检索结果-内蒙古大学图书馆

The impact of backbone selection in YOLOv8 models on brain tumor localization

Iran Journal of Computer Science 2025年 1-23页

作者： Ranjbarzadeh, Ramin Crane, Martin Bendechache, Malika School of Computing Faculty of Engineering and Computing Dublin City University Dublin Ireland ADAPT Research Centre School of Computer Science University of Galway Galway Ireland

This study investigates the utilization of the You Only Look Once (YOLOv8) deep learning framework for accurately identifying the location of brain tumors in medical imaging. We investigate the effects of model size and pretraining on the accuracy and computational efficiency of tumor detection by utilizing different setups of the YOLOv8 model. These setups include various configurations, ranging from very small to large, and can be pretrained on the COCO dataset or not. The experimental results, carried out on Google Colaboratory using NVIDIA Tesla T4 GPUs, show that pretrained models often achieve better performance by utilizing the extensive feature representations learned from the COCO dataset, resulting in increased precision in tumor location. For instance, the YOLOv8-XS model pretrained on COCO achieves an IoU of 0.278 and a Dice coefficient of 0.435, whereas its non-pretrained counterpart attains only 0.241 IoU and 0.388 Dice, indicating a 15% improvement in tumor localization accuracy. Similarly, pretrained YOLOv8-L achieves 0.269 IoU, outperforming standard object detection models such as Mask R-CNN (IoU: 0.212) and Faster R-CNN (IoU: 0.228). These results highlight the impact of pretraining on model performance, particularly for lightweight architectures, while also revealing diminishing returns for larger models. The research uncovers a subtle connection between the complexity of the model, pretraining, and the time required for training. It emphasizes the possible advantages and constraints of pretraining for various sizes of models. © The Author(s) 2025.

关键词： Brain tumor localization COCO dataset Deep learning Object detection Transfer learning

来源：评论

学校读者我要写书评

暂无评论

SGGAA: a robust adversarial generation based on synthesizing point cloud and mesh space

引用

Neural computing and Applications 2025年 1-23页

作者： Hu, Ruihan Yang, Rui Faculty of Computing Harbin Institute of Technology Harbin China School of Electrical and Information Engineering Tianjin University Tianjin China

3D vision recognition offers a significantly more robust tool for achieving machine cognition compared to traditional 2D vision techniques. However, similar to the vulnerabilities present in 2D vision, many 3D vision recognition applications, including point cloud classification models, are susceptible to degradation when confronted with imperceptible outliers. Outliers have a considerable impact on the performance and accuracy of models, compromising their overall effectiveness. Existing 3D point cloud attack and defense strategies neglect outliers and overlook the crucial impact of 3D printability in real-world scenarios. Neglecting this aspect hinders the consideration of potential manipulation or alteration of 3D objects in the physical realm, posing significant implications for the security and robustness of 3D vision systems. This manuscript presents a novel and effective targeted attack framework called synthesizing geometry and geodesic attack (SGGAA). It offers an alternative adversarial training algorithm that bridges the gap between the digital and physical domains. SGGAA introduces a novel generalized distance metric, facilitating the design of powerful and simple techniques for generating adversarial samples and improving the training process. It focuses on the regularizer loss object in synthesizing attack modes for mesh and point cloud spaces, utilizing Hausdorff distance, geometry distance, and geodesic distance losses. Comprehensive evaluations, including hyperparameter tuning, qualitative analysis, and quantitative experiments, were conducted to assess SGGAA’s performance. The results demonstrated its superiority, with high attack success rates, excellent geometric similarity scores, and outperformance of state-of-the-art attack strategies across various 3D defense techniques. © The Author(s), under exclusive licence to Springer-Verlag London Ltd., part of Springer Nature 2025.

