检索结果-内蒙古大学图书馆

HOQF-M: Hybrid Optimization of Quantization Friendly MobileNet Architecture for Vision Based Applications on Edge Devices

引用

SN computer science 2025年第4期6卷 1-16页

作者： Kulkarni, Uday Meena, S.M. Gurlahosur, Sunil V. School of Computer Science and Engineering KLE Technological University Hubballi India

Deep Neural Networks (DNNs) demand extensive memory bandwidth, intermediate storage, and high computational power, limiting their deployment on edge devices with constrained resources. Optimization techniques like network pruning and quantization help adapt DNNs for these environments by removing redundancies and reducing model size. Expanding on our earlier Quantization Friendly MobileNet (QF-MobileNet) architecture, developed to improve inference accuracy while optimizing resource usage, this study introduces the Hybrid Optimization of QF-MobileNet (HOQF-M) model. This two-phase optimization method first applies network slimming heuristics for structured pruning, followed by post-training 4-bit quantization to achieve an optimal balance between resource efficiency and minimal quantization loss. We evaluated HOQF-M through an image classification task using the India Heritage in Digital Space (IHDS) dataset, which comprises over 133,481 images spanning 90 classes. The HOQF-M model encounters a 2% reduction in accuracy compared to the unoptimized QF-MobileNet, HOQF-M attains a 45% reduction in model size, a 55% reduction in time complexity, and a 45% decrease in MAC operations per image. Additionally, we analyze the model’s sensitivity to quantization bit widths and pruning thresholds, revealing how these settings impact accuracy. These findings affirm HOQF-M’s significant potential for edge deployment, providing extensive resource efficiency gains with minimal performance trade-offs, making it highly suitable for environments with limited computational and memory capacities. © The Author(s), under exclusive licence to Springer Nature Singapore Pte Ltd. 2025.

关键词： Deep neural network Edge device Network-slimming Optimization Pruning QF-MobileNet Quantization

来源：评论

学校读者我要写书评

暂无评论

Optimized Placement of Service Function Chains in Edge Cloud with LSTM and ILP

引用

SN computer science 2025年第1期6卷 1-14页

作者： Vishesh, P. Poorva, K. Akshata, H. Ritwik, B. Narayan, D.G. Savanur, Sadaf School of Computer Science and Engineering KLE Technological University Karnataka Hubballi India

Edge computing has emerged as a transformative approach for reducing latency and enhancing network performance by placing computing resources closer to data sources and end users via edge nodes. This approach addresses the delays inherent in traditional cloud computing architectures, where data must travel long distances to centralized data centers. Within this architecture, Network Function Virtualization (NFV) enables network service deployment by executing Virtual Network Functions (VNFs) on virtualized infrastructure, establishing Service Function Chains (SFCs) to manage network traffic efficiently. In this work, we address the challenge of optimizing SFC placement that combines Long Short-Term Memory (LSTM) networks and Integer Linear Programming (ILP). The LSTM model predicts dynamic traffic patterns and resource demands of edge nodes, allowing for a more accurate anticipation of future network conditions. These predictions are then used to formulate an ILP-based optimization model that determines the optimal placement of SFCs to minimize deployment costs while considering constraints like delay, memory, CPU, and bandwidth. We implemented and evaluated the proposed approach in a real-time KubeEdge cluster consisting of one master node and eight edge nodes. The results demonstrate that the LSTM-ILP scheduler significantly improves SFC placement efficiency by effectively identifying underutilized nodes, leading to reduced response times and optimized resource usage. The LSTM-ILP scheduler improves response time by 12.5% compared to the default Kubernetes scheduler when handling 50,000 requests. © The Author(s), under exclusive licence to Springer Nature Singapore Pte Ltd. 2024.

