检索结果-内蒙古大学图书馆

Innovative Video Classification Method based on Deep Learning Approach

Recent Patents on engineering 2025年第2期19卷 95-107页

作者： Hemamalini, V. Jayasutha, D. Vinothini, V.R. Devi, R. Manjula Kumar, A. Anitha, E. Department of Networking and Communications School of Computing SRM Institute of Science and Technology Chennai India Department of Computer Science and Engineering Arjun College of Technology Coimbatore Thamaraikulam India Department of Mathematics Bannari Amman Institute of Technology Sathyamangalam Erode638401 India Department of Computer Science and Design Kongu Engineering College Perundurai Erode638060 India Department of Civil Engineering University Institute of Engineering & Technology Maharshi Dayanand University Rohtak124001 India Department of Artificial Intelligence and Data Science Sri Eshwar College of Engineering Coimbatore641202 India

Background: The automated classification of videos through artificial neural networks is addressed in this work. To explore the concepts and measure the results, the data set UCF101 is used, consisting of video clips taken from YouTube to recognize actions. The study is carried out with the authors' resources to determine the feasibility of independent research in the area. Methods: This work was developed in the Python programming language using the Keras library with Tensorflow as the back-end. The objective is to develop a network that presents performance compatible with the state of the art in terms of classifying videos according to the actions taken. Results: Given the hardware limitations, there is considerable distance between the implementation possibilities in this work and what is known as the state-of-the-art. Conclusion: Throughout the work, some aspects in which this limitation influenced the development are presented, but it is shown that this realization is feasible and that obtaining expressive results is possible 98.6% accuracy is obtained in the UCF101 data set, compared to the 98 percentage points of the best result ever reported, using, however, considerably fewer resources. In addition, the importance of transfer learning in achieving expressive results as well as the different performances of each architecture are reviewed. Thus, this work may open doors to carry patent-based outcomes. © 2025 Bentham science Publishers.

关键词： Convolutional neural networks

来源：评论

学校读者我要写书评

暂无评论

An ensemble approach for classification of diabetic retinopathy in fundus image

引用

Multimedia Tools and Applications 2025年第12期84卷 10567-10586页

作者： Prakash, J. Vinoth kumar, B. Department of Computer Science & Engineering PSG College of Technology Tamil Nadu Coimbatore India Department of Information Technology PSG College of Technology Tamil Nadu Coimbatore India

Diabetes-oriented diabetic retinopathy impacts the blood vessels in the region of the retina to enlarge and leak blood and other fluids. In most cases, diabetic retinopathy affects both eyes. If left untreated, it would eventually lead to blindness. Diabetic Retinopathy is so widespread that around 93 million adults are affected by it on a worldwide scale. Early identification and treatment would reduce the danger of serious diabetic retinopathy visual impairment. To achieve early detection, we can make use of our advancement in image processing to diagnose how severely the patient is affected by diabetic retinopathy. Using a patient’s fundus image and modern deep learning methods, we will be able to classify the result into five categories of severity: unaffected, severe, mild, moderate, and proliferate. Based on the results, the doctor would be able to determine what type of treatment is required to further avoid vision loss. The strategy outlined in this work is to employ five well-known deep learning models, namely Inception ResNet V2, VGG19, Xception, ResNet50, DenseNet112, and integrate their outcomes successfully utilizing ensemble learning. The experimentation outcome displays that the proposed approach classifies all the classes of Diabetic Retinopathy and performs better compared to other methods with a Specificity of 92.02%. © The Author(s), under exclusive licence to Springer science+Business Media, LLC, part of Springer Nature 2024.

