mirror of
https://gitlab.ub.uni-bielefeld.de/sfb1288inf/ocr.git
synced 2024-12-26 04:24:18 +00:00
Update OCR Pipeline
This commit is contained in:
Patrick Jentsch
parent
eb5ccf4e21
commit
36a86887b0
19
hocrtotei
19
hocrtotei
@ -2,25 +2,15 @@
|
||||
# coding=utf-8
|
||||
|
||||
from xml.sax.saxutils import escape
|
||||
import argparse
|
||||
from argparse import ArgumentParser
|
||||
import xml.etree.ElementTree as ET
|
||||
|
||||
parser = argparse.ArgumentParser(
|
||||
description='Merges several hOCR files in order of their occurrence on command line to one TEI result file.'
|
||||
)
|
||||
parser.add_argument(
|
||||
'i',
|
||||
metavar='hOCR-sourcefile',
|
||||
nargs='+'
|
||||
)
|
||||
parser.add_argument(
|
||||
'o',
|
||||
metavar='TEI-destfile',
|
||||
)
|
||||
parser = ArgumentParser(description='Merges hOCR files to one P5 file.')
|
||||
parser.add_argument('i', metavar='hOCR-sourcefile', nargs='+')
|
||||
parser.add_argument('o', metavar='TEI-destfile',)
|
||||
args = parser.parse_args()
|
||||
|
||||
output_file = open(args.o, 'w')
|
||||
|
||||
output_file.write(
|
||||
'<?xml version="1.0" encoding="UTF-8"?>\n'
|
||||
+ '<TEI xmlns="http://www.tei-c.org/ns/1.0" xml:id="dtabf">\n'
|
||||
@ -54,5 +44,4 @@ output_file.write(
|
||||
' </body>\n'
|
||||
+ ' </text>\n'
|
||||
+ '</TEI>')
|
||||
|
||||
output_file.close()
|
||||
|
||||
714
ocr
714
ocr
@ -10,292 +10,206 @@ Author: Patrick Jentsch <p.jentsch@uni-bielefeld.de>
|
||||
"""
|
||||
|
||||
|
||||
import argparse
|
||||
from argparse import ArgumentParser
|
||||
from pyflow import WorkflowRunner
|
||||
import multiprocessing
|
||||
import os
|
||||
import re
|
||||
import sys
|
||||
from pyflow import WorkflowRunner
|
||||
|
||||
|
||||
def parse_arguments():
|
||||
parser = argparse.ArgumentParser(
|
||||
description='''Performs OCR of (historical) documents utilizing OCRopus
|
||||
for preprocessing and Tesseract OCR for OCR. The results
|
||||
are served as hOCR, PDF, raw text and TEI compliant XML
|
||||
files.\n
|
||||
Software requirements: imagemagick, ocropus, pdftoppm,
|
||||
pdfunite, poppler-utils, pyflow, python2.7, python3.5,
|
||||
tesseract'''
|
||||
)
|
||||
parser.add_argument(
|
||||
'-i',
|
||||
dest='input_dir',
|
||||
required=True
|
||||
)
|
||||
parser.add_argument(
|
||||
'-l',
|
||||
choices=[
|
||||
'deu', 'eng', 'enm', 'fra', 'frk', 'frm', 'ita', 'por', 'spa'
|
||||
],
|
||||
dest='lang',
|
||||
required=True
|
||||
)
|
||||
parser.add_argument(
|
||||
'-o',
|
||||
dest='output_dir',
|
||||
required=True
|
||||
)
|
||||
parser.add_argument(
|
||||
'--skip-binarisation',
|
||||
action='store_true',
|
||||
default=False,
|
||||
dest='skip_binarisation',
|
||||
help='skip ocropy binarisation',
|
||||
required=False
|
||||
)
|
||||
parser.add_argument(
|
||||
'--keep-intermediates',
|
||||
action='store_true',
|
||||
default=False,
|
||||
dest='keep_intermediates',
|
||||
help='keep intermediate files',
|
||||
required=False
|
||||
)
|
||||
parser.add_argument(
|
||||
'--nCores',
|
||||
default=min(4, multiprocessing.cpu_count()),
|
||||
dest='n_cores',
|
||||
help='total number of cores available',
|
||||
required=False,
|
||||
type=int
|
||||
)
|
||||
parser.add_argument(
|
||||
'--zip',
|
||||
default='ocr-result-files',
|
||||
dest='zip',
|
||||
type=str,
|
||||
help='package result files in zip bundles and asign an filename prefix',
|
||||
required=False
|
||||
)
|
||||
TESSERACT_MODELS = ['deu', 'eng', 'enm', 'fra', 'frk', 'frm', 'ita', 'por',
|
||||
'spa']
|
||||
|
||||
|
||||
def parse_args():
|
||||
parser = ArgumentParser(description='OCR Pipeline utilizing tesseract.')
