def segmentation_kraken(doc, method=u'segment_kraken', black_colseps=False):
"""
Performs page segmentation using kraken's built-in algorithm and writes a
skeleton TEI file.
Args:
doc (unicode, unicode): The input document tuple
method (unicode): The suffix string append to all output files
black_colseps (bool): Assume black column separator instead of white
ones.
Returns:
Two storage tuples with the first one containing the segmentation and
the second one being the file the segmentation was calculated upon.
"""
input_path = storage.get_abs_path(*doc)
output_path, ext = os.path.splitext(
storage.insert_suffix(input_path, method))
logger.debug('Reading image using PIL')
img = Image.open(input_path)
with open(output_path + '.xml', 'w') as fp:
logger.debug('Initializing TEI with {} ({} {})'.format(
doc[1], *img.size))
tei = OCRRecord()
tei.img = storage.get_url(*doc)
tei.dimensions = img.size
tei.title = os.path.basename(doc[1])
tei.add_respstmt('kraken', 'page segmentation')
for seg in pageseg.segment(img, black_colseps=black_colseps)['boxes']:
logger.debug('Found line at {} {} {} {}'.format(*seg))
tei.add_line(seg)
logger.debug('Write segmentation to {}'.format(fp.name))
tei.write_tei(fp)
return storage.get_storage_path(output_path + '.xml')
def any_to_png(doc, method=u'any_to_png'):
"""
Converts an image (color or otherwise) in any format recognized by pillow
to PNG.
The pillow image library relies on external libraries for loading and
saving Image data. To recognize the most common image formats used for
digital archival you'll need:
- libtiff
- zlib
- libjpeg
- openjpeg (version 2.0 +)
- libwebp
To have access to all formats run (on Debian/Ubuntu):
.. code-block:: console
# apt-get -y install libtiff5-dev libjpeg62-turbo-dev zlib1g-dev \
libwebp-dev libopenjp2-dev
Args:
doc (unicode, unicode): The input document tuple
method (unicode): The suffix string appended to all output files.
Returns:
(unicode, unicode): Storage tuple of the output file
"""
input_path = storage.get_abs_path(*doc)
output_path = os.path.splitext(storage.insert_suffix(input_path, method))[0] + '.png'
return storage.get_storage_path(image.any_to_png(input_path, output_path))
def ocr_tesseract(doc, method=u'ocr_tesseract', languages=None,
extended=False):
"""
Runs tesseract on an input document.
Args:
doc (unicode, unicode): The input document tuple
method (unicode): The suffix string appended to all output files
languages (list): A list of tesseract classifier identifiers
extended (bool): Switch to enable extended hOCR generation containing
character cuts and confidences. Has no effect when
direct or legacy implementation is used.
Returns:
(unicode, unicode): Storage tuple for the output file
"""
image_path = storage.get_abs_path(*doc[1])
# rewrite the segmentation file to lines in UZN format
logger.debug('Rewriting TEI ({}) -> UZN ({})'.format(doc[0][1],
splitext(doc[1][1])[0]
+ '.uzn'))
seg = TEIFacsimile()
with storage.StorageFile(*doc[0]) as fp:
seg.read(fp)
with storage.StorageFile(doc[1][0], splitext(doc[1][1])[0] + '.uzn', mode='wb') as fp:
uzn = UZNWriter(fp)
for line in seg.lines:
uzn.writerow(*line[:4])
if isinstance(languages, basestring):
languages = [languages]
output_path = storage.insert_suffix(image_path, method, *languages)
logger.debug('Invoking tesseract with {} call method'.format(implementation))
if implementation == 'legacy':
result_path = output_path + '.html'
ocr_direct(image_path, output_path, languages)
elif implementation == 'direct':
result_path = output_path + '.hocr'
ocr_direct(image_path, output_path, languages)
elif implementation == 'capi':
result_path = output_path + '.xml'
ocr_capi(image_path, result_path, seg, languages, extended)
else:
raise NidabaTesseractException('Invalid implementation selected',
implementation)
if not result_path[-4:] == '.xml':
logger.debug('Converting hOCR ({}) -> TEI ({})'.format(result_path,
output_path +
'.xml'))
tei = TEIFacsimile()
with open(result_path) as fp:
tei.load_hocr(fp)
os.unlink(result_path)
with open(output_path + '.xml', 'wb') as fp:
tei.write(fp)
result_path = output_path + '.xml'
return storage.get_storage_path(result_path)
def segmentation_kraken(doc, method=u'segment_kraken', black_colseps=False):
"""
Performs page segmentation using kraken's built-in algorithm and writes a
skeleton TEI file.
Args:
doc (unicode, unicode): The input document tuple
method (unicode): The suffix string append to all output files
black_colseps (bool): Assume black column separator instead of white
ones.
Returns:
Two storage tuples with the first one containing the segmentation and
the second one being the file the segmentation was calculated upon.
"""
input_path = storage.get_abs_path(*doc)
output_path, ext = os.path.splitext(storage.insert_suffix(input_path,
method))
logger.debug('Reading image using PIL')
img = Image.open(input_path)
with open(output_path + '.xml', 'w') as fp:
logger.debug('Initializing TEI with {} ({} {})'.format(doc[1], *img.size))
tei = OCRRecord()
tei.img = storage.get_url(*doc)
tei.dimensions = img.size
tei.title = os.path.basename(doc[1])
tei.add_respstmt('kraken', 'page segmentation')
for seg in pageseg.segment(img, black_colseps=black_colseps)['boxes']:
logger.debug('Found line at {} {} {} {}'.format(*seg))
tei.add_line(seg)
logger.debug('Write segmentation to {}'.format(fp.name))
tei.write_tei(fp)
return storage.get_storage_path(output_path + '.xml')
def sauvola(doc, method=u'sauvola', whsize=10, factor=0.35):
"""
Binarizes an input document utilizing Sauvola thresholding as described in
[0]. Expects 8bpp grayscale images as input.
