Add reichsanzeiger gt (#5)

* feat: add Reichsanzeiger-GT to prepare.sh

* add Reichsanzeiger subsets

* add download for Reichsanzeiger subsets

* add first draft for reichsanzeiger gt

* extract sample bundles

* update .gitignore

* clean up

* tidy up project root

* tidy up volumes

* update README

* update data

* build: roll back to version where cis runs still runs

* build: remove default model mounting

.gitignore

@@ -6,7 +6,8 @@ workflows/workspaces
 work/
 workflows/nf-results/*
 workflows/results
-workflows/ocrd-workflows/*.nf
+workflows/ocrd_workflows/*.nf
 models
 .idea
 gt/*
+build/

Dockerfile

-FROM ocrd/all:maximum
+FROM ocrd/all:2023-02-07
 
 WORKDIR /app
 
@@ -12,6 +12,8 @@ RUN apt-get install -y --fix-missing openjdk-11-jre
 COPY src src
 COPY setup.py setup.py
 COPY README.md README.md
+COPY scripts scripts
+COPY data_srcs data_srcs
 
 RUN git init
 RUN git submodule add https://github.com/MehmedGIT/OtoN_Converter submodules/oton
@@ -22,9 +24,6 @@ RUN cd submodules/oton && \
     sed -i "s \$projectDir/ocrd-workspace/ $WORKSPACE_DIR/CURRENT/ g" oton/config.toml && \
     pip install .
 
-COPY prepare.sh prepare.sh
-COPY default_data_sources.txt default_data_sources.txt
-
 RUN pip3 install -r requirements.txt
 RUN pip3 install .
 RUN nextflow

README.md

@@ -26,7 +26,7 @@ Otherwise, the tool will download all `ocrd-tesserocr-recognize` models as well
 - (optional) [customize](#custom-workflows-and-data) QuiVer Benchmarks according to your needs
 - build the image with `docker compose build`
 - spin up a container with `docker compose run -d app`
-- run `docker compose exec app bash prepare.sh`
+- run `docker compose exec app bash scripts/prepare.sh`
 - run `docker compose exec app bash workflows/execute_workflows.sh`
 - the benchmarks and the evaluation results will be available at `data/workflows.json` on your host system
 

data_srcs/reichsanzeiger_full.txt

+https://github.com/UB-Mannheim/reichsanzeiger-gt
\ No newline at end of file

data_srcs/reichsanzeiger_many_ads.list

+053-7972/0044.jp2&CVT=jpeg 1883_55_0044.jpg
+041-7960/0662.jp2&CVT=jpeg 1881_1_0662.jpg
+102-9983/0045.jp2&CVT=jpeg 1871_65_0045.jpg
+101-9982/0017.jp2&CVT=jpeg 1873_1_0017.jpg
+003-7922/0279.jp2&CVT=jpeg 1871_155_0279.jpg

data_srcs/reichsanzeiger_random.list

+104-8023/0452.jp2&CVT=jpeg 1891_33_0452.jpg
+090-8009/0018.jp2&CVT=jpeg 1889_1_0018.jpg
+065-7984/0055.jp2&CVT=jpeg 1885_5_0055.jpg
+093-9053/0488.jp2&CVT=jpeg 1932_300_0488.jpg
+014-8455/0292.jp2&CVT=jpeg 1936_123_0292.jpg
+035-7954/0344.jp2&CVT=jpeg 1879_45_0344.jpg

data_srcs/reichsanzeiger_tables.list

+093-9053/0488.jp2&CVT=jpeg 1932_300_0488.jpg
+014-8455/0292.jp2&CVT=jpeg 1936_123_0292.jpg
+076-9036/0019.jp2&CVT=jpeg 1929_250_0019.jpg
+027-7946/0059.jp2&CVT=jpeg 1877_7_0059.jpg
+053-7972/0044.jp2&CVT=jpeg 1883_55_0044.jpg

data_srcs/reichsanzeiger_title_pages.list

+044-7963/0163.jp2&CVT=jpeg 1881_115_0163.jpg
+079-7998/0444.jp2&CVT=jpeg 1887_134_0444.jpg
+153-9561/0087.jp2&CVT=jpeg 1916_169_0087.jpg
+167-9449/0129.jp2&CVT=jpeg 1918_267_0129.jpg
+065-7984/0054.jp2&CVT=jpeg 1885_5_0054.jpg

