检索结果-内蒙古大学图书馆

Handwritten numeral recognition through superimposition onto printed form

JOURNAL OF KING SAUD UNIVERSITY-COMPUTER AND INFORMATION SCIENCES 2022年第9期34卷 7751-7764页

作者： Shuvo, M. I. R. Akhand, M. A. H. Siddique, Nazmul Khulna Univ Engn & Technol Dept Comp Sci & Engn Khulna 9203 Bangladesh Ulster Univ Sch Comp Engn Intelligent Syst Belfast North Ireland

Handwriting is, eventually, a variation of the printed forms where the characters are little larger, smaller, angled and deformed than the printed forms. The small changes in handwriting define the parameters of the character to be recognized. Handwritten numeral recognition (HNR) poses significant challenges due to the deformations and other variations. This study proposes a new notion of HNR on the hypothesis that the handwritten numerals are distinct deformations of the printed forms, which leads to easier recogni-tion task with higher accuracy when superimposing handwritten numeral images onto the corresponding printed numeral images. In the proposed HNR, auto-encoder and convolutional auto-encoder have been adapted for the superimposing task that transform HNIs into PNIs, while neural network and convolu-tional neural network are employed for classification of PNIs. The superimposition method reduces the computational overhead. Moreover, this method employs simple pre-processing without feature extrac-tion whereas the traditional methods employ pre-processing, feature extraction, and recognition with machine learning tools, which add to the computational overhead. The performance of HNRSP has been evaluated for recognizing handwritten numerals of Bengali, Devanagari, and English on benchmark data -sets and the proposed system achieves 99.68%, 99.73%, and 99.62% recognition accuracy for Bengali, Devanagari, and English handwritten numerals, respectively.(c) 2022 Published by Elsevier B.V. on behalf of King Saud University. This is an open access article under the CC BY-NC-ND license (http://***/licenses/by-nc-nd/4.0/).

关键词： Handwritten numeral recognition Printed numeral Artificial neural network Convolutional neural network auto-encoder Convolutional auto-encoder

来源：评论

学校读者我要写书评

暂无评论

Recognition of Urdu Handwritten Words Using Deep Learning Techniques 20

Recognition of Urdu Handwritten Words Using Deep Learning Te...

引用

20th International Bhurban Conference on Applied Sciences and Technology (IBCAST)

作者： Ahmed, Toqeer Butt, Muhammad Usman Natl Ctr Phys NCP Artificial Intelligence Technol Ctr AITeC Islamabad Pakistan

ISBN: (纸本)9798350308266;9798350308259

Cursive hand script recognition is reported as a problematic task in words recognition due to variations in writing styles from person to person. Urdu script has related challenges comparable to Arabic script due to the complex ligatures. To cope with these issues, we proposed to use deep learning (DL) techniques i.e., transfer learning and ensemble learning. A novel dataset was generated which comprised 2496 images of Urdu handwritten words by 26 different writers and 288 images of computer-generated words. The whole dataset contained 96 classes. The data augmentation technique was used to add diversity and increase the dataset size to 13,920 images. To thoroughly evaluate the results of this study, a comparative analysis of the accomplishments was carried out. Among all the applied DL models, the ensemble learning model produced the best results as compared to state-of-the-art models. The proposed model moves handwritten words recognition to a higher level and can efficiently be automated.

关键词： Urdu handwritten words data augmentation outliers auto-encoder classification

来源：评论

学校读者我要写书评

暂无评论

Fast transformation of discriminators into encoders using pre-trained GANs

引用

PATTERN RECOGNITION LETTERS 2022年 153卷 92-99页

作者： Yu, Cheng Wang, Wenmin Macau Univ Sci & Technol Fac Informat Technol Macau 999078 Peoples R China Macau Univ Sci & Technol Int Inst Next Generat Internet Macau 999078 Peoples R China Macau Univ Sci & Technol State Key Lab Lunar & Planetary Sci Macau 999078 Peoples R China Chongqing Coll Elect Engn Chongqing 401331 Peoples R China