[0] Sauvola, Jaakko, and Matti Pietikäinen. "Adaptive document image
binarization." Pattern recognition 33.2 (2000): 225-236.
Args:
doc (unicode): The input document tuple.
method (unicode): The suffix string appended to all output files
whsize (int): The window width and height that local statistics are
calculated on are twice the value of whsize. The minimal
value is 2.
factor (float): The threshold reduction factor due to variance. 0 =<
factor < 1.
Returns:
(unicode, unicode): Storage tuple of the output file
Raises:
NidabaInvalidParameterException: Input parameters are outside the valid
range.
"""
input_path = storage.get_abs_path(*doc)
output_path = storage.insert_suffix(input_path, method, unicode(whsize),
unicode(factor))
lept_sauvola(input_path, output_path, whsize, factor)
return storage.get_storage_path(output_path)
def otsu(doc, method=u'otsu', thresh=100, mincount=50, bgval=255,
smoothx=2, smoothy=2):
"""
Binarizes an input document utilizing a naive implementation of Otsu's
thresholding.
Args:
doc (unicode, unicode): The input document tuple.
id (unicode): The nidaba batch identifier this task is a part of
method (unicode): The suffix string appended to all output files.
Returns:
(unicode, unicode): Storage tuple of the output file
Raises:
NidabaInvalidParameterException: Input parameters are outside the valid
range.
"""
input_path = storage.get_abs_path(*doc)
output_path = storage.insert_suffix(input_path, method, unicode(thresh),
unicode(mincount), unicode(bgval),
unicode(smoothx), unicode(smoothy))
if smoothx < 0 or smoothy < 0 or bgval < 0 or thresh < 0 or mincount < 0:
raise NidabaInvalidParameterException('Parameters (' + unicode(thresh)
+ ',' + unicode(mincount) + ',' +
unicode(bgval) + ',' +
unicode(smoothx) + ',' +
unicode(smoothy) + ',' +
') outside of valid range')
return storage.get_storage_path(leper.otsu_binarize(input_path,
output_path, thresh,
mincount, bgval,
smoothx, smoothy))
def sauvola(doc, method=u'sauvola', whsize=10, factor=0.35):
"""
Binarizes an input document utilizing Sauvola thresholding as described in
[0]. Expects 8bpp grayscale images as input.
[0] Sauvola, Jaakko, and Matti Pietikäinen. "Adaptive document image
binarization." Pattern recognition 33.2 (2000): 225-236.
Args:
doc (unicode): The input document tuple.
method (unicode): The suffix string appended to all output files
whsize (int): The window width and height that local statistics are
calculated on are twice the value of whsize. The minimal
value is 2.
factor (float): The threshold reduction factor due to variance. 0 =<
factor < 1.
Returns:
(unicode, unicode): Storage tuple of the output file
Raises:
NidabaInvalidParameterException: Input parameters are outside the valid
range.
"""
input_path = storage.get_abs_path(*doc)
output_path = storage.insert_suffix(input_path, method, unicode(whsize),
unicode(factor))
lept_sauvola(input_path, output_path, whsize, factor)
return storage.get_storage_path(output_path)
def any_to_png(doc, method=u'any_to_png'):
"""
Converts an image (color or otherwise) in any format recognized by pillow
to PNG.
The pillow image library relies on external libraries for loading and
saving Image data. To recognize the most common image formats used for
digital archival you'll need:
- libtiff
- zlib
- libjpeg
- openjpeg (version 2.0 +)
- libwebp
To have access to all formats run (on Debian/Ubuntu):
.. code-block:: console
# apt-get -y install libtiff5-dev libjpeg62-turbo-dev zlib1g-dev \
libwebp-dev libopenjp2-dev
Args:
doc (unicode, unicode): The input document tuple
method (unicode): The suffix string appended to all output files.
Returns:
(unicode, unicode): Storage tuple of the output file
"""
input_path = storage.get_abs_path(*doc)
output_path = os.path.splitext(storage.insert_suffix(input_path,
method))[0] + '.png'
return storage.get_storage_path(image.any_to_png(input_path, output_path))
def text_lexicality(doc, method=u'text_lexicality', language=u'', divert=True):
"""
Calculates the lexicality of text in input documents.
Args:
doc (unicode, unicode): The input document tuple
method (unicode): The suffix string appended to the output file.
