sbb_textline_detector: Add a OCR-D interface

pull/1/head
Gerber, Mike 5 years ago
parent 561a6f8a90
commit 0c915c75de

@ -0,0 +1 @@
qurator/sbb_textline_detector/ocrd-tool.json

@ -1 +1,2 @@
from .main import *
from .ocrd_cli import *

@ -35,16 +35,18 @@ __doc__ = \
class textlineerkenner:
def __init__(self, image_dir, dir_out, dir_models):
def __init__(self, image_dir, dir_out, f_name, dir_models):
self.image_dir = image_dir
self.dir_out = dir_out
self.f_name = f_name
if self.f_name is None:
try:
self.f_name = image_dir.split('/')[len(image_dir.split('/')) - 1]
self.f_name = self.f_name.split('.')[0]
print(self.f_name)
except:
self.f_name = self.f_name.split('.')[0]
self.dir_models = dir_models
try:
self.f_name = image_dir.split('/')[len(image_dir.split('/')) - 1]
self.f_name = self.f_name.split('.')[0]
print(self.f_name)
except:
self.f_name = self.f_name.split('.')[0]
self.kernel = np.ones((5, 5), np.uint8)
self.model_page_dir = dir_models + '/model_page.h5'
self.model_region_dir = dir_models + '/model_strukturerkennung.h5'
@ -365,6 +367,7 @@ class textlineerkenner:
return img_r
def get_image_and_scales(self):
print(self.image_dir)
self.image = cv2.imread(self.image_dir)
self.height_org = self.image.shape[0]
self.width_org = self.image.shape[1]
@ -1298,7 +1301,7 @@ class textlineerkenner:
unireg.text = ' '
tree = ET.ElementTree(data)
tree.write(dir_of_image + self.f_name + ".xml")
tree.write(os.path.join(dir_of_image, self.f_name) + ".xml")
def run(self):
self.get_image_and_scales()
@ -1307,7 +1310,7 @@ class textlineerkenner:
boxes,contours=self.get_text_region_contours_and_boxes(text_regions)
self.get_all_image_patches_based_on_text_regions(boxes,image_page)
textline_mask_tot=self.textline_contours(image_page)
self.get_textlines_for_each_textregions(textline_mask_tot,boxes)
self.get_slopes_for_each_text_region(contours)
self.deskew_textline_patches(contours, boxes)
@ -1321,7 +1324,7 @@ class textlineerkenner:
def main(image, out, model):
possibles = globals() # XXX unused?
possibles.update(locals())
x = textlineerkenner(image, out, model)
x = textlineerkenner(image, out, None, model)
x.run()

@ -0,0 +1,19 @@
{
"version": "0.0.1",
"tools": {
"ocrd_sbb_textline_detector": {
"executable": "ocrd_sbb_textline_detector",
"description": "Detect lines",
"steps": ["layout/segmentation/line"],
"input_file_grp": [
"OCR-D-IMG"
],
"output_file_grp": [
"OCR-D-SBB-SEG-LINE"
],
"parameters": {
"model": {"type": "string", "format": "file", "cacheable": true}
}
}
}
}

@ -0,0 +1,61 @@
import json
import os
import click
from ocrd import Processor
from ocrd.decorators import ocrd_cli_options, ocrd_cli_wrap_processor
from ocrd_utils import concat_padded, getLogger
from pkg_resources import resource_string
from qurator.sbb_textline_detector import textlineerkenner
log = getLogger('processor.OcrdSbbTextlineDetectorRecognize')
OCRD_TOOL = json.loads(resource_string(__name__, 'ocrd-tool.json').decode('utf8'))
@click.command()
@ocrd_cli_options
def ocrd_sbb_textline_detector(*args, **kwargs):
return ocrd_cli_wrap_processor(OcrdSbbTextlineDetectorRecognize, *args, **kwargs)
class OcrdSbbTextlineDetectorRecognize(Processor):
def __init__(self, *args, **kwargs):
kwargs['ocrd_tool'] = OCRD_TOOL['tools']['ocrd_sbb_textline_detector']
super(OcrdSbbTextlineDetectorRecognize, self).__init__(*args, **kwargs)
def _make_file_id(self, input_file, input_file_grp, n):
file_id = input_file.ID.replace(input_file_grp, self.output_file_grp)
if file_id == input_file.ID:
file_id = concat_padded(self.output_file_grp, n)
return file_id
def process(self):
for n, page_id in enumerate(self.workspace.mets.physical_pages):
image_file = self.workspace.mets.find_files(fileGrp=self.input_file_grp, pageId=page_id)[0]
log.info("INPUT FILE %i / %s", n, image_file)
file_id = self._make_file_id(image_file, self.output_file_grp, n)
# Process the files
try:
os.mkdir(self.output_file_grp)
except FileExistsError:
pass
model = self.parameter['model']
x = textlineerkenner(image_file.local_filename, self.output_file_grp, file_id, model)
x.run()
self.workspace.add_file(
ID=file_id + '.xml',
file_grp=self.output_file_grp,
pageId=page_id,
mimetype='application/vnd.prima.page+xml',
local_filename=self.output_file_grp + '/' + file_id)
if __name__ == '__main__':
ocrd_sbb_textline_detector()

@ -1,5 +1,5 @@
opencv-python
numpy<=1.14.5
numpy
matplotlib
seaborn
tqdm
@ -9,3 +9,4 @@ scikit-learn
tensorflow-gpu < 2.0
scipy
click
ocrd >= 1.0.0b19

@ -21,6 +21,7 @@ setup(
entry_points={
'console_scripts': [
"sbb_textline_detector=qurator.sbb_textline_detector:main",
"ocrd_sbb_textline_detector=qurator.sbb_textline_detector:ocrd_sbb_textline_detector",
]
},
python_requires='>=3.6.0',

Loading…
Cancel
Save