def test_xywh_from_points(self):
self.assertEqual(xywh_from_points('100,100 200,100 200,200 100,200'), {
'x': 100,
'y': 100,
'w': 100,
'h': 100
})
def process(self):
"""
Performs the line segmentation.
"""
with PyTessBaseAPI(path=TESSDATA_PREFIX) as tessapi:
for (n, input_file) in enumerate(self.input_files):
pcgts = from_file(self.workspace.download_file(input_file))
image_url = pcgts.get_Page().imageFilename
for region in pcgts.get_Page().get_TextRegion():
log.debug("Detecting lines in %s with tesseract",
region.id)
image = self.workspace.resolve_image_as_pil(
image_url,
polygon_from_points(region.get_Coords().points))
tessapi.SetImage(image)
offset = xywh_from_points(region.get_Coords().points)
for (line_no, component) in enumerate(
tessapi.GetComponentImages(RIL.TEXTLINE, True)):
line_id = '%s_line%04d' % (region.id, line_no)
line_xywh = component[1]
line_xywh['x'] += offset['x']
line_xywh['y'] += offset['y']
line_points = points_from_xywh(line_xywh)
region.add_TextLine(
TextLineType(id=line_id,
Coords=CoordsType(line_points)))
ID = concat_padded(self.output_file_grp, n)
self.workspace.add_file(
ID=ID,
file_grp=self.output_file_grp,
basename=ID + '.xml',
mimetype=MIMETYPE_PAGE,
content=to_xml(pcgts).encode('utf-8'),
)
def _process_existing_words(self, words, maxlevel, tessapi):
for word in words:
log.debug("Recognizing text in word '%s'", word.id)
word_xywh = xywh_from_points(word.get_Coords().points)
tessapi.SetRectangle(word_xywh['x'], word_xywh['y'],
word_xywh['w'], word_xywh['h'])
tessapi.SetPageSegMode(PSM.SINGLE_WORD)
if word.get_TextEquiv():
log.warning("Word '%s' already contains text results", word.id)
word_conf = tessapi.AllWordConfidences()
word_conf = word_conf[0] / 100.0 if word_conf else 0.0
word.add_TextEquiv(
TextEquivType(Unicode=tessapi.GetUTF8Text().rstrip("\n\f"),
conf=word_conf))
if maxlevel == 'word':
# maybe add TextEquiv alternatives via ChoiceIterator for WORD?
continue # next word (to avoid indentation below)
## glyph level:
glyphs = word.get_Glyph()
if glyphs:
## external glyph layout:
# raise Exception("existing annotation for Glyph level would clash with OCR results for word '%s'", word.id) # forcing external layout annotation for gylphs is worse with Tesseract
log.warning(
"Word '%s' contains glyphs already, recognition might be suboptimal",
word.id)
self._process_existing_glyphs(glyphs, tessapi)
else:
## internal glyph layout:
self._process_glyphs_in_word(word, tessapi.GetIterator())
def _process_lines(self, textlines, maxlevel, tessapi):
for line in textlines:
log.debug("Recognizing text in line '%s'", line.id)
line_xywh = xywh_from_points(line.get_Coords().points)
# log.debug("xywh: %s", line_xywh)
tessapi.SetRectangle(line_xywh['x'], line_xywh['y'],
line_xywh['w'], line_xywh['h'])
tessapi.SetPageSegMode(
PSM.SINGLE_LINE
) # RAW_LINE fails with Tesseract 3 models and is worse with Tesseract 4 models
if line.get_TextEquiv():
log.warning("Line '%s' already contains text results", line.id)
# tessapi.G
line_conf = tessapi.MeanTextConf(
) / 100.0 # iterator scores are arithmetic averages, too
# add line annotation unconditionally (i.e. even for word or glyph level):
line.add_TextEquiv(
TextEquivType(Unicode=tessapi.GetUTF8Text().rstrip("\n\f"),
conf=line_conf))
if maxlevel == 'line':
# maybe add TextEquiv alternatives via ChoiceIterator for TEXTLINE?
continue # next line (to avoid indentation below)
## word, or glyph level:
words = line.get_Word()
if words:
## external word layout:
# raise Exception("existing annotation for Word level would clash with OCR results for line '%s'", line.id) # forcing external layout annotation for words or glyphs is worse with Tesseract
log.warning(
"Line '%s' contains words already, recognition might be suboptimal",
line.id)
self._process_existing_words(words, maxlevel, tessapi)
else:
## internal word and glyph layout:
self._process_words_in_line(line, maxlevel, tessapi)
def process(self):
"""
Performs the (text) recognition.