Returns:
(unicode, unicode): Storage tuple of the output document
"""
input_path = storage.get_abs_path(*doc[0])
output_path = storage.insert_suffix(input_path, method,
os.path.basename(input_path))
dictionary = storage.get_abs_path(*nidaba_cfg['lang_dicts'][language]['dictionary'])
with storage.StorageFile(*doc) as fp:
tei = OCRRecord()
tei.load_tei(fp)
cnt = 0
err_cnt = 0
for seg_id, segment in facsimile.segments.iteritems():
tok = alg.sanitize(''.join(x['grapheme'] for x in segment['content'].itervalues()))
tok = regex.sub('[^\w]', '', key)
cnt += 1
if not alg.mmap_bin_search(tok, dictionary, entryparser_fn=alg.key_for_single_word):
err_cnt += 1
if not divert:
storage.write_text(*storage.get_storage_path(output_path),
text=unicode(err_cnt / float(cnt)))
return output_path
else:
return {'edit_ratio': err_cnt / float(cnt), 'ground_truth': '', 'doc': doc}
def nlbin(doc, method=u'nlbin', threshold=0.5, zoom=0.5, escale=1.0,
border=0.1, perc=80, range=20, low=5, high=90):
"""
Binarizes an input document utilizing ocropus'/kraken's nlbin algorithm.
Args:
doc (unicode, unicode): The input document tuple.
method (unicode): The suffix string appended to all output files.
threshold (float):
zoom (float):
escale (float):
border (float)
perc (int):
range (int):
low (int):
high (int):
Returns:
(unicode, unicode): Storage tuple of the output file
Raises:
NidabaInvalidParameterException: Input parameters are outside the valid
range.
"""
input_path = storage.get_abs_path(*doc)
output_path = storage.insert_suffix(input_path, method, unicode(threshold),
unicode(zoom), unicode(escale),
unicode(border), unicode(perc),
unicode(range), unicode(low),
unicode(high))
kraken_nlbin(input_path, output_path, threshold, zoom, escale, border,
perc, range, low, high)
return storage.get_storage_path(output_path)
def ocr_tesseract(doc,
method=u'ocr_tesseract',
languages=None,
extended=False):
"""
Runs tesseract on an input document.
Args:
doc (unicode, unicode): The input document tuple
method (unicode): The suffix string appended to all output files
languages (list): A list of tesseract classifier identifiers
extended (bool): Switch to enable extended hOCR generation containing
character cuts and confidences. Has no effect when
direct or legacy implementation is used.
Returns:
(unicode, unicode): Storage tuple for the output file
"""
seg = OCRRecord()
with storage.StorageFile(*doc) as fp:
seg.load_tei(fp)
with storage.StorageFile(doc[0], splitext(doc[1])[0] + '.uzn',
mode='wb') as fp:
uzn = UZNWriter(fp)
for line in seg.lines.itervalues():
uzn.writerow(*line['bbox'])
image_path = storage.get_abs_path(*storage.get_storage_path_url(seg.img))
if isinstance(languages, basestring):
languages = [languages]
output_path = storage.insert_suffix(image_path, method, *languages)
logger.debug(
'Invoking tesseract with {} call method'.format(implementation))
if implementation == 'legacy':
result_path = output_path + '.html'
ocr_direct(image_path, output_path, languages)
elif implementation == 'direct':
result_path = output_path + '.hocr'
ocr_direct(image_path, output_path, languages)
elif implementation == 'capi':
result_path = output_path + '.xml'
ocr_capi(image_path, result_path, seg, languages, extended)
else:
raise NidabaTesseractException('Invalid implementation selected',
implementation)
if not result_path[-4:] == '.xml':
logger.debug('Converting hOCR ({}) -> TEI ({})'.format(
result_path, output_path + '.xml'))
tei = OCRRecord()
with open(result_path) as fp:
tei.load_hocr(fp)
os.unlink(result_path)
with open(output_path + '.xml', 'wb') as fp:
tei.write_tei(fp)
result_path = output_path + '.xml'
return storage.get_storage_path(result_path)
def spell_check(doc,
method=u'spell_check',
language=u'',
filter_punctuation=False):
"""
Adds spelling suggestions to an TEI XML document.
Alternative spellings for each segment will be included in a choice
tagcontaining a series of corr tags with the original segment appearing
beneath a sic element. Correct words, i.e. words appearing verbatim in the
dictionary, are left untouched.
Args:
doc (unicode, unicode): The input document tuple.
method (unicode): The suffix string appended to the output file.
language (unicode): Identifier defined in the nidaba configuration as a
valid dictionary.
filter_punctuation (bool): Switch to filter punctuation inside
``seg``
Returns:
(unicode, unicode): Storage tuple of the output document
"""
input_path = storage.get_abs_path(*doc)
output_path = storage.insert_suffix(input_path, method, language,
unicode(filter_punctuation))
dictionary = storage.get_abs_path(
*nidaba_cfg['lang_dicts'][language]['dictionary'])
del_dictionary = storage.get_abs_path(
*nidaba_cfg['lang_dicts'][language]['deletion_dictionary'])
with storage.StorageFile(*doc) as fp:
logger.debug('Reading TEI ({})'.format(fp.abs_path))
tei = TEIFacsimile()
tei.read(fp)
logger.debug('Performing spell check')
ret = lex.tei_spellcheck(tei, dictionary, del_dictionary,
filter_punctuation)
with storage.StorageFile(*storage.get_storage_path(output_path),
mode='wb') as fp:
logger.debug('Writing TEI ({})'.format(fp.abs_path))
ret.write(fp)
return storage.get_storage_path(output_path)
def deskew(doc, method=u'deskew'):
"""
Removes skew (rotational distortion) from an 1bpp input image.
Args:
doc (unicode, unicode): The input document tuple.
id (unicode): The nidaba batch identifier this task is a part of
method (unicode): The suffix string appended to all output files.
