binarize/dewarp/recognize: move getLogger to processing context / introduce setup method

Author: Robert Sachunsky

ocrd_cis/ocropy/binarize.py

@@ -76,12 +76,16 @@ def __init__(self, *args, **kwargs):
         kwargs['ocrd_tool'] = self.ocrd_tool['tools'][TOOL]
         kwargs['version'] = self.ocrd_tool['version']
         super(OcropyBinarize, self).__init__(*args, **kwargs)
-        LOG = getLogger('processor.OcropyBinarize')
         if hasattr(self, 'output_file_grp'):
-            if self.parameter['grayscale'] and self.parameter['method'] != 'ocropy':
-                LOG.critical('requested method %s does not support grayscale normalized output',
-                             self.parameter['method'])
-                raise Exception('only method=ocropy allows grayscale=true')
+            # processing context
+            self.setup()
+    
+    def setup(self):
+        self.logger = getLogger('processor.OcropyBinarize')
+        if self.parameter['grayscale'] and self.parameter['method'] != 'ocropy':
+            self.logger.critical('requested method %s does not support grayscale normalized output',
+                                 self.parameter['method'])
+            raise Exception('only method=ocropy allows grayscale=true')
 
     def process(self):
         """Binarize (and optionally deskew/despeckle) the pages/regions/lines of the workspace.
@@ -105,13 +109,12 @@ def process(self):
 
         Produce a new output file by serialising the resulting hierarchy.
         """
-        LOG = getLogger('processor.OcropyBinarize')
         level = self.parameter['level-of-operation']
         assert_file_grp_cardinality(self.input_file_grp, 1)
         assert_file_grp_cardinality(self.output_file_grp, 1)
 
         for (n, input_file) in enumerate(self.input_files):
-            LOG.info("INPUT FILE %i / %s", n, input_file.pageId or input_file.ID)
+            self.logger.info("INPUT FILE %i / %s", n, input_file.pageId or input_file.ID)
             file_id = make_file_id(input_file, self.output_file_grp)
 
             pcgts = page_from_file(self.workspace.download_file(input_file))
@@ -139,7 +142,7 @@ def process(self):
                 dpi = page_image_info.resolution
                 if page_image_info.resolutionUnit == 'cm':
                     dpi *= 2.54
-                LOG.info('Page "%s" uses %f DPI', page_id, dpi)
+                self.logger.info('Page "%s" uses %f DPI', page_id, dpi)
                 zoom = 300.0/dpi
             else:
                 zoom = 1
@@ -151,7 +154,7 @@ def process(self):
                 regions = page.get_TextRegion() + (
                     page.get_TableRegion() if level == 'region' else [])
                 if not regions:
-                    LOG.warning('Page "%s" contains no text regions', page_id)
+                    self.logger.warning('Page "%s" contains no text regions', page_id)
                 for region in regions:
                     region_image, region_xywh = self.workspace.image_from_segment(
                         region, page_image, page_xywh, feature_filter='binarized')
@@ -161,7 +164,8 @@ def process(self):
                         continue
                     lines = region.get_TextLine()
                     if not lines:
-                        LOG.warning('Page "%s" region "%s" contains no text lines', page_id, region.id)
+                        self.logger.warning('Page "%s" region "%s" contains no text lines',
+                                            page_id, region.id)
                     for line in lines:
                         line_image, line_xywh = self.workspace.image_from_segment(
                             line, region_image, region_xywh, feature_filter='binarized')
@@ -179,12 +183,11 @@ def process(self):
                 local_filename=file_path,
                 mimetype=MIMETYPE_PAGE,
                 content=to_xml(pcgts))
-            LOG.info('created file ID: %s, file_grp: %s, path: %s',
-                     file_id, self.output_file_grp, out.local_filename)
+            self.logger.info('created file ID: %s, file_grp: %s, path: %s',
+                             file_id, self.output_file_grp, out.local_filename)
 