关键词： Blood vessels

来源：评论

学校读者我要写书评

暂无评论

An improved image compression framework in modern multimedia applications using tetrolet transformation

引用

Multimedia Tools and Applications 2025年第6期84卷 3305-3338页

作者： Yadav, Satya Prakash Department of Computer Science and Engineering Madan Mohan Malaviya University of Technology U.P Gorakhpur India

The fast advancement of the multimedia era has led to an explosion in the use and technology of large amounts of digital snapshots. It has created a developing call for Image compression techniques that can reduce the dimensions of Image without compromising their perceptual excellence. Lossless picture compression techniques have emerged as a popular answer in this regard, as they maintain the precise statistics of the original Image. However, current lossless compression techniques face demanding situations in accomplishing high compression ratios without sacrificing the picture best. This paper proposes a more robust lossless Image compression framework based on the Tetrolet transformation. The framework introduces an adaptive quantization method, which adjusts the quantization step length based totally on the neighborhood Image characteristics, resulting in better compression overall performance for exclusive sorts of Image. Context modeling is also included in the framework to make the most of the inter-Image correlations inside the Image. This method uses context information from the Detroit-converted coefficients, which might be used to estimate the probability of prevalence for every Image value. The proposed model obtained 94.09% accuracy, 92.90% Precision, 95.25% Recall, 92.91% F1-Score and 93.87% Compression Duration. Using as it should be modeling these chances;the framework improves the entropy coding performance and ultimately results in better compression ratios. © The Author(s), under exclusive licence to Springer science+Business Media, LLC, 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

Multilevel characterization of unknown protein sequences using hierarchical long short term memory model

引用

Multimedia Tools and Applications 2025年第9期84卷 6593-6614页

作者： Agrawal, Saurabh Sisodia, Dilip Singh Nagwani, Naresh Kumar Department of Computer Science & Engineering National Institute of Technology Raipur GE Road Chhattisgarh Raipur492010 India

Multilevel characterization of the recently developed Unknown Protein Sequence (UPS) is significant for the drug-designing, disease-diagnosis, and treatment plans. UPS can demonstrate harmful as well as useful characters. Long Short Term Memory (LSTM) based Multilevel Protein Characterization (ML_PC) model using Neuro-Fuzzy Classifier (NFC) and transfer learning has been proposed in this paper for characteristics, species, and function prediction of the UPS. Gram Negative (GN) dataset with known 2286 protein sequences, is used for formation and validation of the ML_PC model. Thereafter NFC is applied at initial stage of the validated ML_PC model to filter the 518 independent protein sequences of the UPS dataset as GN or Non-GN in context with known GN dataset. However, NFC is competent to handle likelihood classification complications of supervised learning. Characteristics, species, and functions of the filtered GN UPS have been predicted in succeeding levels of the ML_PC framework through Protein Characterization (PC) model. LSTM and transfer learning based PC model has been trained and validated using augmented feature of the known GN protein sequence samples, while tested with UPS and filtered GN protein sequences. Transfer learning has been applied as requisite postprocessing to overcome overfitting in the prediction of bacterial protein species with insufficient protein sequence samples. Performance of the multilevel model is improved by 7.52% after combining with transfer learning and NFC. ML_PC model has been compatible to extract relevant biological information of the UPS in context to various categories of bacterial proteins. © The Author(s), under exclusive licence to Springer science+Business Media, LLC, part of Springer Nature 2024.

关键词： Long short-term memory

来源：评论

学校读者我要写书评

暂无评论

Unique hash value based identical video sharing traffic flow optimization across social networks

引用

Multimedia Tools and Applications 2025年第13期84卷 11983-12011页

作者： Konyak, Chenlep Yakha Baydeti, Nagaraju Department of Computer Science and Engineering National Institute of Technology Nagaland Nagaland Chumukedima Dimapur797 103 India

Identical video sharing across social networks results in redundant video data traffic, consuming additional bandwidth, storage resources, and data usage. A novel method is proposed to optimise redundant video-sharing services across social media applications exclusively on the push model, which is based on creating a unique hash value identity (UHVID) to identify the video distinctly. The UHVID is generated based on the centroid of gradient orientation of the key frames. The key frames are extracted from the video based on colour histogram and structural differences. Using UHVID and video metadata, the server application validates and optimizes identical video content flow before actual video sharing. The end users’ privacy is not affected in the proposed framework, as the actual video is not used for comparison, only the hash values of the video. The statistical analysis has shown that in 14 test cases of 2700 traffic initiations and 39,180 transactions, 83.2% of redundant data of identical videos were optimised, resulting in 63.61% data savings, amounting to 19.75 GB. © The Author(s), under exclusive licence to Springer science+Business Media, LLC, part of Springer Nature 2024.

