检索结果-内蒙古大学图书馆

Natural Language processing (ICNLP), International conference on

作者： Chaoqi Song Ying Liu Jinglu He Center for Image and Information Processing Xi’an University of Posts and Telecommunications Xi’an China

ISBN: (数字)9798350349115

ISBN: (纸本)9798350349122

Few-shot image classification is a critical issue in the field of computer vision, facing challenges related to data scarcity and model generalization. Transformer models, representing self-attention mechanisms, have made significant strides in recent years in the domain of few-shot classification. This paper commences with an introduction to the background and challenges of few-shot classification, along with a description of the principles and structure of the Transformer model. Subsequently, the paper categorizes Transformer-based few-shot image classification methods into meta-learning-based, metric-learning-based, fine-tuning-based, and feature-enhancement-based approaches, whose theoretical foundations of each method are expounded and the comparative analysis of representative algorithms are also provided. Furthermore, the paper delves into prospective research directions in this field.

关键词： Surveys Representation learning Learning systems Computational modeling Neural networks Computer architecture Transformers Natural language processing Classification algorithms image classification

来源：评论

学校读者我要写书评

暂无评论

Abnormal image Classification Using Vision Transformer for Smart Agriculture 8

Abnormal Image Classification Using Vision Transformer for S...

引用

8th IEEE International conference on Engineering Technologies and Applied Sciences, ICETAS 2023

作者： Hafeez, Talha Shahid, Muhammad Farrukh Khanzada, Tariq J.S. Fast School of Computing FAST-NUCES Karachi Pakistan King Abdulaziz University Faculty of Computing and Information Technology Department of Information System Jeddah21589 Saudi Arabia Mehran Uet Computer Systems Engineering Departments Jamshoro Pakistan

ISBN: (纸本)9798350327090

Smart agriculture is an emerging innovative sector that harnesses technological advancements like sensor monitoring, image processing, soil quality evaluation, and automated water systems, all of which contribute towards enhanced crop yields and superior quality of produce. With the progression towards automated systems in smart agriculture, the need for a robust data infrastructure that assures data integrity is critical. This demands an integrated approach with anomaly detection algorithms and a robust classification technique to ensure an error-free data pipeline. This study employs two distinct datasets to achieve abnormal image classification using a vision transformer (ViT) and a convolutional neural network (CNN). The primary dataset, obtained through direct collection from cotton fields in Karachi and Balochistan, Pakistan, comprises high-resolution agricultural images captured using specialised cameras and Unmanned Aerial Vehicles (UAVs). The secondary dataset was sourced from a publicly available database. The two classification models, ViT and CNN, were applied to both datasets. These supervised models are primarily used to classify images as normal or abnormal. The results of the two classification models were also compared, with the ViT model surpassing the CNN model in performance and achieving the highest accuracy on both datasets. The Vision Transformer model proved to be robust for supervised data, showing remarkable accuracy of 96.00% on the public dataset, 99.18% on both camera based images and 97.56% on UAV images of collected dataset as compared to the CNN model's accuracy that are 82.00% on the public dataset and 98.36% on camera and 95.12% on UAV images of collected dataset. © 2023 IEEE.

关键词： Classification (of information)

来源：评论

学校读者我要写书评

暂无评论

image processing for Future Machine Learning Algorithm Applications on Infrared Thermography of Building Envelope systems 5th

Image Processing for Future Machine Learning Algorithm Appli...

引用

5th International conference on Building Energy and Environment, COBEE 2022

作者： Gertsvolf, David Horvat, Miljana Khademi, April Aslam, Danesh Berardi, Umberto Department of Architectural Science Toronto Metropolitan University TorontoON Canada Department of Electrical Computer and Biomedical Engineering Toronto Metropolitan University TorontoON Canada

ISBN: (纸本)9789811998218

This paper presents and discusses the development of a procedure required for the future application of machine learning (ML) in the review and investigation of infrared (IR) thermography images for building envelopes. A background on ML and existing applications within the building science industry is presented to capture an understanding of the possibilities available within this innovative method of evaluating deficiencies of building envelopes. Then, a breakdown of the groundwork and image processing required for the application of ML algorithms is reported. A procedure developed using a python coding environment is presented, demonstrating the process of taking an RGB IR image, grey scaling it and normalizing for the application of unsupervised ML algorithm. Similarly, for the purpose of supervised ML, the same approach is presented, followed by a procedure for the creation of segmentation masks using two types of approaches: temperature threshold and manually identified segments. A discussion of possible development of these new techniques concludes this paper. © 2023, The Author(s), under exclusive license to Springer Nature Singapore Pte Ltd.

