mirror of
https://github.com/qurator-spk/eynollah.git
synced 2026-06-16 09:59:13 +02:00
remove (half-implemented) page_alto functionality
This commit is contained in:
kba
parent
9858221724
commit
bed7fe526b
3 changed files with 169 additions and 291 deletions
|
|
@ -56,12 +56,6 @@ from ..utils import is_image_filename
|
||||||
is_flag=True,
|
is_flag=True,
|
||||||
help="if this parameter set to true, vertical textline images will be excluded.",
|
help="if this parameter set to true, vertical textline images will be excluded.",
|
||||||
)
|
)
|
||||||
@click.option(
|
|
||||||
"--page_alto",
|
|
||||||
"-alto",
|
|
||||||
is_flag=True,
|
|
||||||
help="If this parameter is set to True, text line image cropping and text extraction are performed using PAGE/ALTO files. Otherwise, the default method for PAGE XML files is used.",
|
|
||||||
)
|
|
||||||
def linegt_cli(
|
def linegt_cli(
|
||||||
image,
|
image,
|
||||||
dir_in,
|
dir_in,
|
||||||
|
|
@ -70,7 +64,6 @@ def linegt_cli(
|
||||||
pref_of_dataset,
|
pref_of_dataset,
|
||||||
do_not_mask_with_textline_contour,
|
do_not_mask_with_textline_contour,
|
||||||
exclude_vertical_lines,
|
exclude_vertical_lines,
|
||||||
page_alto,
|
|
||||||
):
|
):
|
||||||
assert bool(dir_in) ^ bool(image), "Set --dir-in or --image-filename, not both"
|
assert bool(dir_in) ^ bool(image), "Set --dir-in or --image-filename, not both"
|
||||||
if dir_in:
|
if dir_in:
|
||||||
|
|
@ -86,147 +79,69 @@ def linegt_cli(
|
||||||
dir_xml = os.path.join(dir_xmls, file_name + '.xml')
|
dir_xml = os.path.join(dir_xmls, file_name + '.xml')
|
||||||
img = cv2.imread(dir_img)
|
img = cv2.imread(dir_img)
|
||||||
|
|
||||||
if page_alto:
|
total_bb_coordinates = []
|
||||||
h, w = img.shape[:2]
|
|
||||||
|
|
||||||
tree = ET.parse(dir_xml)
|
tree = ET.parse(dir_xml, parser=ET.XMLParser(encoding="utf-8"))
|
||||||
root = tree.getroot()
|
root = tree.getroot()
|
||||||
|
alltags = [elem.tag for elem in root.iter()]
|
||||||
|
|
||||||
NS = {'alto': root.tag.split('}')[0].strip('{')}#{"alto": "http://www.loc.gov/standards/alto/ns-v4#"}
|
name_space = alltags[0].split('}')[0]
|
||||||
|
name_space = name_space.split('{')[1]
|
||||||
|
|
||||||
results = []
|
region_tags = np.unique([x for x in alltags if x.endswith('TextRegion')])
|
||||||
|
|
||||||
indexer_textlines = 0
|
cropped_lines_region_indexer = []
|
||||||
for line in root.findall(".//alto:TextLine", NS):
|
|
||||||
string_el = line.find("alto:String", NS)
|
|
||||||
textline_text = string_el.attrib["CONTENT"] if string_el is not None else None
|
|
||||||
|
|
||||||
polygon_el = line.find("alto:Shape/alto:Polygon", NS)
|
indexer_text_region = 0
|
||||||
if polygon_el is None:
|
indexer_textlines = 0
|
||||||
continue
|
# FIXME: non recursive, use OCR-D PAGE generateDS API. Or use an existing tool for this purpose altogether
|
||||||
|
for nn in root.iter(region_tags):
|
||||||
|
for child_textregion in nn:
|
||||||
|
if child_textregion.tag.endswith("TextLine"):
|
||||||
|
for child_textlines in child_textregion:
|
||||||
|
if child_textlines.tag.endswith("Coords"):
|
||||||
|
cropped_lines_region_indexer.append(indexer_text_region)
|
||||||
|
p_h = child_textlines.attrib['points'].split(' ')
|
||||||
|
textline_coords = np.array([[int(x.split(',')[0]), int(x.split(',')[1])] for x in p_h])
|
||||||
|
|
||||||
points = polygon_el.attrib["POINTS"].split()
