kur_ara does not have Arabic unicharset.

[Shreeshrii]

Please see details at

tesseract-ocr/tessdata#88 (comment)

tesseract-ocr/langdata#116

tesseract-ocr/tessdata_best#23

@jbreiden @AlexanderP - FYI - regarding problem with packaged traineddata for kur_ara.

[AlexanderP]

@Shreeshrii I understood correctly.
trainedata need to change in packages?
tesseract-ocr-kur-ara -> tesseract-ocr-kur
tesseract-ocr-kur -> tesseract-ocr-kur-ara

[Shreeshrii]

There is no traineddata for kur in tessdata_fast. I will unpack and convert the dawgs to word list and see if it is possible to correct kur_ara files. Please do not make any change yet.

…

On Sat 24 Mar, 2018, 12:06 PM Alexander Pozdnyakov, < ***@***.***> wrote: @Shreeshrii <https://github.com/Shreeshrii> I understood correctly. trainedata need to change in packages? tesseract-ocr-kur-ara -> tesseract-ocr-kur tesseract-ocr-kur -> tesseract-ocr-kur-ara — You are receiving this because you were mentioned. Reply to this email directly, view it on GitHub <#14 (comment)>, or mute the thread <https://github.com/notifications/unsubscribe-auth/AE2_o0HXWqF78k3jpUbGt-n-WlMK9Fwzks5thelYgaJpZM4S4nxt> .

[Shreeshrii]

@AlexanderP

tesseract-ocr-kur-ara -> tesseract-ocr-kur

Yes, the above change can be made. Currently kur_ara has Latin text only.

tesseract-ocr-kur -> tesseract-ocr-kur-ara

This cannot be done since there is no kur traineddata in tessdata_fast.

[Shreeshrii]

@jbreiden @theraysmith

Should I build kur_ara from the ara.traineddata eg. by replacing the wordlist?

Or is there an updated set of Arabic script traineddatas that can be uploaded before 4.0.0 release?

ref: tesseract-ocr/langdata#83 (comment)

I was going to push until I discovered a bug with the RTL word lists.
Then I also need to integrate this issues list, that I haven't looked at in a while, and rerun training.

[amitdo]

Maybe it should be 'kur_lat'.

[AlexanderP]

There is no traineddata for kur in tessdata_fast.
I will unpack and convert the dawgs to word list and see if it is possible
to correct kur_ara files.
Please do not make any change yet.

ok

[stweil]

Was this issue solved by the renaming?

[Shreeshrii]

kmr is Kurdish in Latin script. Renaming has fixed that issue.

kur was Kurdish in Arabic script in Tesseract3. We have still not restored kur or kur_ara.

[stweil]

So you suggest to restore https://github.com/tesseract-ocr/tessdata/blob/3.04.00/kur.traineddata to the master branch of tessdata?

[Shreeshrii]

That will not work since Arabic script in 3.04 relied on cube which is no longer in codebase.

…

On Thu, Dec 19, 2019, 19:54 Stefan Weil ***@***.***> wrote: So you suggest to restore https://github.com/tesseract-ocr/tessdata/blob/3.04.00/kur.traineddata to the master branch of tessdata? — You are receiving this because you were mentioned. Reply to this email directly, view it on GitHub <#14?email_source=notifications&email_token=ABG37IYSSSV6JYSTZESZLS3QZN7YPA5CNFSM4EXCPRW2YY3PNVWWK3TUL52HS4DFVREXG43VMVBW63LNMVXHJKTDN5WW2ZLOORPWSZGOEHJYGLI#issuecomment-567509805>, or unsubscribe <https://github.com/notifications/unsubscribe-auth/ABG37I5JBM674S5VNJE7JVDQZN7YPANCNFSM4EXCPRWQ> .

[Shreeshrii]

I will try to recreate using the wordlist and training text by fine-tuning.

…

On Thu, Dec 19, 2019, 19:59 Shree Devi Kumar ***@***.***> wrote: That will not work since Arabic script in 3.04 relied on cube which is no longer in codebase. On Thu, Dec 19, 2019, 19:54 Stefan Weil ***@***.***> wrote: > So you suggest to restore > https://github.com/tesseract-ocr/tessdata/blob/3.04.00/kur.traineddata > to the master branch of tessdata? > > — > You are receiving this because you were mentioned. > Reply to this email directly, view it on GitHub > <#14?email_source=notifications&email_token=ABG37IYSSSV6JYSTZESZLS3QZN7YPA5CNFSM4EXCPRW2YY3PNVWWK3TUL52HS4DFVREXG43VMVBW63LNMVXHJKTDN5WW2ZLOORPWSZGOEHJYGLI#issuecomment-567509805>, > or unsubscribe > <https://github.com/notifications/unsubscribe-auth/ABG37I5JBM674S5VNJE7JVDQZN7YPANCNFSM4EXCPRWQ> > . >

[Shreeshrii]

https://github.com/Shreeshrii/tesstrain-ckb ckb is the preferred prefix rather than kur_ara My finetuned training gives improved results compared to official ara and script/Arabic traineddata on the synthetic eval set. On Fri, Dec 20, 2019 at 8:05 PM Shree Devi Kumar <shreeshrii@gmail.com> wrote:

…

I will try to recreate using the wordlist and training text by fine-tuning. On Thu, Dec 19, 2019, 19:59 Shree Devi Kumar ***@***.***> wrote: > That will not work since Arabic script in 3.04 relied on cube which is no > longer in codebase. > > > > On Thu, Dec 19, 2019, 19:54 Stefan Weil ***@***.***> wrote: > >> So you suggest to restore >> https://github.com/tesseract-ocr/tessdata/blob/3.04.00/kur.traineddata >> to the master branch of tessdata? >> >> — >> You are receiving this because you were mentioned. >> Reply to this email directly, view it on GitHub >> <#14?email_source=notifications&email_token=ABG37IYSSSV6JYSTZESZLS3QZN7YPA5CNFSM4EXCPRW2YY3PNVWWK3TUL52HS4DFVREXG43VMVBW63LNMVXHJKTDN5WW2ZLOORPWSZGOEHJYGLI#issuecomment-567509805>, >> or unsubscribe >> <https://github.com/notifications/unsubscribe-auth/ABG37I5JBM674S5VNJE7JVDQZN7YPANCNFSM4EXCPRWQ> >> . >> >

--

____________________________________________________________ भजन - कीर्तन - आरती @ http://bhajans.ramparivar.com