关键词： Mesh generation

来源：评论

学校读者我要写书评

暂无评论

A Compact Tri-Port MIMO Antenna With High Isolation for Broadside and Conical Radiation

引用

IEEE Internet of Things Journal 2025年第9期12卷 12219-12225页

作者： Gao, Guo-Ping Bai, Jin-Ming Jin, Li-Jun Guo, Wen-Di Ma, Ge Hu, Bin Lanzhou University Gansu Provincial Key Laboratory of Wearable Computing School of Information Science and Engineering Lanzhou730000 China

Wearable Internet of Things (IoT) devices are widely used in many fields. Such as smart bracelets for health, smart watches for sports, smart safety helmets for industry, and so on. These devices make life easier and more efficient. Multiple-input–multiple-output (MIMO) antenna technology improves data transmission rates and communication reliability by using multiple antennas in a device, so as to meet the demand for real-time data transmission and multidevice connectivity in the IoT. A compact tri-port MIMO antenna with high isolation suitable for wearable applications is proposed. The antenna is designed to work in the frequency band of 2.45 GHz industrial, scientific, and medical (ISM). The antenna has two layers and three ports with low profile suitable for wearable applications. The upper layer is a circular patch with four shorting pins, which excites the TM01 mode with a conical radiation pattern. The lower layer is a circular patch inlaid in an annular patch, which respectively excite the TM11 mode and the TM31 mode with a broadside radiation pattern. To achieve high isolation, characteristic mode analysis (CMA) was conducted on the circular patch, the circular patch with four shorting pins, and the annular patch. The positions of the three ports are reasonably arranged based on mode diversity. The measured bandwidth of the port for conical radiation is 35 MHz (2.43–2.465 GHz), while the measured bandwidths of the two ports for broadside radiation are basically the same which is 50 MHz (2.417–2.467 GHz). The measured overall isolation exceeds 50 dB. © 2014 IEEE.

关键词： Wearable antennas

来源：评论

学校读者我要写书评

暂无评论

Detecting the symptoms of COVID-19 during pandemic environment using smart spectacle thermal images and deep capsule networks

引用

Multimedia Tools and Applications 2025年第8期84卷 4475-4501页

作者： Dwarakanath, B. Pandimurugan, V. Mohandas, R. Sambath, M. Baiju, B.V. Chinnasamy, A. SRM Institute of Science and Technology Ramapuram Chennai India Networking and Communications School of Computing SRMIST kattankulathur Chennai India Computational Intelligence School of Computing SRMIST kattankulathur Chennai India Computer Science and Engineering Hindustan Institute of Technology and Science Chennai India School of Computer Science and Engineering Vellore Institute of Technology Vellore India Department of Data Science and Business Systems School of Computing SRMIST kattankulathur Chennai India

The pandemic creates a more complicated providence of medical assistance and diagnosis procedures. In the world, Covid-19, Severe Acute Respiratory Syndrome Coronavirus-2 (SARS Cov-2), and plague are widely known pandemic disease desperations. Due to the recent COVID-19 pandemic tragedies, various medical diagnosis models and intelligent computing solutions are proposed for medical applications. In this era of computer-based medical environment, conventional clinical solutions are surpassed by many Machine Learning and Deep Learning-based COVID-19 diagnosis models. Anyhow, many existing models are developing lab-based diagnosis environments. Notably, the Gated Recurrent Unit-based Respiratory Data Analysis (GRU-RE), Intelligent Unmanned Aerial Vehicle-based Covid Data Analysis (Thermal Images) (I-UVAC), and Convolutional Neural Network-based Computer Tomography Image Analysis (CNN-CT) are enriched with lightweight image data analysis techniques for obtaining mass pandemic data at real-time conditions. However, the existing models directly deal with bulk images (thermal data and respiratory data) to diagnose the symptoms of COVID-19. Against these works, the proposed spectacle thermal image data analysis model creates an easy and effective way of disease diagnosis deployment strategies. Particularly, the mass detection of disease symptoms needs a more lightweight equipment setup. In this proposed model, each patient's thermal data is collected via the spectacles of medical staff, and the data are analyzed with the help of a complex set of capsule network functions. Comparatively, the conventional capsule network functions are enriched in this proposed model using adequate sampling and data reduction solutions. In this way, the proposed model works effectively for mass thermal data diagnosis applications. In the experimental platform, the proposed and existing models are analyzed in various dimensions (metrics). The comparative results obtained in the experiments just

关键词： COVID-19

来源：评论

学校读者我要写书评

暂无评论

HDeFC: Hierarchical Secure Fuzzy Deduplication Based on Fog computing

引用

IEEE Internet of Things Journal 2025年第9期12卷 12658-12668页

作者： Tang, Zehui Zeng, Shengke Han, Song Liu, Jun He, Mingxing Xihua University School of Computer and Software Engineering Chengdu610039 China Hangzhou City University School of Computer and Computing Science Hangzhou310015 China

Secure deduplication not only optimizes cloud storage but also prevents data leakage. However, traditional schemes are with high computation and communication costs to deal with large-scale multimedia data. To address this problem, we propose a fog computing-based fuzzy deduplication (HDeFC), a two-stage privacy-preserving deduplication protocol. Our approach prevents guessing attacks and single point of failure, besides that also implements noninteractive Proof of Ownership and label consistency verification. In addition, we also achieve the access permission for only legitimate users. Our performance evaluation shows that the hierarchical setting is efficient in deduplication. © 2014 IEEE.

关键词： Information leakage

来源：评论

学校读者我要写书评

暂无评论

A Green Cloud-Based Framework for Energy-Efficient Task Scheduling Using Carbon Intensity Data for Heterogeneous Cloud Servers

引用

IEEE Access 2025年 13卷 73916-73938页

作者： Beena, B.M. Prashanth, C.S.R. Manideep, Thotapalli Sri Surya Saragadam, Sneha Karthik, Garikipati Amrita School of Computing Amrita Vishwa Vidyapeetham Department of Computer Science and Engineering Bengaluru India University of Windsor WindsorON Canada

Managing modern data centre operations is increasingly complex due to rising workloads and numerous interdependent components. Organizations that still rely on outdated, manual data management methods face a heightened risk of human error and struggle to adapt quickly to shifting demands. This inefficiency leads to excessive energy consumption and higher CO2 emissions in cloud data centres. To address these challenges, integrating advanced automation within Infrastructure as a Service (IaaS) has become essential for IT industries, representing a significant step in the ongoing transformation of cloud computing. For data centres aiming to enhance efficiency and reduce their carbon footprint, intelligent automation provides tangible benefits, including optimized resource allocation, dynamic workload balancing, and lower operational costs. As computing resources remain energy-intensive, the growing demand for AI and ML workloads is expected to surge by 160% by 2030 (Goldman Sachs). This heightened focus on energy efficiency has driven the need for advanced scheduling systems that reduce both carbon emissions and operational expenses. This study introduces a deployable cloud-based framework that incorporates real-time carbon intensity data into energy-intensive task scheduling. By utilizing AWS services, the proposed algorithm dynamically adjusts high-energy workloads based on regional carbon intensity fluctuations, using both historical and real-time analytics. This approach enables cloud service providers and enterprises to minimize environmental impact without sacrificing performance. Designed for seamless integration with existing cloud infrastructures—including AWS, Google Cloud, and Azure—this scalable solution utilizes Kubernetes-based scheduling and containerized workloads for intelligent resource management. By combining automation, real-time analytics, and cloud-native technologies, the framework significantly enhances energy efficiency compared to traditional

关键词： Infrastructure as a service (IaaS)

来源：评论

学校读者我要写书评

暂无评论

Detection of diabetic retinopathy and age-related macular degeneration using DenseNet based neural networks

引用

Multimedia Tools and Applications 2025年第1期84卷 289-316页

作者： Singh, Manpinder Dalmia, Saiba Ranjan, Ranjeet Kumar School of Computing DIT University Dehradun248009 India CSE Department Thapar Institute of Engineering & Technology Patiala147004 India

The eyes are the organ of sight and one of the most highly developed sensory organs in our body which covers a larger part of the brain. One of the most common problems that are spreading from kids to adults is an eye disorder which can be defined as abnormal functioning of the eye that can lead to vision disturbance. Examples of major eye problems are cataracts, glaucoma, etc. Artificial Intelligence has helped benefit research in the medical field. Ocular diseases can be detected automatically through Computer Vision and Deep Learning models when high-quality medical eye fundus images are provided to them. Inspired by this, we proposed three deep learning models based on the DenseNet pre-trained model for Ocular Disease Detection. Since the amount of eye scans being performed is rapidly growing at a much faster rate than the interpretation of the scan results, our proposed models could help automate the process thus increasing speed and efficiency. The proposed model performed the detection with an accuracy of 77%, 86%, and 98% for Ocular Disease Detection, Diabetic Retinopathy, and Age-Related Macular Degeneration tasks respectively. © The Author(s), under exclusive licence to Springer Science+Business Media, LLC, part of Springer Nature 2024.

