Scholarlink

<p>Pattern recognition and computer vision struggle with cursive and complex handwriting. Cursive writing, diverse writing styles, and subtle grammatical intricacies make Urdu difficult to read and write. Urdu OCR systems struggle with handwritten letters due to a lack of datasets and optimised recognition algorithms. A carefully selected 2596-page collection of 172,634-handwritten characters written by 649 writers, “MANUU: Handwritten Urdu OCR Dataset,” addresses these gaps. Isolated, starting, middle, and end letters, phrases, and numerals in diverse handwriting styles are included. VGG19, DenseNet201, and EfficientB0 form our unique ensemble model for 89.53% recognition accuracy. This work improves Urdu handwriting OCR efficiency and reliability across applications.</p>

错误：搜索内容不能为空，请输入英文关键词
错误：关键词超出字数限制，请精简
高级检索

Exploring Enhancing Pattern Recognition Techniques for Urdu Handwritten Text: A MANUU Dataset Perspective

  • 期刊论文

  • 发表日期：

    2025年6月16日

查看全文
  • Shaik Moinuddin Ahmed,
  • Abdul Wahid

摘要

Pattern recognition and computer vision struggle with cursive and complex handwriting. Cursive writing, diverse writing styles, and subtle grammatical intricacies make Urdu difficult to read and write. Urdu OCR systems struggle with handwritten letters due to a lack of datasets and optimised recognition algorithms. A carefully selected 2596-page collection of 172,634-handwritten characters written by 649 writers, “MANUU: Handwritten Urdu OCR Dataset,” addresses these gaps. Isolated, starting, middle, and end letters, phrases, and numerals in diverse handwriting styles are included. VGG19, DenseNet201, and EfficientB0 form our unique ensemble model for 89.53% recognition accuracy. This work improves Urdu handwriting OCR efficiency and reliability across applications.