|
||||
parser.add_argument('i')
|
||||
parser.add_argument('o')
|
||||
parser.add_argument('-l', '--language', choices=TESSERACT_MODELS,
|
||||
required=True)
|
||||
parser.add_argument('--binarize', action='store_true',
|
||||
help='use ocropy binarisation')
|
||||
parser.add_argument('--keep-intermediates', action='store_true',
|
||||
help='keep intermediate files')
|
||||
parser.add_argument('--n-cores',
|
||||
default=min(4, multiprocessing.cpu_count()),
|
||||
help='total number of cores available', type=int)
|
||||
parser.add_argument('--log-dir')
|
||||
parser.add_argument('--zip')
|
||||
return parser.parse_args()
|
||||
|
||||
|
||||
class OCRWorkflow(WorkflowRunner):
|
||||
def __init__(self, args):
|
||||
self.jobs = analyze_jobs(args.input_dir, args.output_dir)
|
||||
self.skip_binarisation = args.skip_binarisation
|
||||
self.keep_intermediates = args.keep_intermediates
|
||||
self.lang = args.lang
|
||||
self.n_cores = args.n_cores
|
||||
self.output_dir = args.output_dir
|
||||
self.zip = args.zip
|
||||
class OCRPipelineJob:
|
||||
def __init__(self, file, output_dir):
|
||||
self.file = file
|
||||
self.name = os.path.basename(file).rsplit('.', 1)[0]
|
||||
self.output_dir = output_dir
|
||||
|
||||
|
||||
class OCRPipeline(WorkflowRunner):
|
||||
def __init__(self, binarize, jobs, keep_intermediates, lang, n_cores,
|
||||
output_dir, zip):
|
||||
self.binarize = binarize
|
||||
self.jobs = jobs
|
||||
self.keep_intermediates = keep_intermediates
|
||||
self.lang = lang
|
||||
self.n_cores = n_cores
|
||||
self.output_dir = output_dir
|
||||
self.zip = zip
|
||||
|
||||
def workflow(self):
|
||||
if len(self.jobs) == 0:
|
||||
if not self.jobs:
|
||||
return
|
||||
|
||||
'''
|
||||
' ##################################################
|
||||
' # Create output directories #
|
||||
' # mkdir_jobs #
|
||||
' ##################################################
|
||||
'''
|
||||
create_output_directories_jobs = []
|
||||
for index, job in enumerate(self.jobs):
|
||||
cmd = 'mkdir -p "%s"' % (
|
||||
os.path.join(job['output_dir'], 'tmp')
|
||||
)
|
||||
mkdir_jobs = []
|
||||
for i, job in enumerate(self.jobs):
|
||||
output_dir = os.path.join(job.output_dir, 'tmp')
|
||||
cmd = 'mkdir'
|
||||
cmd += ' -p'
|
||||
cmd += ' "{}"'.format(output_dir)
|
||||
if self.keep_intermediates:
|
||||
cmd += ' "%s" "%s" "%s" "%s"' % (
|
||||
os.path.join(job['output_dir'], 'tmp', 'hocr'),
|
||||
os.path.join(job['output_dir'], 'tmp', 'pdf'),
|
||||
os.path.join(job['output_dir'], 'tmp', 'tiff'),
|
||||
os.path.join(job['output_dir'], 'tmp', 'txt')
|
||||
)
|
||||
if not self.skip_binarisation:
|
||||
cmd += ' "%s"' % (
|
||||
os.path.join(job['output_dir'], 'tmp', 'bin.png')
|
||||
)
|
||||
create_output_directories_jobs.append(
|
||||
self.addTask(
|
||||
command=cmd,
|
||||
label='create_output_directories_job_-_%i' % (index)
|
||||
)
|
||||
)
|
||||
cmd += ' "{}"'.format(os.path.join(output_dir, 'hocr'))
|
||||
cmd += ' "{}"'.format(os.path.join(output_dir, 'pdf'))
|
||||
cmd += ' "{}"'.format(os.path.join(output_dir, 'tiff'))
|
||||
cmd += ' "{}"'.format(os.path.join(output_dir, 'txt'))
|
||||
if self.binarize:
|
||||
cmd += ' "{}"'.format(os.path.join(output_dir, 'bin.png'))
|
||||
cmd += ' "{}"'.format(os.path.join(output_dir, 'nrm.png'))
|
||||
lbl = 'mkdir_job_-_{}'.format(i)
|
||||
mkdir_jobs.append(self.addTask(command=cmd, label=lbl))
|
||||
|
||||
'''
|
||||
' ##################################################
|
||||
' # Split #
|
||||
' # pdftoppm_jobs #
|
||||
' ##################################################
|
||||
'''
|
||||
split_jobs = []
|
||||
split_job_n_cores = min(
|
||||
self.n_cores,
|
||||
max(1, int(self.n_cores / len(self.jobs)))
|
||||
)
|
||||
for index, job in enumerate(self.jobs):
|
||||
if job['filename'].endswith(('.tif', '.tiff')):
|
||||
'''
|
||||
' This command also works for PDF input but ocropus-nlbin
|
||||
' is not able to handle the TIFF output of it.
|
||||
'''
|
||||
cmd = 'convert -density 300 "%s" -compress LZW -scene 1 "%s/page-%%d.tif"' % (
|
||||
job['path'],
|
||||
os.path.join(job['output_dir'], 'tmp')
|
||||
)
|
||||
else:
|
||||
cmd = 'pdftoppm -r 300 -tiff -tiffcompression lzw "%s" "%s"' % (
|
||||
job['path'],
|
||||
os.path.join(job['output_dir'], 'tmp', 'page')
|
||||
)
|
||||
pdftoppm_jobs = []
|
||||
n_cores = min(self.n_cores, max(1, int(self.n_cores / len(self.jobs))))
|
||||
for i, job in enumerate(self.jobs):
|
||||
output_dir = os.path.join(job.output_dir, 'tmp')
|
||||
output_file_base = os.path.join(output_dir, 'page')
|
||||
cmd = 'pdftoppm'
|
||||
cmd += ' -r 300'
|
||||
cmd += ' -tiff'
|
||||
cmd += ' -tiffcompression lzw'
|
||||
cmd += ' "{}" "{}"'.format(job.file, output_file_base)
|
||||
deps = 'mkdir_job_-_{}'.format(i)
|
||||
lbl = 'pdftoppm_job_-_{}'.format(i)
|
||||
pdftoppm_jobs.append(self.addTask(command=cmd, dependencies=deps,
|
||||
label=lbl, nCores=n_cores))
|
||||
|
||||
split_jobs.append(
|
||||
self.addTask(
|
||||
command=cmd,
|
||||
dependencies='create_output_directories_job_-_%i' % (index),
|
||||
label='split_job_-_%i' % (index),
|
||||
nCores=split_job_n_cores
|
||||
)
|
||||
)
|
||||
|
||||
if not self.skip_binarisation:
|
||||
if self.binarize:
|
||||
'''
|
||||
' The binarisation_jobs are created based of the output files of
|
||||
' the split_jobs. So wait until they are finished.
