labels2csv_jhlim

#!/bin/bash
#To prepare a Kaggle sumbission:
# Usage:
#  ./labels2csv_jhlim filelist.txt labelsfile.txt prediction.csv submission.csv

##################
#filelist=/home/jaehyun/kaggle/nationalDataScienceBowl/data/orig/filelist_test.txt
filelist=$1

cat $filelist | sed 's/.*\///' > filelist_test.txt.tmp

cut -d " " -f 1 filelist_test.txt.tmp  > filelist_test.txt.tmp2

sed -i '1s/^/image\n/' filelist_test.txt.tmp2

##################

#labelsfile=/home/jaehyun/kaggle/nationalDataScienceBowl/data/ndsb_labels.txt
labelsfile=$2

declare -A num2label
declare -A label2num
i=0
while read line
do
  index=$(echo $line | cut -d ' ' -f 1)
  label=$(echo $line | cut -d ' ' -f 2)
  num2label[$index]=$label
  label2num[$label]=$index
  #echo num2label\[$index\]=${num2label[$i]}
  #echo label2num\[$label\]=${label2num[$label]}
  let "i+=1"
done < $labelsfile

#echo i = $i
#echo \# of label = ${#num2label[@]}

if [ ${#num2label[@]} != $i ]
then
  echo something wrong with labels. label indices do not match with number of lines in the text file.
  exit
fi

replacementline=""
n=${#num2label[@]}-1
replacementline+=${num2label[0]}
for ((i=1; i<=$n; i++)) do
  replacementline+=','
  replacementline+=${num2label[$i]}
done

#echo $replacementline

#sed '1s/.*/'$replacementline'/' predictions.probs > predictions.probs.tmp
#sed -e '1s/^/'$replacementline'\n/' prediction.csv > prediction.csv.tmp
prediction=$3
sed -e '1s/^/'$replacementline'\n/' $prediction > prediction.csv.tmp

##############################################
#paste -d , filelist_test.txt.tmp prediction.csv.tmp > submission.csv
outfile=$4
paste -d , filelist_test.txt.tmp2 prediction.csv.tmp > $outfile

rm filelist_test.txt.tmp2
rm filelist_test.txt.tmp 
rm prediction.csv.tmp