The source image is really large, maybe try downsizing it to 800x450 or so.
art
From: tesseract-ocr@googlegroups.com [mailto:tesseract-ocr@googlegroups.com] On
Behalf Of S
Sent: Friday, December 02, 2016 5:43 PM
To: tesseract-ocr
Subject: [tesseract-ocr] Re:
Strangely, I've had some luck by removing the "-psm 3" argument and using
"-c tessedit_pageseg_mode=3" instead. There's no more output from OSD, and
I'm getting mostly good hOCR from the problem pages.
As a side note, it looks like the output from tesseract --print-parameters
may be out of
Just a guess, but it looks like the baseline / text angle isn't consistent
on those omitted lines. E.g. in "*S*ome of the resistance *d*uring," the
bottom of the *S* is noticeably higher than the bottom of the *d*, but by
last three words, there's no noticeable slope. By the "no replacements"
I can confirm I am experiencing the same issue described above. Entire
lines in (what should be) very readable images are skipped consistently.
--
You received this message because you are subscribed to the Google Groups
"tesseract-ocr" group.
To unsubscribe from this group and stop receiving
On Thursday, November 10, 2016 at 1:33:49 PM UTC+5:30, peiman F. wrote:
>
> is this work on cube and rtl languages!?
>
> Will be enabled soon. Latest updates listed below
1 Training with font_properties enabled
2 Training Image zoom strengthened
3 Dictionary editing enabled
4 progress-bar
I did a lot of work experimenting with trying to recognise this kind of
text and never got it to a satisfactory level - try Google Cloud Vision.
On 1 December 2016 at 23:44, Ni Min wrote:
> I have download all eng related training data from github and tried with
> the
I have download all eng related training data from github and tried with
the following command but it can't recognize successfully.
I am new to tesseract.
anybody can give some hint?
thanks
tesseract --tessdata-dir ~/tessdata test11.jpg result -l eng -psm 3
--
You received this message
*The immediate problem:*
I'm using tesseract (3.04.1) via command-line to generate hOCR as follows:
tesseract files/test-cases0001.tif files/test1-psm3 -l eng -psm 3 hocr
But I'm seeing output that seems to be from OSD:
> OSD: Weak margin (2.88) for 29022 blob text block, but using orientation
8 matches
Mail list logo