检索结果-内蒙古大学图书馆

FLAG: frequency-based local and global network for face forgery detection

Multimedia Tools and Applications 2025年第2期84卷 647-663页

作者： Zhou, Kai Sun, Guanglu Wang, Jun Wang, Jiahui Yu, Linsen School of Computer Science and Technology Harbin University of Science and Technology Harbin150080 China Department of Information Engineering and Mathematics University of Siena Siena53100 Italy

Deepfake detection aims to mitigate the threat of manipulated content by identifying and exposing forgeries. However, previous methods primarily tend to perform poorly when confronted with cross-dataset scenarios. To address the above issue, we propose an innovative hybrid network called the Frequency-based Local and Global (FLAG) network to explore local and global information with the help of frequency-domain cues for better generalization capability. In consideration of the fact that forged faces often exhibit flaws in the frequency domain, we design a Frequency-based Attention Enhancement Module (FAEM) to enhance the aggregation of CNN and Vision Transformer (ViT). In this design, local features from CNN are attentively enhanced by selected frequency coefficients in FAEM, facilitating generalizable global features learning by the ViT module. The effectiveness of the proposed method is validated via numerous experiments and the generalization performance is improved under cross-dataset scenarios. Especially, the proposed method have obtained an AUC of 99.26% and an ACC of 96.56% using intra-dataset experimental results on FaceForensics++ (C23). © The Author(s), under exclusive licence to Springer Science+Business Media, LLC, part of Springer Nature 2024.

关键词： Face recognition

来源：评论

学校读者我要写书评

暂无评论

Spatial attention-based hybrid VGG-SVM and VGG-RF frameworks for improved cotton leaf disease detection

引用

Neural Computing and Applications 2025年第14期37卷 8309-8329页

作者： Pandiyaraju, V. Anusha, B. Senthil Kumar, A.M. Jaspin, K. Venkatraman, Shravan Kannan, A. School of Computer Science and Engineering Vellore Institute of Technology Tamil Nadu Chennai India Department of Computer Science and Engineering Marian Engineering College Trivandram Kerala Trivandrum India Department of Computer Science and Engineering St. Joseph’s Institute of Technology Tamil Nadu Chennai India Department of Information Science and Technology College of Engineering Guindy Anna University Tamil Nadu Chennai India

The agriculture industry's production and food quality have been impacted by plant leaf diseases in recent years. Hence, it is vital to have a system that can automatically identify and diagnose diseases at an initial stage for enhancing the quality of agricultural output and also for stopping the overall plant extinction. Most state-of-the-art methods do not provide appropriate accuracy of identification when the similarity among different diseases is higher and the input leaf images have complex background. In this work, a transfer learning-based automated crop disease recognition system is proposed to handle this problem. The cotton leaf disease detection framework being proposed here includes two key stages: (1) the feature extraction stage, where the convolutional neural network (CNN) model from the visual geometry group (VGG-16) is employed to obtain the deep features from cotton leaf images with complex background. (2) Cotton leaf disease detection stage: Here, the extracted optimal features are input to the machine learning classifiers, namely random forest (RF) and nonlinear support vector machine (SVM), for the initial classification of the diseases. For providing an optimal cotton crop disease detection performance, Kaggle’s four-class cotton disease dataset with complex background is chosen in this work. For the final decision making and analysis, two hybrid models named spatial attention-based hybrid VGG-RF and hybrid VGG-SVM have been proposed in this work for enhancing the decision accuracy. These hybrid models offered an average accuracy of approximately 98.29% and 99.31%, respectively, which is higher than the accuracy provided by the related work. This proved that the proposed hybrid model, namely spatial attention-based VGG-SVM and VGG-RF models, with spatial attention outperformed the state-of-the-art classifiers. © The Author(s), under exclusive licence to Springer-Verlag London Ltd., part of Springer Nature 2025.

