A web interface for creating ground truth for evaluating and training OCR.
ocr-gt-tools allows editing hOCR files, such as those produced by the tesseract or ocropy OCR frameworks.
- 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.
See INSTALL.md.
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
We are using the wiki to collect transcription hints for unusual glyphs and frequent errors.
Bug fixes, new functions, suggestions for new features and other user feedback are appreciated.
The source code is available from https://github.com/UB-Mannheim/ocr-gt-tools. Please prepare your code contributions also on Github.
Please feel free to open issues 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 AFFERO General Public License (AGPL) version 3 or newer. See LICENSE for details.
This project bundles other free software:
- EB Garamond Font (SIL Open Font License)
- Font Awesome by Dave Gandy (SIL OFL 1.1, MIT)
- bootstrap (MIT)
- clipboard.js (MIT)
- handlebars.js (MIT)
- hocr-extract-images (Apache)
- jQuery (MIT)
- ocropus-gtedit (Apache)
- reset-css (Public Domain)