|
x, y, w, h = cv2.boundingRect(textline_coords)
|
||||||
coords = [
|
|
||||||
(int(points[i]), int(points[i + 1]))
|
|
||||||
for i in range(0, len(points), 2)
|
|
||||||
]
|
|
||||||
|
|
||||||
coords = np.array(coords, dtype=np.int32)
|
if exclude_vertical_lines and h > 1.4 * w:
|
||||||
x, y, w, h = cv2.boundingRect(coords)
|
img_crop = None
|
||||||
|
continue
|
||||||
|
|
||||||
|
total_bb_coordinates.append([x, y, w, h])
|
||||||
|
|
||||||
|
img_poly_on_img = np.copy(img)
|
||||||
|
|
||||||
|
mask_poly = np.zeros(img.shape)
|
||||||
|
mask_poly = cv2.fillPoly(mask_poly, pts=[textline_coords], color=(1, 1, 1))
|
||||||
|
|
||||||
|
mask_poly = mask_poly[y : y + h, x : x + w, :]
|
||||||
|
img_crop = img_poly_on_img[y : y + h, x : x + w, :]
|
||||||
|
|
||||||
|
if not do_not_mask_with_textline_contour:
|
||||||
|
img_crop[mask_poly == 0] = 255
|
||||||
|
|
||||||
|
if img_crop.shape[0] == 0 or img_crop.shape[1] == 0:
|
||||||
|
img_crop = None
|
||||||
|
continue
|
||||||
|
|
||||||
|
|
||||||
if exclude_vertical_lines and h > 1.4 * w:
|
if child_textlines.tag.endswith("TextEquiv"):
|
||||||
img_crop = None
|
for cheild_text in child_textlines:
|
||||||
continue
|
if cheild_text.tag.endswith("Unicode"):
|
||||||
|
textline_text = cheild_text.text
|
||||||
|
if textline_text and img_crop is not None:
|
||||||
|
base_name = os.path.join(
|
||||||
|
dir_out, file_name + '_line_' + str(indexer_textlines)
|
||||||
|
)
|
||||||
|
if pref_of_dataset:
|
||||||
|
base_name += '_' + pref_of_dataset
|
||||||
|
if not do_not_mask_with_textline_contour:
|
||||||
|
base_name += '_masked'
|
||||||
|
|
||||||
img_poly_on_img = np.copy(img)
|
with open(base_name + '.txt', 'w') as text_file:
|
||||||
|
text_file.write(textline_text)
|
||||||
mask_poly = np.zeros(img.shape)
|
cv2.imwrite(base_name + '.png', img_crop)
|
||||||
mask_poly = cv2.fillPoly(mask_poly, pts=[coords], color=(1, 1, 1))
|
indexer_textlines += 1
|
||||||
|
|
||||||
mask_poly = mask_poly[y : y + h, x : x + w, :]
|
|
||||||
img_crop = img_poly_on_img[y : y + h, x : x + w, :]
|
|
||||||
|
|
||||||
if not do_not_mask_with_textline_contour:
|
|
||||||
img_crop[mask_poly == 0] = 255
|
|
||||||
|
|
||||||
if img_crop.shape[0] == 0 or img_crop.shape[1] == 0:
|
|
||||||
img_crop = None
|
|
||||||
continue
|
|
||||||
|
|
||||||
if textline_text and img_crop is not None:
|
|
||||||
base_name = os.path.join(
|
|
||||||
dir_out, file_name + '_line_' + str(indexer_textlines)
|
|
||||||
)
|
|
||||||
if pref_of_dataset:
|
|
||||||
base_name += '_' + pref_of_dataset
|
|
||||||
if not do_not_mask_with_textline_contour:
|
|
||||||
base_name += '_masked'
|
|
||||||
|
|
||||||
with open(base_name + '.txt', 'w') as text_file:
|
|
||||||
text_file.write(textline_text)
|
|
||||||
cv2.imwrite(base_name + '.png', img_crop)
|
|
||||||
indexer_textlines += 1
|
|
||||||
|
|
||||||
|
|
||||||
|
|
||||||
|
|
||||||
|
|
||||||
|
|
||||||
|
|
||||||
|
|
||||||
|
|
||||||
|
|
||||||
|
|
||||||
|
|
||||||
|
|
||||||
|
|
||||||
|
|
||||||
|
|
||||||
else:
|
|
||||||
total_bb_coordinates = []
|
|
||||||
|
|
||||||
tree = ET.parse(dir_xml, parser=ET.XMLParser(encoding="utf-8"))
|
|
||||||
root = tree.getroot()
|
|
||||||
alltags = [elem.tag for elem in root.iter()]
|
|
||||||
|
|
||||||
name_space = alltags[0].split('}')[0]
|
|
||||||
name_space = name_space.split('{')[1]
|
|
||||||
|
|
||||||