docker-compose.yml

@@ -10,7 +10,6 @@ services:
       - ./data:/app/data # this will write the results to your host system
       - ./gt:/app/gt
       # mount your modules, custom workflows and data here
-      - ./models:/usr/local/share/ocrd-resources/
-      #- ./workflows:/app/workflows
-      # - ./workflows/ocrd_workflows:/app/workflows/ocrd_workflows
+      #- ./models/ocrd-tesserocr-recognize:/usr/local/share/tessdata/
+      #- ./models/ocrd-calamari-recognize:/usr/local/share/ocrd-resources/
       # - TODO/custom/data
\ No newline at end of file

prepare.sh

-#!/bin/bash
-
-mkdir gt
-
-while IFS= read -r URL; do
-    OWNER=$(echo "$URL" | cut -d'/' -f4)
-    REPO=$(echo "$URL" | cut -d'/' -f5)
-    if [[ ! -f gt/"$REPO".zip ]]; then
-        echo "Downloading $REPO …"
-        RESULT=$(curl -L \
-        -H "Accept: application/vnd.github+json" \
-        -H "X-GitHub-Api-Version: 2022-11-28" \
-        https://api.github.com/repos/"$OWNER"/"$REPO"/releases/latest)
-        ZIP_URL=$(echo  "$RESULT" | jq -r '.assets | .[].browser_download_url')
-        curl -L -o gt/"$REPO".zip "$ZIP_URL"
-    fi
-done < default_data_sources.txt
-
-cd gt || exit
-# the default data is structured like this:
-# repository_name.zip
-# |___ subordinate_work_1.zip
-# |___ subordinate_work_2.zip
-# |___ ...
-# $ZIP refers to the release itself which is on level "repository_name.zip"
-# the subordinate works are also OCR-D BagIts / zips. these are referred to by $INNER_ZIP.
-for ZIP in *.zip; do
-    NAME=$(echo "$ZIP" | cut -d"." -f1)
-    echo "Processing $NAME"
-    unzip -qq -d "$NAME" "$ZIP"
-    mv "$NAME"/ocrdzip_out/* "$NAME" && rm -r "$NAME"/ocrdzip_out
-    for INNER_ZIP in "$NAME"/*.zip; do
-        echo "Dealing with inner zip files …"
-        INNER_ZIP_NAME=$(basename "$INNER_ZIP" .ocrd.zip)
-        unzip -qq -d "$NAME"/"$INNER_ZIP_NAME" "$INNER_ZIP" && rm "$INNER_ZIP"
-
-        echo "Done."
-    done
-done
-
-echo " … and ready to go!"

scripts/convert-yml-to-json.py

+#!env python3
+
+# This script has been copied from https://github.com/OCR-D/spec/blob/master/scripts/yaml-to-json.py
+
+from yaml import safe_load
+from json import dumps
+from click import command, argument, option
+
+@command()
+@option('--indent', default=2, type=int)
+@argument('src')
+@argument('dst')
+def cli(src, dst, indent):
+    kwargs = {}
+    if indent > 0:
+        kwargs['indent'] = indent
+    with open(src, 'r', encoding='utf-8') as f_in, open(dst, 'w', encoding='utf-8') as f_out:
+        ret = safe_load(f_in)
+        f_out.write(dumps(ret, **kwargs))
+
+if __name__ == '__main__':
+    cli()
\ No newline at end of file

