✨ Use GT4HistOCR_2000000 model from qurator-data for Tesseract

Dockerfile

@@ -2,8 +2,10 @@ FROM ubuntu:18.04
 
 ENV LC_ALL=C.UTF-8 LANG=C.UTF-8
 
-RUN mkdir /var/lib/calamari-models
-COPY data/calamari-models/GT4HistOCR /var/lib/calamari-models/GT4HistOCR
+ENV LEPTONICA_VERSION 1.78.0
+ENV TESSERACT_VERSION 4.1.0
+ENV TESSDATA_BEST_VERSION 4.0.0
+ENV TESSDATA_PREFIX /usr/local/share/tessdata
 
 
 RUN apt-get update && \
@@ -24,12 +26,6 @@ RUN apt-get update && \
 
 # Build Leptonica and Tesseract.
 # XXX Use a multi stage build for this.
-ENV LEPTONICA_VERSION 1.78.0
-ENV TESSERACT_VERSION 4.1.0
-ENV TESSDATA_BEST_VERSION 4.0.0
-
-ENV TESSDATA_PREFIX /usr/local/share/tessdata
-
 RUN curl -sSL -O https://github.com/DanBloomberg/leptonica/releases/download/$LEPTONICA_VERSION/leptonica-$LEPTONICA_VERSION.tar.gz && \
     tar xvzf leptonica-$LEPTONICA_VERSION.tar.gz && \
     cd leptonica-$LEPTONICA_VERSION && \
@@ -45,9 +41,16 @@ RUN curl -sSL -O https://github.com/tesseract-ocr/tesseract/archive/$TESSERACT_V
 RUN curl -sSL -O https://github.com/tesseract-ocr/tessdata_best/archive/$TESSDATA_BEST_VERSION.tar.gz && \
     tar xvzf $TESSDATA_BEST_VERSION.tar.gz && \
     mv tessdata_best-$TESSDATA_BEST_VERSION $TESSDATA_PREFIX && \
-    rm -rf $TESSDATA_BEST_VERSION.tar.gz && \
-    tesseract --list-langs
+    rm -rf $TESSDATA_BEST_VERSION.tar.gz
+
+
+# Copy OCR models
+RUN mkdir /var/lib/calamari-models
+COPY data/calamari-models/GT4HistOCR /var/lib/calamari-models/GT4HistOCR
+RUN mkdir $TESSDATA_PREFIX
+COPY data/tesseract-models/GT4HistOCR/GT4HistOCR_2000000.traineddata $TESSDATA_PREFIX
 
+RUN tesseract --list-langs
 
 
 COPY requirements.txt /tmp

build

@@ -3,8 +3,10 @@ set -e
 
 (
   cd data
-  git annex get calamari-models/GT4HistOCR
-  git annex unlock calamari-models/GT4HistOCR  # So they are not symlinks
+  git annex init
+  git annex upgrade
+  git annex get calamari-models/GT4HistOCR/*.ckpt*
+  git annex get tesseract-models/GT4HistOCR/*.traineddata
 )
 
 docker build -t my_ocrd_workflow .

data

@@ -1 +1 @@
-Subproject commit ca541f5ebb4593fde2aaef217bf606f7c2bf148a
+Subproject commit eb7412a1efbcba53567ec37237732e96e839dbe8

my_ocrd_workflow

@@ -79,7 +79,7 @@ do_linesegmentation() {
 do_ocr() {
   # Perform OCR on the segmented lines
 
-  ocrd_tesserocr_recognize_parameters='{ "model": "frk" }'  # TODO mods:language + fontident → model
+  ocrd_tesserocr_recognize_parameters='{ "model": "GT4HistOCR_2000000" }'  # TODO mods:language + fontident → model
   remove_filegrp OCR-D-OCR-TESS mets.xml
   ocrd-tesserocr-recognize -l $LOG_LEVEL \
     -m mets.xml -I OCR-D-SEG-LINE -O OCR-D-OCR-TESS \