use OCR-D/core PAGE API for reading order and recursive regions

.gitignore

@@ -0,0 +1,2 @@
+*.egg-info
+__pycache__

requirements.txt

@@ -1,5 +1,4 @@
-numpy
+ocrd >= 2.23.2
 pandas
-click
 requests
 matplotlib

tsvtools/cli.py

@@ -1,13 +1,16 @@
+import json
+import glob
+import re
+import os
+from io import StringIO
+
 import numpy as np
 import click
 import pandas as pd
-from io import StringIO
-import os
-import xml.etree.ElementTree as ET
 import requests
-import json
-import glob
-import re
+
+from ocrd_models.ocrd_page import parse
+from ocrd_utils import bbox_from_points
 
 from .ned import ned
 from .ner import ner
@@ -75,12 +78,10 @@ def annotate_tsv(tsv_file, annotated_tsv_file):
 @click.option('--max-confidence', type=float, default=None)
 def page2tsv(page_xml_file, tsv_out_file, purpose, image_url, ner_rest_endpoint, ned_rest_endpoint, noproxy, scale_factor,
              ned_threshold, min_confidence, max_confidence):
-
     if purpose == "NERD":
         out_columns = ['No.', 'TOKEN', 'NE-TAG', 'NE-EMB', 'ID', 'url_id', 'left', 'right', 'top', 'bottom', 'conf']
     elif purpose == "OCR":
         out_columns = ['TEXT', 'url_id', 'left', 'right', 'top', 'bottom', 'conf']
-
         if min_confidence is not None and max_confidence is not None:
             out_columns += ['ocrconf']
     else:
@@ -89,57 +90,36 @@ def page2tsv(page_xml_file, tsv_out_file, purpose, image_url, ner_rest_endpoint,
     if noproxy:
         os.environ['no_proxy'] = '*'
 
-    tree = ET.parse(page_xml_file)
-    xmlns = tree.getroot().tag.split('}')[0].strip('{')
-
     urls = []
     if os.path.exists(tsv_out_file):
         parts = extract_doc_links(tsv_out_file)
-
         urls = [part['url'] for part in parts]
     else:
         pd.DataFrame([], columns=out_columns).to_csv(tsv_out_file, sep="\t", quoting=3, index=False)
 
+    pcgts = parse(page_xml_file)
     tsv = []
     line_info = []
-    for rgn_number, region in enumerate(tree.findall('.//{%s}TextRegion' % xmlns)):
-
-        for text_line in region.findall('.//{%s}TextLine' % xmlns):
-
-            points = [int(scale_factor * float(pos)) for coords in text_line.findall('./{%s}Coords' % xmlns) for p in
-                      coords.attrib['points'].split(' ') for pos in p.split(',')]
-
-            x_points, y_points = points[0::2], points[1::2]
 
-            left, right, top, bottom = min(x_points), max(x_points), min(y_points), max(y_points)
+    for region_idx, region in enumerate(pcgts.get_Page().get_AllRegions(classes=['Text'], order='reading-order')):
+        for text_line in region.get_TextLine():
+            left, top, right, bottom = [int(scale_factor * x) for x in bbox_from_points(text_line.get_Coords().points)]
 
             if min_confidence is not None and max_confidence is not None:
-                conf = np.max([float(text.attrib['conf']) for text in text_line.findall('./{%s}TextEquiv' % xmlns)])
+                conf = np.max([textequiv.conf for textequiv in text_line.get_TextEquiv()])
             else:
                 conf = np.nan
 
             line_info.append((len(urls), left, right, top, bottom, conf))
 
-            for word in text_line.findall('./{%s}Word' % xmlns):
-
-                for text_equiv in word.findall('./{%s}TextEquiv/{%s}Unicode' % (xmlns, xmlns)):
-                    text = text_equiv.text
-
-                    points = []
-
-                    for coords in word.findall('./{%s}Coords' % xmlns):
-
-                        # transform OCR coordinates using `scale_factor` to derive
-                        # correct coordinates for the web presentation image
-                        points += [int(scale_factor * float(pos))
-                                  for p in coords.attrib['points'].split(' ') for pos in p.split(',')]
-
-                    x_points, y_points = points[0::2], points[1::2]
-
-                    left, right, top, bottom = min(x_points), max(x_points), min(y_points), max(y_points)
+            for word in text_line.get_Word():
+                for text_equiv in word.get_TextEquiv():
+                    # transform OCR coordinates using `scale_factor` to derive
+                    # correct coordinates for the web presentation image
+                    left, top, right, bottom = [int(scale_factor * x) for x in bbox_from_points(word.get_Coords().points)]
 
-                    tsv.append((rgn_number, len(line_info)-1, left + (right - left) / 2.0, text,
-                                len(urls), left, right, top, bottom))
+                    tsv.append((region_idx, len(line_info) - 1, left + (right - left) / 2.0,
+                                text_equiv.get_Unicode(), len(urls), left, right, top, bottom))
 
     line_info = pd.DataFrame(line_info, columns=['url_id', 'left', 'right', 'top', 'bottom', 'conf'])