scripts/prepare.sh

+#!/bin/bash
+
+mkdir gt
+
+echo "Prepare OCR-D Ground Truth …"
+
+while IFS= read -r URL; do
+    OWNER=$(echo "$URL" | cut -d'/' -f4)
+    REPO=$(echo "$URL" | cut -d'/' -f5)
+    if [[ ! -f gt/"$REPO".zip ]]; then
+        echo "Downloading $REPO …"
+        RESULT=$(curl -L \
+        -H "Accept: application/vnd.github+json" \
+        -H "X-GitHub-Api-Version: 2022-11-28" \
+        https://api.github.com/repos/"$OWNER"/"$REPO"/releases/latest)
+        ZIP_URL=$(echo  "$RESULT" | jq -r '.assets | .[].browser_download_url')
+        curl -L -o gt/"$REPO".zip "$ZIP_URL"
+    fi
+done < data_srcs/default_data_sources.txt
+
+cd gt || exit
+# the default data is structured like this:
+# repository_name.zip
+# |___ subordinate_work_1.zip
+# |___ subordinate_work_2.zip
+# |___ ...
+# $ZIP refers to the release itself which is on level "repository_name.zip"
+# the subordinate works are also OCR-D BagIts / zips. these are referred to by $INNER_ZIP.
+for ZIP in *.zip; do
+    NAME=$(echo "$ZIP" | cut -d"." -f1)
+    echo "Processing $NAME"
+    if [[ ! -d  $NAME && $NAME != "reichsanzeiger-gt" ]]; then 
+        unzip -qq -d "$NAME" "$ZIP"
+        mv "$NAME"/ocrdzip_out/* "$NAME" && rm -r "$NAME"/ocrdzip_out
+        for INNER_ZIP in "$NAME"/*.zip; do
+            echo "Dealing with inner zip files …"
+            INNER_ZIP_NAME=$(basename "$INNER_ZIP" .ocrd.zip)
+            unzip -qq -d "$NAME"/"$INNER_ZIP_NAME" "$INNER_ZIP" && rm "$INNER_ZIP"
+
+            echo "Recreate required directory structure for $INNER_ZIP_NAME."
+            mkdir "$NAME"/"$INNER_ZIP_NAME"/data/"$INNER_ZIP_NAME"
+            mv "$NAME"/"$INNER_ZIP_NAME"/data/OCR-* "$NAME"/"$INNER_ZIP_NAME"/data/"$INNER_ZIP_NAME"
+            mv "$NAME"/"$INNER_ZIP_NAME"/data/mets.xml "$NAME"/"$INNER_ZIP_NAME"/data/"$INNER_ZIP_NAME"
+            cp "$NAME"/metadata.json "$NAME"/"$INNER_ZIP_NAME"/data/"$INNER_ZIP_NAME"/metadata.json
+
+            echo "Moving $INNER_ZIP_NAME higher in dir structure."
+            mv "$NAME"/"$INNER_ZIP_NAME" .
+            echo "Done."
+        done
+        rm -rf "$NAME"
+    fi
+done
+
+echo "Prepare Reichsanzeiger GT …"
+
+if [[ $1 == "ra-full" ]]; then
+    echo "Preparing the full Reichsanzeiger GT."
+
+    if [ ! -d reichsanzeiger-gt ]; then
+        git clone https://github.com/UB-Mannheim/reichsanzeiger-gt
+    fi
+
+    RA_GT=/app/gt/reichsanzeiger-gt
+    DATA_DIR=/$RA_GT/data
+    cd $DATA_DIR|| exit
+    
+    if [[ -d reichsanzeiger-1820-1939/OCR-D-IMG ]]; then
+        echo "Skip downloading Reichsanzeiger images."
+    else
+        bash download_images.sh
+    fi
+    cd reichsanzeiger-1820-1939 || exit
+    ocrd workspace init
+    mkdir OCR-D-IMG
+    cp ../images/* OCR-D-IMG
+    rm -rf ../images
+    rm -rf ../reichsanzeiger-1820-1939_with-TableRegion
+    cp -r GT-PAGE OCR-D-GT-SEG-LINE
+    
+    echo "Adding images to mets …"
+    
+    FILEGRP="OCR-D-IMG"
+    EXT=".jpg"  # the actual extension of the image files
+    MEDIATYPE='image/jpeg'  # the actual media type of the image files
+    for i in "$FILEGRP"/*"$EXT"; do
+      BASE=$(basename "${i}" $EXT);
+      ocrd workspace add -G $FILEGRP -i ${FILEGRP}_"${BASE}" -g P_"${BASE}" -m $MEDIATYPE "${i}";
+    done
+    
+    python3 /app/scripts/convert-yml-to-json.py --indent 2 $RA_GT/METADATA.yml $RA_GT/metadata.json
+    
+    echo " … and ready to go!"
+   
+else
+    echo "Prepare smaller sets of Reichsanzeiger GT."
+    cd /app || exit
+    bash /app/scripts/prepare_reichsanzeiger_sets.sh
+fi