|
||||
' The ocropus_nlbin_jobs list is created based on the output files
|
||||
' of the pdftoppm_jobs. So wait until they are finished.
|
||||
'''
|
||||
self.waitForTasks()
|
||||
|
||||
'''
|
||||
' ##################################################
|
||||
' # Binarise #
|
||||
' # ocropus_nlbin_jobs #
|
||||
' ##################################################
|
||||
'''
|
||||
binarisation_jobs = []
|
||||
ocropus_nlbin_jobs = []
|
||||
'''
|
||||
' We run ocropus-nlbin with either four or, if there are less then
|
||||
' four cores available for this workflow, the available core
|
||||
' number.
|
||||
'''
|
||||
binarisation_job_n_cores = min(4, self.n_cores)
|
||||
for index, job in enumerate(self.jobs):
|
||||
files = os.listdir(os.path.join(job['output_dir'], 'tmp'))
|
||||
files = filter(lambda x: x.endswith('.tif'), files)
|
||||
n_cores = min(4, self.n_cores)
|
||||
for i, job in enumerate(self.jobs):
|
||||
input_dir = os.path.join(job.output_dir, 'tmp')
|
||||
output_dir = input_dir
|
||||
files = filter(lambda x: x.endswith('.tif'),
|
||||
os.listdir(input_dir))
|
||||
files.sort(key=lambda x: int(re.search(r'\d+', x).group(0)))
|
||||
files = map(
|
||||
lambda x: '"' + os.path.join(job['output_dir'], 'tmp', x) + '"',
|
||||
files
|
||||
)
|
||||
cmd = 'ocropus-nlbin --output "%s" --parallel "%i" %s' % (
|
||||
os.path.join(job['output_dir'], 'tmp'),
|
||||
binarisation_job_n_cores,
|
||||
' '.join(files)
|
||||
)
|
||||
binarisation_jobs.append(
|
||||
self.addTask(
|
||||
command=cmd,
|
||||
dependencies='split_job_-_%i' % (index),
|
||||
label='binarisation_job_-_%i' % (index),
|
||||
nCores=binarisation_job_n_cores
|
||||
)
|
||||
)
|
||||
files = map(lambda x: os.path.join(input_dir, x), files)
|
||||
cmd = 'ocropus-nlbin "{}"'.format('" "'.join(files))
|
||||
cmd += ' -o "{}"'.format(output_dir)
|
||||
cmd += ' -Q "{}"'.format(n_cores)
|
||||
deps = 'pdftoppm_job_-_{}'.format(i)
|
||||
lbl = 'ocropus_nlbin_job_-_{}'.format(i)
|
||||
ocropus_nlbin_jobs.append(
|
||||
self.addTask(command=cmd, dependencies=deps, label=lbl,
|
||||
nCores=n_cores))
|
||||
|
||||
'''
|
||||
' The post_binarisation_jobs are created based of the output files
|
||||
' of the binarisation_jobs. So wait until they are finished.
|
||||
' The post_ocropus_nlbin_jobs are created based on the output files
|
||||
' of the ocropus_nlbin_jobs. So wait until they are finished.
|
||||
'''
|
||||
self.waitForTasks()
|
||||
|
||||
'''
|
||||
' ##################################################
|
||||
' # Normalise file names from binarisation #
|
||||
' # post_ocropus_nlbin_jobs #
|
||||
' ##################################################
|
||||
'''
|
||||
post_binarisation_jobs = []
|
||||
for index, job in enumerate(self.jobs):
|
||||
post_ocropus_nlbin_jobs = []
|
||||
for i, job in enumerate(self.jobs):
|
||||
input_dir = os.path.join(job.output_dir, 'tmp')
|
||||
output_dir = input_dir
|
||||
number = 0
|
||||
files = os.listdir(os.path.join(job['output_dir'], 'tmp'))
|
||||
files = filter(lambda x: x.endswith('.bin.png'), files)
|
||||
files = filter(lambda x: x.endswith('.bin.png'),
|
||||
os.listdir(input_dir))
|
||||
files.sort()
|
||||
for file in files:
|
||||
cmd = 'mv "%s" "%s"' % (
|
||||
os.path.join(job['output_dir'], 'tmp', file),
|
||||
os.path.join(
|
||||
job['output_dir'],
|
||||
'tmp',
|
||||
'page-%i.bin.png' % (int(file.split('.', 1)[0]))
|
||||
)
|
||||
)
|
||||
post_binarisation_jobs.append(
|
||||
self.addTask(
|
||||
command=cmd,
|
||||
dependencies='binarisation_job_-_%i' % (index),
|
||||
label='post_binarisation_job_-_%i-%i' % (
|
||||
index,
|
||||
number
|
||||
)
|
||||
)
|
||||
)
|
||||
# int conversion is done in order to trim leading zeros
|
||||
output_file = os.path.join(output_dir, 'page-{}.bin.png'.format(int(file.split('.', 1)[0]))) # noqa
|
||||
cmd = 'mv "{}" "{}"'.format(os.path.join(output_dir, file),
|
||||
output_file)
|
||||
deps = 'ocropus_nlbin_job_-_{}'.format(i)
|
||||
lbl = 'post_ocropus_nlbin_job_-_{}-{}'.format(i, number)
|
||||
post_ocropus_nlbin_jobs.append(
|
||||
self.addTask(command=cmd, dependencies=deps,
|
||||
label=lbl))
|
||||
number += 1
|
||||
|
||||
'''
|
||||
' The ocr_jobs are created based of the output files of either the
|
||||
' split_jobs or post_binarisation_jobs. So wait until they are
|
||||
' The tesseract_jobs are created based of the output files of either
|
||||
' the pdftoppm_jobs or post_ocropus_nlbin_jobs. So wait until they are
|
||||
' finished.
