mirror of
https://github.com/qurator-spk/ocrd-galley.git
synced 2025-06-07 22:00:12 +02:00
✨ Update to sbb_textline_detector with the fixed AlternativeImage support (= merged PAGE results)
This commit is contained in:
parent
de47a3e5b1
commit
d166077a55
3 changed files with 8 additions and 26 deletions
|
@ -35,9 +35,10 @@ do_validate() {
|
|||
do_binarization() {
|
||||
# Binarize the images
|
||||
|
||||
remove_filegrp OCR-D-IMG-BINPAGE mets.xml
|
||||
remove_filegrp OCR-D-IMG-BIN mets.xml
|
||||
ocrd-olena-binarize -l $LOG_LEVEL \
|
||||
-m mets.xml -I OCR-D-IMG -O OCR-D-IMG-BIN \
|
||||
-m mets.xml -I OCR-D-IMG -O OCR-D-IMG-BINPAGE \
|
||||
-p <(echo '{"impl": "sauvola-ms-split"}')
|
||||
}
|
||||
|
||||
|
@ -66,11 +67,11 @@ do_linesegmentation_tesserocr() {
|
|||
remove_filegrp OCR-D-SEG-REGION mets.xml
|
||||
remove_filegrp OCR-D-SEG-LINE mets.xml
|
||||
#ocrd-ocropy-segment -l $LOG_LEVEL \
|
||||
# -m mets.xml -I OCR-D-IMG-BIN -O OCR-D-SEG-LINE
|
||||
# -m mets.xml -I OCR-D-IMG-BINPAGE -O OCR-D-SEG-LINE
|
||||
# XXX ocrd-ocropy-segment throws an exception for buerger_gedichte_1778.ocrd
|
||||
|
||||
ocrd-tesserocr-segment-region -l $LOG_LEVEL \
|
||||
-m mets.xml -I OCR-D-IMG-BIN -O OCR-D-SEG-REGION
|
||||
-m mets.xml -I OCR-D-IMG-BINPAGE -O OCR-D-SEG-REGION
|
||||
ocrd-tesserocr-segment-line -l $LOG_LEVEL \
|
||||
-m mets.xml -I OCR-D-SEG-REGION -O OCR-D-SEG-LINE
|
||||
|
||||
|
@ -79,11 +80,12 @@ do_linesegmentation_tesserocr() {
|
|||
|
||||
do_linesegmentation_sbb() {
|
||||
# Segment the lines in the images
|
||||
# TODO: Check that this works with the RGB images
|
||||
|
||||
remove_filegrp OCR-D-SEG-REGION mets.xml
|
||||
remove_filegrp OCR-D-SEG-LINE mets.xml
|
||||
ocrd_sbb_textline_detector -l $LOG_LEVEL \
|
||||
-m mets.xml -I OCR-D-IMG -O OCR-D-SEG-LINE \
|
||||
-m mets.xml -I OCR-D-IMG-BINPAGE -O OCR-D-SEG-LINE \
|
||||
-p '{"model": "/var/lib/textline_detection"}'
|
||||
}
|
||||
|
||||
|
@ -120,25 +122,6 @@ page_validate_xml() {
|
|||
done
|
||||
}
|
||||
|
||||
page_fix_image_references_to_bin() {
|
||||
# Make image references point to the binarized images
|
||||
# XXX This is a hack, it is probably better to use alternative images in ocrd_calamari
|
||||
|
||||
filegrp=$1
|
||||
|
||||
local file
|
||||
for file in `ocrd workspace find -G $filegrp`; do
|
||||
# Arrays with filenames to the images
|
||||
imgs=(`ocrd workspace find -G OCR-D-IMG`)
|
||||
imgs_bin=(`ocrd workspace find -G OCR-D-IMG-BIN -m image/png`)
|
||||
|
||||
# Change all image references to point to the corresponding binarized image
|
||||
for i in ${!imgs[@]}; do
|
||||
sed -i "s!imageFilename=.${imgs[$i]}.!imageFilename=\"${imgs_bin[$i]}\"!g" $file
|
||||
done
|
||||
done
|
||||
}
|
||||
|
||||
page_downgrade_to_2018() {
|
||||
# Not used anymore, but kept if needed in the future
|
||||
filegrp=$1
|
||||
|
@ -170,7 +153,6 @@ do_validate
|
|||
|
||||
|
||||
do_linesegmentation_sbb
|
||||
page_fix_image_references_to_bin OCR-D-SEG-LINE
|
||||
page_upgrade_to_2019 OCR-D-SEG-LINE
|
||||
page_validate_xml OCR-D-SEG-REGION
|
||||
page_validate_xml OCR-D-SEG-LINE
|
||||
|
|
|
@ -1,6 +1,6 @@
|
|||
tensorflow-gpu < 2.0 # Needed for sbb_text_linedetector
|
||||
|
||||
ocrd >= 1.0.0
|
||||
ocrd >= 2.0.0
|
||||
|
||||
https://github.com/mikegerber/ocrd_typegroups_classifier/archive/fix/pass-down-page-id.tar.gz # XXX git+https://github.com/seuretm/ocrd_typegroups_classifier.git
|
||||
|
||||
|
@ -11,6 +11,6 @@ ocrd_tesserocr
|
|||
https://github.com/mikegerber/ocrd_calamari/archive/6949876.tar.gz
|
||||
|
||||
|
||||
vendor/sbb_textline_detector-d905c0b.tar
|
||||
vendor/sbb_textline_detector-10bbda9.tar
|
||||
|
||||
https://github.com/qurator-spk/dinglehopper/archive/c305539.tar.gz
|
||||
|
|
Binary file not shown.
Loading…
Add table
Add a link
Reference in a new issue