scripts/prepare_reichsanzeiger_sets.sh

+#!/bin/sh
+
+echo "Download Reichsanzeiger GT repository."
+cd gt || exit
+if [ ! -d reichsanzeiger-gt ]; then
+    git clone https://github.com/UB-Mannheim/reichsanzeiger-gt
+fi
+
+cd .. || exit
+
+
+PREFIX="data_srcs"
+files=(
+    "$PREFIX"/reichsanzeiger_many_ads.list
+    "$PREFIX"/reichsanzeiger_random.list
+    "$PREFIX"/reichsanzeiger_tables.list
+    "$PREFIX"/reichsanzeiger_title_pages.list
+)
+
+for FILE in "${files[@]}"; do
+    NAME=$(basename "$FILE" .list)
+    if [ -d   gt/"$NAME" ]; then
+        echo "Directory gt/$NAME already exists. Skipping download."
+    else
+        echo "Processing $FILE."
+        mkdir -p gt/"$NAME"/data/"$NAME"/OCR-D-IMG
+        mkdir -p gt/"$NAME"/data/"$NAME"/OCR-D-GT-SEG-LINE
+
+        urlbase=$(echo "aHR0cHM6Ly9kaWdpLmJpYi51bmktbWFubmhlaW0uZGUvcmVpY2hzYW56ZWlnZXIuZmNnaT9GSUY9
+L3JlaWNoc2FuemVpZ2VyL2ZpbG0vCg==" | base64 -d)
+        while read -r line; do 
+            wget --limit-rate=500k "${urlbase}${line% *}" -O ./gt/"$NAME"/data/"$NAME"/OCR-D-IMG/"${line#* }"
+            IMG_NAME=$(basename "${line#* }" .jpg)
+            cp gt/reichsanzeiger-gt/data/reichsanzeiger-1820-1939/GT-PAGE/"$IMG_NAME".xml gt/"$NAME"/data/"$NAME"/OCR-D-GT-SEG-LINE/"$IMG_NAME".xml
+        done < "$FILE"
+    fi
+
+    if [ ! -f gt/"$NAME"/mets.xml ]; then
+        echo "Preparing OCR-D workspace for $NAME".
+        cd gt/"$NAME"/data/"$NAME" || exit
+        ocrd workspace init
+        ocrd workspace set-id "$NAME"
+
+        FILEGRP="OCR-D-IMG"
+        FILEGRP_2="OCR-D-GT-SEG-LINE"
+        # add images to mets
+        EXT=".jpg"  # the actual extension of the image files
+        MEDIATYPE='image/jpeg'  # the actual media type of the image files
+        for i in OCR-D-IMG/*"$EXT"; do
+            BASE=$(basename "${i}" $EXT)
+            ocrd workspace add -G $FILEGRP -i "${FILEGRP}"_"${BASE}" -g P_"${BASE}" -m $MEDIATYPE "${i}"
+        done
+
+        # add GT to mets
+        for i in "$FILEGRP_2"/*.xml; do
+            BASE=$(basename "${i}" ".xml")
+            ocrd workspace add -G $FILEGRP_2 -i "${FILEGRP_2}"_"${BASE}" -g P_"${BASE}" -m text/xml "${i}"
+        done
+    fi
+
+    if [ ! -f gt/"$NAME"/data/"$NAME"/metadata.json ]; then
+        cp /app/gt/reichsanzeiger-gt/METADATA.yml /app/gt/"$NAME"/data/"$NAME"/METADATA.yml
+        python3 /app/scripts/convert-yml-to-json.py --indent 2 /app/gt/"$NAME"/data/"$NAME"/METADATA.yml /app/gt/"$NAME"/data/"$NAME"/metadata.json
+    fi
+    cd /app || exit
+done
+
+echo "Preparation of Reichsanzeiger GT subsets done."

src/benchmark_extraction.py

@@ -106,17 +106,8 @@ def get_eval_tool(mets_path: str) -> str:
 
 def get_gt_workspace(workspace_path: str) -> Dict[str, str]:
     current_workspace = get_workspace_name(workspace_path)
-    split_workspace_name = current_workspace.split('_')
-    workspace_name_wo_workflow = split_workspace_name[0] + '_' + split_workspace_name[1] + '_' + split_workspace_name[2]
-    font = ''
-    if split_workspace_name[1] == 'ant':
-        font = 'Antiqua'
-    elif split_workspace_name[1] == 'frak':
-        font = 'Fraktur'
-    else:
-        font = 'Font Mix'
-    url = 'https://github.com/OCR-D/quiver-data/blob/main/' + workspace_name_wo_workflow + '.ocrd.zip'
-    label = f'GT workspace {split_workspace_name[0]}th century {font} {split_workspace_name[2]} layout'
+    url = 'https://github.com/OCR-D/quiver-data/blob/main/' + current_workspace + '.ocrd.zip'
+    label = f'GT workspace {current_workspace}'
     return {
         '@id': url,
         'label': label
@@ -196,7 +187,8 @@ def get_nextflow_completed_process_file(workspace_path: str):
     return file
 
