Hm.
I guess i just ship all 3 of them. *lol*
and add the text of the wiki to the readme.
Greetings,
Simon
Am 04.03.2018 um 18:43 schrieb ShreeDevi Kumar:
The traineddata files in tessdata_best are larger in size and OCR takes
more time. They are supposedly slightly more accurate, but there are
The traineddata files in tessdata_best are larger in size and OCR takes
more time. They are supposedly slightly more accurate, but there are no
definitive results provided by Ray.
tessdata_fast is what has been shipped for Debian and Ubuntu, so that seems
the way to go for doing OCR. These however
Hi ShreeDevi,
I have scraped the cygwin builds.
i am using now the builds i get from the appveyor builds which just
needs me to repackage the resulting stuff.
so tessdata_best isn't like the wiki says for better accuracy?
greetings,
Simon
Am 03.03.2018 um 05:12 schrieb ShreeDevi Kumar:
Hi S
> tessdata repo supports both --oem 0 and --oem 1, but the files are older
and may NOT be fully compatible with current code.
The results may vary depending on language and oem used. I have NOT tested
this much, since newer traineddata give better accuracy for Indian
languages.
ShreeDevi
___
Hi Simon,
If you are planning to package using 4.00alpha from master branch, please
use traineddata files from tessdata_fast. These are the files that have
been shipped for Ubuntu 18.04 and included in Debian. See
https://github.com/tesseract-ocr/tesseract/wiki for some links.
You can update the
Hi all,
Just looked at the git commits for tesseract and read that there has
been changes to the OCR modes.
are the 3 tessdata sets still valid?
tessdata_fast and tessdata_best have been updated so i guess those
reflect the latest developments but tessdata hasn't an update since
september.
i
6 matches
Mail list logo