     def process_page(self, page, page_image, page_xywh, zoom, page_id, file_id):
-        LOG = getLogger('processor.OcropyBinarize')
-        LOG.info("About to binarize page '%s'", page_id)
+        self.logger.info("About to binarize page '%s'", page_id)
         features = page_xywh['features']
         if 'angle' in page_xywh and page_xywh['angle']:
             # orientation has already been annotated (by previous deskewing),
@@ -229,8 +232,7 @@ def process_page(self, page, page_image, page_xywh, zoom, page_id, file_id):
             comments=features))
 
     def process_region(self, region, region_image, region_xywh, zoom, page_id, file_id):
-        LOG = getLogger('processor.OcropyBinarize')
-        LOG.info("About to binarize page '%s' region '%s'", page_id, region.id)
+        self.logger.info("About to binarize page '%s' region '%s'", page_id, region.id)
         features = region_xywh['features']
         if 'angle' in region_xywh and region_xywh['angle']:
             # orientation has already been annotated (by previous deskewing),
@@ -277,9 +279,8 @@ def process_region(self, region, region_image, region_xywh, zoom, page_id, file_
             comments=features))
 
     def process_line(self, line, line_image, line_xywh, zoom, page_id, region_id, file_id):
-        LOG = getLogger('processor.OcropyBinarize')
-        LOG.info("About to binarize page '%s' region '%s' line '%s'",
-                 page_id, region_id, line.id)
+        self.logger.info("About to binarize page '%s' region '%s' line '%s'",
+                         page_id, region_id, line.id)
         features = line_xywh['features']
         bin_image, angle = binarize(line_image,
                                     method=self.parameter['method'],
@@ -294,8 +295,8 @@ def process_line(self, line, line_image, line_xywh, zoom, page_id, region_id, fi
         #orientation = -angle
         #orientation = 180 - (180 - orientation) % 360 # map to [-179.999,180]
         #line.set_orientation(orientation) # does not exist on line level!
-        LOG.warning("cannot add orientation %.2f to page '%s' region '%s' line '%s'",
-                    -angle, page_id, region_id, line.id)
+        self.logger.warning("cannot add orientation %.2f to page '%s' region '%s' line '%s'",
+                            -angle, page_id, region_id, line.id)
         bin_image = remove_noise(bin_image,
                                  maxsize=self.parameter['noise_maxsize'])
         if self.parameter['noise_maxsize']:

ocrd_cis/ocropy/dewarp.py

@@ -76,9 +76,14 @@ def __init__(self, *args, **kwargs):
         kwargs['version'] = self.ocrd_tool['version']
         super(OcropyDewarp, self).__init__(*args, **kwargs)
         if hasattr(self, 'output_file_grp'):
-            # defaults from ocrolib.lineest:
-            range_ = self.parameter['range']
-            self.lnorm = lineest.CenterNormalizer(params=(range_, 1.0, 0.3))
+            # processing context
+            self.setup()
+    
+    def setup(self):
+        # defaults from ocrolib.lineest:
+        range_ = self.parameter['range']
+        self.lnorm = lineest.CenterNormalizer(params=(range_, 1.0, 0.3))
+        self.logger = getLogger('processor.OcropyDewarp')
 
     def process(self):
         """Dewarp the lines of the workspace.
@@ -99,12 +104,11 @@ def process(self):
 
         Produce a new output file by serialising the resulting hierarchy.
         """
-        LOG = getLogger('processor.OcropyDewarp')
         assert_file_grp_cardinality(self.input_file_grp, 1)
         assert_file_grp_cardinality(self.output_file_grp, 1)
 
         for (n, input_file) in enumerate(self.input_files):
-            LOG.info("INPUT FILE %i / %s", n, input_file.pageId or input_file.ID)
+            self.logger.info("INPUT FILE %i / %s", n, input_file.pageId or input_file.ID)
             file_id = make_file_id(input_file, self.output_file_grp)
 
             pcgts = page_from_file(self.workspace.download_file(input_file))
@@ -132,36 +136,36 @@ def process(self):
                 dpi = page_image_info.resolution
                 if page_image_info.resolutionUnit == 'cm':
                     dpi *= 2.54
-                LOG.info('Page "%s" uses %f DPI', page_id, dpi)
+                self.logger.info('Page "%s" uses %f DPI', page_id, dpi)
                 zoom = 300.0/dpi
             else:
                 zoom = 1
 
             regions = page.get_TextRegion()
             if not regions:
-                LOG.warning('Page "%s" contains no text regions', page_id)
+                self.logger.warning('Page "%s" contains no text regions', page_id)
             for region in regions:
                 region_image, region_xywh = self.workspace.image_from_segment(
                     region, page_image, page_xywh)
 
                 lines = region.get_TextLine()
                 if not lines:
-                    LOG.warning('Region %s contains no text lines', region.id)
+                    self.logger.warning('Region %s contains no text lines', region.id)
                 for line in lines:
                     line_image, line_xywh = self.workspace.image_from_segment(
                         line, region_image, region_xywh)
 
