Tessercat 4.0 korean detecting chinese

[Fanatico]

I'm running tesseract with the "-l kor" param but it is detecting some chinese characters, the image really have 3 chinese characters but none of them is returning correctly (and I'm not expecting them to return correctly) but the others korean characters are being recognized as chinese characters

tesseract teste_kor.tif teste_kor -l kor --oem 3 --psm 6

Any idea of how to fix it?

Result:

1 화

서 05)

수 마 0 뜨 \) 에 사 로 잡혀 눈 을 도 저

히 뜰 수가 없다.

힘 을 내 도 겨우 반 개 하는 것이 고

작 . 그 이상 움직일 수가 없었다.

" 아 ‥…. 7

苗 朮 習 趾 葉 刁 估 舍 點 選 們 同 對 刀

려 소 리 를 낸다. 하지만 신 음 에 가

까운 목 소 리 만 홀 러 나 올 뿐이었다.

“장로 Q 全 程 ::: 가 시 면 ‥.”

[ShreeDevi Kumar]

Which traineddata are you using?

Use combine_tessdata and extract the config file to see if chinese is included as sub language.

Also look at the lstm-unicharset to see if the Chinese characters are included in it.

--
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-oc...@googlegroups.com.
To post to this group, send email to tesser...@googlegroups.com.
Visit this group at https://groups.google.com/group/tesseract-ocr.
To view this discussion on the web visit https://groups.google.com/d/msgid/tesseract-ocr/1e5142e1-d198-46d3-95ee-1a3206d1a2c4%40googlegroups.com.
For more options, visit https://groups.google.com/d/optout.

[Fanatico]

I used one traineddata that I created on removing the top layer from the kor.traineddata from "tessdata_best", after this I replaced this traineddata with the one from "tessdata_best" and got the same problem.

Yes, it include chi_tra as sublanguage

tessedit_load_sublangs chi_tra

lstm-unicharset only has corean characters

[ShreeDevi Kumar]

Please remove the sub language line from config file, and use combine tessdata to overwrite it.

Right now it seems to be using chi_tra also.

--

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-oc...@googlegroups.com.
To post to this group, send email to tesser...@googlegroups.com.
Visit this group at https://groups.google.com/group/tesseract-ocr.

To view this discussion on the web visit https://groups.google.com/d/msgid/tesseract-ocr/0d50ee2b-b5d4-4c73-a45b-d5245403ad04%40googlegroups.com.

[Fanatico]

It worked, thanks.

Any reason for this chi_tra there?

[ShreeDevi Kumar]

Leftover from 3.04, my guess.

To view this discussion on the web visit https://groups.google.com/d/msgid/tesseract-ocr/8496ad57-f7eb-426c-a4ae-5d365c56bc96%40googlegroups.com.

[ShreeDevi Kumar]

For Korean, please check whether adding the following lines to config, improves your results further.

#Fixes https://github.com/tesseract-ocr/tesseract/issues/1009

preserve_interword_spaces 1

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

On Mon, Apr 9, 2018 at 1:45 PM, ShreeDevi Kumar <shree...@gmail.com> wrote:

Leftover from 3.04, my guess.

On Mon 9 Apr, 2018, 12:52 PM Fanatico, <fanati...@gmail.com> wrote:

It worked, thanks.

Any reason for this chi_tra there?


On Monday, 9 April 2018 03:24:44 UTC-3, shree wrote:

Please remove the sub language line from config file, and use combine tessdata to overwrite it.

Right now it seems to be using chi_tra also.

On Mon 9 Apr, 2018, 11:48 AM Fanatico, <fanati...@gmail.com> wrote:

I used one traineddata that I created on removing the top layer from the kor.traineddata from "tessdata_best", after this I replaced this traineddata with the one from "tessdata_best" and got the same problem.

Yes, it include chi_tra as sublanguage

tessedit_load_sublangs chi_tra

lstm-unicharset only has corean characters

--
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-oc...@googlegroups.com.
To post to this group, send email to tesser...@googlegroups.com.
Visit this group at https://groups.google.com/group/tesseract-ocr.
To view this discussion on the web visit https://groups.google.com/d/msgid/tesseract-ocr/0d50ee2b-b5d4-4c73-a45b-d5245403ad04%40googlegroups.com.
For more options, visit https://groups.google.com/d/optout.

--
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+unsubscribe@googlegroups.com.

[Fanatico]

The conf from kor did already have it

[Fanatico]

After some research in Korean I found that they do use Chinese characters in their language, so it is correct to set Chinese as a sublanguage, the problem is that the kor.training_text doesn't have chinede letters, so the code is only training Korean and ignoring the Chinese, so if I tesseract on an image that has Korean and Chinese it is going to recognize some Korean characters as Chinese and some Chinese characters as Korean.