Re training with Tesseract 3.x

[Sarasi Lalithsena]

Hi,

Is there a way to retrain an existing model in version 3.04? Let' say some one want to retrain an existing model with some specific data. I see this option is available for Tesseract 4.00.

Just wondering whether there is similar feature for version 3.043 or 3.04.

In case there is no such option, Is there any repository which contains all files and details such as fonts it be being trained on to recreate the existing model?

Thank you

Sarasi Lalithsena

[Shree Devi Kumar]

see

https://github.com/tesseract-ocr/langdata

https://github.com/tesseract-ocr/tesseract/blob/master/src/training/language-specific.sh

https://github.com/tesseract-ocr/tesseract/wiki/Training-Tesseract-3.03%E2%80%933.05

--
You received this message because you are subscribed to the Google Groups "tesseract-ocr" group.
To unsubscribe from this group and stop receiving emails from it, send an email to tesseract-oc...@googlegroups.com.
To post to this group, send email to tesser...@googlegroups.com.
Visit this group at https://groups.google.com/group/tesseract-ocr.
To view this discussion on the web visit https://groups.google.com/d/msgid/tesseract-ocr/49827128-9f88-4860-99bd-f600c9808d71%40googlegroups.com.
For more options, visit https://groups.google.com/d/optout.

--

____________________________________________________________
भजन - कीर्तन - आरती @ http://bhajans.ramparivar.com

[Sarasi Lalithsena]

Thanks a lot.

I followed the last link you posted to create simple models. While the 'langdata' has the training text, I was confused with how can I get to know all the fonts a model is trained to recreate the existing model before improving given that the number of fonts limited to 64. By looking at the language-specific.sh file, can I assume that the existing model is trained on the LATIN_FONTS? 

Thanks

Regards

Sarasi Lalithsena

To view this discussion on the web visit https://groups.google.com/d/msgid/tesseract-ocr/CAG2NduUR09wW4N8qBxJGVM%3Dd7YDMeLx5TF_QUBNJ-tzsj2MvEw%40mail.gmail.com.