关键词： Computer vision

来源：评论

学校读者我要写书评

暂无评论

Emotion detection: Facial emotion detection using ML 4th

Emotion detection: Facial emotion detection using ML

引用

4th International Conference on Computational Methods in Science and Technology, ICCMST 2024

作者： Sharma, Sharad Chauhan, Ritik Vajpayee, Pooja School of Computing Science and Engineering Galgotias University Greater Noida India

ISBN: (纸本)9781032911571

Cognitive perception of images is an intense task, like guessing the truth of a thought or a mystery. In this process, we use different methods to solve the need to know the job. In recent years, emotional intelligence has received widespread attention due to its important role in improving human-computer interaction and emotional analysis. Because of its processes for extracting visual features, deep neural networks-particularly convolutional neural network (CNN)-are frequently employed in FER. A lot of research has been done on CNN, but only a few areas have solved the facial recognition (FER) problem. After many reports and studies, the shortcoming of most existing models is that they often consider the front profile and ignore the side profile of the current image, although different angular views are also important for the perspective model. This helps you make informed decisions about personal needs, providing support, or security threats. However, this proved to be too complex for computers, and feature extraction needed a wide range of image processing methods. Numerous machine learning techniques are appropriate for this particular purpose. To use machine learning for any kind of detection, an algorithm must first be trained and then tested on relevant data. This article examines some machine learning algorithms and strategies for extraction that aid in the analysis of human emotion. This review contains a comprehensive review of a variety of materials, methods, and measures commonly used in psychology research. It also highlights the challenges and limitations of current models, including dealing with uncertainties and cultural differences. Face recognition in large datasets poses additional challenges due to the diversity of images as well as the differences between side and front views. © 2025 the Author(s).

关键词： Convolutional neural networks

来源：评论

学校读者我要写书评

暂无评论

CRIS-b: A High-Speed Unified Modulo Reduction Algorithm and Hardware Architecture for CRYSTALS-Kyber 38

CRIS-b: A High-Speed Unified Modulo Reduction Algorithm and ...

引用

38th International Conference on VLSI Design, VLSID 2025

作者： Majumdar, Alip Shrestha, Rahul School of Computing and Electrical Engineering Indian Institute of Technology Mandi India

ISBN: (纸本)9798331522445

This paper presents highly reliable algorithm and high-speed hardware architecture for a unified modulo reduction for CRYSTALS-Kyber. This new architecture for modulo reduction is capable of operating at a maximum clock frequency of 351.62 MHz on Artix-7 FPGA platform, consuming a latency of 5 clock cycles, which is approximately 35.24% faster than state-of-the-art implementations. In this work, we have alleviated the size expansion problem of intermediate value of the conventional Barrett algorithm (which is one of the fastest and most reliable algorithm for modulo reduction, reported in literature) by suggesting a new technique, referred as 'Conditional R-index Switching for b-value Minimization (CRIS-b)'. We propose a unified modulo reduction design handling both positive and negative spectrum numbers for all arithmetic modulo operations in a butterfly unit of Number Theoretic Transform(NTT). © 2025 IEEE.

关键词： Quantum cryptography

来源：评论

学校读者我要写书评

暂无评论

Finding Identity through Gait 6

Finding Identity through Gait

引用

6th IEEE International Conference on Image Processing, Applications and Systems, IPAS 2025

作者： Srinivasan, Lavanya University of West London School of Computing and Engineering London United Kingdom

ISBN: (纸本)9798331506520

Human gait identification is the recognition of a person from a series of walking images. In contrast to fingerprint or iris-based identification methods, gait identification offers the significant benefit of remote execution. Gait detection is emerging as one of the most promising biometric identification techniques. Traditional methods to identify the identity of gait are Background subtraction, Gait Energy Image, Gait Entropy, etc., human gait identification is a potential new tool for identifying individuals beyond traditional methods. The objectives of this study were to develop an automated gait detection system and examine the unique aspects of gait. This study encompassed the following: gait predictions and estimation based on HRNet 17 joint positions (locating body joints of humans in an image/video), better predictions with higher confidence scores by eliminating poor predictions, and the 14 joint coordinates and 17 joint points are fed into Graph Convolutional Network by eliminating low confidence pose to classify labels. Semi-supervised learning on graph-structured data can be accomplished via Graph Convolutional Networks. Graph Convolutional Networks is a very effective version of convolutional neural networks, which function directly on graphs. The experimental results show that the Graphical Convolutional Network identifies gait subjects with higher accuracy of normal walking and carrying a backpack with 98% and 96% respectively. This work will be extended in the future by fusing face and gait to identify various gait subjects and the gait in forensic identifications. © 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

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：