diff options
Diffstat (limited to 'snippets')
| -rw-r--r-- | snippets/tessfortesting | 22 |
1 files changed, 22 insertions, 0 deletions
diff --git a/snippets/tessfortesting b/snippets/tessfortesting new file mode 100644 index 0000000..daec106 --- /dev/null +++ b/snippets/tessfortesting @@ -0,0 +1,22 @@ +#!/bin/sh + +# rename .gt.txt files to .txt +for i in *gt.txt; do n=`echo "$i"|sed 's/\.gt\.txt/.txt/g'`; mv "$i" "$n"; done + +# create boxes for all lines +for i in *png; do b=`basename "$i" .png`; python ../../generate_line_box.py -i "$b.png" -t "$b.txt" > "$b.box"; done + +# back up ground truth .txt (the lstm.train step wipes it) +for i in *txt; do cp "$i" "$i.bak"; done + +# create lstmf counterparts to all files +for i in *png; do b=`basename "$i" .png`; tesseract "$i" "$b" --psm 6 lstm.train; done + +# restore ground truth .txt +for i in *.bak; do b=`basename "$i" .bak`; mv "$i" "$b"; done + +# create list of lstmf files +ls *lstmf > all-files.txt + +# do the evaluation (note: verbosity 2 means that correct lines are printed as well as incorrect ones) +lstmeval --model ../rescribealphav4.traineddata --eval_listfile all-files.txt --verbosity 2 2>&1 | tee evaluation |