A typical generative adversarial network (GAN) consists of a generator and a discriminator. Currently, finely tuned deep GANs can synthesize high-quality (HQ) images via their generators. However, the discriminator in typical GANs is only able to distinguish true or fake images in the training process. Moreover, some synthesized images from GANs are imperfect, and we can not reconstruct images via GANs. In this paper, we revisit pre-trained GANs and offer a self-supervised method to quickly transform GAN's discriminators into encoders. We reuse parameters of the GAN's discriminator and replace its output layer, so it can be transformed into an encoder and output reformed latent vectors. The transformation makes GAN architecture more symmetrical and allows for better performance. Based on the method, GANs can be made to reconstruct synthesized images via GAN encoders. Compared to synthesized images, these reconstructions can maintain or even attain higher quality. The code and pre-trained models are available at https://***/disanda/GAN-encoder- Sym . (c) 2021 Elsevier B.V. All rights reserved.

关键词： Generative adversarial net (GAN) auto-encoder Image synthesis Image reconstruction

来源：评论

学校读者我要写书评

暂无评论

Predictive Data Replication for XR Applications in Multi-Connectivity Enabled mmWave Networks

Predictive Data Replication for XR Applications in Multi-Con...

引用

International Balkan Conference on Communications and Networking (BalkanCom)

作者： Javed, Muhammad Affan Liu, Pei Panwar, Shivendra S. NYU Tandon Sch Engn Dept Elect & Comp Engn Brooklyn NY 11201 USA

ISBN: (纸本)9798350339109

Emerging applications such as Extended Reality (XR) require a fundamental change in the way network architecture and functions are designed and optimized due to strict Quality of Service (QoS) requirements, especially with respect to hard deadlines. Fortunately, multi-connectivity enabled mmWave networks provide us with the capability of catering to such stringent constraints. In a multi-connectivity enabled network where users (UEs) connect to multiple base stations (gNBs) that can simultaneously and rapidly switch data connection between them for optimal data delivery, it is vital to carefully select which gNBs the data should be placed at pre-emptively. Selectively placing data at multiple base stations (gNBs) can lead to a network which is more resilient to blockages and which minimizes data plane interruptions. In this paper, we use a Deep Learning agent that encodes a complex system state and then uses a Deep Q-Network (DQN) to find the optimal selection of gNBs where the UEs' data should be placed. Our results show that using our Deep Learning agent, which essentially uses a vast amount of state information to pre-emptively predict the best selection of gNBs for future transmissions, delivers markedly better performance than other heuristic selection algorithms.

关键词： multi-connectivity mmWave handover blockages low latency XR applications deep reinforcement learning DQN auto-encoder

来源：评论

学校读者我要写书评

暂无评论

Deep Learning Based Solar Power Forecasting with Improved Representation

Deep Learning Based Solar Power Forecasting with Improved Re...

引用

2023 International Conference on Power Energy, Environment and Intelligent Control, PEEIC 2023

作者： Correa, Frederick Sidney Centre for Interdisciplinary Research in Business and Technology Chitkara University Institute of Engineering and Technology Chitkara University Punjab India

ISBN: (纸本)9798350357769

The growing number of solar power plants is making it a viable option for renewable energy sources to meet the energy needs of communities. Moreover, accurate prediction of output improves integration of these plants into the grid. Deep learning models, which can take advantage of high-performance processors, and the data, have the potential to improve solar power prediction. This study proposes an auto-encoder and Gated Recurrent Unit based method for improved forecasting. The effectiveness of this approach is compared to other state of art for predicting solar power over different time periods. The performance of these models is evaluated using different performance parameters. The outcome of the analysis validates the effectiveness of the proposed model. © 2023 IEEE.

关键词： auto-encoder Deep learning gated recurrent unit Machine learning solar power

来源：评论

学校读者我要写书评

暂无评论

De-Glared: Eyeglasses Glare and Reflection Removal Using Deep Neural Networks 26

De-Glared: Eyeglasses Glare and Reflection Removal Using Dee...