关键词： Anomaly detection

来源：评论

学校读者我要写书评

暂无评论

Implementation of a Deep Learning Technique for Detection of Objects on Roads for Autonomous Vehicles

Implementation of a Deep Learning Technique for Detection of...

引用

International conference on Applied Electromagnetics, Signal processing and Communication (AESPC)

作者： Kakarla. Deepti S. Aruna Deepti Ch. Neetu K.R. Deepthi Dept of ECE Vasavi college of Engineering Hyderabad India

ISBN: (数字)9798331528171

ISBN: (纸本)9798331528188

Autonomous driving and vehicle safety things profoundly depend on object detection to guarantee precise an robust right-time performance under different environmental conditions. The effectiveness of this systems depends on precise and efficient processing algorithms. The manuscript aim is to contrast the correctness of three notable image processing algorithms in the context of auto driving: Convolutional Neural Networks (CNN), Retina Net, and You Only Look Once v8 (YOLOv8). The assessment is performed using the Roboflow and kaggle dataset. The first algorithm under assessment is the Convolutional Neural Networks(CNN), Retina Net, and You Only Look Once v8 (YOLOv8). Each algorithm's unique approach to object detection is reviewed, observing their strengths and weaknesses in accurately identifying things in complex road situations. A comparative analysis is performed using the Robo flow and Kaggle dataset, covering various real-world contexts. The algorithms are evaluated based on correctness, considering occurrences with and without occlusion and truncation. Computation time,a vital aspect for real-time applications, is also considered. The precision-recall curve is applied as a metric to determine the algorithms' overall performance. Retina Net, with its focal loss mechanism, is reviewed for its capability to manage occluded and truncated things and its influence on overall correctness. YOLOv8 emerges as a standout performer, displaying superior correctness in discovering challenging road target things under complex situations and adverse weather conditions. Its right-time deployment capabilities, combined with high correctness, position it as a compelling choice for autonomous driving systems. In conclusion, this comparative analysis shines a light on the strengths and deficiencies of notable image processing algorithms for object discovery in autonomous driving systems. YOLOv8 stands outas the most accurate and efficient algorithm, excelling in challenging real-

关键词： YOLO Accuracy Roads image processing Simulation Vehicle safety Retina Real-time systems Convolutional neural networks Autonomous vehicles

来源：评论

学校读者我要写书评

暂无评论

Resource-Constrained image processing for Scientific CubeSats 6

Resource-Constrained Image Processing for Scientific CubeSat...

引用

6th International conference on algorithms, Computing and systems, ICACS 2022

作者： Kozaris, Ioannis Gioulekas, Fotios Katsaros, Panagiotis School of Informatics Aristotle University of Thessaloniki Thessaloniki Greece Directorate of Informatics 5th Regional Health Authority Larissa Greece

ISBN: (纸本)9781450397407

CubeSat systems are preferred for short-term space missions, like those with a scientific objective, since they make it possible to complete them in an affordable cost. On the other hand, the resource-constrained nature of such systems poses significant challenges in software design, for the system payload. We focus on image compression and processing for scientific space missions, a vital functionality, due to the limited onboard storage capacity and the infrequent time periods, in which a satellite in orbit can send images to the ground, through a limited bandwidth connection. For CubeSat systems, the performance of image processing depends - even more than other space systems - on trade-offs, which are influenced by size, power and complexity constraints. In this context, we present the design challenges and the image compression/processing solution developed, for a university CubeSat built to carry on a biological experiment. All changes of the performance capability in the different parts of the payload image data chain (image resolution, onboard storage capacity, communication channel bandwidth and error characteristics) are taken into account. Moreover, any decision that increases the risk of incorrect reception of images from the experiment is weighted against the benefits of improved performance. We provide experimental results that show how our image processing solution resolves the associated trade-offs, while fulfilling the mission requirements. © 2022 ACM.

