检索结果-内蒙古大学图书馆

IEEE Annual Conference on Pervasive Computing and Communications Workshops (PerCom)

作者： Manoj Lenka Ayon Chakraborty Department of Computer Science and Engineering SENSE LAB IIT Madras

Recent innovations in Wi-Fi sensing capitalizes on a host of powerful deep neural network architectures that make inferences based on minute spatio-temporal dynamics in the wireless channel. Many of such inference techniques being resource intensive, conventional wisdom recommends offloading them to the network edge for further processing. In this paper, we argue that edge based sensing is often not a viable option for many applications (cost, bandwidth, latency etc). Rather, we explore the paradigm of on-device Wi-Fi sensing where inference is carried out locally on resource constrained IoT platforms. We present extensive benchmark results characterizing the resource consumption (memory, energy) and the performance (accuracy, inference rate) of some typical sensing tasks. We propose Wisdom, a framework that, depending on capabilities of the hardware platform and application’s requirements, can compress the inference model. Such context aware compression aims to improve the overall utility of the system - maximal inference performance at minimal resource costs. We demonstrate that models obtained using the Wisdom framework achieve higher utility compared to baseline models in more than 85% of cases.

关键词：

来源：评论

学校读者我要写书评

暂无评论

LesNet: An Automated Skin Lesion Deep Convolutional Neural Network Classifier through Augmentation and Transfer Learning

引用

Procedia computer science 2024年 235卷 112-121页

作者： Aqib Nazir Mir Iqra Nissar Danish Raza Rizvi Ankush Kumar Research Lab Department of Computer Engineering Jamia Millia Islamia New Delhi 110025India Department of Computer Science and Engineering Punjabi University Punjab147002 India

Skin cancer is one of the most prevalent forms of cancer around the world. Initial diagnosis relies on visual assessment of the affected area, followed by detailed dermoscopic analysis. The development of an automated system for classifying skin lesions poses a considerable challenge due to inherent noise and subtle variability in lesion images. Deep convolutional neural networks have demonstrated exceptional proficiency in image classification tasks spanning diverse domains. This paper demonstrates an end-to-end classification architecture built on top of transfer learning with data augmentation. The proposed model capitalizes on pre-trained architectures such as DenseNet, VGG-16, and Inception for feature extraction, and employs fully connected dense layers for categorizing seven distinct types of lesions. Several data augmentation techniques are also used to handle the class imbalance problem. Extensive experimentation encompassing various hyperparameters, imbalanced data scenarios, and balanced datasets was conducted to refine the automated skin lesion system. The proposed approach achieved a notable accuracy of 98% on the HAM10000 dataset and 94% on the ISIC-2019 dataset. Importantly, the experimental findings surpassed the performance of current state-of-the-art models for lesion classification. Furthermore, this paper examines the impact of class imbalance and data augmentation on the model's accuracy.

关键词： Skin lesion Transfer learning Augmentation CNN Classification

来源：评论

学校读者我要写书评

暂无评论

An Efficient Soil Image Classification Framework Using Transfer Learning Models for Intelligent Geotechnical Applications 2nd

An Efficient Soil Image Classification Framework Using Trans...

引用

2nd International Conference on Sustainability: Developments and Innovations, ICSDI 2024

作者： Hemdan, Ezz El-Din Al-Atroush, M.E. Structure and Materials Research Lab Prince Sultan University Riyadh Saudi Arabia Computer Science and Engineering Department Faculty of Electronic Engineering Menoufia University Menoufia Egypt Department of Engineering Management College of Engineering Prince Sultan University P.O. Box 66833 Riyadh Saudi Arabia

ISBN: (纸本)9789819787111

Recently, the application of deep learning approaches has significantly contributed to the success of soil investigations, particularly in the categorization of soil images. This, in turn, facilitates informed decision-making in geotechnical engineering projects. This paper introduces an efficient soil classification method utilizing transfer learning models based on Convolutional Neural Networks (CNNs) to categorize soil images. Various pre-trained networks, including Alexnet, Vgg16, Googlenet, Resnet50, Mobilenetv2, and Squeezenet, are employed in this study. The highest accuracy, reaching 90.32%, is achieved by Googlenet, while Alexnet follows closely with an accuracy of 89.24%. Conversely, Squeezenet yields the lowest accuracy at 79.56%. This proposed framework has the potential for integration into Internet of Things (IoT) systems, leading to the development of an innovative, portable IoT-based device for soil type recognition at remote geo-sites, catering to the needs of geotechnical engineers. The results obtained highlight that the Googlenet-based framework outperforms others, offering a superior performance in soil classification. This framework serves as a valuable tool for implementing an intelligent soil image classification system, supporting real-time decision-making in geotechnical engineering applications. © The Author(s), under exclusive license to Springer Nature Singapore Pte Ltd. 2025.