关键词： Edge cloud Integer Linear Programming KubeEdge Long Short-Term Memory Network Function Virtualization Service Function Chains

来源：评论

学校读者我要写书评

暂无评论

Deep Recurrent Residual U-Net with Semi-Supervised Learning for Deforestation Change Detection

引用

SN computer science 2024年第7期5卷 1-15页

作者： Bidari, Indira Chickerur, Satyadhyan School of Computer Science and Engineering KLE Technological University Hubballi India

“Deforestation” refers to the systematic removal of trees from forests to facilitate significant human activities. Expansion of agriculture, infrastructure and logging are the main causes of deforestation. Deforestation causes greenhouse gas emissions, water pollution and loss of biodiversity. The forest area is gradually decreasing with time. To overcome the above challenges, detection of forest change is crucial. Recently, due to advances in satellite technology, deforestation images have become accessible. In this study, we proposed a deep recurrent residual network to detect how much forest has been cleared over a period of time. Initially, an image of a particular forest area at two different time periods is provided as input to the change detection technique. The input image is iteratively segmented using a novel recurrent residual U-net. It then aggregates these segmented results and provides input to a semi-supervised learning method. From this, the semi supervised MLP method detects how many changes the current image has undergone from the previous image. Our proposed work is implemented on the platform of Python. Our proposed task is evaluated in terms of precision, accuracy, Fscore, recall, specificity, and sensitivity. Our proposed work achieved a high accuracy rate of 94.26%. © The Author(s), under exclusive licence to Springer Nature Singapore Pte Ltd. 2024.

关键词： Change detection Deep learning Deforestation Python Segmentation Semi supervised MLP

来源：评论

学校读者我要写书评

暂无评论

Sentimental Analysis of Amazon Reviews using Map Reducer Technique in Hadoop 18

Sentimental Analysis of Amazon Reviews using Map Reducer Tec...

引用

18th INDIAcom;11th International Conference on Computing for Sustainable Global Development, INDIACom 2024

作者： Byali, Manasi S. Kanakaraddi, Suvarna Badagi, Tulsa Chikaraddi, Ashoka Kle Technological University Department of Computer Science Engineering Hubli India Kle Technological University School of Computer Science Engineering Hubli India

ISBN: (纸本)9789380544519

Sentiment analysis (SA) is an active and dynamic aspect of text mining that focuses on the automated analysis of subjectivity, opinions, and sentiments in textual material. The study explores new SA domains such as resource development, emotion recognition, and transfer learning. The study combines rule-based classification, machine learning, and supervised learning in a novel hybrid technique. The usefulness of this hybrid method is demonstrated through experiments on multiple datasets, including MySpace comments, product reviews, and movie reviews, resulting in improved classification efficacy as judged by both macro and micro-averaged criteria. The study also supports an additional semi-automatic approach in which classifiers cooperate to enhance overall performance. In this article, we have applied SVM, Decision tree, and Linear SVM to the Amazon review dataset and got the output in the form of performance metrics such as CPU-time and maximum memory usage. and we found that Depending on the linearity of the data, a linear SVM or a Decision Tree is more appropriate;select a linear SVM for clearly defined separations and a Decision Tree for intricate, non-linear interactions, customizing the model to the features of the dataset and the goals of the research. © 2024 Bharati Vidyapeeth, New Delhi.

关键词： Hadoop Linear Regression MapReduce RDBMS (Relational Database Management System) SVM (Support Vector Machine)

来源：评论

学校读者我要写书评

暂无评论

Multiclass Classification of Breast Lesion Using Histopathology Images 5th

Multiclass Classification of Breast Lesion Using Histopathol...

引用

5th International Conference on Computing and Network Communications, CoCoNet 2023

作者： Kulkarni, Swetha Desai, Shrinivas Baligar, Vishwanath P. Nirmala, S.R. School of Computer Science and Engineering KLE Technological University Hubballi India

ISBN: (纸本)9789819747108

Breast cancer ranks as the second most common cancer affecting women. Early prediction of breast lesions significantly enhances patient survival rates. In India, however, obtaining an early medical diagnosis poses considerable challenges. Often, by the time the initial screening takes place, the lesion has progressed to a grade 2 or higher level. The classification of different stages of breast lesion histopathology images is a complex task, as it may lead to confusion in interpreting certain lesion images. Most of the existing literature has extensively explored various neural network models for multiclass classification of different stages of breast lesion histopathology images. In such cases, pre-trained deep neural network models have been commonly utilized, but they have yielded comparatively lower accuracy. Therefore, in this paper, we propose a novel approach involving the fine-tuning of the standard Inception V3 model and optimizing hyperparameters for categorizing histopathology breast lesion images stained with the BACH dataset. To enhance the deep neural networks (DNN) performance, several data augmentation techniques are employed. Our proposed network architecture, utilizing the fine-tuned Inception V3 model, achieves an impressive 97% test accuracy, a 76% F1-score, 87% precision, and 80% recall. When compared with recently reported literature, our model exhibits slightly superior performance. © The Author(s), under exclusive license to Springer Nature Singapore Pte Ltd. 2025.

