new page extraction model integration

Dockerfile

@@ -36,6 +36,8 @@ COPY . .
 COPY ocrd-tool.json .
 # prepackage ocrd-tool.json as ocrd-all-tool.json
 RUN ocrd ocrd-tool ocrd-tool.json dump-tools > $(dirname $(ocrd bashlib filename))/ocrd-all-tool.json
+# prepackage ocrd-all-module-dir.json
+RUN ocrd ocrd-tool ocrd-tool.json dump-module-dirs > $(dirname $(ocrd bashlib filename))/ocrd-all-module-dir.json
 # install everything and reduce image size
 RUN make install EXTRAS=OCR && rm -rf /build/eynollah
 # smoke test

Makefile

@@ -3,8 +3,9 @@ PIP ?= pip3
 EXTRAS ?=
 
 # DOCKER_BASE_IMAGE = artefakt.dev.sbb.berlin:5000/sbb/ocrd_core:v2.68.0
-DOCKER_BASE_IMAGE = docker.io/ocrd/core-cuda-tf2:v3.3.0
+DOCKER_BASE_IMAGE ?= docker.io/ocrd/core-cuda-tf2:latest
-DOCKER_TAG = ocrd/eynollah
+DOCKER_TAG ?= ocrd/eynollah
+DOCKER ?= docker
 
 #SEG_MODEL := https://qurator-data.de/eynollah/2021-04-25/models_eynollah.tar.gz
 #SEG_MODEL := https://qurator-data.de/eynollah/2022-04-05/models_eynollah_renamed.tar.gz
@@ -117,7 +118,7 @@ coverage:
 
 # Build docker image
 docker:
-	docker build \
+	$(DOCKER) build \
 	--build-arg DOCKER_BASE_IMAGE=$(DOCKER_BASE_IMAGE) \
 	--build-arg VCS_REF=$$(git rev-parse --short HEAD) \
 	--build-arg BUILD_DATE=$$(date -u +"%Y-%m-%dT%H:%M:%SZ") \

src/eynollah/ocrd-tool.json

@@ -38,7 +38,7 @@
          "textline_light": {
           "type": "boolean",
           "default": true,
-          "description": "Light version need textline light"
+          "description": "Light version need textline light. If this parameter set to true, this tool will try to return contoure of textlines instead of rectangle bounding box of textline with a faster method."
         },
         "tables": {
           "type": "boolean",
@@ -65,11 +65,6 @@
           "default": false,
           "description": "if this parameter set to true, this tool would check that input image need resizing and enhancement or not."
         },
-        "textline_light": {
-          "type": "boolean",
-          "default": false,
-          "description": "if this parameter set to true, this tool will try to return contoure of textlines instead of rectangle bounding box of textline with a faster method."
-        },
         "right_to_left": {
           "type": "boolean",
           "default": false,
@@ -79,6 +74,11 @@
           "type": "boolean",
           "default": false,
           "description": "ignore the special role of headings during reading order detection"
+        },
+        "reading_order_machine_based": {
+          "type": "boolean",
+          "default": false,
+          "description": "use data-driven (rather than rule-based) reading order detection"
         }
       },
       "resources": [

src/eynollah/processor.py

@@ -14,15 +14,17 @@ class EynollahProcessor(Processor):
         return 'ocrd-eynollah-segment'
 
     def setup(self) -> None:
-        if self.parameter['textline_light'] and not self.parameter['light_version']:
+        assert self.parameter
-            raise ValueError("Error: You set parameter 'textline_light' to enable light textline detection, "
+        if self.parameter['textline_light'] != self.parameter['light_version']:
-                             "but parameter 'light_version' is not enabled")
+            raise ValueError("Error: You must set or unset both parameter 'textline_light' (to enable light textline detection), "
+                             "and parameter 'light_version' (faster+simpler method for main region detection and deskewing)")
         self.eynollah = Eynollah(
             self.resolve_resource(self.parameter['models']),
             logger=self.logger,
             allow_enhancement=self.parameter['allow_enhancement'],
             curved_line=self.parameter['curved_line'],
             right2left=self.parameter['right_to_left'],
+            reading_order_machine_based=self.parameter['reading_order_machine_based'],
             ignore_page_extraction=self.parameter['ignore_page_extraction'],
             light_version=self.parameter['light_version'],
             textline_light=self.parameter['textline_light'],
@@ -56,6 +58,8 @@ class EynollahProcessor(Processor):
         - If ``ignore_page_extraction``, then attempt no cropping of the page.
         - If ``curved_line``, then compute contour polygons for text lines
           instead of simple bounding boxes.
+        - If ``reading_order_machine_based``, then detect reading order via
+          data-driven model instead of geometrical heuristics.
 
         Produce a new output file by serialising the resulting hierarchy.
         """