✨ sbb_textline_detection: Preserve input PAGE info by merging segmentation results

ocrd_sbb_textline_detection used the output XML by main.py as is, and
– by doing this – threw away any input data from the input PAGE,
including the critical pc:AlternativeImage and the less important
pc:MetadataItem.

Fix this by merging the segmentation results into a file created from
the input file.

Also add a pc:MetadataItem processingStep about the segmentation
operation.

qurator/sbb_textline_detector/ocrd_cli.py

@@ -1,11 +1,14 @@
 import json
 import os
+import tempfile
 
 import click
+import ocrd_models.ocrd_page
 from ocrd import Processor
 from ocrd.decorators import ocrd_cli_options, ocrd_cli_wrap_processor
 from ocrd_modelfactory import page_from_file
 from ocrd_models import OcrdFile
+from ocrd_models.ocrd_page_generateds import MetadataItemType, LabelsType, LabelType
 from ocrd_utils import concat_padded, getLogger, MIMETYPE_PAGE
 from pkg_resources import resource_string
 
@@ -22,10 +25,14 @@ def ocrd_sbb_textline_detector(*args, **kwargs):
     return ocrd_cli_wrap_processor(OcrdSbbTextlineDetectorRecognize, *args, **kwargs)
 
 
+TOOL = 'ocrd_sbb_textline_detector'
+
+
 class OcrdSbbTextlineDetectorRecognize(Processor):
 
     def __init__(self, *args, **kwargs):
-        kwargs['ocrd_tool'] = OCRD_TOOL['tools']['ocrd_sbb_textline_detector']
+        kwargs['ocrd_tool'] = OCRD_TOOL['tools'][TOOL]
+        kwargs['version'] = OCRD_TOOL['version']
         super(OcrdSbbTextlineDetectorRecognize, self).__init__(*args, **kwargs)
 
     def _make_file_id(self, input_file, input_file_grp, n):
@@ -49,7 +56,6 @@ class OcrdSbbTextlineDetectorRecognize(Processor):
             log.info("INPUT FILE %i / %s", n, input_file)
 
             file_id = self._make_file_id(input_file, self.output_file_grp, n)
-            image_file = self._resolve_image_file(input_file)
 
             # Process the files
             try:
@@ -57,16 +63,47 @@ class OcrdSbbTextlineDetectorRecognize(Processor):
             except FileExistsError:
                 pass
 
-            model = self.parameter['model']
-            x = textlineerkenner(image_file, self.output_file_grp, file_id, model)
-            x.run()
+            with tempfile.TemporaryDirectory() as tmp_dirname:
+                # Segment the image
+                image_file = self._resolve_image_file(input_file)
+                model = self.parameter['model']
+                x = textlineerkenner(image_file, tmp_dirname, file_id, model)
+                x.run()
+
+                # Read segmentation results
+                tmp_filename = os.path.join(tmp_dirname, file_id) + '.xml'
+                tmp_pcgts = ocrd_models.ocrd_page.parse(tmp_filename)
+                tmp_page = tmp_pcgts.get_Page()
+
+            # Create a new PAGE file from the input file
+            pcgts = page_from_file(self.workspace.download_file(input_file))
+            page = pcgts.get_Page()
+
+            # Merge results → PAGE file
+            page.set_PrintSpace(tmp_page.get_PrintSpace())
+            page.set_ReadingOrder(tmp_page.get_ReadingOrder())
+            page.set_TextRegion(tmp_page.get_TextRegion())
+
+            # Save metadata about this operation
+            metadata = pcgts.get_Metadata()
+            metadata.add_MetadataItem(
+                MetadataItemType(type_="processingStep",
+                                 name=self.ocrd_tool['steps'][0],
+                                 value=TOOL,
+                                 Labels=[LabelsType(
+                                     externalModel="ocrd-tool",
+                                     externalId="parameters",
+                                     Label=[LabelType(type_=name, value=self.parameter[name])
+                                            for name in self.parameter.keys()])]))
 
             self.workspace.add_file(
                  ID=file_id,
                  file_grp=self.output_file_grp,
                  pageId=page_id,
                  mimetype='application/vnd.prima.page+xml',
-                 local_filename=os.path.join(self.output_file_grp, file_id) + '.xml')
+                 local_filename=os.path.join(self.output_file_grp, file_id) + '.xml',
+                 content=ocrd_models.ocrd_page.to_xml(pcgts)
+            )
 
 
 if __name__ == '__main__':

requirements.txt

@@ -9,5 +9,4 @@ scikit-learn
 tensorflow-gpu < 2.0
 scipy
 click
-ocrd >= 1.0.0b19
-
+ocrd >= 2.0.0