-                    LOG.info("About to dewarp page '%s' region '%s' line '%s'",
-                             page_id, region.id, line.id)
+                    self.logger.info("About to dewarp page '%s' region '%s' line '%s'",
+                                     page_id, region.id, line.id)
                     try:
                         dew_image = dewarp(line_image, self.lnorm, check=True,
                                            max_neighbour=self.parameter['max_neighbour'],
                                            zoom=zoom)
                     except InvalidLine as err:
-                        LOG.error('cannot dewarp line "%s": %s', line.id, err)
+                        self.logger.error('cannot dewarp line "%s": %s', line.id, err)
                         continue
                     except InadequateLine as err:
-                        LOG.warning('cannot dewarp line "%s": %s', line.id, err)
+                        self.logger.warning('cannot dewarp line "%s": %s', line.id, err)
                         # as a fallback, simply pad the image vertically
                         # (just as dewarping would do on average, so at least
                         #  this line has similar margins as the others):
@@ -188,5 +192,5 @@ def process(self):
                 local_filename=file_path,
                 mimetype=MIMETYPE_PAGE,
                 content=to_xml(pcgts))
-            LOG.info('created file ID: %s, file_grp: %s, path: %s',
-                     file_id, self.output_file_grp, out.local_filename)
+            self.logger.info('created file ID: %s, file_grp: %s, path: %s',
+                             file_id, self.output_file_grp, out.local_filename)

ocrd_cis/ocropy/recognize.py

@@ -89,6 +89,19 @@ def __init__(self, *args, **kwargs):
         kwargs['ocrd_tool'] = self.ocrd_tool['tools'][TOOL]
         kwargs['version'] = self.ocrd_tool['version']
         super(OcropyRecognize, self).__init__(*args, **kwargs)
+        if hasattr(self, 'output_file_grp'):
+            # processing context
+            self.setup()
+    
+    def setup(self):
+        self.logger = getLogger('processor.OcropyRecognize')
+        # from ocropus-rpred:
+        self.network = load_object(self.get_model(), verbose=1)
+        for x in self.network.walk():
+            x.postLoad()
+        for x in self.network.walk():
+            if isinstance(x, lstm.LSTM):
+                x.allocate(5000)
 
     def get_model(self):
         """Search for the model file.  First checks if
@@ -106,7 +119,6 @@ def get_model(self):
         return model
 
     def process(self):
-
         """Recognize lines / words / glyphs of the workspace.
 
         Open and deserialise each PAGE input file and its respective image,
@@ -131,24 +143,13 @@ def process(self):
 
         Produce a new output file by serialising the resulting hierarchy.
         """
-        LOG = getLogger('processor.OcropyRecognize')
-
         assert_file_grp_cardinality(self.input_file_grp, 1)
         assert_file_grp_cardinality(self.output_file_grp, 1)
-
-        # from ocropus-rpred:
-        self.network = load_object(self.get_model(), verbose=1)
-        for x in self.network.walk():
-            x.postLoad()
-        for x in self.network.walk():
-            if isinstance(x, lstm.LSTM):
-                x.allocate(5000)
-
         maxlevel = self.parameter['textequiv_level']
 
-        # LOG.info("Using model %s in %s for recognition", model)
+        # self.logger.info("Using model %s in %s for recognition", model)
         for (n, input_file) in enumerate(self.input_files):
-            LOG.info("INPUT FILE %i / %s", n, input_file.pageId or input_file.ID)
+            self.logger.info("INPUT FILE %i / %s", n, input_file.pageId or input_file.ID)
             pcgts = page_from_file(self.workspace.download_file(input_file))
             page_id = pcgts.pcGtsId or input_file.pageId or input_file.ID # (PageType has no id)
             page = pcgts.get_Page()
@@ -169,11 +170,11 @@ def process(self):
             page_image, page_coords, _ = self.workspace.image_from_page(
                 page, page_id)
 
