I'm sure I'm making a beginner mistake here, but I'm struggling quite a bit.
I've built straight from source, both version 4.1.1 and 5.0.0 on Ubuntu
18.04, and Ubuntu 20.04(fresh install, never used, but properly updated).
All exhibit the same behavior. I installed all the dependencies following
Are any of these vertical fonts?
Encoding errors could be if the characters in training text are not in the
unicharset.
On Fri, Jan 8, 2021, 00:46 Kamui 7 wrote:
> Looks like that fixed bug #1. Now it is able to successfully create 400
> pages. Do you have any ideas as to why the other 2 errors
(base) ubuntu@tesseract-ocr-1:~/TEST$ tesseract rus.png - -l rus+eng
--tessdata-dir ~/tessdata_best
D 20:22 Э
5IN AROW 5IN AROW 5IN AROW
Translate this sentence Translate this sentence Translate this sentence
(0) Вопросы есть? (0) Вопросы есть? Вопросы есть?
Апу questions Any questions Any questi
I still fail to understand why Tesseract performs so poorly. Isn’t it made
for OCR in screenshots? Doesn’t it understand Russian at all?
On Monday, November 2, 2020 at 5:45:41 PM UTC+1 d-ka wrote:
> Well, that’d require much additional logic because the general layout
> entails quite a diverse
Looks like that fixed bug #1. Now it is able to successfully create 400
pages. Do you have any ideas as to why the other 2 errors are occurring?
On Thursday, January 7, 2021 at 11:28:12 AM UTC-6 shree wrote:
> Your training text file is only 175 lines, so the rendered image fits in 4
> pages. Yo
Unfortunately I am not aware of (maintained) python leptonica support (any
volunteers?), but you can directly use leptonica&tesseract via cffi in
python.
See some examples :
https://sk-spell.sk.cx/building-minimalistic-tesseract
https://github.com/zdenop/SimpleTesseractPythonWrapper/blob/master/Sim
I replaced the training text with the one from the official langdata repo
and now it seems to only produce 30 pages. Is there any place to get the
training text that the official jpn.traineddata was trained on?
I have also checked to make sure the fonts support english and japanese as
well
On
Your training text file is only 175 lines, so the rendered image fits in 4
pages. You need to use a larger text if you want more pages.
Also check that your fonts support both English and Japanese as the text
seems to have samples of both languages.
On Thu, Jan 7, 2021, 22:40 Kamui 7 wrote:
> I
Or you may have an old version of data/ben/checkpoints/ben_checkpoint
--
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-ocr+unsubscr...@googlegroups.com
I did a find command in the root directory and searched for the tesstrain
script. It could only find the script that i pulled from the latest
tesseract git repo. My training script calls that specific tesstrain script
using a relative path so it couldn't be an older version
On Thursday, January
Old versions of tesstrain.sh used to limit training to 3 pages. Looks like
you may have an old version in the path somewhere.
On Thu, Jan 7, 2021 at 10:17 PM Kamui 7 wrote:
> I have a script to train tesseract and I ran it on Arch Linux, Debian, and
> even a docker container and they all produce
I have a script to train tesseract and I ran it on Arch Linux, Debian, and
even a docker container and they all produce the same errors. I checked to
make sure the script is correct as well.
Bug 1:
This happens when tesstrain runs text2image. The max pages parameter does
not work at all. It en
Segmentation fault is usually if you are not using the tessdata_best model
as Start_model
On Thu, Jan 7, 2021, 20:13 Soumik Ranjan Dasgupta
wrote:
> Sorry, I attached the wrong log file. Please find the new one attached.
>
> On Thu, Jan 7, 2021 at 8:09 PM Soumik Ranjan Dasgupta <
> ranjansou...@
ModuleNotFoundError: No module named 'bidi
Install python-bidi
On Thu, Jan 7, 2021, 15:45 Soumik Ranjan Dasgupta
wrote:
> Hi Shreeshrii,
>
> I took your command exactly as it is and ran it (made sure the
> tessdata_best directory is present in $HOME
> with best ben.traineddata) and ran into an
can you suggest me with an alternate for leptonica for "python & windows"
On Thursday, January 7, 2021 at 1:42:28 AM UTC+5:30 zdenop wrote:
> try to play with the leptonica pixAutoPhotoinvert function[1].
> quick test with following C code snippets provided attached result:
>
> pix = leptonica.pi
Hi Shreeshrii,
I took your command exactly as it is and ran it (made sure the
tessdata_best directory is present in $HOME
with best ben.traineddata) and ran into an extremely weird error.
Here is the log:
find data/ben-ground-truth -name '*.gt.txt' | xargs cat | sort | uniq >
"data/ben/all-gt"
16 matches
Mail list logo
