Open Vocabulary Word Recognition From Transcribed Bangla Texts

📄 arXiv: 2610.01134v1 📥 PDF

作者: Faias Satter, Sk. Md. Masudul Ahsan

分类: cs.CV, cs.LG

发布日期: 2026-10-01

备注: 6 pages, 4 figures, 5 tables. Accepted version of the paper published in the 2023 26th International Conference on Computer and Information Technology (ICCIT). Code: https://github.com/FaiasPromit/Optical-Character-Recognition-From-Handwritten-Bangla-Texts

期刊: 2023 26th International Conference on Computer and Information Technology (ICCIT), Cox's Bazar, Bangladesh, 2023

DOI: 10.1109/ICCIT60459.2023.10441393


💡 一句话要点

提出深度学习模型以解决孟加拉手写文本识别问题

🎯 匹配领域: 支柱三:空间感知与语义 (Perception & Semantics)

关键词: 光学字符识别 手写文本识别 深度学习 目标检测 孟加拉语 集成学习 非极大值抑制

📋 核心要点

  1. 现有的OCR系统在识别孟加拉手写文本时存在较大挑战,尤其是缺乏可靠的解决方案。
  2. 本研究提出了基于深度学习的手写孟加拉单词识别方法,采用三种目标检测模型进行训练和测试。
  3. 实验结果表明,集成模型在识别性能上表现最佳,F1-score达到92.61%,单词级别识别率为96.12%。

📝 摘要(中文)

光学字符识别(OCR)技术能够扫描纸质文档并提取文本,极大地简化了人们的工作。然而,现有的OCR系统在孟加拉语的应用上仍存在显著不足,尤其是在手写文本的识别方面。本研究旨在通过深度学习技术识别手写孟加拉单词图像。研究中采用了三种目标检测模型:基于MobileNetV2的SSD、基于InceptionResNetV2的Faster R-CNN,以及这两者的集成模型。为提高模型效果,研究引入了一种改进的非极大值抑制方法。通过对9841幅多样化手写风格的孟加拉单词图像进行训练和测试,结果显示集成模型的F1-score达92.61%,在单词级别的识别率为96.12%。

🔬 方法详解

问题定义:本研究旨在解决现有OCR系统在手写孟加拉文本识别中的不足,尤其是在识别准确性和可靠性方面的挑战。现有方法在处理多样化的手写风格时表现不佳,导致识别率低下。

核心思路:本研究通过深度学习技术,特别是目标检测模型,来提高手写孟加拉单词的识别精度。采用SSD和Faster R-CNN模型,并结合集成学习的方法,以期提升整体识别性能。

技术框架:研究的整体架构包括数据集的构建、模型的训练与测试、以及后处理阶段。首先,构建了包含9841幅手写单词图像的定制数据集;其次,训练了三种不同的目标检测模型;最后,应用改进的非极大值抑制方法来优化模型输出。

关键创新:本研究的主要创新在于引入了集成模型的概念,将SSD与Faster R-CNN结合,显著提升了识别性能。这种方法在处理复杂的手写文本时,能够更好地捕捉到不同书写风格的特征。

关键设计:在模型设计中,采用了改进的非极大值抑制算法,以减少误检和漏检的情况。此外,模型的训练过程中使用了特定的损失函数,以优化识别精度和召回率。

🖼️ 关键图片

img_0
img_1
img_2

📊 实验亮点

实验结果显示,集成模型的F1-score达92.61%,在单词级别的识别率为96.12%。与单独模型相比,集成模型在识别准确性上有显著提升,展示了深度学习在手写文本识别中的潜力。

🎯 应用场景

该研究的成果可广泛应用于教育、文档数字化和语言处理等领域,尤其是在需要处理孟加拉语手写文本的场景中。通过提高手写文本的识别率,可以有效促进信息获取和知识传播,具有重要的社会价值和实际意义。

📄 摘要(原文)

An optical character recognition (OCR) can scan a paper and extract text using technology, making people's jobs easier. While various OCR systems are available in the software industry, finding a reliable equivalent solution for Bangla takes much work. When it comes to handwritten texts, the situation is much more unusual. Recognizing words from word images is the most critical stage in any OCR process. It is the second stage after segmenting words from text pictures. If this stage fails, the overall performance of the OCR will be poor, regardless of how well the other phases perform. This study aims to recognize words using deep learning in a handwritten Bangla word image. Three object detection models, SSD with MobileNetV2, Faster R-CNN with InceptionResNetV2, and an ensemble model of these two, have been used to train and test handwritten word images. A modified Non-Maximum Suppression has been introduced to enhance the effectiveness of the models' results. A customized dataset of 9841 handwritten Bangla word images has been compiled, featuring diverse handwriting styles from various individuals. All three models' performances have been checked against the test dataset, and the ensemble model has been the most impressive, with an F1-score of 92.61%. Also, at the word level, the ensemble model correctly recognizes 96.12% of the words to some extent. The system can be further improved by introducing a post-processing phase to correct errors generated by the system.