关键词： Convolutional neural networks

来源：评论

学校读者我要写书评

暂无评论

An overview on IRS-enabled sensing and communications for 6G: architectures, fundamental limits, and joint beamforming designs

引用

Science China(information Sciences) 2025年第5期68卷 170-193页

作者： Xianxin SONG Yuan FANG Feng WANG Zixiang REN Xianghao YU Ye ZHANG Fan LIU Jie XU Derrick Wing Kwan NG Rui ZHANG Shuguang CUI School of Science and Engineering (SSE) Shenzhen Future Network of Intelligence Institute (FNii-Shenzhen)and Guangdong Provincial Key Laboratory of Future Networks of Intelligence The Chinese University of Hong Kong Department of Electrical Engineering City University of Hong Kong School of Information Engineering Guangdong University of Technology Key Laboratory of Wireless-Optical Communications Chinese Academy of Sciences School of Information Science and Technology University of Science and Technology of China Computer School Beijing Information Science and Technology University National Mobile Communications Research Laboratory School of Information Science and Engineering Southeast University School of Electrical Engineering and Telecommunications University of New South Wales School of Science and Engineering Shenzhen Research Institute of Big Data The Chinese University of Hong Kong Department of Electrical and Computer Engineering National University of Singapore

This study presents an overview on intelligent reflecting surface(IRS)-enabled sensing and communication for the forthcoming sixth-generation(6G) wireless networks, in which IRSs are strategically deployed to proactively reconfigure wireless environments to improve both sensing and communication(S&C) performance. First, we exploit a single IRS to enable wireless sensing in the base station's(BS's) non-line-of-sight(NLoS) area. In particular, we present three IRS-enabled NLoS target sensing architectures with fully-passive, semi-passive, and active IRSs, respectively. We compare their pros and cons by analyzing the fundamental sensing performance limits for target detection and parameter estimation. Next, we consider a single IRS to facilitate integrated sensing and communication(ISAC), in which the transmit signals at the BS are used for achieving both S&C functionalities, aided by the IRS through reflective beamforming. We present joint transmit signal and receiver processing designs for realizing efficient ISAC, and jointly optimize the transmit beamforming at the BS and reflective beamforming at the IRS to balance the fundamental performance tradeoff between S&C. Furthermore, we discuss multi-IRS networked ISAC, by particularly focusing on multi-IRS-enabled multi-link ISAC, multi-region ISAC, and ISAC signal routing, respectively. Finally, we highlight various promising research topics in this area to motivate future work.

关键词： integrated sensing and communication (ISAC) intelligent reflecting surface (IRS) non-line-of-sight (NLoS) sensing sensing and communication tradeoff

来源：评论

学校读者我要写书评

暂无评论

Advanced deep learning models for automatic detection of driver’s facial expressions, movements, and alertness in varied lighting conditions: a comparative analysis

引用

Multimedia Tools and Applications 2025年第6期84卷 2947-2983页

作者： Das, Shiplu Pratihar, Sanjoy Pradhan, Buddhadeb Computer Science and Engineering Indian Institute of Information Technology West Bengal Kalyani India University of Engineering and Management Kolkata West Bengal New Town India

Emotion recognition is vital in understanding human behavior and improving the driving experience. We present a systematic analysis of automatic detection of driver’s facial expressions and movements while crying, feeling sad, experiencing anger, drowsiness, fear and surprise to reduce road accident using a deep learning method in varied lighting conditions. The proposed approach aims to accurately detect and analyze driver emotions in real time, it would help in advancing the growth of smart driver assistance systems. To conduct our analysis, we create a comprehensive dataset comprising annotated facial images of drivers exhibiting crying, sadness, anger, drowsiness, fear and surprise. The dataset is carefully designed to encompass diverse lighting conditions, facial orientations, and driver demographics, ensuring the generalizability and robustness of the proposed method. The CNN-BiLSTM model is compared against the CNN, LSTM, CNN-LSTM, BiLSTM, and CNN-LSTM models in the research to perform precise localization of facial expressions and movements. The trial findings show how well our technique accurately identifies and evaluates a driver’s facial expressions. © The Author(s), under exclusive licence to Springer Science+Business Media, LLC, part of Springer Nature 2024.