"""
print(self.parameter)
with PyTessBaseAPI(path=TESSDATA_PREFIX, lang=DEFAULT_MODEL) as tessapi:
log.info("Using model %s in %s for recognition", get_languages()[0], get_languages()[1][-1])
for (n, input_file) in enumerate(self.input_files):
log.info("INPUT FILE %i / %s", n, input_file)
pcgts = from_file(self.workspace.download_file(input_file))
# TODO use binarized / gray
pil_image = self.workspace.resolve_image_as_pil(pcgts.get_Page().imageFilename)
tessapi.SetImage(pil_image)
# TODO slow
# tessapi.SetPageSegMode(PSM.SINGLE_LINE)
log.info("page %s", pcgts)
for region in pcgts.get_Page().get_TextRegion():
textlines = region.get_TextLine()
log.info("About to recognize text in %i lines of region '%s'", len(textlines), region.id)
for line in textlines:
log.debug("Recognizing text in line '%s'", line.id)
xywh = xywh_from_points(line.get_Coords().points)
tessapi.SetRectangle(xywh['x'], xywh['y'], xywh['w'], xywh['h'])
# log.debug("xywh: %s", xywh)
line.add_TextEquiv(TextEquivType(Unicode=tessapi.GetUTF8Text()))
# tessapi.G
# print(tessapi.AllWordConfidences())
ID = mets_file_id(self.output_file_grp, n)
self.add_output_file(
ID=ID,
file_grp=self.output_file_grp,
basename=ID + '.xml',
mimetype=MIMETYPE_PAGE,
content=to_xml(pcgts).encode('utf-8'),
)
def test_xywh_from_points_unordered(self):
self.assertEqual(
xywh_from_points('500,500 100,100 200,100 200,200 100,200'), {
'x': 100,
'y': 100,
'w': 400,
'h': 400
})
def _process_existing_glyphs(self, glyphs, tessapi):
for glyph in glyphs:
log.debug("Recognizing glyph in word '%s'", glyph.id)
glyph_xywh = xywh_from_points(glyph.get_Coords().points)
tessapi.SetRectangle(glyph_xywh['x'], glyph_xywh['y'],
glyph_xywh['w'], glyph_xywh['h'])
tessapi.SetPageSegMode(PSM.SINGLE_CHAR)
if glyph.get_TextEquiv():
log.warning("Glyph '%s' already contains text results",
glyph.id)
glyph_conf = tessapi.AllWordConfidences()
glyph_conf = glyph_conf[0] / 100.0 if glyph_conf else 0.0
glyph.add_TextEquiv(
TextEquivType(Unicode=tessapi.GetUTF8Text().rstrip("\n\f"),
conf=glyph_conf))
def _process_regions(self, regions, maxlevel, tessapi):
for region in regions:
log.debug("Recognizing text in region '%s'", region.id)
if maxlevel == 'region':
region_xywh = xywh_from_points(region.get_Coords().points)
tessapi.SetRectangle(region_xywh['x'], region_xywh['y'],
region_xywh['w'], region_xywh['h'])
tessapi.SetPageSegMode(PSM.SINGLE_BLOCK)
if region.get_TextEquiv():
log.warning("Region '%s' already contains text results",
region.id)
region.add_TextEquiv(
TextEquivType(
Unicode=tessapi.GetUTF8Text().rstrip("\n\f")))
continue # next region (to avoid indentation below)
## line, word, or glyph level:
textlines = region.get_TextLine()
if not textlines:
log.warning("Region '%s' contains no text lines", region.id)
else:
self._process_lines(textlines, maxlevel, tessapi)
def process(self):
"""
Performs the line segmentation.
