arrow
Return

Modular OCR Using Web Scraping Data

delete2026-01-01
delete0
PRE
AI
G
Guy Gisfan *
E
Eli David
N
Nathan S. Netanyahu
DOI:10.1007/978-3-032-09371-4_12delete
deleteOriginal
deleteShare
deleteSave
Abstract

Abstract

En 中文
This paper explores the effectiveness of traditional (modular) Optical Character Recognition (OCR) pipelines compared to Large Vision Language Models (LVLMs), particularly in terms of robustness, accuracy, and resource efficiency. We propose leveraging web-generated datasets to train OCR systems, highlighting the rich diversity of layouts, styles, and linguistic variations offered by web content. Our approach shows that web scraping with smart augmentations can generate diverse OCR training datasets for training modular OCR. Acquiring website screenshots for modular OCR training has yet to be explored and requires precise word localization for training the word detection model. This differs from many LVLMs, which are mostly trained end-to-end to extract full text. Experimental evaluations demonstrate that even though we trained our OCR pipeline on design website templates for developers rather than real public websites, we achieved competitive and even superior results with the state-of-the-art LVLM-based models and superior results in noisy and distorted scenarios, while requiring fewer computational resources for training and inference. These findings underscore the long-lasting relevance of modular OCR systems in diverse and resource-constrained settings. Although LVLMs present advantages in handling diverse and generalized tasks, their use is unnecessary for the OCR task, where modular pipelines excel in terms of efficiency and performance.
Keywords:
Optical Character Recognition (OCR)
Large Vision Language Model (LVLM)
Word detection
Word recognition
Web-based data generation

Journal

D
DOCUMENT ANALYSIS AND RECOGNITION - ICDAR 2025 WORKSHOPS, PT II
IF:
0
Papers:
22
Citations:
1

Organization

B
bar-ilan-university
Scholars:
2
Papers: 1
Citations: 0
Cited Papers

Cited Papers

A Synthetic Recipe for OCR
err2019-09-01
err0
PREAI
errDavid Etter; Stephen Rawls; Cameron Carpenter; Gregory Sell
errShare
errSave
EAST: An Efficient and Accurate Scene Text Detector
err2017-07-01
err0
errOAAI
errXinyu Zhou; Cong Yao; He Wen; Yuzhi Wang; Shuchang Zhou; Weiran He; Jiajun Liang
errShare
errSave
U-Net: Convolutional Networks for Biomedical Image Segmentation
err2015-11-18
err0
PREAI
errOlaf Ronneberger; Philipp Fischer; Thomas Brox
errShare
errSave
errShare
errSave
What Is Wrong With Scene Text Recognition Model Comparisons? Dataset and Model Analysis
err2019-10-01
err0
errOAAI
errJeonghun Baek; Geewook Kim; Junyeop Lee; Sungrae Park; Dongyoon Han; Sangdoo Yun; Seong Joon Oh; Hwalsuk Lee
errShare
errSave
Character Region Awareness for Text Detection
err2019-06-01
err0
errOAAI
errYoungmin Baek; Bado Lee; Dongyoon Han; Sangdoo Yun; Hwalsuk Lee
errShare
errSave
Enhancing Vision-Language Pre-Training with Rich Supervisions
err2024-06-16
err0
PREAI
errGao,Yuan; Shi,Kunyu; Zhu,Pengkai; Belval,Edouard; Nuriel,Oren; Appalaraju,Srikar; Ghadar,Shabnam; Tu,Zhuowen; Mahadevan,Vijay; Soatto,Stefano
errShare
errSave
researcher View more