关键词： Convolutional neural networks

来源：评论

学校读者我要写书评

暂无评论

Deep Learning Model for computer Vision: Sustainable Image Classification Using Machine Learning 1st

Deep Learning Model for Computer Vision: Sustainable Image C...

引用

1st International Conference on Sustainable Development through Machine Learning, Artificial Intelligence, and IoT, ICSD 2023

作者： Alankar, Bhavya Chauhan, Ritu Kaur, Harleen Ebadati, E. Omid Mahdi Department of Computer Science and Engineering School of Engineering Sciences and Technology Jamia Hamdard New Delhi India Artificial Intelligence and IoT Lab Center for Computational Biology and Bioinformatics Amity University UP Noida India Department of Mathematics and Computer Science Kharazmi University Tehran Iran

ISBN: (纸本)9783031470547

The core challenge within computer vision is image classification, which involves assigning an image a label and categorizing it using the features that have been extracted from it. Even though this work appears rudimentary, it requires the application involving artificial intelligence, specifically Deep Learning, a subfield of Machine Learning. In this paper, we have discussed how Convolutional Neural Networks (CNN), the most modern technique for image classification, are used in Deep Learning to classify images. We have discussed the applications of image classification and drawn the conclusion that using deep learning is preferable to using machine learning. Additionally, we have also presented a comparative study of various CNN models to observe how they could potentially be employed as pre-trained models for the image classification job. © 2023, The Author(s), under exclusive license to Springer Nature Switzerland AG.

关键词： Convolution

来源：评论

学校读者我要写书评

暂无评论

Enhancing Explainable Matrix Factorization with Tags for Multi-Style Explanations 15

Enhancing Explainable Matrix Factorization with Tags for Mul...

引用

15th International Conference on Knowledge Discovery and Information Retrieval, KDIR 2023 as part of the 15th International Joint Conference on Knowledge Discovery, Knowledge engineering and Knowledge Management, IC3K 2023

作者： Seton, Olurotimi Haghighi, Pegah Sagheb Alshammari, Mohammed Nasraoui, Olfa Knowledge Discovery and Web Mining Lab Computer Science and Engineering Dept University of Louisville United States Department of Computer Science Faculty of Computing and Information Technology Northern Border University Arar Saudi Arabia

ISBN: (纸本)9789897586712

Black-box AI models tend to be more accurate but less transparent and scrutable than white-box models. This poses a limitation for recommender systems that rely on black-box models, such as Matrix Factorization (MF). Explainable Matrix Factorization (EMF) models are "explainable" extensions of Matrix Factorization, a state of the art technique widely used due to its flexibility in learning from sparse data and accuracy. EMF can incorporate explanations derived, by design, from user or item neighborhood graphs, among others, into the model training process, thereby making their recommendations explainable. So far, an EMF model can learn a model that produces only one explanation style, and this in turn limits the number of recommendations with computable explanation scores. In this paper, we propose a framework for EMFs with multiple styles of explanation, based on ratings and tags, by incorporating EMF algorithms that use scores derived from tag-centric graphs to connect rating neighborhood-based EMF techniques to tag-based explanations. We used precalculated explainability scores that have been previously validated in user studies that evaluated user satisfaction with each style individually. Our evaluation experiments show that our proposed methods provide accurate recommendations while providing multiple explanation styles, without sacrificing the accuracy of the recommendations. Copyright © 2023 by SCITEPRESS - science and Technology Publications, Lda. Under CC license (CC BY-NC-ND 4.0)

关键词： Matrix algebra

来源：评论

学校读者我要写书评

暂无评论

Holistic and Historical Instance Comparison for Cervical Cell Detection

Holistic and Historical Instance Comparison for Cervical Cel...

