You cannot select more than 25 topics Topics must start with a letter or number, can include dashes ('-') and can be up to 35 characters long.
ocrd-galley/my_ocrd_workflow

100 lines
3.0 KiB
Plaintext

5 years ago
#!/bin/bash
set -e # Abort on error
set -x
remove_filegrp() {
filegrp_use=$1
mets=$2
xmlstarlet ed --inplace \
-N mets=http://www.loc.gov/METS/ \
-d "//mets:fileGrp[@USE='$filegrp_use']" $mets
}
do_fontident() {
ocrd_typegroups_classifier_parameters='
{
"network": "/home/mike/devel/OCR-D/monorepo/ocrd_typegroups_classifier/ocrd_typegroups_classifier/models/classifier.tgc",
"stride":143
}'
remove_filegrp OCR-D-FONTIDENT mets.xml
ocrd-typegroups-classifier -l DEBUG \
-m mets.xml -I OCR-D-IMG -O OCR-D-FONTIDENT \
-p <(echo $ocrd_typegroups_classifier_parameters)
# XXX does DEFAULT have any meaning? /buerger_gedichte_1778.ocrd does not have
# any DEFAULT, yet -I DEFAULT seems to work for ocrd-typegroups-classifier
#ocrd workspace validate mets.xml
# XXX <error>Unspecified USE category 'FONTIDENT' in fileGrp 'OCR-D-FONTIDENT'</error>
# XXX <error>File 'OCR-D-FONTIDENT_OCR-D-IMG_0002' does not manifest any physical page.</error>
# XXX <notice>Won't download remote image <OCR-D-IMG/OCR-D-IMG_0002></notice>
}
do_linesegmentation() {
remove_filegrp OCR-D-SEG-LINE mets.xml
ocrd-ocropy-segment -l DEBUG \
-m mets.xml -I OCR-D-IMG -O OCR-D-SEG-LINE
#ocrd workspace validate mets.xml
# XXX This leaves copies of the images at the top level of the workspace, because it "downloads" the "remote" files.
# Clean it up.
find . -maxdepth 1 -name "OCR-D-IMG*" -type f -exec rm -v {} \;
# XXX ocrd-tesserocr-segment-line does not seem to produce any line segmentation
# XXX mv {ocrd-ocropy-segment,-line}
}
5 years ago
do_ocr() {
ocrd_tesserocr_recognize_parameters='{ "model": "eng" }' # TODO mods:language + fontident → model
remove_filegrp OCR-D-OCR-TESS mets.xml
ocrd-tesserocr-recognize -l DEBUG \
-m mets.xml -I OCR-D-SEG-LINE -O OCR-D-OCR-TESS \
-p <(echo $ocrd_tesserocr_recognize_parameters)
#ocrd workspace validate mets.xml
}
page_fix_xml() {
# XXX core does not produce valid XML https://github.com/OCR-D/core/issues/242
filegrp=$1
local file
for file in `ocrd workspace find -G $filegrp`; do
sed -i 's#pagecontent/2017-07-15#pagecontent/2018-07-15#g' $file
done
}
page_validate_xml() {
filegrp=$1
local file
for file in `ocrd workspace find -G $filegrp`; do
xmllint --noout --schema `dirname $0`/xsd/pagecontent.2018-07-15.xsd $file
done
}
page_fix_image_references() {
# Make image references relative to the PAGE XML file. The rest of OCR-D probably isn't going to like it, but it
# is a. correct and b. makes PAGE Viewer open the image file automatically.
filegrp=$1
local file
for file in `ocrd workspace find -G $filegrp`; do
sed -i 's#imageFilename="OCR-D-IMG#imageFilename="../OCR-D-IMG#g' $file
done
}
do_fontident
do_linesegmentation
5 years ago
do_ocr
page_fix_xml OCR-D-OCR-TESS
page_validate_xml OCR-D-OCR-TESS # This also makes sure PAGE Viewer can open it
page_fix_image_references OCR-D-OCR-TESS
5 years ago
# XXX Multiple calls create multiple identical mets:agent elements
# XXX Global -l DEBUG
# vim:tw=120: