# Normalization mode - 2, 1 - for unicharset_extractor and Pass through Recoder for combine_lang_model ifeq ($(LANG_TYPE),Indic) NORM_MODE =2 RECODER =--pass_through_recoder
On Thu, Apr 9, 2020 at 12:29 PM Shree Devi Kumar <[email protected]> wrote: > Unicharset will look like the following: > > द 1 34,72,192,192,100,122,0,0,99,114 Devanagari 11 0 11 द # द [926 ]x > र 1 58,64,192,192,84,119,0,0,81,110 Devanagari 12 0 12 र # र [930 ]x > ् 0 3,32,61,197,12,181,0,0,0,1 Devanagari 13 17 13 ् # ् [94d ] > श 1 61,64,192,195,128,148,0,12,130,147 Devanagari 14 0 14 श # श [936 ]x > य 1 63,64,192,192,114,142,0,0,111,133 Devanagari 15 0 15 य # य [92f ]x > त 1 61,64,192,192,112,135,0,0,110,126 Devanagari 16 0 16 त # त [924 ]x > ि 0 62,65,228,253,132,279,0,0,40,65 Devanagari 17 0 17 ि # ि [93f ] > प 1 63,64,192,192,98,126,0,0,97,119 Devanagari 18 0 18 प # प [92a ]x > ू 0 1,35,67,197,33,193,0,0,0,1 Devanagari 19 17 19 ू # ू [942 ] > ज 1 63,64,192,192,138,165,0,0,128,157 Devanagari 20 0 20 ज # ज [91c ]x > > You can unpack any of the existing traineddatas from tessdata_best or > tessdata_fast and check. > > combine_tessdata -u > > and looks at the lstm-unicharset in the components > > On Thu, Apr 9, 2020 at 12:15 PM Piyush Chandra <[email protected]> > wrote: > >> Thank you Shree for giving the overview. >> >> Could you please help me understand your last point? Your unicharset >> should have Unicode codepoints. what does that mean? any example would be >> helpful. I was actually using akshara (attached box fiile image) . >> >> >> >> On Thursday, 9 April 2020 09:02:43 UTC+5:30, shree wrote: >>> >>> devenagari.unicharset, Latin.unicharset and radical-stroke.txt >>> >>> The script unicharset are useful in setting character properties. For >>> most scripts they are already available in langadata_lstm. I don't think >>> they are mandatory for lstm training but by copying them once you can avoid >>> the warning messages. >>> >>> radical-stroke.txt is used only for CJK languages, but tesseract checks >>> for it during training process, so you need to make it available. >>> >>> For chattisgarhi, if training for as written in Devanagari, I will >>> suggest training from script/Devanagari.traineddata rather than English. >>> >>> Please note if you are starting from scratch, then you don't need a >>> starting traineddata. If you use one, then you are finetuning. >>> >>> Finally, you need to use the correct mode for Indic language with >>> unicharset_extractor. Your unicharset should have Unicode codepoints, not >>> akshara (consanant vowel sign combination). >>> >>> >>> >>> >>> -- >> 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 view this discussion on the web visit >> https://groups.google.com/d/msgid/tesseract-ocr/338f0a8e-d998-4411-bcb6-8d49dfbb4ab6%40googlegroups.com >> <https://groups.google.com/d/msgid/tesseract-ocr/338f0a8e-d998-4411-bcb6-8d49dfbb4ab6%40googlegroups.com?utm_medium=email&utm_source=footer> >> . >> > > > -- > > ____________________________________________________________ > भजन - कीर्तन - आरती @ http://bhajans.ramparivar.com > -- ____________________________________________________________ भजन - कीर्तन - आरती @ http://bhajans.ramparivar.com -- 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 view this discussion on the web visit https://groups.google.com/d/msgid/tesseract-ocr/CAG2NduWP%2B9G6%2BBPFxmDPhwLdYEy4OMAadS9%3DuV6gtuYW79wRiQ%40mail.gmail.com.