关键词： Diseases

来源：评论

学校读者我要写书评

暂无评论

Machine Learning Approach to Intrusion Detection: Performance Evaluation 2

Machine Learning Approach to Intrusion Detection: Performanc...

引用

2nd International Conference on Machine Learning and Data engineering, ICMLDE 2023

作者： Giraddi, Vishal Giraddi, Shantala Narayan, D.G. Bidaragaddi, Anupama Kanakareddi, Suvarna G. School of Computer Science and Engineering KLE Technological University Hubli India

Software-Defined Networking (SDN) simplifies network administration as well as reduces the need for manual configuration on individual devices. SDN is expected to remain a crucial technology in shaping the future of network infrastructure. However due to key vulnerabilities like Open Flow Protocol Weaknesses, Flow Table Poisoning, Weak authentication and authorization mechanisms, it is prone to intrusion attacks. In this paper, authors aim to build framework based on various machine learning models (ML) to capture anomaly packets. Four models Decision tree, Support Vector Classifier, Naïve Bayes Classifier and artificial neural networks are designed and trained to capture anomaly packets. While SVC achieved an impressive accuracy of 99%, Artificial Neural Networks (ANN) demonstrated satisfactory speed and efficiency, despite their accuracy being slightly lower than SVM's. ANN is then deployed in SDN simulated using Mininet. The model is able to detect all DOS traffic. Other types of attacks SQL Injection, U2R Attack, Buffer Overflow Attack, Probe Attack, Brute Force Attack could not be tested as there is provision to simulated these using Mininet. © 2024 Elsevier B.V.. All rights reserved.

关键词： Intrusion detection

来源：评论

学校读者我要写书评

暂无评论

Beyond Words: Exploring Co-Attention with BERT in Visual Question Answering 5

Beyond Words: Exploring Co-Attention with BERT in Visual Que...

引用

5th IEEE International Conference for Emerging Technology, INCET 2024

作者： Vibhashree, B.S. Kamble, Nisarga Karamadi, Sagarika Varur, Sneha Desai, Padmashree D. School of Computer Science and Engineering KLE Technological University Hubballi India

ISBN: (纸本)9798350361155

The Visual Question Answering project described in this work uses a multi-modal approach that combines Bidirectional Encoder Representations from Transformers for complex natural language question understanding, You Only Look Once for robust object detection, and a Hierarchical Co-Attention Mechanism for efficient integration of textual and visual data. The goal is to develop a thorough Visual Question Answering system that can respond to inquiries about visual information with precision and contextual relevance. While Bidirectional Encoder Representations from Transformers captures subtle question semantics, the You Only Look Once model makes efficient object detection possible. In order to handle the complexities of multi-modal information processing, the Hierarchical Co-Attention Mechanism improves the fusion of local and global characteristics. The evaluation criteria include co-attention fusion scores, object detection metrics, Natural Language Processing metrics for Bidirectional Encoder Representations from Transformers, and overall Visual Question Answering accuracy, among other pertinent dimensions. By taking into account both response correctness and the efficiency of individual components, the study seeks to obtain a sophisticated view of the system's performance. Open-ended evaluation criteria allow for a comprehensive assessment because of the Visual Question Answering assignment's varied nature. The results show that the system can intelligently answer to questions regarding visual content and demonstrate its potential for a wide range of practical applications. The project is open-sourced, encouraging community involvement and collaboration to push the boundaries of multi-modal question answering systems even farther. © 2024 IEEE.

关键词： Semantics

来源：评论

学校读者我要写书评

暂无评论

Severity Calculation of Tar Disease in Sorghum Using Deep Learning 9

Severity Calculation of Tar Disease in Sorghum Using Deep Le...

