A Python/Cython wrapper for the Omnipage CSDK. It offers a simple python API to get OCR results.
First, set LD_LIBRARY_PATH to point to the CSDK shared libraries.
For example:
$ export LD_LIBRARY_PATH=/usr/local/lib/nuance-omnipage-csdk-lib64-20.3:$LD_LIBRARY_PATHNext, clone this repo, enter the directory and
$ python3 setup.py installIf the install fails, you may have to edit the paths pointing to the SDK include files and shared libraries. They can be found in setup.py.
import os
from pycsdk import *
from PIL import Image
csdk = CSDK('my_company', 'my_product', license_file='/path/to/license_file', code='license_code')
csdk.set_setting('Kernel.OcrMgr.TradeOff', TO_ACCURATE)
csdk.set_setting('Kernel.OcrMgr.PDF.TradeOff', TO_ACCURATE)
csdk.set_setting('Kernel.OcrMgr.PDF.ProcessingMode', PDF_PM_AUTO)
csdk.set_setting('Kernel.OcrMgr.DefaultRecognitionModule', RM_OMNIFONT_PLUS3W)
csdk.set_language(LANG_FRE)
csdk.add_language(LANG_ENG)
# note each file can be handled by a child process with e.g. os.fork()
# (the CSDK does not support multithreading)
with csdk.open_file('myfile.pdf') as f:
for page_id in range(f.nb_pages):
with f.open_page(page_id) as p:
p.process(despeckle_method=DESPECKLE_MEDIAN, remove_rule_lines=True)
print('found {} zones, {} letters'.format(len(p.zones), len(p.letters)))
# do something interesting with zones (and cells in zones) and letters
print(p.zones)
print(p.letters)
# save image for page
p.image.save('myfile_{:04}.png'.format(page_id + 1), 'PNG')