"""
with PyTessBaseAPI(
psm=PSM.SINGLE_LINE,
path=TESSDATA_PREFIX,
) as tessapi:
for (n, input_file) in enumerate(self.input_files):
pcgts = from_file(self.workspace.download_file(input_file))
image_url = pcgts.get_Page().imageFilename
for region in pcgts.get_Page().get_TextRegion():
for line in region.get_TextLine():
log.debug("Detecting words in line '%s'", line.id)
image = self.workspace.resolve_image_as_pil(
image_url,
polygon_from_points(line.get_Coords().points))
tessapi.SetImage(image)
offset = xywh_from_points(line.get_Coords().points)
for (word_no, component) in enumerate(
tessapi.GetComponentImages(RIL.WORD, True)):
word_id = '%s_word%04d' % (line.id, word_no)
word_xywh = component[1]
word_xywh['x'] += offset['x']
word_xywh['y'] += offset['y']
line.add_Word(
WordType(id=word_id,
Coords=CoordsType(
points_from_xywh(word_xywh))))
ID = concat_padded(self.output_file_grp, n)
self.add_output_file(
ID=ID,
file_grp=self.output_file_grp,
basename=ID + '.xml',
mimetype=MIMETYPE_PAGE,
content=to_xml(pcgts).encode('utf-8'),
)
def process(self):
"""
Performs the (text) recognition.
"""
print(self.parameter)
if self.parameter['textequiv_level'] not in ['line', 'glyph']:
raise Exception("currently only implemented at the line/glyph level")
model = get_languages()[1][-1] # last installed model
if 'model' in self.parameter:
model = self.parameter['model']
if model not in get_languages()[1]:
raise Exception("configured model " + model + " is not installed")
with PyTessBaseAPI(path=TESSDATA_PREFIX, lang=model) as tessapi:
log.info("Using model %s in %s for recognition", model, get_languages()[0])
for (n, input_file) in enumerate(self.input_files):
log.info("INPUT FILE %i / %s", n, input_file)
pcgts = from_file(self.workspace.download_file(input_file))
# TODO use binarized / gray
pil_image = self.workspace.resolve_image_as_pil(pcgts.get_Page().imageFilename)
tessapi.SetImage(pil_image)
# TODO slow
# tessapi.SetPageSegMode(PSM.SINGLE_LINE)
log.info("page %s", pcgts)
for region in pcgts.get_Page().get_TextRegion():
textlines = region.get_TextLine()
log.info("About to recognize text in %i lines of region '%s'", len(textlines), region.id)
for line in textlines:
log.debug("Recognizing text in line '%s'", line.id)
xywh = xywh_from_points(line.get_Coords().points)
tessapi.SetRectangle(xywh['x'], xywh['y'], xywh['w'], xywh['h'])
tessapi.SetPageSegMode(PSM.SINGLE_LINE)
# log.debug("xywh: %s", xywh)
line.add_TextEquiv(TextEquivType(Unicode=tessapi.GetUTF8Text()))
# tessapi.G
# print(tessapi.AllWordConfidences())
if self.parameter['textequiv_level'] == 'glyph':
for word in line.get_Word():
log.debug("Recognizing text in word '%s'", word.id)
xywh = xywh_from_points(word.get_Coords().points)
tessapi.SetRectangle(xywh['x'], xywh['y'], xywh['w'], xywh['h'])
tessapi.SetPageSegMode(PSM.SINGLE_WORD)
word.add_TextEquiv(TextEquivType(Unicode=tessapi.GetUTF8Text()))
result_it = tessapi.GetIterator()
for (result_no, result) in enumerate(iterate_level(result_it, RIL.SYMBOL)):
#symb = result.GetUTF8Text(RIL.SYMBOL) # is first choice?
#conf = result.Confidence(RIL.SYMBOL) # is first choice?
bbox = result.BoundingBox(RIL.SYMBOL)
if bbox == None:
continue
glyph_id = '%s_glyph%04d' % (word.id, result_no)
log.debug("Recognizing text in glyph '%s'", glyph_id)
glyph = GlyphType(id=glyph_id, Coords=CoordsType(points_from_x0y0x1y1(bbox)))
word.add_Glyph(glyph)
choice_it = result.GetChoiceIterator()
for (choice_no, choice) in enumerate(choice_it):
alternative_symb = choice.GetUTF8Text()
alternative_conf = choice.Confidence()
glyph.add_TextEquiv(TextEquivType(index=choice_no, conf=alternative_conf, Unicode=alternative_symb))
ID = concat_padded(self.output_file_grp, n)
self.add_output_file(
ID=ID,
file_grp=self.output_file_grp,
basename=ID + '.xml',
mimetype=MIMETYPE_PAGE,
content=to_xml(pcgts),
)
