tessearact-ocr ignore I and 0

[Vladimir Radnovic]

I have problem whit I and 0 ( zerro )
on picture you can see

segmentation clean filter ignore 0 ( zerro ) nad I

can someone help me you figure out what is problem... ?

thanks

Vladimir

[zdenko podobny]

If send screenshot like this, I guess nobody is intersting even testing of your problem...

Send final image you try to, describe how you try to OCR etc...

Zdenko

--
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 http://groups.google.com/group/tesseract-ocr.
To view this discussion on the web visit https://groups.google.com/d/msgid/tesseract-ocr/f3ab8d1d-d71f-4ea2-a7de-5ffc6015962f%40googlegroups.com.
For more options, visit https://groups.google.com/d/optout.

[Vladimir Radnovic]

Hi Zdenko
In attachme is real image...

I put screnshoot just to show what is problem

I make treaindata for Latin ( Cro,BIH,Serbian letters ) Š Đ Ž Č Ć 

Bat I have problem whit I and 0 ( in Serbian numberplates thay have special zero like you can see on image )

Can you send me you mail on pp

Thanks

Vladimir

[zdenko podobny]

1. You posted 2 times original (not pre-processed) images. Please post image you try to OCR. It is pretty boring and discouraging to get output like you.
2. I just cropped plate in image editor, cover state symbol and run (tesseract 3.03):
   tesseract fiat_.png - -psm 7 -l srp \
     -c tessedit_char_whitelist=KIČ0123456789 2>/dev/null
   output was this:
   KI 513 IČ
   As you can see on special 0 is problem if you remove noise. 

Unfortunately tesseract is not able to "add" additional symbol to existing, so you need to play a little with with training... Maybe you can try to train only digits, but not "one-by-one". Try to make groups of digits "013  513 310" as they on plate. Be careful so there are no columns on training image (vertical spaces). I have wrong experiences with such images.

If you create such training (e.g.srp_digits.traineddata), try to this command to see if it helps:

tesseract fiat_.png - -psm 7 -l srp+srp_digits \

  -c tessedit_char_whitelist=KIČ0123456789

And please share your experience ;-)

Zdenko

To view this discussion on the web visit https://groups.google.com/d/msgid/tesseract-ocr/5ff6a1c1-d535-4320-86b9-1d4c11759d72%40googlegroups.com.

[Vladimir Radnovic]

Hi Zdenko
I tray to train and in attach is txt file what I tray (tablice.txt) and srb.treaineddata

I make new "font" 100% like on plate... bat in segmentation I have problem he ignore I and 0 like on my first post

What I make wrong ?

does this tipe of picture is ok or I need to remove all noise from picture ? ( picture in attachment )

Thanks

Vladimir