region_tags = np.unique([x for x in alltags if x.endswith('TextRegion')])
|
|
||||||
|
|
||||||
cropped_lines_region_indexer = []
|
|
||||||
|
|
||||||
indexer_text_region = 0
|
|
||||||
indexer_textlines = 0
|
|
||||||
# FIXME: non recursive, use OCR-D PAGE generateDS API. Or use an existing tool for this purpose altogether
|
|
||||||
for nn in root.iter(region_tags):
|
|
||||||
for child_textregion in nn:
|
|
||||||
if child_textregion.tag.endswith("TextLine"):
|
|
||||||
for child_textlines in child_textregion:
|
|
||||||
if child_textlines.tag.endswith("Coords"):
|
|
||||||
cropped_lines_region_indexer.append(indexer_text_region)
|
|
||||||
p_h = child_textlines.attrib['points'].split(' ')
|
|
||||||
textline_coords = np.array([[int(x.split(',')[0]), int(x.split(',')[1])] for x in p_h])
|
|
||||||
|
|
||||||
x, y, w, h = cv2.boundingRect(textline_coords)
|
|
||||||
|
|
||||||
if exclude_vertical_lines and h > 1.4 * w:
|
|
||||||
img_crop = None
|
|
||||||
continue
|
|
||||||
|
|
||||||
total_bb_coordinates.append([x, y, w, h])
|
|
||||||
|
|
||||||
img_poly_on_img = np.copy(img)
|
|
||||||
|
|
||||||
mask_poly = np.zeros(img.shape)
|
|
||||||
mask_poly = cv2.fillPoly(mask_poly, pts=[textline_coords], color=(1, 1, 1))
|
|
||||||
|
|
||||||
mask_poly = mask_poly[y : y + h, x : x + w, :]
|
|
||||||
img_crop = img_poly_on_img[y : y + h, x : x + w, :]
|
|
||||||
|
|
||||||
if not do_not_mask_with_textline_contour:
|
|
||||||
img_crop[mask_poly == 0] = 255
|
|
||||||
|
|
||||||
if img_crop.shape[0] == 0 or img_crop.shape[1] == 0:
|
|
||||||
img_crop = None
|
|
||||||
continue
|
|
||||||
|
|
||||||
|
|
||||||
if child_textlines.tag.endswith("TextEquiv"):
|
|
||||||
for cheild_text in child_textlines:
|
|
||||||
if cheild_text.tag.endswith("Unicode"):
|
|
||||||
textline_text = cheild_text.text
|
|
||||||
if textline_text and img_crop is not None:
|
|
||||||
base_name = os.path.join(
|
|
||||||
dir_out, file_name + '_line_' + str(indexer_textlines)
|
|
||||||
)
|
|
||||||
if pref_of_dataset:
|
|
||||||
base_name += '_' + pref_of_dataset
|
|
||||||
if not do_not_mask_with_textline_contour:
|
|
||||||
base_name += '_masked'
|
|
||||||
|
|
||||||
with open(base_name + '.txt', 'w') as text_file:
|
|
||||||
text_file.write(textline_text)
|
|
||||||
cv2.imwrite(base_name + '.png', img_crop)
|
|
||||||
indexer_textlines += 1
|
|
||||||
|
|
|
||||||
|
|
@ -73,14 +73,8 @@ def main():
|
||||||
is_flag=True,
|
is_flag=True,
|
||||||
help="if this parameter set to true, generated labels and in the case of provided org images cropping will be imposed and cropped labels and images will be written in output directories.",
|
help="if this parameter set to true, generated labels and in the case of provided org images cropping will be imposed and cropped labels and images will be written in output directories.",
|
||||||
)
|
)
|
||||||
@click.option(
|
|
||||||
"--page_alto",
|
|
||||||
"-alto",
|
|
||||||
is_flag=True,
|
|
||||||
help="If this parameter is set to True, textline label generation is performed using PAGE/ALTO files. Otherwise, the default method for PAGE XML files is used.",
|
|
||||||
)
|
|
||||||
|
|
||||||
def pagexml2label(dir_xml,dir_out,type_output,config, printspace, dir_images, dir_out_images, page_alto):
|
def pagexml2label(dir_xml,dir_out,type_output,config, printspace, dir_images, dir_out_images):
|
||||||
if config:
