def extract_image(opt, page_nr, filename):
try:
width, height = image_size(page_nr, filename)
subsample = 1
while (width*height) / subsample > (1 << 20) * 50:
subsample += 1
subsample = min(subsample, 12)
except Exception:
utils.print_traceback("Unable to get image size, subsample=1", filename)
subsample = 1
if subsample != 1:
print "subsample", subsample
tiff_name = opt.out_dir + 'page_%04d.tif' % page_nr
ddjvu = djvulibre_path + 'ddjvu'
ls = subprocess.Popen([ ddjvu, "-format=tiff", "-page=%d" % page_nr, "-subsample=%d" % subsample, filename, tiff_name], stdout=subprocess.PIPE, preexec_fn=setrlimits, close_fds = True)
text = utils.safe_read(ls.stdout)
if text:
print text
ls.wait()
if ls.returncode != 0:
print >> sys.stderr, "extract_image fail: ", ls.returncode, filename, page_nr
return None
return tiff_name
def ocr(filename, out_basename, lang, config = ''):
if tesseract_data_prefix:
os.environ['TESSDATA_PREFIX'] = tesseract_data_prefix
ls = subprocess.Popen([ tesseract_path, filename, out_basename, "-l", lang, config], stdout=subprocess.PIPE, preexec_fn=setrlimits, close_fds = True)
text = utils.safe_read(ls.stdout)
if text:
print text,
ls.wait()
if config == '':
out_filename = out_basename + ".txt"
else:
out_filename = out_basename + ".hocr"
if not os.path.exists(out_filename) or ls.returncode:
# in case returncode == 0
print >> sys.stderr, "ocr.ocr() fail to exec tesseract:", ls.returncode, filename
fd = open(out_filename, 'w')
fd.write('An error occurred during ocr processing: ' + filename)
fd.close()
fd = open(out_filename)
txt = fd.read()
fd.close()
if tesseract_data_prefix:
del os.environ['TESSDATA_PREFIX']
if ls.returncode != 0:
print >> sys.stderr, "ocr.ocr() fail to exec tesseract:", ls.returncode, filename
return None
return txt
def ocr(filename, out_basename, lang, config = ''):
if tesseract_data_prefix:
os.environ['TESSDATA_PREFIX'] = tesseract_data_prefix
ls = subprocess.Popen([ tesseract_path, filename, out_basename, "-l", lang, config], stdout=subprocess.PIPE, preexec_fn=setrlimits, close_fds = True)
text = utils.safe_read(ls.stdout)
if text:
print text,
ls.wait()
if config == '':
out_filename = out_basename + ".txt"
else:
out_filename = out_basename + ".html"
if not os.path.exists(out_filename) and not ls.returncode:
# in case returncode == 0
print >> sys.stderr, "ocr.ocr() fail to exec tesseract:", ls.returncode, filename
fd = open(out_filename, 'w')
fd.write('An error occurred during ocr processing: ' + filename)
fd.close()
fd = open(out_filename)
txt = fd.read()
fd.close()
if tesseract_data_prefix:
del os.environ['TESSDATA_PREFIX']
if ls.returncode != 0:
print >> sys.stderr, "ocr.ocr() fail to exec tesseract:", ls.returncode, filename
return None
return txt
def extract_image(opt, page_nr, filename):
try:
width, height = image_size(page_nr, filename)
subsample = 1
while (width*height) / subsample > (1 << 20) * 50:
subsample += 1
subsample = min(subsample, 12)
except Exception:
utils.print_traceback("Unable to get image size, subsample=1", filename)
subsample = 1
if subsample != 1:
print "subsample", subsample
tiff_name = opt.out_dir + 'page_%04d.tif' % page_nr
ddjvu = djvulibre_path + 'ddjvu'
ls = subprocess.Popen([ ddjvu, "-format=tiff", "-page=%d" % page_nr, "-subsample=%d" % subsample, filename, tiff_name], stdout=subprocess.PIPE, preexec_fn=setrlimits, close_fds = True)
text = utils.safe_read(ls.stdout)
if text:
print text
ls.wait()
if ls.returncode != 0:
print >> sys.stderr, "extract_image fail: ", ls.returncode, filename, page_nr
return None
return tiff_name
def get_nr_pages_djvu(filename):
djvused = djvulibre_path + 'djvused'
ls = subprocess.Popen([ djvused, "-e", "n", filename], stdout=subprocess.PIPE, preexec_fn=setrlimits, close_fds = True)
text = utils.safe_read(ls.stdout)
ls.wait()
if ls.returncode != 0:
print >> sys.stderr, "Error: djvused fail to exec", ls.returncode
return None
return int(text)
def get_nr_pages_djvu(filename):
djvused = djvulibre_path + 'djvused'
ls = subprocess.Popen([ djvused, "-e", "n", filename], stdout=subprocess.PIPE, preexec_fn=setrlimits, close_fds = True)
text = utils.safe_read(ls.stdout)
ls.wait()
if ls.returncode != 0:
print >> sys.stderr, "Error: djvused fail to exec", ls.returncode
return None
return int(text)
def image_size(page_nr, filename):
djvused = djvulibre_path + 'djvused'
ls = subprocess.Popen([ djvused, "-e", "select %d; size" % page_nr, filename], stdout=subprocess.PIPE, preexec_fn=setrlimits, close_fds = True)
text = utils.safe_read(ls.stdout)
ls.wait()
if ls.returncode != 0:
print >> sys.stderr, "Error: djvused fail to exec", ls.returncode
return None
match = re.search('width=(\d+) height=(\d+)', text)
return int(match.group(1)), int(match.group(2))
def image_size(page_nr, filename):
djvused = djvulibre_path + 'djvused'
ls = subprocess.Popen([ djvused, "-e", "select %d; size" % page_nr, filename], stdout=subprocess.PIPE, preexec_fn=setrlimits, close_fds = True)
text = utils.safe_read(ls.stdout)
ls.wait()
if ls.returncode != 0:
print >> sys.stderr, "Error: djvused fail to exec", ls.returncode
return None
match = re.search('width=(\d+) height=(\d+)', text)
return int(match.group(1)), int(match.group(2))
'''
extract the speech from the whole dataset, and save each file under the same folder as the
'''
from utils import get_all_transcript_paths, safe_read
from extract_speech import extract_speech_string
from tqdm import tqdm
if __name__ == '__main__':
for path in tqdm(list(get_all_transcript_paths())):
transcript = extract_speech_string(safe_read(path))
with open(f'transcripts/{path.parts[-1]}', 'w') as f:
f.write(transcript)