Any way to prevent contextual digits<->letters flipping ?

57 views
Skip to first unread message

Damien D

unread,
Jul 10, 2014, 2:15:10 AM7/10/14
to tesser...@googlegroups.com
Hi everyone,

tesseract seems to sometime use the closest characters to "guess" what will be the next one.

Let me explain that with an exemple:

I want to parse a picture that contains the following sequence of characters
SE3P-104168
but most of the time the output will be
SESP-104168

I believe tesseract is flipping the "3" to a "S" because the closest characters are alphabetical and it seems more "logical" to have a chain of alphabetical characters but I would like to keep the "3" !
I am also attaching the picture I am parsing so you can see that it really looks like a "3" and it is not close to a "S".

If anyone has an idea about solving that problem it would be quite helpful.
I am playing with psm 8 FYI.

Thank you.
blured.tif

Nick White

unread,
Jul 10, 2014, 10:56:54 AM7/10/14
to tesser...@googlegroups.com
Hi,

I haven't tried it, but quickly grepping around the source code
suggests setting the config variable "crunch_include_numerals" to
true might do the job.

Please let us know if that works.

Nick
> --
> 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/6620e9a4-55d7-49b4-91d5-5401906d1177%40googlegroups.com.
> For more options, visit https://groups.google.com/d/optout.


Reply all
Reply to author
Forward
0 new messages