关键词： image compression

来源：评论

学校读者我要写书评

暂无评论

Simple Linear Iterative Clustering with Efficiency 1

引用

17th International conference on Intelligent Information Hiding and Multimedia Signal processing (IIH-MSP) / 14th International conference on Frontiers of Information Technology, Applications and Tools (FITAT)

作者： Liao, Nannan Liu, Hui Li, Cheng Ren, Xia Guo, Baolong Xidian Univ Inst Intelligent Control & Image Engn Xian Peoples R China

ISBN: (数字)9789811910579

ISBN: (纸本)9789811910579;9789811910562

As an instructive work to generate satisfactory superpixels, simple linear iterative clustering (SLIC), has become fundamental and popular in various computer vision tasks. In this work, the algorithm is reconsidered, and an integrated framework is proposed to further improve the running speed and segmentation performance. In the first stage, a fast convergence strategy for clustering is presented on SLIC to redistribute the seeds efficiently. This is done to initialize a set of clustering centers that is fairly representative of local information on the image plane. Then, a followup work of SLIC termed simple non-iterative clustering (SNIC) is utilized to process more accurate segmentation without any post-processing to enforce connectivity. Experimental results show that the framework could generate a synergetic effect and performs better than previous superpixel algorithms in a limited computational time.

关键词： Clustering Initialization Redistribution Superpixels

来源：评论

学校读者我要写书评

暂无评论

Leveraging NVIDIA Clara for Real-Time Cardiac image Segmentation and Diagnosis

Leveraging NVIDIA Clara for Real-Time Cardiac Image Segmenta...

引用

Innovative Computing, Intelligent Communication and Smart Electrical systems (ICSES), International conference on

作者： S. Navaneethan Goalla Hemanth R. Nikilkumar Department of Electronics and Communication Engineering Saveetha Engineering College Chennai Tamilnadu India

ISBN: (数字)9798331543624

ISBN: (纸本)9798331543631

Medical imaging is crucial for heart diagnosis, but outdated algorithms and hardware result in delayed processing and low accuracy. Using NVIDIA Clara, a GPU-accelerated platform, the study proposes real-time cardiac picture segmentation and diagnosis. The proposed system uses advanced DL algorithms built for cardiac imaging to enhance processing speed and accuracy. Employing rigorous data preprocessing and deep learning model able to perform precise partitioning of cardiac structures and positioning defects that lead to quick heart arrhythmia diagnose in a timely manner. It also substantially reduces processing times compared to older systems while improving the contour quality and diagnosability. The system achieves the following DSC values: 0.93 for the left ventricle; 0.92 for right; 0.90 for left atrium; 0.89 for right ventricle and 0.91 for the major blood vessels. With diagnostic accuracy scores higher than the current system Sensitivity 94.2%, Specificity 92.8%, and Accuracy 93.5%. The method has potential to enhance cardiac care by facilitating earlier interventions leading better patient outcomes.

关键词： Heart image segmentation Accuracy Sensitivity Protocols Hospitals Real-time systems Partitioning algorithms Standards Monitoring

来源：评论

学校读者我要写书评

暂无评论

Research on the Design of Robot image Sensing System Based on Artificial Intelligence

Research on the Design of Robot Image Sensing System Based o...