Returns:
(unicode, unicode): Storage tuple of the output file
"""
input_path = storage.get_abs_path(*doc)
output_path = storage.insert_suffix(input_path, method)
return storage.get_storage_path(leper.deskew(input_path, output_path))
def rgb_to_gray(doc, method=u'rgb_to_gray'):
"""
Converts an arbitrary bit depth image to grayscale and writes it back
appending a suffix.
Args:
doc (unicode, unicode): The input document tuple
method (unicode): The suffix string appended to all output files.
Returns:
(unicode, unicode): Storage tuple of the output file
"""
input_path = storage.get_abs_path(*doc)
output_path = storage.insert_suffix(input_path, method)
return storage.get_storage_path(image.rgb_to_gray(input_path, output_path))
def deskew(doc, method=u'deskew'):
"""
Removes skew (rotational distortion) from an 1bpp input image.
Args:
doc (unicode, unicode): The input document tuple.
method (unicode): The suffix string appended to all output files.
Returns:
(unicode, unicode): Storage tuple of the output file
"""
input_path = storage.get_abs_path(*doc)
output_path = storage.insert_suffix(input_path, method)
lept_deskew(input_path, output_path)
return storage.get_storage_path(output_path)
def deskew(doc, method=u'deskew'):
"""
Removes skew (rotational distortion) from an 1bpp input image.
Args:
doc (unicode, unicode): The input document tuple.
method (unicode): The suffix string appended to all output files.
Returns:
(unicode, unicode): Storage tuple of the output file
"""
input_path = storage.get_abs_path(*doc)
output_path = storage.insert_suffix(input_path, method)
lept_deskew(input_path, output_path)
return storage.get_storage_path(output_path)
def spell_check(doc, method=u'spell_check', language=u'',
filter_punctuation=False):
"""
Adds spelling suggestions to an TEI XML document.
Alternative spellings for each segment will be included in a choice
tagcontaining a series of corr tags with the original segment appearing
beneath a sic element. Correct words, i.e. words appearing verbatim in the
dictionary, are left untouched.
Args:
doc (unicode, unicode): The input document tuple.
method (unicode): The suffix string appended to the output file.
language (unicode): Identifier defined in the nidaba configuration as a
valid dictionary.
filter_punctuation (bool): Switch to filter punctuation inside
``seg``
Returns:
(unicode, unicode): Storage tuple of the output document
"""
input_path = storage.get_abs_path(*doc)
output_path = storage.insert_suffix(input_path, method, language,
unicode(filter_punctuation))
dictionary = storage.get_abs_path(*nidaba_cfg['lang_dicts'][language]['dictionary'])
del_dictionary = storage.get_abs_path(*nidaba_cfg['lang_dicts'][language]['deletion_dictionary'])
with storage.StorageFile(*doc) as fp:
logger.debug('Reading TEI ({})'.format(fp.abs_path))
tei = OCRRecord()
tei.load_tei(fp)
logger.debug('Performing spell check')
ret = lex.tei_spellcheck(tei, dictionary, del_dictionary,
filter_punctuation)
with storage.StorageFile(*storage.get_storage_path(output_path), mode='wb') as fp:
logger.debug('Writing TEI ({})'.format(fp.abs_path))
ret.write_tei(fp)
return storage.get_storage_path(output_path)
def rgb_to_gray(doc, method=u'rgb_to_gray'):
"""
Converts an arbitrary bit depth image to grayscale and writes it back
appending a suffix.
Args:
doc (unicode, unicode): The input document tuple
method (unicode): The suffix string appended to all output files.
Returns:
(unicode, unicode): Storage tuple of the output file
"""
input_path = storage.get_abs_path(*doc)
output_path = storage.insert_suffix(input_path, method)
return storage.get_storage_path(image.rgb_to_gray(input_path, output_path))
def dewarp(doc, method=u'dewarp'):
"""
Removes perspective distortion (as commonly exhibited by overhead scans)
from an 1bpp input image.
Args:
doc (unicode, unicode): The input document tuple.
method (unicode): The suffix string appended to all output files.
Returns:
(unicode, unicode): Storage tuple of the output file
"""
input_path = storage.get_abs_path(*doc)
output_path = storage.insert_suffix(input_path, method)
lept_dewarp(input_path, output_path)
return storage.get_storage_path(output_path)
def dewarp(doc, method=u'dewarp'):
"""
Removes perspective distortion (as commonly exhibited by overhead scans)
from an 1bpp input image.
Args:
doc (unicode, unicode): The input document tuple.
method (unicode): The suffix string appended to all output files.
Returns:
(unicode, unicode): Storage tuple of the output file
"""
input_path = storage.get_abs_path(*doc)
output_path = storage.insert_suffix(input_path, method)
lept_dewarp(input_path, output_path)
return storage.get_storage_path(output_path)
def otsu(doc, method=u'otsu'):
"""
Binarizes an input document utilizing a naive implementation of Otsu's
thresholding.
Args:
doc (unicode, unicode): The input document tuple.
method (unicode): The suffix string appended to all output files.
Returns:
(unicode, unicode): Storage tuple of the output file
"""
input_path = storage.get_abs_path(*doc)
output_path = storage.insert_suffix(input_path, method)
return storage.get_storage_path(image.otsu(input_path, output_path))
def ocr_tesseract(doc, method=u'ocr_tesseract', languages=None):
"""
Runs tesseract on an input document.
