images with prompts oriented toward human-object interaction (HOI). The difficulty in HOI generation arises from two aspects. Firstly, the complexity and diversity of human poses challenge plausible human generation. Furthermore, untrustworthy generation of interaction boundary regions may lead to deficiency in HOI semantics. To tackle the problems, we propose a Semantic-Aware HOI generation framework SA-HOI. It utilizes human pose quality and interaction boundary region information as guidance for denoising process, thereby encouraging refinement in these regions to produce more reasonable HOI images. Based on it, we establish an iterative inversion and image refinement pipeline to continually enhance generation quality. Further, we introduce a comprehensive benchmark for HOI generation, which comprises a dataset involving diverse and fine-grained HOI categories, along with multiple custom-tailored evaluation metrics for HOI generation. Experiments demonstrate that our method significantly improves generation quality under both HOI-specific and conventional image evaluation metrics. The code is available at https://***/XZPKU/***. Copyright 2024 by the author(s)

关键词： Semantics

来源：评论

学校读者我要写书评

暂无评论

A structural variation genotyping algorithm enhanced by CNV quantitative transfer

引用

Frontiers of computer science 2022年第6期16卷 159-165页

作者： Tian ZHENG Xinyang QIAN Jiayin WANG Department of Computer Science and Technology School of Electronic and Information EngineeringXi’an Jiaotong UniversityXi’an710049China Institute of Data Science and Information Quality Shaanxi Engineering Research Center of Medical and Health Big DataXi’an Jiaotong UniversityXi’an710049China

Genotyping of structural variations considering copy number variations(CNVs)is an infancy and challenging ***,a prevalent form of critical genetic variations that cause abnormal copy numbers of large genomic regions in cells,often affect transcription and contribute to a variety of *** characteristics of CNVs often lead to the ambiguity and confusion of existing genotyping features and algorithms,which may cause heterozygous variations to be erroneously genotyped as homozygous variations and seriously affect the accuracy of downstream *** the allelic copy number increases,the error rate of genotyping increases *** instances with different copy numbers play an auxiliary role in the genotyping classification problem,but some will seriously interfere with the accuracy of the *** by these,we propose a transfer learning-based method to genotype structural variations accurately considering *** method first divides the instances with different allelic copy numbers and trains the basic machine learning framework with different genotype *** maximizes the weights of the instances that contribute to classification and minimizes the weights of the instances that hinder correct *** adjusting the weights of the instances with different allelic copy numbers,the contribution of all the instances to genotyping can be maximized,and the genotyping errors of heterozygote variations caused by CNVs can be *** applied the proposed method to both the simulated and real datasets,and compared it to some popular algorithms including GATK,Facets and *** experimental results demonstrate that the proposed method outperforms the others in terms of accuracy,stability and *** source codes have been uploaded at github/TrinaZ/CNVtransfer for academic use only.

关键词： genotyping copy number variations transfer earning

来源：评论

学校读者我要写书评

暂无评论

Symbolic Regression Enhanced Decision Trees for Classification Tasks 38

Symbolic Regression Enhanced Decision Trees for Classificati...

引用

38th AAAI Conference on Artificial Intelligence, AAAI 2024

作者： Fong, Kei Sen Motani, Mehul Department of Electrical and Computer Engineering National University of Singapore Singapore Institute of Data Science National University of Singapore Singapore

ISBN: (纸本)1577358872

We introduce a conceptually simple yet effective method to create small, compact decision trees – by using splits found via Symbolic Regression (SR). Traditional decision tree (DT) algorithms partition a dataset on axis-parallel splits. When the true boundaries are not along the feature axes, DT is likely to have a complicated structure and a dense decision boundary. In this paper, we introduce SR-Enhanced DT (SREDT) – a method which utilizes SR to increase the richness of the class of possible DT splits. We evaluate SREDT on both synthetic and real-world datasets. Despite its simplicity, our method produces surprisingly small trees that outperform both DT and oblique DT (ODT) on supervised classification tasks in terms of accuracy and F-score. We show empirically that SREDTs decrease inference time (compared to DT and ODT) and argue that they allow us to obtain more explainable descriptions of the decision process. SREDT also performs competitively against state-of-the-art tabular classification methods, including tree ensembles and deep models. Finally, we introduce a local search mechanism to improve SREDT and evaluate it on 56 PMLB datasets. This mechanism shows improved performance on 77.2% of the datasets, outperforming DT and ODT. In terms of F-Score, local SREDT outperforms DT and ODT in 82.5% and 73.7% of the datasets respectively and in terms of inference time, local SREDT requires 25.8% and 26.6% less inference time than DT and ODT respectively. Copyright © 2024, Association for the Advancement of Artificial Intelligence (***). All rights reserved.

