search for forum/issue tracker - there is explanation why LSTM can not
exact character box coordinates.
If you need exact character boxes IMO you need to use legacy engine (but
it could have other problems)
Zdenko
pi 12. 6. 2020 o 12:31 'Tariq Ahmad' via tesseract-ocr <
Many thanks for your reply - useful to know.
I now find that pytesseract is returning the wrong coordinates for
individual characters. For example, for this image (which has a 10pixel
border):
image_to_boxes returns:
A: 17 32 10 22
L: 17 32 24 33
etc
etc
These I believe are interpreted as
https://github.com/tesseract-ocr/tessdoc/blob/master/ImproveQuality.md#missing-borders
Zdenko
st 10. 6. 2020 o 18:50 'Tariq Ahmad' via tesseract-ocr <
tesseract-ocr@googlegroups.com> napĂsal(a):
> I cannot understand whyTessaract fails on this (cropped) image:
>
>
> Yet if i add a random
I cannot understand whyTessaract fails on this (cropped) image:
Yet if i add a random white border it works:
Can anyone shed any light please?
--
You received this message because you are subscribed to the Google Groups
"tesseract-ocr" group.
To unsubscribe from this group and stop
4 matches
Mail list logo