关键词： Advanced driver assistance systems

来源：评论

学校读者我要写书评

暂无评论

Stationary wavelet transform and SpinalNet trained light spectrum Tasmanian devil optimization enabled DR detection using fundus images

引用

Multimedia Tools and Applications 2025年第4期84卷 1597-1626页

作者： Tamboli, Mubin Shoukat Dhumane, Amol Prasad, Rajesh Prasad, Jayashree Rajesh Ranjan, Nihar M. Computer Engineering Pimpri Chinchwad College of Engineering Pradhikaran Maharashtra Pune411044 India Computer Science and Engineering Symbiosis Institute of Technology Maharashtra Pune India Computer Science and Engineering School of Computing MIT Art Design and Technology University Loni Kalbhor Maharashtra Pune412201 India Information Technology JSPM’s Rajarshi Shahu College of Engineering Tathawade Maharashtra Pune411033 India

Diabetes is a common disease that causes complications in the eyes known as Diabetic Retinopathy (DR). It aids in discovering the DR and tends to be the main cause behind people’s blindness amidst the previous decades. For prevent the complications of DR and to control the level of blood Glucose, the timely and crucial diagnosis is essential. A model is provided for detecting the DR. Here, Light Spectrum Tasmanian Devil Optimization (LSTDO) is utilized with SpinalNet for DR detection. The method undergoes an adaptive median filter and Region of interest mining process to generate a pre-processed image. The Sparse Fuzzy C-Means (FCM) is adapted for segmenting the optic disks followed by feature extraction. Then, the Hough line transformation with Tyler Coye algorithm (HLTCA) is used for blood vessel segmentation. Concurrently, the Stationary Wavelet Transform (SWT) is adapted to input image to overwhelm the deficiency of translation-invariance issue and then the feature mining is done. The feature vector is generated by unifying the outputs of the three-feature extraction phases and is employed to detect DR. The accuracy, sensitivity, and specificity obtained by the LSTDO-based SpinalNet are 91.6%, 89.9% and 92.1%, respectively. © The Author(s), under exclusive licence to Springer Science+Business Media, LLC, part of Springer Nature 2024.

关键词： Feature extraction

来源：评论

学校读者我要写书评

暂无评论

Traffic sign recognition model based on scale sequence features and high-order spatial interactions

引用

Neural Computing and Applications 2025年第5期37卷 3605-3620页

作者： Zhang, Gan Wang, Yafei Li, Wenju Fu, Xianping School of Information Science and Technology Dalian Maritime University Dalian116026 China School of Computer Science and Information Engineering Shanghai Institute of Technology Shanghai201418 China Pengcheng Laboratory Shenzhen518055 China

Traffic sign recognition is an integral part of driver assistance systems play a crucial role in enhancing road safety. Due to a large number of challenging targets, such as occlusion, distortion, and small targets in actual scenes, existing methods still have bottlenecks in the accuracy of detection and recognition. Therefore, this paper proposes a real-time traffic sign recognition model based on scale sequence features (SSFs) and high-order spatial interactions (HOSIs) on the basis of YOLOv5. This model extracted scale-invariant SSF using a high-dimensional convolution on the underlying feature map to improve the detection effect of multi-scale targets. At the same time, recursive gated convolution modules are integrated into the feature pyramid network, expanding the interactions to arbitrary order and providing richer feature information for high-order convolutions. In addition, a cross-stage partial (CSP) structure is incorporated into the spatial pooling pyramid, which improves the structure’s performance without introducing excessive computational complexity. The experimental results show that the proposed model achieves 95.5% and 99.92% of mAP on Tsinghua–Tencent 100K (TT100K) and German Traffic Sign Recognition Benchmark (GTSRB), respectively, which surpass the performance of the most advanced model and meets the requirements of real-time detection. © The Author(s), under exclusive licence to Springer-Verlag London Ltd., part of Springer Nature 2024.