关键词： Decision trees

来源：评论

学校读者我要写书评

暂无评论

A Seat Headrest Loudspeaker System with Personalized Sound Zone Capabilities

A Seat Headrest Loudspeaker System with Personalized Sound Z...

引用

2024 IEEE International Conference on Consumer Electronics, ICCE 2024

作者： Fukui, Masahiro Kobayashi, Kazunori Kamado, Noriyoshi Computer and Data Science Laboratories NTT Corporation Tokyo Japan NTT Sonority Inc Product Department Tokyo Japan

ISBN: (纸本)9798350324136

This paper proposes a novel local reproduction method called 'personalized sound zone,' which can confine a sound field by using a combination of software processing and hardware innovations that remove loudspeaker enclosures from an array, and discusses the development of a headrest loudspeaker prototype unit not to leak sound into adjacent seats for airline seating. The proposed method is based on a two-channel fixed beamforming algorithm that does not require extensive equipment, and its prototype is compact enough to fit in a seat headrest. The personalized sound zone software algorithm is implemented on a DSP circuit board and processed in real time. © 2024 IEEE.

关键词： enclosureless loudspeaker fixed beamforming array local sound field reproduction seat headrest

来源：评论

学校读者我要写书评

暂无评论

Federated Class-Incremental Learning via Weighted Aggregation and Distillation

引用

IEEE Internet of Things Journal 2025年第12期12卷 22489-22503页

作者： Wu, Feng Ziying Tan, Alysa Feng, Siwei Yu, Han Deng, Tao Zhao, Libang Chen, Yuanlu Soochow University School of Computer Science and Technology Suzhou215000 China Nanyang Technological University College of Computing and Data Science Nanyang Ave 639798 Singapore Southwest Jiaotong University Ministry of Education Key Laboratory of Photonic-Electronic Integration and Communication-Sensing Convergence Chengdu610031 China

Federated Class-Incremental Learning (FCIL) aims to design privacy-preserving collaborative training methods to continuously learn new classes from distributed datasets. In these scenarios, federated clients face the challenge of encountering new classes while being constrained by limited memory capacity, which can lead to catastrophic forgetting in the resulting global model. Existing FCIL approaches tend to overlook the challenges posed by the heterogeneity of dataset label distribution among clients, thereby constraining the generalization capacity of the global model they learn. Some of these methods also suffer from excessive computational burdens when addressing catastrophic forgetting problems. Furthermore, certain approaches are constrained to handling only straightforward data, posing significant difficulties in managing complex datasets and tackling more intricate scenarios. In this article, we propose the Weighted Aggregation and Distillation-based FCIL (WAD-FCIL) method to address these limitations. To address data heterogeneity arising from class imbalance, we first introduce a task-aware client clustering method to identify clients with extreme class deviations before global model aggregation to eliminate potential impact on the global model. Then, we propose a multisampling weighted aggregation approach during the global FL model update that integrates knowledge from different clients and dynamically adjusts the weight of each client model to facilitate model update. To mitigate catastrophic forgetting, we propose a multimodel distillation strategy that involves selecting multiple teacher models for knowledge distillation. Extensive experiments comparing WAD-FCIL with ten state-of-the-art methods demonstrate that it significantly outperforms the baselines by 0.8%–3.2% in terms of average test accuracy on three representative benchmark datasets. The code of this work is available at https://***/wufeng10010/WAD-FCIL. © 2014 IEEE.

关键词： Federated learning

来源：评论

学校读者我要写书评

暂无评论

Adaptive Successive POI Recommendation via Trajectory Sequences Processing and Long Short-Term Preference Learning

引用

computers, Materials & Continua 2024年第10期81卷 685-706页

作者： Yali Si Feng Li Shan Zhong Chenghang Huo Jing Chen Jinglian Liu School of Computer Science and Engineering Changshu Institute of TechnologyChangshu215500China Suzhou Industrial Intelligence and Big Data Laboratory Suzhou215000China School of Information Science and Engineering Yanshan UniversityQinhuangdao066004China College of Electronic and Information Engineering Guangdong Ocean UniversityZhanjiang524088China

Point-of-interest(POI)recommendations in location-based social networks(LBSNs)have developed rapidly by incorporating feature information and deep learning ***,most studies have failed to accurately reflect different users’preferences,in particular,the short-term preferences of inactive *** better learn user preferences,in this study,we propose a long-short-term-preference-based adaptive successive POI recommendation(LSTP-ASR)method by combining trajectory sequence processing,long short-term preference learning,and spatiotemporal ***,the check-in trajectory sequences are adaptively divided into recent and historical sequences according to a dynamic time ***,an adaptive filling strategy is used to expand the recent check-in sequences of users with inactive check-in behavior using those of similar active *** further propose an adaptive learning model to accurately extract long short-term preferences of users to establish an efficient successive POI recommendation system.A spatiotemporal-context-based recurrent neural network and temporal-context-based long short-term memory network are used to model the users’recent and historical checkin trajectory sequences,*** experiments on the Foursquare and Gowalla datasets reveal that the proposed method outperforms several other baseline methods in terms of three evaluation *** specifically,LSTP-ASR outperforms the previously best baseline method(RTPM)with a 17.15%and 20.62%average improvement on the Foursquare and Gowalla datasets in terms of the Fβmetric,respectively.