|
if config:
|
||||||
with open(config) as f:
|
with open(config) as f:
|
||||||
config_params = json.load(f)
|
config_params = json.load(f)
|
||||||
|
|
@ -88,7 +82,7 @@ def pagexml2label(dir_xml,dir_out,type_output,config, printspace, dir_images, di
|
||||||
print("passed")
|
print("passed")
|
||||||
config_params = None
|
config_params = None
|
||||||
gt_list = get_content_of_dir(dir_xml)
|
gt_list = get_content_of_dir(dir_xml)
|
||||||
get_images_of_ground_truth(gt_list,dir_xml,dir_out,type_output, config, config_params, printspace, dir_images, dir_out_images, page_alto)
|
get_images_of_ground_truth(gt_list,dir_xml,dir_out,type_output, config, config_params, printspace, dir_images, dir_out_images)
|
||||||
|
|
||||||
@main.command()
|
@main.command()
|
||||||
@click.option(
|
@click.option(
|
||||||
|
|
|
||||||
|
|
@ -686,7 +686,7 @@ def get_layout_contours_for_visualization(xml_file):
|
||||||
co_noise.append(np.array(c_t_in))
|
co_noise.append(np.array(c_t_in))
|
||||||
return co_text, co_graphic, co_sep, co_img, co_table, co_map, co_music, co_noise, y_len, x_len
|
return co_text, co_graphic, co_sep, co_img, co_table, co_map, co_music, co_noise, y_len, x_len
|
||||||
|
|
||||||
def get_images_of_ground_truth(gt_list, dir_in, output_dir, output_type, config_file, config_params, printspace, dir_images, dir_out_images, page_alto=False):
|
def get_images_of_ground_truth(gt_list, dir_in, output_dir, output_type, config_file, config_params, printspace, dir_images, dir_out_images):
|
||||||
"""
|
"""
|
||||||
Reading the page xml files and write the ground truth images into given output directory.
|
Reading the page xml files and write the ground truth images into given output directory.
|
||||||
"""
|
"""
|
||||||
|
|
@ -699,94 +699,81 @@ def get_images_of_ground_truth(gt_list, dir_in, output_dir, output_type, config_
|
||||||
print(gt_list[index])
|
print(gt_list[index])
|
||||||
|
|
||||||
try:
|
try:
|
||||||
if page_alto:
|
tree1 = ET.parse(dir_in+'/'+gt_list[index], parser = ET.XMLParser(encoding='utf-8'))
|
||||||
tree = ET.parse(dir_in+'/'+gt_list[index])
|
root1=tree1.getroot()
|
||||||
root = tree.getroot()
|
alltags=[elem.tag for elem in root1.iter()]
|
||||||
|
link=alltags[0].split('}')[0]+'}'
|
||||||
NS = {'alto': root.tag.split('}')[0].strip('{')}#{"alto": "http://www.loc.gov/standards/alto/ns-v4#"}
|
|
||||||
x_len, y_len = 0, 0
|
|
||||||
|
|
||||||
page = root.find('.//alto:Page', NS)
|
|
||||||
|
|
||||||
x_len = int( page.get("WIDTH") )
|
|
||||||
y_len = int( page.get("HEIGHT") )
|
|
||||||
|
|
||||||
else:
|
|
||||||
tree1 = ET.parse(dir_in+'/'+gt_list[index], parser = ET.XMLParser(encoding='utf-8'))
|
|
||||||
root1=tree1.getroot()
|
|
||||||
alltags=[elem.tag for elem in root1.iter()]
|
|
||||||
link=alltags[0].split('}')[0]+'}'
|
|
||||||
|
|
||||||
|
|
||||||
x_len, y_len = 0, 0
|
x_len, y_len = 0, 0
|
||||||
for jj in root1.iter(link+'Page'):
|
for jj in root1.iter(link+'Page'):
|
||||||
y_len=int(jj.attrib['imageHeight'])
|
y_len=int(jj.attrib['imageHeight'])
|
||||||
x_len=int(jj.attrib['imageWidth'])
|
x_len=int(jj.attrib['imageWidth'])
|
||||||
|
|
||||||
if 'columns_width' in list(config_params.keys()):
|
if 'columns_width' in list(config_params.keys()):
|
||||||
columns_width_dict = config_params['columns_width']
|
columns_width_dict = config_params['columns_width']
|
||||||