关键词： Convolution

来源：评论

学校读者我要写书评

暂无评论

HGTFM: Hierarchical Gating-Driven Transformer Fusion Model for Robust Multimodal Sentiment Analysis

引用

IEEE Access 2025年 13卷 74430-74445页

作者： Yang, Chengcheng Liang, Zhiyao Yan, Dashun Hu, Zeng Wu, Ting Macau University of Science and Technology School of Computer Science and Engineering Avenida Wai Long Taipa China Zhongkai University of Agriculture and Engineering School of Information Science and Technology Guangzhou510550 China

In multimodal sentiment analysis, a significant challenge lies in quantifying the contribution of each modality and achieving effective modality fusion. This paper presents a Hierarchical Gating-Driven Transformer Fusion Model (HGTFM), which effectively achieves multimodal data fusion through an advanced transformer architecture and a Hierarchical Gated Fusion Module (HGFM). The model first encodes audio and video features using LSTM and CNN, respectively, while employing BERT to encode text features. This approach effectively enhances the model ability to manage long-range temporal dependencies. Furthermore, the model introduces HGFM, which dynamically integrates multimodal features across different levels after processing by the modality interaction network (MIN), thus intelligently adjusting the contribution weights of each modality. In particular, the model incorporates a Unimodal Label Generation Module (ULGM), which considers the differences between the modalities. This enables HGFM to more accurately identify the contributions of samples with significant modal variations, thus optimizing the module potential. The experimental results on several established benchmark datasets demonstrate that HGTFM achieves comparable performance improvements in multimodal sentiment analysis tasks. © 2013 IEEE.

关键词： Sentiment analysis Transformers Analytical models Data models Adaptation models Tensors Accuracy Noise measurement Training Robustness

来源：评论

学校读者我要写书评

暂无评论

Enhancing Accuracy of Handwritten Character Segmentation Using Region-Based Convolutional Neural Networks (R-CNN) 1

Enhancing Accuracy of Handwritten Character Segmentation Usi...

引用

1st International Conference on AIML-Applications for engineering and technology, ICAET 2025

作者： Varpe, Kanchan Sakhare, Sachin Joshi, Rahul Vishwakarma Institute of Information Technology Department of Computer Engineering Pune India Department of Computer Science and Engineering Pune India

ISBN: (纸本)9798350355611

Handwritten character segmentation plays a pivotal role in the performance of Optical Character Recognition (OCR) systems. This paper introduces an innovative approach to enhancing segmentation accuracy using Region-Based Convolutional Neural Networks (R-CNN). Traditional methods often struggle with issues such as varying character sizes, overlapping, and inconsistent writing styles, leading to suboptimal recognition results. The proposed R-CNN model addresses these challenges by effectively localizing and segmenting individual characters from handwritten text images. The methodology involves training the R-CNN on a comprehensive dataset of handwritten characters, annotated to capture diverse styles and complexities. The model leverages deep learning's feature extraction capabilities and region-based localization to accurately segment characters, even in challenging scenarios like connected or touching characters. Comparative experiments show that the proposed approach outperforms conventional techniques in terms of segmentation accuracy and robustness. Results demonstrate significant improvements, with the R-CNN model achieving higher precision and recall rates across multiple handwritten datasets, including different languages and scripts. This advancement holds promise for enhancing OCR systems used in applications such as automated document processing and digital archiving. Future work will focus on integrating this segmentation method with advanced recognition models to further boost overall handwritten text recognition performance. © 2025 IEEE.

