Repository navigation
A Hebrew traineddata for ocrmypdf -l heb_rivok (Hebrew + English + digits)
#1784
RivoksLab
started this conversation in
Show and tell
Replies: 1 comment
|
Update: heb_rivok 1.0.1 is out and fixes the column order, so plain New SHA-256: |
0 replies
Sign up for free
to join this conversation on GitHub.
Already have an account?
Sign in to comment
Uh oh!
There was an error while loading. Please reload this page.
Uh oh!
There was an error while loading. Please reload this page.
For anyone OCRing Hebrew PDFs: we released
heb_rivok, a Tesseract model fine-tuned from tessdata_best'sscript/Hebrewon Hebrew and mixed Hebrew/English lines. Drop the file into your tessdata folder, as the language docs describe, then run:On whole pages (our measurements):
heb, 91.4% for tessdata_bestscript/Hebrew)heb+eng)On real document lines the character error rate drops from 4.7% (Ubuntu's
heb) to 1.4%.One Apache-2.0 file of 5.7 MB: https://huggingface.co/Rivok/tesseract-heb-rivok. SHA-256:
dd9ba506b4ec6e62119a1334949824137692eada686687e51ac10dbb2f7a3df8.Limits:
I'm the author (Rivok Labs); this post was written with AI help. Happy to answer questions.
Edited 2026-11-07: updated for version 1.0.1, which fixed the column order on multi-column pages.
All reactions