Args:
doc (unicode, unicode): The input document tuple
id (unicode): The nidaba batch identifier this task is a part of
method (unicode): The suffix string appended to all output files
languages (list of unicode): A list of languages for
the tesseract language model
Returns:
(unicode, unicode): Storage tuple for the output file
"""
input_path = storage.get_abs_path(*doc)
output_path = storage.insert_suffix(input_path, method, *languages)
return storage.get_storage_path(tesseract.ocr(input_path, output_path,
languages))
def nlbin(doc,
method=u'nlbin',
threshold=0.5,
zoom=0.5,
escale=1.0,
border=0.1,
perc=80,
range=20,
low=5,
high=90):
"""
Binarizes an input document utilizing ocropus'/kraken's nlbin algorithm.
Args:
doc (unicode, unicode): The input document tuple.
method (unicode): The suffix string appended to all output files.
threshold (float):
zoom (float):
escale (float):
border (float)
perc (int):
range (int):
low (int):
high (int):
Returns:
(unicode, unicode): Storage tuple of the output file
Raises:
NidabaInvalidParameterException: Input parameters are outside the valid
range.
"""
input_path = storage.get_abs_path(*doc)
output_path = storage.insert_suffix(input_path, method, unicode(threshold),
unicode(zoom), unicode(escale),
unicode(border), unicode(perc),
unicode(range), unicode(low),
unicode(high))
img = Image.open(input_path)
o_img = binarization.nlbin(img, threshold, zoom, escale, border, perc,
range, low, high)
o_img.save(output_path)
return storage.get_storage_path(output_path)
def ocr_ocropus(doc, method=u'ocr_ocropus', model=None):
"""
Runs ocropus on an input document.
Args:
doc (unicode, unicode): The input document tuple
method (unicode): The suffix string appended to all output files
model (unicode): Identifier for the font model to use
Returns:
(unicode, unicode): Storage tuple for the output file
"""
image_path = storage.get_abs_path(*doc[1])
segmentation_path = storage.get_abs_path(*doc[0])
output_path = os.path.splitext(
storage.insert_suffix(image_path, method, model))[0] + '.xml'
model = storage.get_abs_path(*(nidaba_cfg['ocropus_models'][model]))
return storage.get_storage_path(
ocr(image_path, segmentation_path, output_path, model))
def ocr_ocropus(doc, method=u'ocr_ocropus', model=None):
"""
Runs ocropus on an input document.
Args:
doc (unicode, unicode): The input document tuple
id (unicode): The nidaba batch identifier this task is a part of
method (unicode): The suffix string appended to all output files
model (unicode): Identifier for the font model to use
Returns:
(unicode, unicode): Storage tuple for the output file
"""
input_path = storage.get_abs_path(*doc)
output_path = os.path.splitext(storage.insert_suffix(input_path, method,
model))[0] + '.html'
model = storage.get_abs_path(*(nidaba_cfg['ocropus_models'][model]))
return storage.get_storage_path(ocropus.ocr(input_path, output_path,
model))
def text_lexicality(doc, method=u'text_lexicality', language=u'', divert=True):
"""
Calculates the lexicality of text in input documents.
Args:
doc (unicode, unicode): The input document tuple
method (unicode): The suffix string appended to the output file.
Returns:
(unicode, unicode): Storage tuple of the output document
"""
input_path = storage.get_abs_path(*doc[0])
output_path = storage.insert_suffix(input_path, method,
os.path.basename(input_path))
dictionary = storage.get_abs_path(
*nidaba_cfg['lang_dicts'][language]['dictionary'])
with storage.StorageFile(*doc) as fp:
tei = OCRRecord()
tei.load_tei(fp)
cnt = 0
err_cnt = 0
for seg_id, segment in facsimile.segments.iteritems():
tok = alg.sanitize(''.join(x['grapheme']
for x in segment['content'].itervalues()))
tok = regex.sub('[^\w]', '', key)
cnt += 1
if not alg.mmap_bin_search(
tok, dictionary, entryparser_fn=alg.key_for_single_word):
err_cnt += 1
if not divert:
storage.write_text(*storage.get_storage_path(output_path),
text=unicode(err_cnt / float(cnt)))
return output_path
else:
return {
'edit_ratio': err_cnt / float(cnt),
'ground_truth': '',
'doc': doc
}
def blend_hocr(docs, method=u'blend_hocr', language=u''):
"""
Blends multiple hOCR files using the algorithm from Bruce Robertsons
rigaudon. It requires a working spell checking for the input document's
language; otherwise all matched bboxes will be bunched together without any
scoring.
Args:
docs [(id, path), ...]: A list of storage module tupels that will be
merged into a single output document.
id (unicode): The nidaba batch identifier this task is a part of
language (unicode): Language used for spell-checking
based scoring. If not defined no scoring will be used.
method (unicode): The suffix string appended to the output file.
Returns:
(unicode, unicode): Storage tuple of the output document
"""
# create the output document path from the first input document
input_path = storage.get_abs_path(*docs[0])
output_path = storage.insert_suffix(input_path, method)
return merge_hocr.merge(docs, language,
storage.get_storage_path(output_path))
def text_rep_confidence(doc, method=u'text_rep_confidence', divert=True):
"""
Extracts self reported confidence values from input documents.
Args:
doc (unicode, unicode): The input document tuple
method (unicode): The suffix string appended to the output file.
