Text Detection based on Deep Learning

2023 International Conference on Innovative Data Communication Technologies and Application (ICIDCA) Pub Date : 2023-03-14 DOI:10.1109/ICIDCA56705.2023.10099672

A. Thilagavathy, Karuturi Hemanth Suresh, Katari Tejesh Chowdary, Meka Tejash, Vijayarao Lakshmi Chakradhar

{"title":"Text Detection based on Deep Learning","authors":"A. Thilagavathy, Karuturi Hemanth Suresh, Katari Tejesh Chowdary, Meka Tejash, Vijayarao Lakshmi Chakradhar","doi":"10.1109/ICIDCA56705.2023.10099672","DOIUrl":null,"url":null,"abstract":"Optical Character Recognition (OCR), a system for automatic recognition, is used in a variety of application sectors to convert text or images into editable data. With the distinct outline and size, printable or typed letters are easy to identify. However, this is never the case while dealing with handwritten text since each person's handwriting is unique. Handwritten text is difficult for OCR to read. This study presents a two-phase method for recognizing and classifying the input data. Convolutional Neural Networks (CNNs) were used as a framework for categorization. The process starts with the recognition of input text. The second stage is to determine the language used to generate the input number. Further, Python is used to analyze the handwritten characters in MNIST database. The simulation results show an error-free recognition rate and extremely high efficiency. The proposed work has achieved a 1.4% training loss, 99% testing accuracy and 99.6% training accuracy.","PeriodicalId":108272,"journal":{"name":"2023 International Conference on Innovative Data Communication Technologies and Application (ICIDCA)","volume":"14 1","pages":"0"},"PeriodicalIF":0.0000,"publicationDate":"2023-03-14","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":"0","resultStr":null,"platform":"Semanticscholar","paperid":null,"PeriodicalName":"2023 International Conference on Innovative Data Communication Technologies and Application (ICIDCA)","FirstCategoryId":"1085","ListUrlMain":"https://doi.org/10.1109/ICIDCA56705.2023.10099672","RegionNum":0,"RegionCategory":null,"ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":null,"EPubDate":"","PubModel":"","JCR":"","JCRName":"","Score":null,"Total":0}

引用次数: 0

Abstract

Optical Character Recognition (OCR), a system for automatic recognition, is used in a variety of application sectors to convert text or images into editable data. With the distinct outline and size, printable or typed letters are easy to identify. However, this is never the case while dealing with handwritten text since each person's handwriting is unique. Handwritten text is difficult for OCR to read. This study presents a two-phase method for recognizing and classifying the input data. Convolutional Neural Networks (CNNs) were used as a framework for categorization. The process starts with the recognition of input text. The second stage is to determine the language used to generate the input number. Further, Python is used to analyze the handwritten characters in MNIST database. The simulation results show an error-free recognition rate and extremely high efficiency. The proposed work has achieved a 1.4% training loss, 99% testing accuracy and 99.6% training accuracy.

查看原文本刊更多论文

基于深度学习的文本检测

光学字符识别(OCR)是一种自动识别系统，用于各种应用领域将文本或图像转换为可编辑的数据。可打印或打印的信件轮廓和大小都很明显，很容易识别。然而，在处理手写文本时，情况并非如此，因为每个人的笔迹都是独一无二的。手写文本很难被OCR识别。本文提出了一种两阶段识别和分类输入数据的方法。使用卷积神经网络(cnn)作为分类框架。这个过程从识别输入文本开始。第二阶段是确定用于生成输入数字的语言。进一步，使用Python对MNIST数据库中的手写字符进行分析。仿真结果表明，该方法具有良好的无错误识别率和极高的效率。所提出的工作实现了1.4%的训练损失，99%的测试准确率和99.6%的训练准确率。

本文章由计算机程序翻译，如有差异，请以英文原文为准。

求助全文

约1分钟内获得全文求助全文

来源期刊

2023 International Conference on Innovative Data Communication Technologies and Application (ICIDCA)

自引率

0.00%

发文量