|
||||
'''
|
||||
self.waitForTasks()
|
||||
|
||||
'''
|
||||
' ##################################################
|
||||
' # Optical Character Recognition #
|
||||
' # tesseract_jobs #
|
||||
' ##################################################
|
||||
'''
|
||||
ocr_jobs = []
|
||||
tesseract_jobs = []
|
||||
'''
|
||||
' Tesseract runs fastest with four cores. So we run it with either four
|
||||
' or, if there are less then four cores available for this workflow,
|
||||
' the available core number.
|
||||
'''
|
||||
ocr_job_n_cores = min(4, self.n_cores)
|
||||
for index, job in enumerate(self.jobs):
|
||||
files = os.listdir(os.path.join(job['output_dir'], 'tmp'))
|
||||
if self.skip_binarisation:
|
||||
files = filter(lambda x: x.endswith('.tif'), files)
|
||||
else:
|
||||
files = filter(lambda x: x.endswith('.bin.png'), files)
|
||||
n_cores = min(4, self.n_cores)
|
||||
for i, job in enumerate(self.jobs):
|
||||
input_dir = os.path.join(job.output_dir, 'tmp')
|
||||
output_dir = input_dir
|
||||
files = filter(lambda x: x.endswith('.bin.png' if self.binarize else '.tif'), # noqa
|
||||
os.listdir(input_dir))
|
||||
files.sort(key=lambda x: int(re.search(r'\d+', x).group(0)))
|
||||
files = map(
|
||||
lambda x: os.path.join(job['output_dir'], 'tmp', x),
|
||||
files
|
||||
)
|
||||
files = map(lambda x: os.path.join(input_dir, x), files)
|
||||
number = 0
|
||||
for file in files:
|
||||
cmd = 'tesseract "%s" "%s" -l "%s" hocr pdf txt' % (
|
||||
file,
|
||||
os.path.join(
|
||||
job['output_dir'],
|
||||
'tmp',
|
||||
file.rsplit('.', 1 if self.skip_binarisation else 2)[0]
|
||||
),
|
||||
self.lang
|
||||
)
|
||||
if self.skip_binarisation:
|
||||
ocr_job_dependencies = 'split_job_-_%i' % (index)
|
||||
output_file_base = os.path.join(output_dir, file.rsplit('.', 2 if self.binarize else 1)[0]) # noqa
|
||||
cmd = 'tesseract "{}" "{}"'.format(file, output_file_base)
|
||||
cmd += ' -l "{}"'.format(self.lang)
|
||||
cmd += ' hocr pdf txt'
|
||||
if self.binarize:
|
||||
deps = 'post_ocropus_nlbin_job_-_{}-{}'.format(i, number)
|
||||
else:
|
||||
ocr_job_dependencies = filter(
|
||||
lambda x: x == 'post_binarisation_job_-_%i-%i' % (
|
||||
index,
|
||||
number
|
||||
),
|
||||
post_binarisation_jobs
|
||||
)
|
||||
ocr_jobs.append(
|
||||
self.addTask(
|
||||
command=cmd,
|
||||
dependencies=ocr_job_dependencies,
|
||||
label='ocr_job_-_%i-%i' % (index, number),
|
||||
nCores=ocr_job_n_cores
|
||||
)
|
||||
)
|
||||
deps = 'pdftoppm_job_-_{}'.format(i)
|
||||
label = 'tesseract_jobs_-_{}-{}'.format(i, number)
|
||||
tesseract_jobs.append(
|
||||
self.addTask(command=cmd, dependencies=deps, label=label,
|
||||
nCores=n_cores))
|
||||
number += 1
|
||||
|
||||
'''
|
||||
@ -306,251 +220,191 @@ class OCRWorkflow(WorkflowRunner):
|
||||
|
||||
'''
|
||||
' ##################################################
|
||||
' # Create TEI P5 files #
|
||||
' # hocrtotei_jobs #
|
||||
' ##################################################
|
||||
'''
|
||||
hocr_to_tei_jobs = []
|
||||
for index, job in enumerate(self.jobs):
|
||||
files = os.listdir(os.path.join(job['output_dir'], 'tmp'))
|
||||
files = filter(lambda x: x.endswith('.hocr'), files)
|
||||
hocrtotei_jobs = []
|
||||
for i, job in enumerate(self.jobs):
|
||||
input_dir = os.path.join(job.output_dir, 'tmp')
|
||||
files = filter(lambda x: x.endswith('.hocr'),
|
||||
os.listdir(input_dir))
|
||||
files.sort(key=lambda x: int(re.search(r'\d+', x).group(0)))
|
||||
files = map(
|
||||
lambda x: '"' + os.path.join(job['output_dir'], 'tmp', x) + '"',
|
||||
files
|
||||
)
|
||||
cmd = 'hocrtotei %s "%s"' % (
|
||||
' '.join(files),
|
||||
os.path.join(
|
||||
job['output_dir'],
|
