#!/bin/bash
# BASH SCRIPT FOR ADDING CHARACTERS
################################################
# directory with training scripts
tesstrain_dir=/usr/local/bin/tesseract-training/

# directory with the old 'best' training set
tessdata_dir=~/Desktop/tesseract/tessdata

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

# fonts to use for training - not a huge set but we hope enough to
#    add the extra chars
#fonts_for_training="Magic R Regular"

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

# IMPORTANT - ADD THE NEW CHARS TO langdata/eng/eng.training_text with
#    about 15 instances per char
langdata_dir=~/Desktop/langdata_lstm

# output directories for this run
train_output_dir=~/Desktop/tesstutorial/trainplusminus
eval_output_dir=~/Desktop/tesstutorial/evalplusminus

# maxpages one wants to create on training
maxpages=1000

# maximum iterations for finetuning
maxiterations=3600

# 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=yes
MakeLSTM=yes
RunTraining=yes
BuildFinalTrainedFile=yes
################################################################

################################################
if [ $MakeTraining = "yes" ]; then
    echo "###### MAKING TRAINING DATA ######"
# GENERATE TRAINING DATA
rm $train_output_dir/*
~/../../usr/local/bin/tesstrain.sh \
  --fonts_dir $fonts_dir \
  --lang eng \
  --linedata_only \
  --langdata_dir $langdata_dir \
  --tessdata_dir $tessdata_dir \
  --fontlist "Arial" \
  --noextract_font_properties\
  --exposures "0" \
  --maxpages $maxpages \
  --save_box_tiff \
  --output_dir $train_output_dir
fi
################################################

################################################
if [ $MakeEval = "yes" ]; then
 echo "###### MAKING EVAL DATA ######"
# GENERATE TRAINING DATA
rm $eval_output_dir/*
~/../../usr/local/bin/tesstrain.sh \
  --fonts_dir $fonts_dir \
  --lang eng \
  --linedata_only \
  --langdata_dir $langdata_dir \
  --tessdata_dir $tessdata_dir \
  --fontlist "Times New Roman" \
  --noextract_font_properties \
  --exposures "0" \
  --maxpages $maxpages \
  --save_box_tiff \
  --output_dir $eval_output_dir
fi
################################################

################################################
# EXTRACT THE CURRENT MODEL OF THE BEST TRAINING DATA SET
if [ $MakeLSTM = "yes" ]; then
    echo "#### combine_testdata to make lstm from previous trained set ####"
~/../../usr/local/bin/combine_tessdata \
-e $tessdata_dir/eng.traineddata $train_output_dir/eng.lstm
fi
################################################

################################################
# FINETUNE THE CURRENT MODEL VIA THE NEW TRAINING DATA
if [ $RunTraining = "yes" ]; then
    echo "#### training from previous optimum  #####"
~/../../usr/local/bin/lstmtraining \
	--continue_from $train_output_dir/eng.lstm \
	--model_output $train_output_dir/plusminus \
	--traineddata $train_output_dir/eng/eng.traineddata \
	--old_traineddata $tessdata_dir/eng.traineddata \
	--train_listfile $train_output_dir/eng.training_files.txt \
	--max_iterations $maxiterations
fi
################################################

################################################
# COMBINE THE NEW BEST TRAINING DATA
if [ $BuildFinalTrainedFile = "yes" ] ; then
    echo "#### Building final trained file $final_trained_data_file ####"
~/../../usr/local/bin/lstmtraining --stop_training\
  --continue_from $train_output_dir/plusminus_checkpoint\
  --traineddata $train_output_dir/eng/eng.traineddata\
  --old_traineddata $tessdata_dir/eng.traineddata\
  --U  $train_output_dir/eng/eng.unicharset\
  --model_output $train_output_dir/eng.traineddata
fi
################################################