引用

IEEE International conference on Power, Intelligent Computing and systems (ICPICS)

作者： Li Kang Zhengshuo Liu Shandong Institute of Commerce and Technology Shandong Jinan China Lower Kent Ridge Road National University of Singapore Singapore

ISBN: (数字)9798350374315

ISBN: (纸本)9798350374322

This study focuses on the design and implementation of an artificial intelligence-driven robotic image perception system, aimed at enhancing robots' visual perception capabilities in complex environments. By exploring a cylindrical projection-based global calibration method, this research investigates its model, distortion correction algorithms, and viewpoint switching algorithms, offering an efficient and precise image processing solution. Furthermore, this paper provides an in-depth analysis of image anti-aliasing algorithms, including nearest neighbor interpolation, bilinear interpolation, and bicubic interpolation, and conducts a detailed comparison among these three algorithms. The section on panoramic image rapid synthesis achieves efficient stitching and blending of images through SURF feature point extraction and matching. In terms of system design and testing, specific functional design schemes are proposed and comprehensively tested and analyzed, validating the effectiveness and practicality of the system. This study not only provides theoretical support and technological implementation for robotic vision systems but also serves as a robust reference for future developments in robotic vision technology.

关键词： Interpolation Machine vision Switches Robot sensing systems Feature extraction Stability analysis Sensors System analysis and design Testing Visual perception

来源：评论

学校读者我要写书评

暂无评论

Human-Centric Multimodal Machine Learning: Recent Advances and Testbed on AI-Based Recruitment

引用

SN Computer Science 2023年第5期4卷 434页

作者： Peña, Alejandro Serna, Ignacio Morales, Aythami Fierrez, Julian Ortega, Alfonso Herrarte, Ainhoa Alcantara, Manuel Ortega-Garcia, Javier Universidad Autónoma de Madrid Madrid 28049 Spain

The presence of decision-making algorithms in society is rapidly increasing nowadays, while concerns about their transparency and the possibility of these algorithms becoming new sources of discrimination are arising. There is a certain consensus about the need to develop AI applications with a Human-Centric approach. Human-Centric Machine Learning needs to be developed based on four main requirements: (i) utility and social good;(ii) privacy and data ownership;(iii) transparency and accountability;and (iv) fairness in AI-driven decision-making processes. All these four Human-Centric requirements are closely related to each other. With the aim of studying how current multimodal algorithms based on heterogeneous sources of information are affected by sensitive elements and inner biases in the data, we propose a fictitious case study focused on automated recruitment: FairCVtest. We train automatic recruitment algorithms using a set of multimodal synthetic profiles including image, text, and structured data, which are consciously scored with gender and racial biases. FairCVtest shows the capacity of the Artificial Intelligence (AI) behind automatic recruitment tools built this way (a common practice in many other application scenarios beyond recruitment) to extract sensitive information from unstructured data and exploit it in combination to data biases in undesirable (unfair) ways. We present an overview of recent works developing techniques capable of removing sensitive information and biases from the decision-making process of deep learning architectures, as well as commonly used databases for fairness research in AI. We demonstrate how learning approaches developed to guarantee privacy in latent spaces can lead to unbiased and fair automatic decision-making process. Our methodology and results show how to generate fairer AI-based tools in general, and in particular fairer automated recruitment systems. © 2023, The Author(s).

关键词： Automated recruitment Bias Biometrics Computer vision Deep learning FairCV Fairness Multimodal Natural language processing

来源：评论

学校读者我要写书评

暂无评论

2023 IEEE Open conference of Electrical, Electronic and Information Sciences, eStream 2023 - Proceedings

2023 IEEE Open Conference of Electrical, Electronic and Info...

引用

10th IEEE Open conference of Electrical, Electronic and Information Sciences, eStream 2023

ISBN: (纸本)9798350303834

The proceedings contain 19 papers. The topics discussed include: electromagnetic interferences resistant PI/I controller for electrical energy converters;opportunities for automated e-learning path generation in adaptive e-learning systems;development and research of miniature high precision modular rotary encoder kit based on dual optical sensors;dynamic characteristics study of a newly developed suppression system;methodology of low orbit satellites signal processing in the electromagnetically noisy environment;theoretical and experimental research of device for ultrafine particulate matter agglomeration;comparative analysis of the applicability of five clustering algorithms for market segmentation;research and development of information technology for determining shoe size by image;deposition of polydisperse ultrafine particles in a gas flow using an advanced electro-cyclone apparatus;toward bee motion pattern identification on hive landing board;and performance analysis of different ANN-based weight updating algorithms in forecasting short-term load demands in cluster microgrids.

关键词：

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：