Same question here. I see that the documentation on training Tesseract 4
makes some reference to manuscripts:
As with base Tesseract, there is a choice between rendering synthetic
training data from fonts, or labeling some pre-existing images (like
ancient manuscripts for example).
So, if I understand correctly, there is no support yet for training with
labelled pre-existing images ? The concept of font does not makes sense
with manuscripts, and what we can use in this case is just pairs of images
and text (transcription).
Best,
Jean-Baptiste Camps
Le lundi 12 mars 2018 10:59:41 UTC+1, shree a écrit :
>
> >I have an image and a text file with the line content for each line of
> manuscript text. The doc says what to do, but not how.
>
> >I first thought I'd need img/box files pairs, but it seems it was for
> Tesseract 3 and is now irrelevant...
>
> Tesseract4.0.0beta.1 does not officially support LSTM training from
> box/tif pairs.
>
> It uses box/tif pairs generated using the synthetic training data
> generation pipeline using a training_text and set of fonts, for making the
> lstmf files that are used by lstmtraining.
>
> langdata refers to the langdata repository under tesseract-ocr github
> repo. The files in it have not been updated for 4.0.0
>
>
>
>
> ShreeDevi
> ____________________________________________________________
> भजन - कीर्तन - आरती @ http://bhajans.ramparivar.com
>
> On Mon, Mar 12, 2018 at 2:00 PM, ShreeDevi Kumar <[email protected]
> <javascript:>> wrote:
>
>> Please try tesseract 4.0.0beta.1 with languages such as
>>
>> *enm* (English, Middle (1100-1500))
>>
>> and
>>
>> Fraktur script
>>
>> Also, look at the following project from a few years back
>>
>> http://emop.tamu.edu/outcomes/Franken-Plus
>>
>> ShreeDevi
>> ____________________________________________________________
>> भजन - कीर्तन - आरती @ http://bhajans.ramparivar.com
>>
>> On Mon, Mar 12, 2018 at 4:32 AM, Guillaume Desforges <[email protected]
>> <javascript:>> wrote:
>>
>>> Hi
>>>
>>> I want to try using Tesseract 4 for old manuscript languages ("The Song
>>> of Roland" and such).
>>>
>>> I have looked at
>>> https://github.com/tesseract-ocr/tesseract/wiki/TrainingTesseract-4.00
>>> but the steps are very unclear.
>>>
>>> I have an image and a text file with the line content for each line of
>>> manuscript text. The doc says what to do, but not how.
>>>
>>> I first thought I'd need img/box files pairs, but it seems it was for
>>> Tesseract 3 and is now irrelevant...
>>>
>>> So I guess my starting point is here :
>>> https://github.com/tesseract-ocr/tesseract/wiki/TrainingTesseract-4.00#tutorial-guide-to-lstmtraining
>>>
>>> There is no tool to create the lstm-recoder directly. Instead there is
>>>> a new tool, combine_lang_model which takes as input an input_unicharset
>>>> and script_dir(script_dir points to the langdata directory) and
>>>> optional word list files. It creates the lstm-recoder from the
>>>> input_unicharset and creates all the dawgs, if wordlists are provided,
>>>> putting everything together into a traineddata file.
>>>
>>>
>>> I don't really get this part. How do I make input_unicharset ? What is
>>> langdata?
>>>
>>> Thanks
>>>
>>> Guillaume Desforges
>>>
>>> --
>>> 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 [email protected] <javascript:>.
>>> To post to this group, send email to [email protected]
>>> <javascript:>.
>>> 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/fe1d68a2-76ce-4005-98ea-672710365517%40googlegroups.com
>>>
>>> <https://groups.google.com/d/msgid/tesseract-ocr/fe1d68a2-76ce-4005-98ea-672710365517%40googlegroups.com?utm_medium=email&utm_source=footer>
>>> .
>>> 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 [email protected].
To post to this group, send email to [email protected].
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/200db744-d010-4555-a4b7-86c64ba0b9bf%40googlegroups.com.
For more options, visit https://groups.google.com/d/optout.
