#!/bin/sh #ppn="PPN719671574" #ppn="PPN726351202" # Einblattdruck ppn=PPN787969397 # Folz, Hans: Abenteuerliches Klopfan, 1481 set -e self_dir=`dirname $0` self_dir=`realpath $self_dir` export no_proxy=localhost,digital.staatsbibliothek-berlin.de workspace=$ppn mkdir "$workspace" cd "$workspace" pwd oai_identifier="oai%3Adigital.staatsbibliothek-berlin.de%3A$ppn" oai_url="https://digital.staatsbibliothek-berlin.de/oai?verb=GetRecord&metadataPrefix=mets&identifier=$oai_identifier" echo "$oai_url" curl "$oai_url" | xmlstarlet sel -t -c '//*[local-name()="mets"]' > mets.xml sed -i 's#file:///goobi/tiff001/sbb/#file:///srv/digisam_images/sbb/#g' mets.xml ocrd workspace validate mets.xml | grep -v "Won't download remote image" $self_dir/run-docker-hub -I PRESENTATION --skip-validation # TODO # my_ocrd_workflow # ---------------- # * Need option to add volumes e.g. /srv/digisam_images # * Use run-docker-hub # File bugs in OCR-D # ------------------ # * PAGE-XML OCR-D-IMG-BINPAGE/OCR-D-IMG-BINPAGE_0001.xml : imageFilename '/srv/digisam_images/sbb/PPN719671574/00000001.tif' not found in METS # -> had to use relative file names # * Should be able to disable in workspace validate Won't download remote image # sbb_textline_detector # --------------------- # * sbb_textline_detector slow # -> Support loading the models once so the OCR-D processor can profit from processing multiple pages # * Check what happens with the skewed textlines in SEG_LINE_0019