关键词： Digital storage

来源：评论

学校读者我要写书评

暂无评论

Double Averaging and Gradient Projection: Convergence Guarantees for Decentralized Constrained Optimization

引用

IEEE Transactions on Automatic Control 2025年第5期70卷 3433-3440页

作者： Shahriari-Mehr, Firooz Panahi, Ashkan Chalmers University of Technology University of Gothenburg Department of Computer Science and Engineering SE-412 96 Gothenburg Sweden

We consider a generic decentralized constrained optimization problem over static, directed communication networks, where each agent has exclusive access to only one convex, differentiable, local objective term and one convex constraint set. For this setup, we propose a novel decentralized algorithm, called double averaging and gradient projection (DAGP). We achieve global optimality through a novel distributed tracking technique we call distributed null projection. Further, we show that DAGP can be used to solve unconstrained problems with nondifferentiable objective terms with a problem reduction scheme. Assuming only smoothness of the objective terms, we study the convergence of DAGP and establish sublinear rates of convergence in terms of feasibility, consensus, and optimality, with no extra assumption (e.g., strong convexity). For the analysis, we forego the difficulties of selecting Lyapunov functions by proposing a new methodology of convergence analysis, which we refer to as aggregate lower-bounding. To demonstrate the generality of this method, we also provide an alternative convergence proof for the standard gradient descent algorithm with smooth functions. © 1963-2012 IEEE.

关键词： Constrained optimization

来源：评论

学校读者我要写书评

暂无评论

Enhanced Neutrosophic Set and Machine Learning Approach for Kidney Disease Prediction

引用

Neutrosophic Sets and Systems 2025年 80卷 465-477页

作者： Al-Doori, Humam M. Alkellezli, Tareef S. Abdelhafeez, Ahmed Eassaa, Mohamed Sawah, Mohamed S. El-Douh, Ahmed A. Cybersecurity Sciences Department College of Science Ashur University Baghdad Iraq Cybersecurity Engineering Department College of Engineering Ashur University Baghdad Iraq Computer Science Department Faculty of Information System and Computer Science October 6 University Giza12585 Egypt Applied Science Research Center Applied Science Private University Amman Jordan Department of Computer Science Faculty of Information Technology Ajloun National University P.O.43 Ajloun26810 Jordan Cybersecurity Technology Engineering Department College of Engineering Technology Ashur University Baghdad Iraq

Kidney disease (KD) is a gradually increasing global health concern. It is a chronic illness linked to higher rates of morbidity and mortality, a higher risk of cardiovascular disease and numerous other illnesses, and expensive medical expenses. The machine learning (ML) models are applied for KD prediction with higher accuracy and precision. The KD dataset has uncertainty and vague information, so, we used the neutrosophic set (NS) to deal with vague and uncertainty information in the KD dataset. The KD dataset is converted into the N-KD dataset with three membership functions: truth, indeterminacy, and falsity. Three ML models are used in this study such as logistic regression (LR), support vector machine (SVM), and k-nearest neighbor (KNN). These ML models are applied to the N-KD dataset. The results show the LR has higher accuracy and precision on the N-KD dataset than the original KD dataset. © 2025, University of New Mexico. All rights reserved.

关键词： Logistic regression

来源：评论

学校读者我要写书评

暂无评论

Analyzing functional, technical and bibliometric trends of blockchain applications in education: A systematic review

引用

Multimedia Tools and Applications 2025年第8期84卷 4003-4048页

作者： Choudhary, Ankita Chawla, Meenu Tiwari, Namita Department of Computer Science and Engineering Maulana Azad National Institute of Technology Madhya Pradesh Bhopal462003 India

In an educational ecosystem, accurate and complete management of student’s records is a complex and error-prone process. Blockchain is a promising solution to add substantial value around enhanced efficiency, privacy protection, technological advancement, access control, and secure exchange of student course records. This work presents a systematic literature review to analyze the applications of blockchain technology in the educational domain addressing the gaps identified in the previous reviews. The technical papers published in English were searched by key terms related to blockchain and education on IEEE Xplore, science Direct, and six other databases. The risk of bias was assessed by considering all the crucial search terms preventing the accidental exclusion of any article. Many blockchain frameworks in the educational domain have been proposed;including some prototypes and implementations. The PRISMA framework was used for the evaluation of 150 distinct models, 124 of which were prototypes, pilots, or full implementation and the remaining 26 were only proposals. This work contributes a structured bibliometric, functional, and technical examination of these articles, assessing privacy, cost, performance, security, architecture, and standards to identify and bridge gaps. The study concludes an avalanche of proposals, with approximately half of them being prototypes, indicating that implementation and proof-of-concept pilots are expected. Blockchain is found to solve various record management issues at significant trade-offs between performance, storage, and cost. However, scalability issues, poor overall performance, and increased price are major limitations for implementing blockchain in education solutions. © The Author(s), under exclusive licence to Springer science+Business Media, LLC, part of Springer Nature 2024.

关键词： Records management

来源：评论

学校读者我要写书评

暂无评论

Multi-channel mantis search spatial attention convolutional neural network based hyperspectral image change detection

引用

Multimedia Tools and Applications 2025年 1-22页

作者： Divya, V.M. Department of Computer Science and Engineering East West Institute of Technology Visvesvaraya Technological University Belagavi560018 India

Hyperspectral imaging has transformed remote sensing by offering detailed spectral data for monitoring environmental changes, like urban development and terrestrial cover variations. However, existing change detection techniques often face challenges like underfitting and overfitting. In this research, a novel approach is introduced for hyperspectral image change detection using a Multi-Channel Mantis Search (MCMS). This innovative methodology leverages the power of a Multi-Channel Convolutional Neural Network (MCNN) to process each spectral band independently, facilitating the capture of unique features within each band. The input hyperspectral images are gathered from the Pavia University Dataset and the Indian Pines Dataset. This approach leads to significantly improved feature extraction, enhancing the discrimination of different materials in hyperspectral images. To further enhance the performance of the change detection model, the Mantis Search Algorithm (MSA) is integrated. The MSA is employed to optimize the model's hyperparameters, fine-tuning its architecture and training parameters to achieve superior results. The combination of MCNN and MSA improves hyperspectral image change detection accuracy, achieving 98.21% and 99.31% with the Indian Pines and Pavia University datasets correspondingly. © The Author(s), under exclusive licence to Springer science+Business Media, LLC, 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

A cancelable biometric authentication scheme based on geometric transformation

引用

Multimedia Tools and Applications 2025年第15期84卷 15407-15428页

作者： Champaneria, Vivek H. Patel, Sankita J. Zaveri, Mukesh A. Department of Computer Science And Engineering Sardar Vallabhbhai National Institute of Technology Icchanath Gujarat Surat395007 India

In digital identification technologies, fingerprint-based biometric systems are extensively utilized for authentication. Fingerprint authentication systems usually store the original minutiae points as user templates in the database which consist of minutiae bifurcation and ridge end points. Storing original minutiae points in the database can create a potential security risk to a person’s identity. An adversary might attack the database and compromise the user template. Various studies have demonstrated the feasibility of reconstructing the original fingerprint by utilizing the minutiae points. Existing research methods still need to be improved, and therefore, it is essential to provide security and robustness to fingerprint-based biometric systems. In this paper, a novel template protection technique using non-invertible transformation is proposed to create a secure and robust template. The transformation function uses the original minutiae positions, orientation details, and user key set values. The results demonstrated equal error rates of 1.5%, 0.85%, and 2.71% for the FVC2002 DB1, DB2, and DB3 fingerprint databases. The proposed method has demonstrated noteworthy performance improvement compared to other notable contributions. The outcome of the proposed method demonstrates practicality and effectiveness. © The Author(s), under exclusive licence to Springer science+Business Media, LLC, part of Springer Nature 2024.

关键词： Database systems

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：