关键词： Location-based social networks adaptive successive point-of-interest recommendation long short-term preference trajectory sequences

来源：评论

学校读者我要写书评

暂无评论

VAEPose: 6D Pose Estimation with Visual Appearance Enhancement for Low-Light Conditions 25

VAEPose: 6D Pose Estimation with Visual Appearance Enhanceme...

引用

25th IEEE International Conference on Industrial Technology, ICIT 2024

作者： Gao, Yan Chen, Lu Liu, Jingyang Insti. Of Big Data Science & Industry School Of Computer & Information Tech. Shanxi University Taiyuan China

ISBN: (纸本)9798350340266

6D pose estimation is a key technology in the field of computer vision, and has great application potential in the fields of virtual reality, augmented reality, robot operation, and intelligent driving. When using deep neural network technology to accomplish this advanced vision task, the quality of the data is demanding. Currently, the task achieves excellent results in normal lighting conditions, but the task presents serious challenges due to the significant degradation of data quality in low-light conditions. The results are not satisfactory when the object pose is directly estimated using a one-stage processing mode in low-light conditions. Therefore, in view of this observation, we choose a two-stage processing mode to solve the challenges. In this work, we propose a Low-light Pose Estimation Network based on the Visual Appearance Enhancement module (VAEPose). Among them, the Low-light Appearance Enhancement Module (LAEM) is designed to improve data quality and show clearer details, helping the network model to be more adaptable to light and less dependent on data. In the LAEM module, an attention mechanism is introduced to achieve the fusion of channel and spatial attention of the extracted RGB features, resulting in finer color and texture information. The Virtual Low-light dataset (VLD) used in the experiment is created by us in virtual environment. Experiments verify the effectiveness of the proposed method. © 2024 IEEE.

关键词： Image enhancement

来源：评论

学校读者我要写书评

暂无评论

A Method of Selecting Negative Examples from Clinical EEG by Combining Ensemble Learning and Uncertainty Estimation 17

A Method of Selecting Negative Examples from Clinical EEG by...

引用

17th International Congress on Image and Signal Processing, BioMedical Engineering and Informatics, CISP-BMEI 2024

作者： Liu, Jianwen Han, Liyong Huang, Zexin Lin, Zhixiong Wang, Chenghua Huang, Zhihua College of Computer and Data Science Fuzhou University Fuzhou China Fujian Sanbo Funeng Brain Hospital Fuzhou China

ISBN: (纸本)9798331507398

Electroencephalogram (EEG) analysis is a critical tool for diagnosing various neurological disorders. Intelligent EEG models facilitate the analysis and diagnosis of these conditions. However, the development of such models relies on expert labeling of EEG signals. Due to limited time and resources, experts typically only label positive examples, leaving the selection of negative examples to be handled by algorithms. This paper proposes a novel method for selecting EEG negative samples by integrating ensemble learning with uncertainty estimation. The method employs a Stacking ensemble strategy, with AtcNet, USleep, and EEG Transformer as base learners, and Support Vector Machines (SVM) as meta-learners. By integrating two uncertainty metrics, the Variance Metric (VAR) and the Kullback-Leibler Divergence Metric (KL), this method selects high-quality EEG negative samples. Experimental results demonstrate that this approach significantly improves the model's performance. the quality of negative example samples selected by this method is better than that selected by any one of the models of AtcNet, USleep, and EEG Transformer alone. Compared to the baseline method, the proposed method enhances the macro-average F1 score by up to 5.67%. This study confirms the superiority of the proposed negative sample selection method and the effectiveness of the constructed dataset. © 2024 IEEE.

关键词： Support vector machines

来源：评论

学校读者我要写书评

暂无评论

Optimization Models for Teaching Load Formation of University Teachers 1

Optimization Models for Teaching Load Formation of Universit...

引用

IEEE AITU: 1st International Student Conference: "Digital Generation", AITU 2024

作者： Chigambayeva, Diana Goryakin, Makar Batova, Polina Department of Computational and Data Science Astana IT University Astana Kazakhstan Depatment of Computer Engineering Astana IT University Astana Kazakhstan

ISBN: (纸本)9798350364378

Functional and mathematical models for the distribution of academic workload at the stage of preparing the educational process at a university are considered, which make it possible to largely determine the uniformity of teachers’ workload with various types of academic work when fulfilling the criteria and restrictions adopted for the distribution. ©2024 IEEE.

关键词： functional model mathematical model study load educational process educational work types of educational work

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：