Public | Automated Build

Last pushed: 2 years ago
Short Description
Ergonomic line-by-line transcription of scanned text.
Full Description


A web interface for creating ground truth for evaluating and training OCR.

Table of Contents


ocr-gt-tools allows editing
hOCR files,
such as those produced by the
tesseract or
ocropy OCR frameworks.


<img src="./doc/screenshots/screenshot.png" height="300"/>


  • Editing transcriptions of lines
  • Commenting on line and page level
  • Use standardized comment tags to mark common problems
  • Cheatsheet
  • Zoom in / Zoom out
  • Filter visible elements
  • Select multiple lines and apply tags.



About the code

The server-side code is written in Perl.

The frontend is written in HTML and Javascript.


  • Open 'ocr-gt-tools/index.html' with a browser
  • open in a second Window 'Page Previews' from Kitodo
  • Search the book from which you created the hOCR file
  • Drag and drop a image from the Kitodo 'Page Preview' Window to the Window with 'ocr-gt-tools/index.html'
  • The perl script ocr-gt-tools.cgi will create in the background all files, which takes a few seconds
  • with ajax a json objects will be returned to index.html
  • index.html will load with ajax the created 'correction.html' and 'anmerkungen.txt' inline
  • 'Speichern' will get active if you have written a comment or a text line


Expand the wiki

We are using the wiki to collect transcription hints for unusual
and frequent errors.

Pull Requests

Bug fixes, new functions, suggestions for new features and other user feedback
are appreciated.

The source code is available from
Please prepare your code contributions also on Github.

Bug reports

Please feel free to open
for any bug you
encounter and features you'd like to have.


This is free software. You may use it under the terms of the GNU General Public
License Version 3 (GPLv3). See LICENSE for details.

This project bundles other free software:

Docker Pull Command
Source Repository