||||
os.path.join(job['output_dir'], job['name'] + '.xml')
|
||||
)
|
||||
)
|
||||
hocr_to_tei_jobs.append(
|
||||
self.addTask(
|
||||
command=cmd,
|
||||
dependencies=filter(
|
||||
lambda x: x.startswith('ocr_job_-_%i' % (index)),
|
||||
ocr_jobs
|
||||
),
|
||||
label='hocr_to_tei_job_-_%i' % (index)
|
||||
)
|
||||
)
|
||||
files = map(lambda x: os.path.join(input_dir, x), files)
|
||||
output_file = os.path.join(job.output_dir,
|
||||
'{}.xml'.format(job.name))
|
||||
cmd = 'hocrtotei "{}" "{}"'.format('" "'.join(files), output_file)
|
||||
deps = filter(lambda x: x.startswith('ocr_job_-_{}'.format(i)),
|
||||
tesseract_jobs)
|
||||
lbl = 'hocrtotei_job_-_{}'.format(i)
|
||||
hocrtotei_jobs.append(self.addTask(command=cmd, dependencies=deps,
|
||||
label=lbl))
|
||||
|
||||
'''
|
||||
' ##################################################
|
||||
' # Merge PDF files #
|
||||
' # pdfunite_jobs #
|
||||
' ##################################################
|
||||
'''
|
||||
pdf_merge_jobs = []
|
||||
for index, job in enumerate(self.jobs):
|
||||
files = os.listdir(os.path.join(job['output_dir'], 'tmp'))
|
||||
files = filter(lambda x: x.endswith('.pdf'), files)
|
||||
pdfunite_jobs = []
|
||||
for i, job in enumerate(self.jobs):
|
||||
input_dir = os.path.join(job.output_dir, 'tmp')
|
||||
files = filter(lambda x: x.endswith('.pdf'), os.listdir(input_dir))
|
||||
files.sort(key=lambda x: int(re.search(r'\d+', x).group(0)))
|
||||
files = map(
|
||||
lambda x: '"' + os.path.join(job['output_dir'], 'tmp', x) + '"',
|
||||
files
|
||||
)
|
||||
cmd = 'pdfunite %s "%s"' % (
|
||||
' '.join(files),
|
||||
os.path.join(
|
||||
job['output_dir'],
|
||||
os.path.join(job['output_dir'], job['name'] + '.pdf')
|
||||
)
|
||||
)
|
||||
pdf_merge_jobs.append(
|
||||
self.addTask(
|
||||
command=cmd,
|
||||
dependencies=filter(
|
||||
lambda x: x.startswith('ocr_job_-_%i' % (index)),
|
||||
ocr_jobs
|
||||
),
|
||||
label='pdf_merge_job_-_%i' % (index)
|
||||
)
|
||||
)
|
||||
files = map(lambda x: os.path.join(input_dir, x), files)
|
||||
output_file = os.path.join(job.output_dir,
|
||||
'{}.pdf'.format(job.name))
|
||||
cmd = 'pdfunite "{}" "{}"'.format('" "'.join(files), output_file)
|
||||
deps = filter(lambda x: x.startswith('ocr_job_-_{}'.format(i)),
|
||||
tesseract_jobs)
|
||||
lbl = 'pdfunite_job_-_{}'.format(i)
|
||||
pdfunite_jobs.append(self.addTask(command=cmd, dependencies=deps,
|
||||
label=lbl))
|
||||
|
||||
'''
|
||||
' ##################################################
|
||||
' # Merge text files #
|
||||
' # cat_jobs #
|
||||
' ##################################################
|
||||
'''
|
||||
txt_merge_jobs = []
|
||||
for index, job in enumerate(self.jobs):
|
||||
files = os.listdir(os.path.join(job['output_dir'], 'tmp'))
|
||||
files = filter(lambda x: x.endswith('.txt'), files)
|
||||
cat_jobs = []
|
||||
for i, job in enumerate(self.jobs):
|
||||
input_dir = os.path.join(job.output_dir, 'tmp')
|
||||
files = filter(lambda x: x.endswith('.txt'), os.listdir(input_dir))
|
||||
files.sort(key=lambda x: int(re.search(r'\d+', x).group(0)))
|
||||
files = map(
|
||||
lambda x: '"' + os.path.join(job['output_dir'], 'tmp', x) + '"',
|
||||
files
|
||||
)
|
||||
cmd = 'cat %s > "%s"' % (
|
||||
' '.join(files),
|
||||
os.path.join(
|
||||
job['output_dir'],
|
||||
os.path.join(job['output_dir'], job['name'] + '.txt')
|
||||
)
|
||||
)
|
||||
txt_merge_jobs.append(
|
||||
self.addTask(
|
||||
command=cmd,
|
||||
dependencies=filter(
|
||||
lambda x: x.startswith('ocr_job_-_%i' % (index)),
|
||||
ocr_jobs
|
||||
),
|
||||
label='txt_merge_job_-_%i' % (index)
|
||||
)
|
||||
)
|
||||
|
||||
if self.zip:
|
||||
all_zip_jobs = []
|
