Problem facing with tessearct training 4 with arabic

378 views
Skip to first unread message

Amir Raouf

unread,
Apr 26, 2018, 1:00:47 AM4/26/18
to tesseract-ocr
First The arabic is read by tesseract with good accuracy but NO DIGITS read so I decided to train only numbers with specific font I need

This is the question https://stackoverflow.com/questions/50029477/issue-with-training-tesseract-4-0

Any advice

ShreeDevi Kumar

unread,
Apr 26, 2018, 1:18:39 AM4/26/18
to tesser...@googlegroups.com
You are trying to train only digits but then using the unicharset which has these numbers only for compressing the wordlist (which uses Arabic alphabet)  to a 'dawg'.

The command you have used only creates the starter traineddata for LSTM training. Please follow the instructions given in the wiki page related to training tesseract4.


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

--
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-ocr+unsubscribe@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/8d352529-8cdf-4e83-ba96-691abbd74423%40googlegroups.com.
For more options, visit https://groups.google.com/d/optout.

abdelsal...@gmail.com

unread,
May 2, 2018, 4:57:53 AM5/2/18
to tesseract-ocr

Hi , I'm using tesseract 4 with vs 2017 . i have used with English characters first , now i started to include arabic as well . the thing is got weird characters even when i change the eng.traineddata to ara.traineddata.

i found out that it's the characters when the UTF8 code is treated as Hex code .
image
this is the image i want to recognize .
the result is
image
i convert the letters in arabic to UTF8 code using this website
https://r12a.github.io/app-conversion/
and when i take this code and converted as a hex code to character i get the same characters that tesseract showed me the first time.
"عبدالسلام مدي عبدالعزيز"

I think the problem is with UTF8 or might be my visual studio can't recognize Arabic letters or what

Reply all
Reply to author
Forward
0 new messages