Hi,

One thing you could do is you could open up the actually box file and
split the boxes manually.  There's a description on the main tesseract
training page.  [1]  Here is the excerpt

As you can see, the low double quote character has been expressed as
two single commas. The bounding boxes must be merged as follows:

    * First number (left) take the minimum of the two lines (197)
    * Second number (bottom) take the minimum of the two lines (496)
    * Third number (right) take the maximum of the two lines (214)
    * Fourth number (top) take the maximum of the two lines (508)

So I think in this case you would just have to keep your second and
third numbers the same, and then your lefts and right would be based
on the first box and then add the character width.

I see that your using trainingTesseract.py.  I'm using the same
script.  It works great and added a bit of functionality to spit out a
PDF with text placed over the image.  The split box functionality
though is not really working on my system it gives an error.  I would
like to fix it and then I could use it.  I"ll keep you posted.

[1] - http://code.google.com/p/tesseract-ocr/wiki/TrainingTesseract

On Sep 25, 7:07 am, Tenzin Dendup <[email protected]> wrote:
> Hi once again,
>
> I am trying to train tesseract for Dzongkha
> (http://en.wikipedia.org/wiki/Dzongkha_language) language. I am using
> tesseract-2.03 on Debian Squeeze. I have created a training image file
> in gray-scale. When I try to create the box file using "tesseract
> dzo1.tif dzo1 batch.nochop makebox" , I get some output which says:
>
> Using substitute bounding box at (562,2547)->(1376,2629)
> Using substitute bounding box at (988,2372)->(1830,2440)
> Using substitute bounding box at (560,1924)->(1374,2006)
> Using substitute bounding box at (222,1841)->(2135,1916)
> Using substitute bounding box at (220,1216)->(2132,1291)
> Using substitute bounding box at (291,675)->(1370,757)
>
> Afterwards, while trying to edit the box file using
> tesseractTrainer.py, I notice that some 20 clearly separated blocks
> are grouped into one box (Please see the attached file 1.png).
>
> Also, Some of the image blocks (characters) are not exactly fitting
> inside the bounding box (attached file 2.png).
>
> It would be very helpful if i could get some solutions to this or if
> this has been discussed earlier in the list, i would be grateful to
> get a link to it.
>
> Regards
> Tenzin
>
>  1.png
> 484KViewDownload
>
>  2.png
> 235KViewDownload
--~--~---------~--~----~------------~-------~--~----~
You received this message because you are subscribed to the Google Groups 
"tesseract-ocr" group.
To post to this group, send email to [email protected]
To unsubscribe from this group, send email to 
[email protected]
For more options, visit this group at 
http://groups.google.com/group/tesseract-ocr?hl=en
-~----------~----~----~----~------~----~------~--~---

Reply via email to