 def get_nextflow_time(workspace_path: str, time_type: str) -> float:
-    files = listdir(workspace_path)
+    highest_workspace_dir = '/'.join(workspace_path.split('/')[:-2])
+    files = listdir(highest_workspace_dir)
     logs = []
     for file in files:
         if '.command.log' in file:
@@ -204,7 +196,7 @@ def get_nextflow_time(workspace_path: str, time_type: str) -> float:
 
     time_per_workflow_step = []
     for log in logs:
-        with open(workspace_path + '/' + log, 'r', encoding='utf-8') as l:
+        with open(highest_workspace_dir + '/' + log, 'r', encoding='utf-8') as l:
             log_file = l.read()
             no_sec_s = re.search(rf'([0-9]+?\.[0-9]+?)s \({time_type}\)', log_file).group(1)
             time_per_workflow_step.append(float(no_sec_s))

workflows/execute_workflows.sh

@@ -39,9 +39,9 @@ convert_ocrd_wfs_to_NextFlow() {
 
 download_models() {
     echo "Download the necessary models if not available"
-    if [[ ! -f /usr/local/share/ocrd-resources/ocrd-tesserocr-recognize/ ]]
+    if [[ ! -d /usr/local/share/tessdata ]]
     then
-        mkdir -p /usr/local/share/ocrd-resources/
+        #mkdir -p /usr/local/share/ocrd-resources/
         ocrd resmgr download ocrd-tesserocr-recognize '*'
     fi
     if [[ ! -d /usr/local/share/ocrd-resources/ocrd-calamari-recognize/qurator-gt4histocr-1.0 ]]
@@ -58,32 +58,15 @@ create_wf_specific_workspaces() {
 
     # create workspace for all OCR workflows.
     # each workflow has a separate workspace to work with.
-    echo "Create workflow specific workspaces for each dir in ./gt …"
     for DIR in "$ROOT"/gt/*/; do
         DIR_NAME=$(basename "$DIR")
-        if grep -q "multivolume work" <<< "$(cat $DIR/mets.xml)"; then
-            echo "$DIR_NAME is a multivolume work"
-
-            for WORKFLOW in "$OCRD_WORKFLOW_DIR"/*ocr.txt.nf
-            do
-                WF_NAME=$(basename -s .txt.nf "$WORKFLOW")
-                for SUB_WORK in $DIR/*/; do
-                    SUB_WORK_DIR_NAME=$(basename "$SUB_WORK")
-                    TARGET="$WORKSPACE_DIR"/tmp/"$DIR_NAME"_"$SUB_WORK_DIR_NAME"_"$WF_NAME"
-                    cp -r "$ROOT"/gt/"$DIR_NAME"/"$SUB_WORK_DIR_NAME" "$TARGET"
-                    if [[ -f "$ROOT"/gt/"$DIR_NAME"/metadata.json ]]; then
-                        cp -r "$ROOT"/gt/"$DIR_NAME"/metadata.json "$TARGET"/metadata.json
-                    fi
-                    cp "$WORKFLOW" "$TARGET"/data/
-                done
-
-            done
-        else
+        if [[ ! $DIR_NAME == "reichsanzeiger-gt" ]]; then
+            echo "Create workflow specific workspace for $DIR_NAME."
             for WORKFLOW in "$OCRD_WORKFLOW_DIR"/*ocr.txt.nf
             do
                 WF_NAME=$(basename -s .txt.nf "$WORKFLOW")
                 cp -r "$ROOT"/gt/"$DIR_NAME" "$WORKSPACE_DIR"/tmp/"$DIR_NAME"_"$WF_NAME"
-                cp "$WORKFLOW" "$WORKSPACE_DIR"/tmp/"$DIR_NAME"_"$WF_NAME"/
+                cp "$WORKFLOW" "$WORKSPACE_DIR"/tmp/"$DIR_NAME"_"$WF_NAME"/data/*/
             done
         fi
     done
@@ -93,11 +76,10 @@ clean_up_tmp_dirs() {
     echo "Clean up intermediate dirs …"
     for DIR in "$WORKSPACE_DIR"/tmp/*
     do
+        echo "Cleaning up $DIR."
         DIR_NAME=$(basename "$DIR")
         mv "$DIR" "$WORKSPACE_DIR"/"$DIR_NAME"
-        cp "$OCRD_WORKFLOW_DIR"/*eval.txt.nf "$WORKSPACE_DIR"/"$DIR_NAME"
-        ls "$WORKSPACE_DIR"/"$DIR_NAME"
-        cp -r "$WORKSPACE_DIR"/"$DIR_NAME"/data/* "$WORKSPACE_DIR"/"$DIR_NAME"/
+        cp "$OCRD_WORKFLOW_DIR"/*eval.txt.nf "$WORKSPACE_DIR"/"$DIR_NAME"/data/*/
     done
 
     rm -rf "$WORKSPACE_DIR"/tmp
@@ -114,16 +96,16 @@ execute_wfs_and_extract_benchmarks() {
 