||||
all_zip_job_dependencies = (hocr_to_tei_jobs
|
||||
+ pdf_merge_jobs
|
||||
+ txt_merge_jobs)
|
||||
cmd = 'cd "%s" && zip "%s"-all-ocr-files.zip */*.{pdf,txt,xml} -x "pyflow.data*" && cd -' % (
|
||||
self.output_dir,
|
||||
self.zip
|
||||
)
|
||||
all_zip_jobs.append(
|
||||
self.addTask(
|
||||
command=cmd,
|
||||
dependencies=all_zip_job_dependencies,
|
||||
label='all_zip_job'
|
||||
)
|
||||
)
|
||||
|
||||
pdf_zip_jobs = []
|
||||
pdf_zip_job_dependencies = all_zip_jobs
|
||||
cmd = 'cd "%s" && zip -m "%s"-ocr-pdf.zip */*.pdf -x "pyflow.data*" && cd -' % (
|
||||
self.output_dir,
|
||||
self.zip
|
||||
)
|
||||
pdf_zip_jobs.append(
|
||||
self.addTask(
|
||||
command=cmd,
|
||||
dependencies=pdf_zip_job_dependencies,
|
||||
label='pdf_zip_job'
|
||||
)
|
||||
)
|
||||
|
||||
txt_zip_jobs = []
|
||||
txt_zip_job_dependencies = all_zip_jobs
|
||||
cmd = 'cd "%s" && zip -m "%s"-ocr-txt.zip */*.txt -x "pyflow.data*" && cd -' % (
|
||||
self.output_dir,
|
||||
self.zip
|
||||
)
|
||||
txt_zip_jobs.append(
|
||||
self.addTask(
|
||||
command=cmd,
|
||||
dependencies=txt_zip_job_dependencies,
|
||||
label='txt_zip_job'
|
||||
)
|
||||
)
|
||||
|
||||
xml_zip_jobs = []
|
||||
xml_zip_job_dependencies = all_zip_jobs
|
||||
cmd = 'cd "%s" && zip -m "%s"-ocr-xml.zip */*.xml -x "pyflow.data*" && cd -' % (
|
||||
self.output_dir,
|
||||
self.zip
|
||||
)
|
||||
xml_zip_jobs.append(
|
||||
self.addTask(
|
||||
command=cmd,
|
||||
dependencies=xml_zip_job_dependencies,
|
||||
label='xml_zip_job'
|
||||
)
|
||||
)
|
||||
files = map(lambda x: os.path.join(input_dir, x), files)
|
||||
output_file = os.path.join(job.output_dir,
|
||||
'{}.txt'.format(job.name))
|
||||
cmd = 'cat "{}" > "{}"'.format('" "'.join(files), output_file)
|
||||
deps = filter(lambda x: x.startswith('ocr_job_-_{}'.format(i)),
|
||||
tesseract_jobs)
|
||||
lbl = 'cat_job_-_{}'.format(i)
|
||||
cat_jobs.append(self.addTask(command=cmd, dependencies=deps,
|
||||
label=lbl))
|
||||
|
||||
'''
|
||||
' ##################################################
|
||||
' # Cleanup #
|
||||
' # zip_jobs #
|
||||
' ##################################################
|
||||
'''
|
||||
cleanup_jobs = []
|
||||
zip_jobs = []
|
||||
if self.zip is not None:
|
||||
cmd = 'cd "{}"'.format(self.output_dir)
|
||||
cmd += ' && '
|
||||
cmd += 'zip'
|
||||
cmd += ' -r'
|
||||
cmd += ' "{}_-_all" .'.format(self.zip)
|
||||
cmd += ' -x "pyflow.data*" "*tmp*"'
|
||||
cmd += ' -i "*.pdf" "*.txt" "*.xml"'
|
||||
cmd += ' && '
|
||||
cmd += 'cd -'
|
||||
deps = hocrtotei_jobs + pdfunite_jobs + cat_jobs
|
||||
lbl = 'zip_job_-_all'
|
||||
zip_jobs.append(self.addTask(command=cmd, dependencies=deps,
|
||||
label=lbl))
|
||||
|
||||
cmd = 'cd "{}"'.format(self.output_dir)
|
||||
cmd += ' && '
|
||||
cmd += 'zip'
|
||||
cmd += ' -m'
|
||||
cmd += ' -r'
|
||||
cmd += ' "{}_-_pdf" .'.format(self.zip)
|
||||
cmd += ' -x "pyflow.data*" "*tmp*"'
|
||||
cmd += ' -i "*.pdf"'
|
||||
cmd += ' && '
|
||||
cmd += 'cd -'
|
||||
deps = 'zip_job_-_all'
|
||||
lbl = 'zip_job_-_pdf'
|
||||
zip_jobs.append(self.addTask(command=cmd, dependencies=deps,
|
||||
label=lbl))
|
||||
|
||||
cmd = 'cd "{}"'.format(self.output_dir)
|
||||
cmd += ' && '
|
||||
cmd += 'zip'
|
||||
cmd += ' -m'
|
||||
cmd += ' -r'
|
||||
cmd += ' "{}_-_txt" .'.format(self.zip)
|
||||
cmd += ' -x "pyflow.data*" "*tmp*"'
|
||||
cmd += ' -i "*.txt"'
|
||||
cmd += ' && '
|
||||
cmd += 'cd -'
|
||||
deps = 'zip_job_-_all'
|
||||
lbl = 'zip_job_-_txt'
|
||||
zip_jobs.append(self.addTask(command=cmd, dependencies=deps,
|
||||
label=lbl))
|
||||
|
||||
cmd = 'cd "{}"'.format(self.output_dir)
|
||||
cmd += ' && '
|
||||
cmd += 'zip'
|
||||
cmd += ' -m'
|