引用

2024 IEEE International Conference on Bioinformatics and Biomedicine, BIBM 2024

作者： Jiang, Hao Liu, Runsheng Zhou, Yanning Lin, Huangjing Chen, Hao Tencent Ai Lab Shenzhen China Imsight Ai Research Lab Shenzhen China The Hong Kong University of Science and Technology Department of Computer Science and Engineering Hong Kong The Hong Kong University of Science and Technology Department of Chemical and Biological Engineering Hong Kong Hkust Shenzhen-Hong Kong Collaborative Innovation Research Institute Futian Shenzhen China

ISBN: (纸本)9798350386226

Cytology screening from Papanicolaou (Pap) smears is a common and effective tool for the preventive clinical management of cervical cancer, where abnormal cell detection from whole slide images serves as the foundation for reporting cervical cytology. However, cervical cell detection remains challenging due to 1) hazily-defined cell types (e.g., ASC-US) with subtle morphological discrepancies caused by the dynamic cancerization process, i.e., cell class ambiguity, and 2) imbalanced class distributions of clinical data may cause missed detection, especially for minor categories, i.e., cell class imbalance. To this end, we propose a holistic and historical instance comparison approach for cervical cell detection. Specifically, we first develop a holistic instance comparison scheme enforcing both RoI-level and class-level cell discrimination. This coarse-to-fine cell comparison encourages the model to learn foreground-distinguishable and class-wise representations. To emphatically improve the distinguishability of minor classes, we then introduce a historical instance comparison scheme with a confident sample selection-based memory bank, which involves comparing current embeddings with historical embeddings for better cell instance discrimination. Extensive experiments and analysis on two large-scale cytology datasets including 42,592 and 114,513 cervical cells demonstrate the effectiveness of our method. The code is available at https://***/hjiangaz/HERO. © 2024 IEEE.

关键词： Contrastive Learning

来源：评论

学校读者我要写书评

暂无评论

Generalized Multi-protocol label Switching based Virtualization for Cloud Computing 1

Generalized Multi-protocol Label Switching based Virtualizat...

引用

1st International Conference on Connected Systems and Intelligence, CSI 2022

作者： Saxena, Rajat Patel, Ajay Singh Shri Vaishnav Institute of Information Technology Shri Vaishnav Vidyapeeth Vishwavidyalaya Cloud computing Lab Department of Computer Science and Engineering Indore India Cloud computing Lab Department of Computer Science and Engineering Indore India

ISBN: (纸本)9781665458153

Generalized Multi Protocol label Switching (GMPLS) or Multi-Protocol Lambda Switching is a new technology which produce amplification in Multi Protocol label Switching (MPLS) to provide support network switching, space switching, and packet switching for time and wavelength. Thus, we can say that GMPLS is extension of MPLS which provides resilience and restoration by automatic switching. In this paper, we provide a Docker based simu-1ation environment which emulates a complex network and create insight functioning of GMPLS. The Docker based simulation environment is light-weight and scalable. It emulates large and complex network with minimal specifications. We have tested this Docker based GMPLS simulation testbed that depends on scaled network resources. Our method has shown tremendous improvement over the other virtualization methods. In this paper, we do a comparative study based on UML, Virtual Box, and Docker and found that Docker consumes very less resources when compared to UML and Virtual Box. In a system with 1 TB Hard disk space and 16 GB RAM, we can design and study a topology with 30-40 nodes using Virtual Box and a topology with 10-15 nodes using UML. Whereas, with Docker a topology with 300-nodes can be designed and tested. Thus scalability is limited by system specifications. © 2022 IEEE.

关键词： multi-protocol label switching

来源：评论

学校读者我要写书评

暂无评论

VMAP: Matching-based Efficient Offloading in IoT-Fog Environments with Variable Resources 17

VMAP: Matching-based Efficient Offloading in IoT-Fog Environ...

引用

17th IEEE International Conference on Advanced Networks and Telecommunications Systems, ANTS 2023

作者： Morey, Jui Satpathy, Anurag Addya, Sourav Kanti Cloud and Smart System Services Lab Department of Computer Science and Engineering National Institute of Technology Karnataka Surathkal India Missouri University of Science and Technology United States

ISBN: (纸本)9798350307672

Fog computing is a promising technology for critical, resource-intensive, and time-sensitive applications. In this regard, a significant challenge is generating an offloading solution that minimizes the latency, energy, and number of outages for a dense IoT-Fog environment. However, the existing solutions either focus on a single objective or mainly dedicate fixed-sized resources as virtual resource units (VRUs). Moreover, these solutions are restrictive and not comprehensive, resulting in poor performance. To overcome these challenges, this paper proposes a VMAP model addressing the lacunas above. Offloading problem is abstracted as a one-to-many matching game between two sets of entities - tasks and fog nodes (FNs) by considering both preferences. Moreover, the preferences and weights of the parameters are generated using the Analytic Hierarchy Process (AHP). Exhaustive simulations indicate that the proposed strategy outperforms the baseline algorithms, considering average task latency and energy consumption by 35% and 22.2%, respectively. Additionally, resource utilization also experiences a boost by 28.57%, and 97.98% of tasks complete their execution within the deadlines. © 2023 IEEE.