引用

26th International Conference on Computer and Information Technology, ICCIT 2023

作者： Sadik-E-Tawheed, Md. Ahmed, Tasnia Bhuiyan, Masud Hossain, Abir Shatabda, Swakkhar United International University Department of Computer Science and Engineering Bangladesh Bangladesh

ISBN: (纸本)9798350359015

Eliminating reflections from eyeglasses in digital images presents a formidable challenge necessitating the deployment of deep learning techniques. When photographs featuring individuals wearing eyeglasses are captured, reflections and glare from various angles can obscure the eyes, rendering them indistinct. This impairment adversely affects image quality and poses significant hurdles for computer vision tasks such as eye identification, segmentation, and facial recognition. Our research has devised an effective neural network based on U-Net, proficiently designed to eradicate glare and reflections from digital eyeglass-clad images, thereby yielding crisp and unobstructed eye visuals. In constructing our dataset, we meticulously assembled 2541 images as we encountered a scarcity of extensive datasets containing human images both with and without eyeglass reflections. This study involved a comprehensive analysis of various algorithms and these algorithms underwent rigorous training using this dataset. Our comprehensive analysis unveiled that U-Net exhibited unparalleled performance in contrast to alternative algorithms, boasting an impressive PSNR value of 38.48 dB. Furthermore, it has demonstrated proficiency in effectively mitigating reflection glare in images. © 2023 IEEE.

关键词： auto-encoder Eyeglasses Glare Removal Inception-Net Reflection Removal Res-Net U-Net

来源：评论

学校读者我要写书评

暂无评论

Reconstruction of sea surface temperature data from Sea satellite observation based on convolutional automatic encoder

Reconstruction of sea surface temperature data from Sea sate...

引用

Conference on Earth and Space - From Infrared to Terahertz (ESIT)

作者： Li, Yuheng Cao, Kaixiang Li, Yuxin Sun, Weifu Piesat Informat Technol Co Ltd Beijing 100195 Peoples R China State Ocean Adm Inst Oceanog 1 Qingdao 266061 Peoples R China SOA Beihai Offshore Engn Survey Inst Qingdao 266061 Peoples R China Hebei Univ Engn Sch Earth Sci & Engn Handan 056000 Peoples R China

ISBN: (纸本)9781510661196;9781510661202

Sea surface temperature (SST) is a key parameter for monitoring the ocean environment and understanding various ocean phenomena, and is a key indicator of climate change. Satellite remote sensing data is an important technical tool for SST research, but the availability of data is reduced due to the influence of clouds and aerosols, which generate a large amount of missing data. The data interpolation empirical orthogonal function (DINEOF) method has usability and accuracy in reconstructing missing grid points of remote sensing datasets. In this study, we use a convolutional self-encoder neural network, modified for model skip connection and fully connected layers, and introduce an attention mechanism to extract spatio-temporal features of SST data, called attention data interpolation convolutional autoencoder (A-DINCAE), to achieve the reconstruction of infrared radiometer SST data and compare A-DINCAE with DINCAE and DINEOF Reconstruction accuracy. The accuracy of the reconstruction results is quantitatively evaluated using cross-validation datasets and actual measurement data, and the study area is selected as the South China Sea with the boundaries of 103-121 degrees E and 0-23 degrees N. The validation results show that the reconstruction effect of the A-DINCAE model on the SST missing data is better than that of DINCAE, the accuracy of the reconstruction results is much higher than that of DINEOF, and the reconstruction results restore the main SST of the sea area physical features of the sea area. This paper confirms that the attention mechanism can improve the DINCAE spatio-temporal feature extraction ability, and the small-scale features of the missing data are restored under the same data reconstruction conditions, and the A-DINCAE is more efficient than DINEOF, and The accuracy of the improved model has been improved.

关键词： Sea surface temperature DINCAE DINEOF auto-encoder

来源：评论

学校读者我要写书评

暂无评论

Using Deep Siamese networks for trajectory analysis to extract motion patterns in videos

引用

ELECTRONICS LETTERS 2022年第9期58卷 356-359页

作者： Boyle, Jonathan Nawaz, Tahir Ferryman, James Univ Reading Sch Math Phys & Computat Sci Dept Comp Sci Reading Berks England Natl Univ Sci & Technol Dept Mechatron Engn Islamabad Pakistan

This paper investigates the use of Siamese networks for trajectory similarity analysis in surveillance tasks. Specifically, the proposed approach uses an auto-encoder as a part of training a discriminative twin (Siamese) network to perform trajectory similarity analysis, thus presenting an end-to-end framework to perform an online motion pattern extraction in the scene with an ability to incorporate new incoming trajectory(ies) incrementally. The effectiveness of the proposed method is evaluated on four challenging public real-world datasets containing both vehicle and person targets, and compared with five existing methods. The proposed method consistently shows better or comparable performance than the existing methods on all datasets.