引用

9th IEEE International Conference for Convergence in Technology, I2CT 2024

作者： Varur, Sneha Maidur, Tarun Sharad, C.K. Maladkar, Swapnil Manjunath Bhajantri, Suresh H. Kle Technological University School of Computer Science And Engineering Hubballi India

ISBN: (纸本)9798350394474

Sorghum, a crucial cereal crop, is confronted with challenges from diseases such as Tar spot, Anthracnose and Rust impacting production. In response to the scarcity of relevant data, our project proposes a Deep Learning solution utilizing the ResNet architecture to classify Tar disease severity in sorghum leaves. Facing the challenge of limited data, we meticulously curated a benchmark dataset comprising healthy and affected sorghum leaves. To enhance our system's intelligence, we selected ResNet, a specialized form of Deep Learning known for understanding intricate leaf patterns and performing well with a small dataset. The proposed methodology incorporates image masking to emphasize disease-related features and severity assessment through automated detection and quantification of black spots. The project aims to deliver a robust and reliable tool for accurately identifying and assessing Tar disease severity in sorghum, providing valuable insights for effective agricultural management. © 2024 IEEE.

关键词： Tar

来源：评论

学校读者我要写书评

暂无评论

Revolutionizing IPL Ticketing with NFTs

Revolutionizing IPL Ticketing with NFTs

引用

2024 IEEE Conference on engineering Informatics, ICEI 2024

作者： Appannavar, Khushi M. Deshpande, Ananya Hiremath, Risheek V. Kurkuri, Sanjana Hiremath, Uma KLE Technological University School of Computer Science and Engineering Hubballi India

ISBN: (纸本)9798331505776

The Indian Premier League (IPL) is a widely popular cricket league hosted in India, known for its high-energy matches and star-studded players. Previously, getting IPL match tickets required long lines and physical visits to counters. With the advent of online platforms, fans can now easily purchase tickets through the official IPL website, making the ticketing process more convenient and accessible. However, the IPL ticketing system still faces significant challenges, including counterfeit tickets and scalping, which undermine the authenticity and integrity of the booking process. To address these issues, the study proposes a solution by integrating blockchain technology, particularly through NonFungible Tokens (NFTs), into the IPL ticket booking platform. The system utilizes the Ethereum blockchain to ensure the uniqueness, verifiability and transferability of IPL tickets. By leveraging smart contracts, organizers can create and manage unique NFTs for each ticket, preventing counterfeiting and ensuring a reliable record of transactions. A decentralized approach allows organizers, vendors and fans to interact directly with the blockchain, enhancing security and transparency. The system is designed to be efficient, with automated ticket issuance and transfer facilitated by smart contracts. Additionally, it promotes responsible and sustainable practices by reducing the need for physical tickets and encouraging digital event management. The following research contributes to the understanding of how blockchain and NFTs can transform the ticketing industry, fostering a more sustainable future. © 2024 IEEE.

关键词： Ethereum

来源：评论

学校读者我要写书评

暂无评论

Brain Tumor Detection and Classification Using Deep Learning Models

Brain Tumor Detection and Classification Using Deep Learning...

引用

International Conference on Power engineering and Intelligent Systems, PEIS 2024

作者： Pujar, Manjunath Kavanashree, H. Jitendra, M. Halemani, Shankaraling Handur, Vidya School of Computer Science and Engineering KLE Technological University Karnataka Hubballi India

ISBN: (纸本)9789819767090

Brain tumors represent a significant health concern globally, contributing to a high mortality rate. Timely detection and intervention are critical for improving patient outcomes. The proposed work focuses on brain tumor detection and classification using deep learning models. The system comprises key stages of data collection, preprocessing, feature extraction, and classification. Several deep learning models were evaluated for their performance. The results show that the DenseNet-121 model outperformed others with an accuracy of 97.69%. This proposed system offers a reliable and efficient approach to predicting brain tumor likelihood, serving as a valuable screening tool for early detection. This innovative system holds great promise as a reliable screening tool for early detection, poised to significantly enhance survival rates and mitigate the mortality associated with brain tumors. By offering medical practitioners an efficient tool for early disease detection, this intelligent model becomes a vital asset in the continuous efforts to address the impact of brain tumors on global health. © The Author(s), under exclusive license to Springer Nature Singapore Pte Ltd. 2025.

关键词： Contrastive 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：