关键词： Fog

来源：评论

学校读者我要写书评

暂无评论

Improving DeepFake Video Detection Performance with a Noval Deep Learning Approach 3

Improving DeepFake Video Detection Performance with a Noval ...

引用

3rd IEEE International Conference on Electronic engineering, ICEEM 2023

作者： Fouda, Mona A. El-Shafai, Walid El-Rabaie, El-Sayed M. Menoufia University Faculty of Electronic Engineering Department of Electronics and Electrical Communications Engineering Menouf32952 Egypt Prince Sultan University Security Engineering Lab Computer Science Department Riyadh11586 Saudi Arabia

ISBN: (纸本)9798350323511

With the rise in both the quantity and sophistication of deepfake videos, the need for robust detection systems to identify potentially misleading content on social media and the internet has become paramount. However, current automated face forgery detection systems still face limitations, often demonstrating bias towards the training dataset. This research paper addresses this issue by proposing a novel approach for detecting deepfake media. We introduce a custom Visual Geometry Group (VGG16) deepfake detection method that leverages convolutional neural network architectures. To evaluate the effectiveness of our approach, we utilize the deepfake detection challenge (DFDC) dataset on Kaggle to build network models and compare the performance of our custom VGG16 method against the standard VGG16. Additionally, we investigate the impact of data augmentation techniques on the performance of Convolutional Neural Network (CNN)-based deepfake detectors, examining their effect on both VGG16 and our custom VGG16 approach using the DFDC dataset. Our results demonstrate a high level of accuracy, with precision, recall, and f1-score values of 0.983, 0.975, and 0.979, respectively, and an overall accuracy of 0.986 for deepfake detection. This study presents a promising approach to enhance the accuracy of deepfake video detection, representing a crucial step towards mitigating the potential negative impacts of deepfake technology. © 2023 IEEE.

关键词： Convolutional neural networks

来源：评论

学校读者我要写书评

暂无评论

Proposed Different Signal Processing Tools for Efficient Optical Wireless Communications

引用

computers, Materials & Continua 2022年第5期71卷 3293-3318页

作者： Hend Ibrahim Abeer D.Algarni Mahmoud Abdalla Walid El-Shafai Fathi E.Abd El-Samie Naglaa F.Soliman Department of Electronics and Communications Engineering Faculty of EngineeringZagazig UniversityZagazig44519Egypt Department of Information Technology College of Computer and Information SciencesPrincess Nourah Bint Abdulrahman UniversityRiyadhSaudi Arabia Department of Electronics and Electrical Communications Engineering Faculty of Electronic EngineeringMenoufia UniversityMenoufia32952Egypt Security Engineering Lab Computer Science DepartmentPrince Sultan UniversityRiyadh 11586Saudi Arabia

Optical Wireless Communication(OWC)is a new trend in communication systems to achieve large bandwidth,high bit rate,high security,fast deployment,and low *** basic idea of the OWC is to transmit data on unguided media with *** system requires multi-carrier modulation such as Orthogonal Frequency Division Multiplexing(OFDM).This paper studies optical OFDM performance based on Intensity Modulation with Direct Detection(IM/DD)*** system requires a non-negativity *** paper presents a framework for wireless optical OFDM system that comprises(IM/DD)with different forms,Direct Current biased Optical OFDM(DCO-OFDM),Asymmetrically Clipped Optical OFDM(ACO-OFDM),Asymmetrically DC-biased Optical OFDM(ADO-OFDM),and *** also considers channel coding as a tool for error control,channel equalization for reducing deterioration due to channel effects,and investigation of the turbulence *** evaluation results of the proposed framework reveal enhancement of *** performance of the IM/DD-OFDM system is investigated over different channel models such as AWGN,log-normal turbulence channel model,and ceiling bounce channel *** simulation results show that the BER performance of the IM/DD-OFDM communication system is enhanced while the fading strength is *** results reveal also that Hamming codes,BCH codes,and convolutional codes achieve better BER ***,two algorithms of channel estimation and equalization are considered and *** algorithms include the Least Squares(LS)and the Minimum Mean Square Error(MMSE).The simulation results show that the MMSE algorithm gives better BER performance than the LS algorithm.

关键词： Optical communication systems OWC IM/DD OFDM MMSE LS ADO-OFDM DCO-OFDM ACO-OFDM

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：