关键词： Convolutional neural networks

来源：评论

学校读者我要写书评

暂无评论

Enhancing the efficiency of lung cancer screening: predictive models utilizing deep learning from CT scans

引用

Neural Computing and Applications 2025年 1-19页

作者： Tawfeek, Medhat A. Alrashdi, Ibrahim Alruwaili, Madallah Shaban, Warda M. Talaat, Fatma M. Department of Computer Science Faculty of Computers and Information Menoufia University Shebin Elkom32511 Egypt Department of Computer Science College of Computer and Information Sciences Jouf University Sakakah Saudi Arabia Department of Computer Engineering and Networks College of Computer and Information Sciences Jouf University Sakakah Saudi Arabia Department of Communication and Electronics Engineering Nile Higher Institute for Engineering and Technology Mansoura Egypt Faculty of Artificial Intelligence Kafrelsheikh University Kafrelsheikh Egypt Faculty of Computer Science and Engineering New Mansoura University Gamasa35712 Egypt

Lung cancer is the most lethal form of cancer. This paper introduces a novel framework to discern and classify pulmonary disorders such as pneumonia, tuberculosis, and lung cancer by analyzing conventional X-ray and CT scan images called lung cancer risk prediction (LCRP) model. LCRP has four modules, namely data collection and preprocessing, data augmentation module, image segmentation module, and prediction module. Actually, LCRP employs three deep learning models;sequential model, functional model, and transfer model on publicly available training datasets. Convolutional neural networks (CNNs) have emerged as a highly effective field in machine learning, particularly for image datasets in the field of biomedical applications. The primary goal is to validate these models by comparing their performance with other models in order to determine their effectiveness in addressing challenging datasets. Our research has revealed a noteworthy enhancement in the efficiency of binary and multi-class classification using mask R-CNN image segmentation. During the model training process, a combination of Adam and stochastic gradient descent dual optimizers has been used to improve performance. LCRP have outperformed current pre-trained models by minimizing training parameters, computational costs, and overhead. It introduces 98.5% accuracy, 88.7% specificity, 89% sensitivity, 89.2% precision, and 89.09% F-measure. © The Author(s), under exclusive licence to Springer-Verlag London Ltd., part of Springer Nature 2025.

关键词： Lung cancer

来源：评论

学校读者我要写书评

暂无评论

Exploring diversity and time-aware recommendations: an LSTM-DNN model with novel bidirectional dynamic time warping algorithm

引用

Soft Computing 2025年第4期29卷 2003-2013页

作者： Li, Te Chen, Liqiong Sun, Huaiying Hou, Mengxia Lei, Yunjie Zhi, Kaiwen School of Computer and Engineering East China University of Science and Technology Meilong road Shanghai200237 China School of Computer Science and Information Engineering Shanghai Institute of Technology Haiquan road Shanghai201400 China School of Chemistry and Environmental Engineering Shanghai Institute of Technology Haiquan road Shanghai201400 China

With the advent of the Web 3.0 era, the amount and types of data in the network have sharply increased, and the application scenarios of recommendation algorithms are continuously expanding. Location recommendation has gradually become one of the popular application scenarios in recommendation algorithms. Traditional recommendation algorithms not only ignore the temporal attribute of data when recommending information to users, but also blindly pursue the recommendation accuracy, which will cause certain "information cocoon room" problems. Therefore, this article treats user historical data as a time series and proposes an LSTM-DNN model based on the novel bidirectional Dynamic Time Warping (DTW) algorithm. Firstly, in response to the issue of different users consuming different amounts of information, this article proposes a novel bidirectional DTW algorithm to calculate the similarity between different users. Secondly, this article supplements the user dataset from three perspectives: "utilization" and "exploration" of information, and spatiotemporal attributes of data, which alleviates the problem of data sparsity and cold start in the dataset to a certain extent. Moreover, it effectively enhances the diversity of recommendation results. Finally, this paper constructs an Long Short-Term Memory-Deep Neural Networks (LSTM-DNN) to dynamically obtain user interests and preferences, and proposes a new metric Cumulative Self-System Diversity (CSSD) to measure the diversity of algorithm recommendation results. Experiments have shown that the model effectively enhances the diversity of recommendation results while ensuring recommendation accuracy. © The Author(s), under exclusive licence to Springer-Verlag GmbH Germany, part of Springer Nature 2025.

关键词： Time series

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：