Returns:
(unicode, unicode): Storage tuple of the output document
"""
input_path = storage.get_abs_path(*doc[0])
output_path = storage.insert_suffix(input_path, method,
os.path.basename(input_path))
with storage.StorageFile(*doc) as fp:
tei = OCRRecord()
tei.load_tei(fp)
edist = numpy.mean([x['confidence'] for x in tei.graphemes.itervalues()])
if not divert:
storage.write_text(*storage.get_storage_path(output_path),
text=unicode(edist))
return output_path
else:
return {'edit_ratio': edist, 'ground_truth': '', 'doc': doc}
def text_rep_confidence(doc, method=u'text_rep_confidence', divert=True):
"""
Extracts self reported confidence values from input documents.
Args:
doc (unicode, unicode): The input document tuple
method (unicode): The suffix string appended to the output file.
Returns:
(unicode, unicode): Storage tuple of the output document
"""
input_path = storage.get_abs_path(*doc[0])
output_path = storage.insert_suffix(input_path, method,
os.path.basename(input_path))
with storage.StorageFile(*doc) as fp:
tei = OCRRecord()
tei.load_tei(fp)
edist = numpy.mean([x['confidence'] for x in tei.graphemes.itervalues()])
if not divert:
storage.write_text(*storage.get_storage_path(output_path),
text=unicode(edist))
return output_path
else:
return {'edit_ratio': edist, 'ground_truth': '', 'doc': doc}
def segmentation_tesseract(doc, method=u'segment_tesseract'):
"""
Performs page segmentation using tesseract's built-in algorithm and writes
a TEI XML segmentation file.
Args:
doc (unicode, unicode): The input document tuple
method (unicode): The suffix string appended to all output files.
Returns:
Two storage tuples with the first one containing the segmentation and
the second one being the file the segmentation was calculated upon.
"""
input_path = storage.get_abs_path(*doc)
output_path = splitext(storage.insert_suffix(input_path, method))[0] + '.xml'
ver = tesseract.TessVersion()
if int(ver.split('.')[0]) < 3 or int(ver.split('.')[1]) < 2:
raise NidabaTesseractException('libtesseract version is too old. Set '
'implementation to direct.')
# tesseract has a tendency to crash arbitrarily on some inputs
# necessitating execution in a separate process to ensure the worker
# doesn't just die. We use fork as the multiprocessing module thinks
# programmers are too stupid to reap their children.
logger.info('Forking before entering unstable ctypes code')
pid = os.fork()
if pid != 0:
try:
logger.info('Waiting for child to complete')
_, status = os.waitpid(pid, 0)
except OSError as e:
if e.errno not in (errno.EINTR, errno.ECHILD):
raise
return storage.get_storage_path(output_path), doc
if os.WIFSIGNALED(status):
raise NidabaTesseractException('Tesseract killed by signal: {0}'.format(os.WTERMSIG(status)))
return storage.get_storage_path(output_path), doc
api = tesseract.TessBaseAPICreate()
rc = tesseract.TessBaseAPIInit3(api, tessdata.encode('utf-8'), None)
if (rc):
tesseract.TessBaseAPIDelete(api)
raise NidabaTesseractException('Tesseract initialization failed.')
# only do segmentation and script detection
logger.debug('Setting page set mode to 2')
tesseract.TessBaseAPISetPageSegMode(api, 2)
logger.debug('Reading {} using leptonica'.format(input_path))
pix = leptonica.pixRead(input_path.encode('utf-8'))
logger.debug('Setting PIX as input image')
tesseract.TessBaseAPISetImage2(api, pix)
logger.debug('Analyzing page layout')
it = tesseract.TessBaseAPIAnalyseLayout(api)
logger.debug('Destroying PIX')
leptonica.pixDestroy(ctypes.byref(pix))
x0, y0, x1, y1 = (ctypes.c_int(), ctypes.c_int(), ctypes.c_int(),
ctypes.c_int())
w, h = Image.open(input_path).size
logger.info('Initializing TEI XML file with {}x{} {}/{}'.format(w, h, *doc))
tei = TEIFacsimile()
tei.document((w, h), os.path.join(*doc))
tei.title = os.path.basename(doc[1])
tei.add_respstmt('tesseract', 'page segmentation')
while True:
if tesseract.TessPageIteratorIsAtBeginningOf(it, RIL_TEXTLINE):
tesseract.TessPageIteratorBoundingBox(it,
RIL_TEXTLINE,
ctypes.byref(x0),
ctypes.byref(y0),
ctypes.byref(x1),
ctypes.byref(y1))
tei.add_line((x0.value, y0.value, x1.value, y1.value))
logger.debug('Segmenter found new line at {} {} {} {}'.format(x0.value,
y0.value,
x1.value,
y1.value))
if tesseract.TessPageIteratorIsAtBeginningOf(it, RIL_WORD):
tesseract.TessPageIteratorBoundingBox(it,
RIL_WORD,
ctypes.byref(x0),
ctypes.byref(y0),
ctypes.byref(x1),
ctypes.byref(y1))
tei.add_segment((x0.value, y0.value, x1.value, y1.value))
logger.debug('Segmenter found new word at {} {} {} {}'.format(x0.value,
y0.value,
x1.value,
y1.value))
tesseract.TessPageIteratorBoundingBox(it,
RIL_SYMBOL,
ctypes.byref(x0),
ctypes.byref(y0),
ctypes.byref(x1),
ctypes.byref(y1))
tei.add_graphemes([(None, (x0.value, y0.value, x1.value, y1.value))])
logger.debug('Segmenter found new symbol at {} {} {} {}'.format(x0.value,
y0.value,
x1.value,
y1.value))
if not tesseract.TessPageIteratorNext(it, RIL_SYMBOL):
logger.debug('No more elements on page')
break
logger.debug('Deleting page iterator and base API')
tesseract.TessPageIteratorDelete(it)
tesseract.TessBaseAPIEnd(api)
tesseract.TessBaseAPIDelete(api)
logger.info('Writing segmentation to {}'.format(output_path))
with open(output_path, 'w') as fp:
tei.write(fp)
logger.info('Quitting child process')
os._exit(os.EX_OK)
return storage.get_storage_path(output_path), doc
def segmentation_tesseract(doc, method=u'segment_tesseract'):
"""
Performs page segmentation using tesseract's built-in algorithm and writes
a TEI XML segmentation file.