关键词： image motion analysis Deep Siamese networks trajectory analysis online motion pattern extraction trajectory similarity analysis end-to-end framework surveillance tasks feature extraction object detection motion patterns auto-encoder image representation incoming trajectory discriminative twin network

来源：评论

学校读者我要写书评

暂无评论

Model-Driven End-to-End Learning for Integrated Sensing and Communication

Model-Driven End-to-End Learning for Integrated Sensing and ...

引用

IEEE International Conference on Communications (IEEE ICC)

作者： Mateos-Ramos, Jose Miguel Hager, Christian Keskin, Musa Furkan Le Magoarou, Luc Wymeersch, Henk Chalmers Univ Technol Dept Elect Engn Gothenburg Sweden Univ Rennes INSA Rennes CNRS UMR 6164IETR Rennes France

ISBN: (纸本)9781538674628

Integrated sensing and communication (ISAC) is envisioned to be one of the pillars of 6G. However, 6G is also expected to be severely affected by hardware impairments. Under such impairments, standard model-based approaches might fail if they do not capture the underlying reality. To this end, data-driven methods are an alternative to deal with cases where imperfections cannot be easily modeled. In this paper, we propose a model-driven learning architecture for joint single-target multi-input multi-output (MIMO) sensing and multi-input single-output (MISO) communication. We compare it with a standard neural network approach under complexity constraints. Results show that under hardware impairments, both learning methods yield better results than the model-based standard baseline. If complexity constraints are further introduced, model-driven learning outperforms the neural-network-based approach. Model-driven learning also shows better generalization performance for new unseen testing scenarios.

关键词： auto-encoder integrated sensing and communication joint radar and communications model-driven machine learning

来源：评论

学校读者我要写书评

暂无评论

An Advanced Learning Approach for Early Stage Diabetes Detection

An Advanced Learning Approach for Early Stage Diabetes Detec...

引用

2023 International Conference on Sustainable Communication Networks and Application, ICSCNA 2023

作者： Raju, Tella Kamalakar Kumar, A.V.Senthil Bharathiar University Tamil Nadu Coimbatore India Hindusthan College of Arts and Science Department of Mca Tamil Nadu Coimbatore India

ISBN: (纸本)9798350313987

Diabetes mellitus is a chronic metabolic disorder affecting millions of people worldwide. Early detection and intervention are critical for effectively managing diabetes and preventing complications. This study presents an Advanced Learning Approach (ALA) for the early detection of diabetes based on abnormal health conditions in patients. To identify potential diabetic patients in the early stages of the disease, the proposed method combines machine learning techniques with health data analysis. The study uses a diverse dataset of electronic health records gathered from various online sources. Feature extraction and selection techniques are used to process and refine the data, ensuring the inclusion of relevant health indicators. Machine learning (ML) models, such as Logistic Regression (LR) used as a training model and proposed approach is auto-encoders, are used to extract intricate patterns and relationships from data. Integrating multiple data modalities, including physiological, genetic, and lifestyle data, is the core innovation of this approach. This comprehensive approach enables a more accurate and complete assessment of a person's health status. Furthermore, it makes it easier to detect subtle anomalies that may indicate the early stages of diabetes, even before clinical symptoms appear. The proposed system's evaluation demonstrates its efficacy in early diabetes detection, with high accuracy rates in distinguishing diabetic from non-diabetic individuals. Furthermore, the model provides interpretable insights into the specific health factors influencing classification decisions, which can help healthcare professionals, make informed diagnostic and therapeutic recommendations. © 2023 IEEE.

关键词： Advanced Learning Approach (ALA) auto-encoder Logistic Regression (LR) Machine Learning (ML)

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：