             DIR_NAME=$(basename $WS_DIR)
 
-            run "$WS_DIR"/*ocr.txt.nf "$DIR_NAME" "$WS_DIR"
-            run "$WS_DIR"/*eval.txt.nf "$DIR_NAME" "$WS_DIR"
+            run "$WS_DIR"/data/*/*ocr.txt.nf "$DIR_NAME" "$WS_DIR"
+            run "$WS_DIR"/data/*/*eval.txt.nf "$DIR_NAME" "$WS_DIR"
 
             # create a result JSON according to the specs          
             echo "Get Benchmark JSON …"
-            quiver benchmarks-extraction "$WS_DIR" "$WORKFLOW"
+            quiver benchmarks-extraction "$WS_DIR"/data/* "$WORKFLOW"
             echo "Done."
 
             # move data to results dir
-            mv "$WS_DIR"/*.json "$WORKFLOW_DIR"/results
+            mv "$WS_DIR"/data/*/*.json "$WORKFLOW_DIR"/results
         fi
     done
     cd "$ROOT" || exit
@@ -132,7 +114,7 @@ execute_wfs_and_extract_benchmarks() {
 adjust_workflow_settings() {
     # $1: $WORKFLOW
     # $2: $DIR_NAME
-    sed -i "s CURRENT app/workflows/workspaces/$2 g" "$1"
+    sed -i "s CURRENT app/workflows/workspaces/$2/data/*/ g" "$1"
 }
 
 rename_and_move_nextflow_result() {
@@ -142,13 +124,13 @@ rename_and_move_nextflow_result() {
     WORKFLOW_NAME=$(basename -s .txt.nf "$1")
     rm "$WORKFLOW_DIR"/nf-results/*process_completed.json
     mv "$WORKFLOW_DIR"/nf-results/*_completed.json "$WORKFLOW_DIR"/results/"$2"_"$WORKFLOW_NAME"_completed.json
-    if [ $WORKFLOW_NAME != "dinglehopper_eval" ]; then
+    if [ "$WORKFLOW_NAME" != "dinglehopper_eval" ]; then
         for DIR in "$WORKSPACE_DIR"/work/*
         do
-            WORK_DIR_NAME=$(basename $DIR)
-            for SUB_WORK_DIR in $DIR/*
+            WORK_DIR_NAME=$(basename "$DIR")
+            for SUB_WORK_DIR in "$DIR"/*
             do
-                SUB_WORK_DIR_NAME=$(basename $SUB_WORK_DIR)
+                SUB_WORK_DIR_NAME=$(basename "$SUB_WORK_DIR")
                 mv "$WORKSPACE_DIR"/work/"$WORK_DIR_NAME"/"$SUB_WORK_DIR_NAME"/.command.log "$WORKSPACE_DIR"/"$2"/"$WORK_DIR_NAME"_"$SUB_WORK_DIR_NAME".command.log
             done
             
@@ -165,7 +147,7 @@ run() {
     adjust_workflow_settings "$1" "$2"
     nextflow run "$1" -with-weblog http://127.0.0.1:8000/nextflow/
     rename_and_move_nextflow_result "$1" "$2"
-    save_workspaces "$3" "$2" "$1"
+    save_workspaces "$3"/data "$2" "$1"
 }
 
 save_workspaces() {
@@ -173,7 +155,7 @@ save_workspaces() {
     # $2: $DIR_NAME
     # $3: $WORKFLOW
     echo "Zipping workspace $1"
-    ocrd zip bag -d $1 -i $1 $1
+    ocrd zip bag -d "$DIR_NAME"/data/* -i "$DIR_NAME"/data/* "$DIR_NAME"
     WORKFLOW_NAME=$(basename -s .txt.nf "$3")
     mv "$WORKSPACE_DIR"/"$2".zip "$WORKFLOW_DIR"/results/"$2"_"$WORKFLOW_NAME".zip
 }