Args:
doc (unicode, unicode): The input document tuple
method (unicode): The suffix string appended to all output files.
Returns:
Two storage tuples with the first one containing the segmentation and
the second one being the file the segmentation was calculated upon.
"""
input_path = storage.get_abs_path(*doc)
output_path = splitext(storage.insert_suffix(input_path,
method))[0] + '.xml'
ver = tesseract.TessVersion()
if int(ver.split('.')[0]) < 3 or int(ver.split('.')[1]) < 2:
raise NidabaTesseractException('libtesseract version is too old. Set '
'implementation to direct.')
# tesseract has a tendency to crash arbitrarily on some inputs
# necessitating execution in a separate process to ensure the worker
# doesn't just die. We use fork as the multiprocessing module thinks
# programmers are too stupid to reap their children.
logger.info('Forking before entering unstable ctypes code')
pid = os.fork()
if pid != 0:
try:
logger.info('Waiting for child to complete')
_, status = os.waitpid(pid, 0)
except OSError as e:
if e.errno not in (errno.EINTR, errno.ECHILD):
raise
return storage.get_storage_path(output_path)
if os.WIFSIGNALED(status):
raise NidabaTesseractException(
'Tesseract killed by signal: {0}'.format(os.WTERMSIG(status)))
return storage.get_storage_path(output_path)
api = tesseract.TessBaseAPICreate()
rc = tesseract.TessBaseAPIInit3(api, tessdata.encode('utf-8'), None)
if (rc):
tesseract.TessBaseAPIDelete(api)
raise NidabaTesseractException('Tesseract initialization failed.')
# only do segmentation and script detection
logger.debug('Setting page set mode to 2')
tesseract.TessBaseAPISetPageSegMode(api, 2)
logger.debug('Reading {} using leptonica'.format(input_path))
pix = leptonica.pixRead(input_path.encode('utf-8'))
logger.debug('Setting PIX as input image')
tesseract.TessBaseAPISetImage2(api, pix)
logger.debug('Analyzing page layout')
it = tesseract.TessBaseAPIAnalyseLayout(api)
logger.debug('Destroying PIX')
leptonica.pixDestroy(ctypes.byref(pix))
x0, y0, x1, y1 = (ctypes.c_int(), ctypes.c_int(), ctypes.c_int(),
ctypes.c_int())
w, h = Image.open(input_path).size
logger.info('Initializing TEI XML file with {}x{} {}/{}'.format(
w, h, *doc))
tei = OCRRecord()
tei.dimensions = (w, h)
tei.img = storage.get_url(*doc)
tei.title = os.path.basename(doc[1])
tei.add_respstmt('tesseract', 'page segmentation')
while True:
if tesseract.TessPageIteratorIsAtBeginningOf(it, RIL_TEXTLINE):
tesseract.TessPageIteratorBoundingBox(it, RIL_TEXTLINE,
ctypes.byref(x0),
ctypes.byref(y0),
ctypes.byref(x1),
ctypes.byref(y1))
tei.add_line((x0.value, y0.value, x1.value, y1.value))
logger.debug('Segmenter found new line at {} {} {} {}'.format(
x0.value, y0.value, x1.value, y1.value))
if tesseract.TessPageIteratorIsAtBeginningOf(it, RIL_WORD):
tesseract.TessPageIteratorBoundingBox(it, RIL_WORD,
ctypes.byref(x0),
ctypes.byref(y0),
ctypes.byref(x1),
ctypes.byref(y1))
tei.add_segment((x0.value, y0.value, x1.value, y1.value))
logger.debug('Segmenter found new word at {} {} {} {}'.format(
x0.value, y0.value, x1.value, y1.value))
tesseract.TessPageIteratorBoundingBox(it, RIL_SYMBOL, ctypes.byref(x0),
ctypes.byref(y0),
ctypes.byref(x1),
ctypes.byref(y1))
tei.add_graphemes([{
'grapheme': '',
'bbox': (x0.value, y0.value, x1.value, y1.value)
}])
logger.debug('Segmenter found new symbol at {} {} {} {}'.format(
x0.value, y0.value, x1.value, y1.value))
if not tesseract.TessPageIteratorNext(it, RIL_SYMBOL):
logger.debug('No more elements on page')
break
logger.debug('Deleting page iterator and base API')
tesseract.TessPageIteratorDelete(it)
tesseract.TessBaseAPIEnd(api)
tesseract.TessBaseAPIDelete(api)
logger.info('Writing segmentation to {}'.format(output_path))
with open(output_path, 'w') as fp:
tei.write_tei(fp)
logger.info('Quitting child process')
os._exit(os.EX_OK)
return storage.get_storage_path(output_path)
def text_diff_ratio(doc, method=u'text_diff_ratio', ground_truth=None,
xml_in=True, gt_format=u'tei', clean_in=True, clean_gt=True,
divert=True):
"""
Calculates the similarity of the input documents and a given ground truth
using the algorithm of python's difflib SequenceMatcher. The result is a
value between 0.0 (no commonality) and 1.0 (identical strings).
