Getting the model output from my trained data

[ZenMaster181]

Hi, I am new to this training tesseract.

 I have annotated the images and created .tif and .box files from JTessboxEditor  and also created my traineddata and unicharset files from the same tool.

As I have tried to get the net_specs from the below command

training/combine_tessdata -d tessdata/xxx.traineddata

However, It just gave me the results as 

/src/training$ combine_tessdata -d 1.traineddataVersion string:v4.0.0.201810301:unicharset:size=4447, offset=1923:inttemp:size=344340, offset=46394:pffmtable:size=511, offset=3489795:normproto:size=785, offset=34949013:shapetable:size=76, offset=35027523:version:size=15, offset=350351
How do I get the net_specs to generate the model out put?
I do have tried to create traineddata from the tesseract  git page commands, However, I do get the following errorWarning: properties incomplete for index 72 = JWarning: properties incomplete for index 73 = &Warning: properties incomplete for index 74 = %Warning: properties incomplete for index 75 = $Warning: properties incomplete for index 76 = ’Warning: properties incomplete for index 77 = #Warning: properties incomplete for index 78 = ~Warning: properties incomplete for index 79 = ^Config file is optional, continuing...Failed to read data from: tesstutorial/tessdata/my/my.configFailed to read data from: tesstutorial/tessdata/radical-stroke.txt
I have attached my traineddata and listed created from jtessbox editor
Please suggest me a methord to solve this issue. Thank you in advance

[Shree Devi Kumar]

You are mixing legacy Tesseract training and LSTM training.

The traineddata and other files from jtessboxeditor seem to be for the legacy engine.

--
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 view this discussion on the web visit https://groups.google.com/d/msgid/tesseract-ocr/5ac1f0de-a926-4d3f-8a1f-fcf1482ef91a%40googlegroups.com.

[ZenMaster181]

I am sorry to be a drag but, I do not know the difference, I do have .tif, .box, .lstmf and a ttf font file with me

Is it possible to create traineddata file from the above given files Using the below code?

src/training/tesstrain.sh --fonts_dir /usr/share/fonts --lang eng --linedata_only \
  --noextract_font_properties --langdata_dir ../langdata \
  --tessdata_dir ./tessdata --output_dir ~/tesstutorial/engtrain

Sorry for being a noob

[ZenMaster181]

I have also cloned the git as per your instructions in the

https://github.com/tesseract-ocr/tesseract/wiki/TrainingTesseract-4.00#fine-tuning-for-impact

"Follow the instructions below to do the first time setup for TessTutorial"

[Shree Devi Kumar]

If you have the box and tiff files from jtesseditor, you can use https://github.com/tesseract-ocr/tesstrain for training

However, training is needed only in special cases.

Have you tried with existing traineddata files?

--

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 view this discussion on the web visit https://groups.google.com/d/msgid/tesseract-ocr/9472e872-434e-4f38-b21c-aa39f648b636%40googlegroups.com.

--

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

[ZenMaster181]

Yes, I have tried to generate model output from the code mentioned on the page, However, the traineddata file which I have created from JtessboxEditor does not have the Net_ specifications.

I have tried to this code to get the version and net specs of my trained data file. this is the result

combine_tessdata -d 1.traineddata

Version string:v4.0.0.201810301:unicharset:size=4447,
 offset=1923:inttemp:size=344340, 
offset=46394:pffmtable:size=511, 
offset=3489795:normproto:size=785, 
offset=34949013:shapetable:size=76,
 offset=35027523:version:size=15, 
offset=350351

 offset=46394:pffmtable:size=511,
This does not specify net_specs, for temp I have used the net specs of eng.trained data however, It me another error

Deserialized Header failed : EBOS- 4718210.pdf-1.tif.lstmf(which was loaded from lstmf file list)

Please check the above file

[Shree Devi Kumar]

Please see https://github.com/tesseract-ocr/tesseract/blob/master/doc/combine_tessdata.1.asc

https://github.com/tesseract-ocr/tesseract/wiki/Data-Files-in-tessdata_fast#version-string--40000alpha--network-specification-for-tessdata_fast 

Your trainedadata file is of legacy format. It will NOT work for LSTM training. 

--

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 view this discussion on the web visit https://groups.google.com/d/msgid/tesseract-ocr/cdc136e4-b740-4ff0-aacf-564fb847bf69%40googlegroups.com.

[ZenMaster181]

I got that, However, When I tried to create the traineddata file from the code:

src/training/tesstrain.sh\

--fonts_dir /usr/share/fonts\
--lang eng --linedata_only \
 --noextract_font_properties \
--langdata_dir ../langdata \

 --tessdata_dir ./tessdata \
-output_dir ~/tesstutorial/engtrain
ERROR: Could not find training text file tesstutorial/langdata/eng/eng/eng.training_text

I have checked that the file exists in langdata directory
But I have a doubt, How do the trained data file will work on my .tif file If I am not giving any information to the above code?

[Shree Devi Kumar]

You are mixing many different approaches for training.

If you have box/tiff pairs, use makefile from tesseract-ocr/tesstrain

If you want to train from text and fonts, use tesseract-ocr/tesseract/src/tesstrain.sh

--

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 view this discussion on the web visit https://groups.google.com/d/msgid/tesseract-ocr/a8264f54-a7c5-4d2c-bfce-1c8b7e236a6e%40googlegroups.com.

[ZenMaster181]

Please make sure the TESSDATA_PREFIX environment variable is set to your "tessdata" directory.Failed loading language 'eng'Tesseract couldn't load any languages!Could not initialize tesseract.ERROR: Program tesseract failed. Abort.

Sorry this was the error after I have run the script
Should I upload my own font file or create it with any tool?

[ZenMaster181]

Thank's for the reply on this regard. I will try to follow the procedure.