Add format-for-hocr-pdf.sh script

1 files changed, 39 insertions, 0 deletions

diff --git a/format-for-hocr-pdf.sh b/format-for-hocr-pdf.sh
new file mode 100644
index 0000000..89d4dd8
--- /dev/null
+++ b/format-for-hocr-pdf.sh
@@ -0,0 +1,39 @@
+#!/bin/sh
+usage="Usage: $0 indir outdir
+
+Creates a new directory with image and hocr files appropriately named
+and formatted for use with the the hocr-pdf tool from hocr-tools."
+
+test $# -ne 2 && echo "$usage" && exit 1
+
+# All possible training files to check for
+TRAININGS="rescribealphav4 rescribealphav5"
+
+if ! test -d "$1"; then
+	echo "Error: $1 does not exist"
+	exit 1
+fi
+mkdir -p "$2" || exit 1
+
+echo "Copying hocrs and converting pngs from $1 to $2"
+find "$1" -maxdepth 1 -type f -name '*.binarized.png' | while read i; do
+	b=`basename "$i" .binarized.png`
+
+	hocr=""
+	for t in $TRAININGS; do
+		n=`echo "$i" | sed "s/.binarized.png/_$t.hocr/"`
+		test -f "$n" && hocr="$n"
+	done
+	if test -z "$hocr"; then
+		echo "Warning: no corresponding hocr file found for $i, skipping."
+		continue
+	fi
+
+	gm convert "$i" "$2/$b.jpg" || exit 1
+	cp "$hocr" "$2/$b.hocr" || exit 1
+done
+
+nhocr=`find "$2" -type f -name '*hocr'|wc -l`
+njpg=`find "$2" -type f -name '*jpg'|wc -l`
+echo "Done. There are $nhocr hocr files and $njpg jpg files in $2"
+