Args:
doc (unicode, unicode): The input document tuple
method (unicode): The suffix string appended to the output file.
ground_truth (unicode): Ground truth location tuple or a list of ground
truths to choose from. When more than one is
given, the file sharing the longest prefix with
the input document is chosen.
xml_in (bool): Switch to treat input as an TEI-XML document.
gt_format (unicode): Switch to select ground truth format. Valid values
are 'tei', 'hocr', and 'text'.
clean_in (bool): Normalize to NFD and strip input data. (DO NOT DISABLE!)
clean_gt (bool): Normalize to NFD and strip ground truth. (DO NOT DISABLE!)
divert (bool): Switch selecting output diversion. If enabled the output
will be added to the tracking arguments and the input
document will be returned as the result of the task. Use
this to insert a statistical measure into a chain
without affecting the results.
Returns:
(unicode, unicode): Storage tuple of the output document
"""
input_path = storage.get_abs_path(*doc[0])
output_path = storage.insert_suffix(input_path, method,
os.path.basename(input_path))
if not isinstance(ground_truth[0], basestring):
ground_truth = find_matching(doc, ground_truth)
with storage.StorageFile(*ground_truth) as fp:
if gt_format == 'tei':
tei = TEIFacsimile()
tei.read(fp)
t = StringIO.StringIO()
tei.write_text(t)
gt = t.getvalue()
elif gt_format == 'hocr':
gt = html.parse(fp).text_content()
elif gt_format == 'text':
gt = fp.read()
else:
raise NidabaInvalidParameterException('Input format ' + gt_format + ' unknown.')
with storage.StorageFile(*doc) as fp:
if xml_in:
tei = TEIFacsimile()
tei.read(fp)
t = StringIO.StringIO()
tei.write_text(t)
text = t.getvalue()
else:
text = fp.read()
if clean_in:
text = cleanup(text)
if clean_gt:
gt = cleanup(gt)
logger.debug('Recognition result: \n{}'.format(text))
logger.debug('Ground truth: \n{}'.format(gt))
sm = difflib.SequenceMatcher()
sm.set_seqs(text, gt)
logger.debug('Accuracy: {}'.format(sm.ratio()))
if not divert:
storage.write_text(*storage.get_storage_path(output_path),
text=unicode(sm.ratio()))
return output_path
else:
return {'diff_ratio': sm.ratio(), 'ground_truth': ground_truth, 'doc': doc}
def text_diff_ratio(doc,
method=u'text_diff_ratio',
ground_truth=None,
xml_in=True,
gt_format=u'tei',
clean_in=True,
clean_gt=True,
divert=True):
"""
Calculates the similarity of the input documents and a given ground truth
using the algorithm of python's difflib SequenceMatcher. The result is a
value between 0.0 (no commonality) and 1.0 (identical strings).
Args:
doc (unicode, unicode): The input document tuple
method (unicode): The suffix string appended to the output file.
ground_truth (unicode): Ground truth location tuple or a list of ground
truths to choose from. When more than one is
given, the file sharing the longest prefix with
the input document is chosen.
xml_in (bool): Switch to treat input as an TEI-XML document.
gt_format (unicode): Switch to select ground truth format. Valid values
are 'tei', 'hocr', and 'text'.
clean_in (bool): Normalize to NFD and strip input data. (DO NOT DISABLE!)
clean_gt (bool): Normalize to NFD and strip ground truth. (DO NOT DISABLE!)
divert (bool): Switch selecting output diversion. If enabled the output
will be added to the tracking arguments and the input
document will be returned as the result of the task. Use
this to insert a statistical measure into a chain
without affecting the results.
Returns:
(unicode, unicode): Storage tuple of the output document
"""
input_path = storage.get_abs_path(*doc[0])
output_path = storage.insert_suffix(input_path, method,
os.path.basename(input_path))
if not isinstance(ground_truth[0], basestring):
ground_truth = find_matching(doc, ground_truth)
with storage.StorageFile(*ground_truth) as fp:
if gt_format == 'tei':
tei = TEIFacsimile()
tei.read(fp)
t = StringIO.StringIO()
tei.write_text(t)
gt = t.getvalue()
elif gt_format == 'hocr':
gt = html.parse(fp).text_content()
elif gt_format == 'text':
gt = fp.read()
else:
raise NidabaInvalidParameterException('Input format ' + gt_format +
' unknown.')
with storage.StorageFile(*doc) as fp:
if xml_in:
tei = TEIFacsimile()
tei.read(fp)
t = StringIO.StringIO()
tei.write_text(t)
text = t.getvalue()
else:
text = fp.read()
if clean_in:
text = cleanup(text)
if clean_gt:
gt = cleanup(gt)
logger.debug('Recognition result: \n{}'.format(text))
logger.debug('Ground truth: \n{}'.format(gt))
sm = difflib.SequenceMatcher()
sm.set_seqs(text, gt)
logger.debug('Accuracy: {}'.format(sm.ratio()))
if not divert:
storage.write_text(*storage.get_storage_path(output_path),
text=unicode(sm.ratio()))
return output_path
else:
return {
'diff_ratio': sm.ratio(),
'ground_truth': ground_truth,
'doc': doc
}
