#!/bin/bash

# based on https://github.com/tesseract-ocr/tesseract/wiki/TrainingTesseract-4.00#fine-tuning-for--a-few-characters

# directory with training scripts - this is not the usual place
#   because they are not installed by default
tesstrain_dir=/usr/local/bin/tesseract-training/

# directory with the old 'best' training set
tessdata_dir=./tessdata

# fonts directory for this system
#fonts_dir=/Library/Fonts

# using macports pango in this case
fonts_dir=/opt/local/share/fonts


# fonts to use for training - not a huge set but we hope enough to
#    add the extra chars
fonts_for_training="'Helvetica' 'Helvetica Bold' 'Helvetica Medium Oblique' 'Helvetica Bold Oblique'  'Times'  'Times Bold' 'Times Medium Italic'  'Times Bold Italic' 'Courier' 'Courier Bold' 'Courier Italic' 'Utopia' 'Utopia Bold' 'Utopia Bold Italic' 'Terminal' 'Terminal Bold'  'New Century Schoolbook' 'New Century Schoolbook Bold' 'New Century Schoolbook Medium Italic' 'New Century Schoolbook Bold Italic' 'Lucida Medium' 'Lucida Medium Italic' 'Lucida Bold' 'Lucida Bold Italic'"

# a minimal set for fast tests
#fonts_for_training="'Helvetica' 'Times' 'Courier Bold'"

# fonts for computing evals of best fit model
fonts_for_eval="Helvetica"


# downloaded directory with language data -
#   like XXX.unicharset, eng directory, etc
# IMPORTANT - ADD THE NEW CHARS TO langdata/eng/eng.training_text with
#    about 15 instances per char
langdata_dir=./langdata



# output directories for this run
train_output_dir=./trained_plus_chars
eval_output_dir=./eval_plus_chars

# the output trained data file to drop into tesseract
final_trained_data_file=$train_output_dir/eng.traineddata.NEW

# fatal bug workaround for pango
export  PANGOCAIRO_BACKEND=fc

################################################################
# variables to set tasks performed
MakeTraining=no
MakeEval=no
MakeLSTM=no
RunTraining=no
BuildFinalTrainedFile=yes
################################################################




if [ $MakeTraining = "yes" ]; then
    echo "###### MAKING TRAINING DATA ######"

# the EVAL handles the quotes in the font list
eval $tesstrain_dir/tesstrain.sh  \
     --fonts_dir $fonts_dir \
     --fontlist $fonts_for_training \
     --lang eng \
     --linedata_only\
     --noextract_font_properties \
     --langdata_dir $langdata_dir \
     --tessdata_dir $tessdata_dir \
     --output_dir $train_output_dir
fi

# at this point, $train_output_dir should have eng.FontX.exp0.lstmf
# and eng.training_files.txt


# eval data
if [ $MakeEval = "yes" ]; then
 echo "###### MAKING EVAL DATA ######"
eval $tesstrain_dir/tesstrain.sh \
     --fonts_dir $fonts_dir\
     --fontlist $fonts_for_eval \
     --lang eng \
     --linedata_only \
     --noextract_font_properties \
     --langdata_dir  $langdata_dir \
     --tessdata_dir  $tessdata_dir \
     --output_dir $eval_output_dir

fi

# at this point, $eval_output_dir should have similar viles as
# $train_output_dir but for different font set

if [ $MakeLSTM = "yes" ]; then
    echo "#### combine_testdata to make lstm from previous trained set ####"
    $tesstrain_dir/combine_tessdata \
      -e $tessdata_dir/eng.traineddata  \
       $train_output_dir/eng.lstm
fi

# at this point, we should have $train_output_dir/eng.lstm

if [ $RunTraining = "yes" ]; then
    echo "#### training from previous optimum  #####"
    $tesstrain_dir/lstmtraining\
    --model_output    $train_output_dir/pluschars \
    --continue_from   $train_output_dir/eng.lstm \
    --traineddata     $train_output_dir/eng/eng.traineddata \
    --old_traineddata $tessdata_dir/eng.traineddata \
    --max_iterations 3600 \
    --train_listfile $train_output_dir/eng.training_files.txt
fi


if [ $BuildFinalTrainedFile = "yes" ] ; then
    echo "#### Building final trained file $final_trained_data_file d####"
    $tesstrain_dir/lstmtraining \
    --stop_training \
    --continue_from $train_output_dir/pluschars_checkpoint \
    --traineddata $train_output_dir/eng/eng.traineddata \
    --U  $train_output_dir/eng/eng.unicharset \
    --model_output $final_trained_data_file

fi


# now  $final_trained_data_file is substituted for installed
# trained file /usr/local/share/tessdata/eng.traineddata
# (but new chars are not recognized)