||||
cmd += ' -r'
|
||||
cmd += ' "{}_-_xml" .'.format(self.zip)
|
||||
cmd += ' -x "pyflow.data*" "*tmp*"'
|
||||
cmd += ' -i "*.xml"'
|
||||
cmd += ' && '
|
||||
cmd += 'cd -'
|
||||
deps = 'zip_job_-_all'
|
||||
lbl = 'zip_job_-_xml'
|
||||
zip_jobs.append(self.addTask(command=cmd, dependencies=deps,
|
||||
label=lbl))
|
||||
|
||||
'''
|
||||
' ##################################################
|
||||
' # mv_jobs #
|
||||
' ##################################################
|
||||
'''
|
||||
mv_jobs = []
|
||||
if self.keep_intermediates:
|
||||
for index, job in enumerate(self.jobs):
|
||||
cleanup_job_dependencies = [
|
||||
'hocr_to_tei_job_-_%i' % (index),
|
||||
'pdf_merge_job_-_%i' % (index),
|
||||
'txt_merge_job_-_%i' % (index)
|
||||
]
|
||||
cmd = 'mv "%s"/*.hocr "%s"' % (
|
||||
os.path.join(job['output_dir'], 'tmp'),
|
||||
os.path.join(job['output_dir'], 'tmp', 'hocr'),
|
||||
)
|
||||
cmd += ' && mv "%s"/*.pdf "%s"' % (
|
||||
os.path.join(job['output_dir'], 'tmp'),
|
||||
os.path.join(job['output_dir'], 'tmp', 'pdf'),
|
||||
)
|
||||
cmd += ' && mv "%s"/*.tif "%s"' % (
|
||||
os.path.join(job['output_dir'], 'tmp'),
|
||||
os.path.join(job['output_dir'], 'tmp', 'tiff'),
|
||||
)
|
||||
cmd += ' && mv "%s"/*.txt "%s"' % (
|
||||
os.path.join(job['output_dir'], 'tmp'),
|
||||
os.path.join(job['output_dir'], 'tmp', 'txt'),
|
||||
)
|
||||
if not self.skip_binarisation:
|
||||
cmd += ' && mv "%s"/*.bin.png "%s"' % (
|
||||
os.path.join(job['output_dir'], 'tmp'),
|
||||
os.path.join(job['output_dir'], 'tmp', 'bin.png'),
|
||||
)
|
||||
cmd += ' && rm "%s"/*.nrm.png' % (
|
||||
os.path.join(job['output_dir'], 'tmp')
|
||||
)
|
||||
cleanup_jobs.append(
|
||||
self.addTask(
|
||||
command=cmd,
|
||||
dependencies=cleanup_job_dependencies,
|
||||
label='cleanup_job_-_%i' % (index)
|
||||
)
|
||||
)
|
||||
for i, job in enumerate(self.jobs):
|
||||
input_dir = os.path.join(job.output_dir, 'tmp')
|
||||
output_dir = input_dir
|
||||
cmd = 'mv "{}"/*.hocr "{}"'.format(
|
||||
input_dir, os.path.join(output_dir, 'hocr'))
|
||||
cmd += ' && '
|
||||
cmd += 'mv "{}"/*.pdf "{}"'.format(input_dir, os.path.join(output_dir, 'pdf')) # noqa
|
||||
cmd += ' && '
|
||||
cmd += 'mv "{}"/*.tif "{}"'.format(input_dir, os.path.join(output_dir, 'tiff')) # noqa
|
||||
cmd += ' && '
|
||||
cmd += 'mv "{}"/*.txt "{}"'.format(input_dir, os.path.join(output_dir, 'txt')) # noqa
|
||||
if self.binarize:
|
||||
cmd += ' && '
|
||||
cmd += 'mv "{}"/*.bin.png "{}"'.format(input_dir, os.path.join(output_dir, 'bin.png')) # noqa
|
||||
cmd += ' && '
|
||||
cmd += 'mv "{}"/*.nrm.png "{}"'.format(input_dir, os.path.join(output_dir, 'nrm.png')) # noqa
|
||||
deps = ['hocrtotei_job_-_{}'.format(i),
|
||||
'pdfunite_job_-_{}'.format(i),
|
||||
'cat_job_-_{}'.format(i)]
|
||||
lbl = 'mv_job_-_{}'.format(i)
|
||||
mv_jobs.append(self.addTask(command=cmd, dependencies=deps,
|
||||
label=lbl))
|
||||
else:
|
||||
for index, job in enumerate(self.jobs):
|
||||
cleanup_job_dependencies = [
|
||||
'hocr_to_tei_job_-_%i' % (index),
|
||||
'pdf_merge_job_-_%i' % (index),
|
||||
'txt_merge_job_-_%i' % (index)
|
||||
]
|
||||
cmd = 'rm -r "%s"' % (
|
||||
os.path.join(job['output_dir'], 'tmp')
|
||||
)
|
||||
cleanup_jobs.append(
|
||||
self.addTask(
|
||||
command=cmd,
|
||||
dependencies=cleanup_job_dependencies,
|
||||
label='cleanup_job_-_%i' % (index)
|
||||
)
|
||||
)
|
||||
for i, job in enumerate(self.jobs):
|
||||
input_dir = os.path.join(job.output_dir, 'tmp')
|
||||
cmd = 'rm -r "{}"'.format(input_dir)
|
||||
deps = ['hocrtotei_job_-_{}'.format(i),
|
||||
'pdfunite_job_-_{}'.format(i),
|
||||
'cat_job_-_{}'.format(i)]
|
||||
lbl = 'mv_job_-_{}'.format(i)
|
||||