-            LOG.info("Recognizing text in page '%s'", page_id)
+            self.logger.info("Recognizing text in page '%s'", page_id)
             # region, line, word, or glyph level:
             regions = page.get_TextRegion()
             if not regions:
-                LOG.warning("Page '%s' contains no text regions", page_id)
+                self.logger.warning("Page '%s' contains no text regions", page_id)
             self.process_regions(regions, maxlevel, page_image, page_coords)
 
             # update METS (add the PAGE file):
@@ -187,21 +188,20 @@ def process(self):
                 local_filename=file_path,
                 mimetype=MIMETYPE_PAGE,
                 content=to_xml(pcgts))
-            LOG.info('created file ID: %s, file_grp: %s, path: %s',
-                     file_id, self.output_file_grp, out.local_filename)
+            self.logger.info('created file ID: %s, file_grp: %s, path: %s',
+                             file_id, self.output_file_grp, out.local_filename)
 
     def process_regions(self, regions, maxlevel, page_image, page_coords):
-        LOG = getLogger('processor.OcropyRecognize')
         edits = 0
         lengs = 0
         for region in regions:
             region_image, region_coords = self.workspace.image_from_segment(
                 region, page_image, page_coords)
 
-            LOG.info("Recognizing text in region '%s'", region.id)
+            self.logger.info("Recognizing text in region '%s'", region.id)
             textlines = region.get_TextLine()
             if not textlines:
-                LOG.warning("Region '%s' contains no text lines", region.id)
+                self.logger.warning("Region '%s' contains no text lines", region.id)
             else:
                 edits_, lengs_ = self.process_lines(textlines, maxlevel, region_image, region_coords)
                 edits += edits_
@@ -212,28 +212,27 @@ def process_regions(self, regions, maxlevel, page_image, page_coords):
                                         else u'' for line in textlines)
             region.set_TextEquiv([TextEquivType(Unicode=region_unicode)])
         if lengs > 0:
-            LOG.info('CER: %.1f%%', 100.0 * edits / lengs)
+            self.logger.info('CER: %.1f%%', 100.0 * edits / lengs)
 
     def process_lines(self, textlines, maxlevel, region_image, region_coords):
-        LOG = getLogger('processor.OcropyRecognize')
         edits = 0
         lengs = 0
         for line in textlines:
             line_image, line_coords = self.workspace.image_from_segment(
                 line, region_image, region_coords)
 
-            LOG.info("Recognizing text in line '%s'", line.id)
+            self.logger.info("Recognizing text in line '%s'", line.id)
             if line.get_TextEquiv():
                 linegt = line.TextEquiv[0].Unicode
             else:
                 linegt = ''
-            LOG.debug("GT  '%s': '%s'", line.id, linegt)
+            self.logger.debug("GT  '%s': '%s'", line.id, linegt)
             # remove existing annotation below line level:
             line.set_TextEquiv([])
             line.set_Word([])
 
             if line_image.size[1] < 16:
-                LOG.debug("ERROR: bounding box is too narrow at line %s", line.id)
+                self.logger.debug("ERROR: bounding box is too narrow at line %s", line.id)
                 continue
             # resize image to 48 pixel height
             final_img, scale = resize_keep_ratio(line_image)
@@ -243,9 +242,9 @@ def process_lines(self, textlines, maxlevel, region_image, region_coords):
                 linepred, clist, rlist, confidlist = recognize(
                     final_img, self.pad, self.network, check=True)
             except Exception as err:
-                LOG.debug('ERROR: error processing line "%s": %s', line.id, err)
+                self.logger.debug('error processing line "%s": %s', line.id, err)
                 continue
-            LOG.debug("OCR '%s': '%s'", line.id, linepred)
+            self.logger.debug("OCR '%s': '%s'", line.id, linepred)
             edits += Levenshtein.distance(linepred, linegt)
             lengs += len(linegt)