mv_jobs.append(self.addTask(command=cmd, dependencies=deps,
|
||||
label=lbl))
|
||||
|
||||
|
||||
def analyze_jobs(input_dir, output_dir):
|
||||
def collect_jobs(input_dir, output_dir):
|
||||
jobs = []
|
||||
|
||||
for file in os.listdir(input_dir):
|
||||
if os.path.isdir(os.path.join(input_dir, file)):
|
||||
jobs += analyze_jobs(
|
||||
os.path.join(input_dir, file),
|
||||
os.path.join(output_dir, file)
|
||||
)
|
||||
elif file.endswith(('.pdf', '.tif', '.tiff')):
|
||||
jobs.append(
|
||||
{
|
||||
'filename': file,
|
||||
'name': file.rsplit('.', 1)[0],
|
||||
'output_dir': os.path.join(output_dir, file),
|
||||
'path': os.path.join(input_dir, file)
|
||||
}
|
||||
)
|
||||
|
||||
jobs += collect_jobs(os.path.join(input_dir, file),
|
||||
os.path.join(output_dir, file))
|
||||
elif file.endswith('.pdf'):
|
||||
jobs.append(OCRPipelineJob(os.path.join(input_dir, file),
|
||||
os.path.join(output_dir, file)))
|
||||
return jobs
|
||||
|
||||
|
||||
def main():
|
||||
args = parse_arguments()
|
||||
|
||||
wflow = OCRWorkflow(args)
|
||||
|
||||
retval = wflow.run(dataDirRoot=args.output_dir, nCores=args.n_cores)
|
||||
|
||||
args = parse_args()
|
||||
jobs = collect_jobs(args.i, args.o)
|
||||
ocr_pipeline = OCRPipeline(args.binarize, jobs, args.keep_intermediates,
|
||||
args.language, args.n_cores, args.o, args.zip)
|
||||
retval = ocr_pipeline.run(dataDirRoot=(args.log_dir or args.o),
|
||||
nCores=args.n_cores)
|
||||
sys.exit(retval)
|
||||
|
||||
|
||||
|
||||
44
wrapper/ocr
44
wrapper/ocr
@ -1,39 +1,29 @@
|
||||
#!/usr/bin/env python3
|
||||
# coding=utf-8
|
||||
|
||||
import argparse
|
||||
from argparse import ArgumentParser
|
||||
import os
|
||||
import subprocess
|
||||
|
||||
container_image = 'gitlab.ub.uni-bielefeld.de:4567/sfb1288inf/ocr:latest'
|
||||
container_input_dir = '/input'
|
||||
container_output_dir = '/output'
|
||||
uid = str(os.getuid())
|
||||
gid = str(os.getgid())
|
||||
CONTAINER_IMAGE = 'gitlab.ub.uni-bielefeld.de:4567/sfb1288inf/ocr:latest'
|
||||
CONTAINER_INPUT_DIR = '/input'
|
||||
CONTAINER_OUTPUT_DIR = '/output'
|
||||
UID = str(os.getuid())
|
||||
GID = str(os.getgid())
|
||||
|
||||
parser = argparse.ArgumentParser(add_help=False)
|
||||
parser.add_argument(
|
||||
'-i',
|
||||
dest='input_dir',
|
||||
required=False
|
||||
)
|
||||
parser.add_argument(
|
||||
'-o',
|
||||
dest='output_dir',
|
||||
required=False
|
||||
)
|
||||
parser = ArgumentParser(add_help=False)
|
||||
parser.add_argument('-i')
|
||||
parser.add_argument('-o')
|
||||
args, remaining_args = parser.parse_known_args()
|
||||
|
||||
cmd = ['docker', 'run', '--rm', '-it', '-u', uid + ':' + gid]
|
||||
if args.input_dir is not None:
|
||||
host_input_dir = os.path.abspath(args.input_dir)
|
||||
cmd += ['-v', host_input_dir + ':' + container_input_dir]
|
||||
remaining_args += ['-i', container_input_dir]
|
||||
if args.output_dir is not None:
|
||||
host_output_dir = os.path.abspath(args.output_dir)
|
||||
cmd += ['-v', host_output_dir + ':' + container_output_dir]
|
||||
remaining_args += ['-o', container_output_dir]
|
||||
cmd.append(container_image)
|
||||
cmd = ['docker', 'run', '--rm', '-it', '-u', '{}:{}'.format(UID, GID)]
|
||||
if args.o is not None:
|
||||
cmd += ['-v', '{}:{}'.format(os.path.abspath(args.o), CONTAINER_OUTPUT_DIR)]
|
||||
remaining_args.insert(0, CONTAINER_OUTPUT_DIR)
|
||||
if args.i is not None:
|
||||
cmd += ['-v', '{}:{}'.format(os.path.abspath(args.i), CONTAINER_INPUT_DIR)]
|
||||
remaining_args.insert(0, CONTAINER_INPUT_DIR)
|
||||
cmd.append(CONTAINER_IMAGE)
|
||||
cmd += remaining_args
|
||||
|
||||
subprocess.run(cmd)
|
||||
|
||||
Loading…
Reference in New Issue
Block a user