metadata_element = root1.find(link+'Metadata')
|
metadata_element = root1.find(link+'Metadata')
|
||||||
num_col = None
|
num_col = None
|
||||||
for child in metadata_element:
|
for child in metadata_element:
|
||||||
tag2 = child.tag
|
tag2 = child.tag
|
||||||
if tag2.endswith('}Comments') or tag2.endswith('}comments'):
|
if tag2.endswith('}Comments') or tag2.endswith('}comments'):
|
||||||
text_comments = child.text
|
text_comments = child.text
|
||||||
num_col = int(text_comments.split('num_col')[1])
|
num_col = int(text_comments.split('num_col')[1])
|
||||||
|
|
||||||
if num_col:
|
if num_col:
|
||||||
x_new = columns_width_dict[str(num_col)]
|
x_new = columns_width_dict[str(num_col)]
|
||||||
y_new = int ( x_new * (y_len / float(x_len)) )
|
y_new = int ( x_new * (y_len / float(x_len)) )
|
||||||
|
|
||||||
if printspace or "printspace_as_class_in_layout" in list(config_params.keys()):
|
if printspace or "printspace_as_class_in_layout" in list(config_params.keys()):
|
||||||
region_tags = np.unique([x for x in alltags if x.endswith('PrintSpace') or x.endswith('Border')])
|
region_tags = np.unique([x for x in alltags if x.endswith('PrintSpace') or x.endswith('Border')])
|
||||||
co_use_case = []
|
co_use_case = []
|
||||||
|
|
||||||
for tag in region_tags:
|
for tag in region_tags:
|
||||||
tag_endings = ['}PrintSpace','}Border']
|
tag_endings = ['}PrintSpace','}Border']
|
||||||
|
|
||||||
if tag.endswith(tag_endings[0]) or tag.endswith(tag_endings[1]):
|
if tag.endswith(tag_endings[0]) or tag.endswith(tag_endings[1]):
|
||||||
for nn in root1.iter(tag):
|
for nn in root1.iter(tag):
|
||||||
c_t_in = []
|
c_t_in = []
|
||||||
sumi = 0
|
sumi = 0
|
||||||
for vv in nn.iter():
|
for vv in nn.iter():
|
||||||
# check the format of coords
|
# check the format of coords
|
||||||
if vv.tag == link + 'Coords':
|
if vv.tag == link + 'Coords':
|
||||||
coords = bool(vv.attrib)
|
coords = bool(vv.attrib)
|
||||||
if coords:
|
if coords:
|
||||||
p_h = vv.attrib['points'].split(' ')
|
p_h = vv.attrib['points'].split(' ')
|
||||||
c_t_in.append(
|
c_t_in.append(
|
||||||
np.array([[int(x.split(',')[0]), int(x.split(',')[1])] for x in p_h]))
|
np.array([[int(x.split(',')[0]), int(x.split(',')[1])] for x in p_h]))
|
||||||
break
|
|
||||||
else:
|
|
||||||
pass
|
|
||||||
|
|
||||||
if vv.tag == link + 'Point':
|
|
||||||
c_t_in.append([int(float(vv.attrib['x'])), int(float(vv.attrib['y']))])
|
|
||||||
sumi += 1
|
|
||||||
elif vv.tag != link + 'Point' and sumi >= 1:
|
|
||||||
break
|
break
|
||||||
co_use_case.append(np.array(c_t_in))
|
else:
|
||||||
|
pass
|
||||||
|
|
||||||
img = np.zeros((y_len, x_len, 3))
|
if vv.tag == link + 'Point':
|
||||||
|
c_t_in.append([int(float(vv.attrib['x'])), int(float(vv.attrib['y']))])
|
||||||
|
sumi += 1
|
||||||
|
elif vv.tag != link + 'Point' and sumi >= 1:
|
||||||
|
break
|
||||||
|
co_use_case.append(np.array(c_t_in))
|
||||||
|
|
||||||
img_poly = cv2.fillPoly(img, pts=co_use_case, color=(1, 1, 1))
|
img = np.zeros((y_len, x_len, 3))
|
||||||
|
|
||||||
img_poly = img_poly.astype(np.uint8)
|
img_poly = cv2.fillPoly(img, pts=co_use_case, color=(1, 1, 1))
|
||||||
|
|
||||||
imgray = cv2.cvtColor(img_poly, cv2.COLOR_BGR2GRAY)
|
img_poly = img_poly.astype(np.uint8)
|
||||||
_, thresh = cv2.threshold(imgray, 0, 255, 0)
|
|
||||||
|
|
||||||
contours, _ = cv2.findContours(thresh, cv2.RETR_TREE, cv2.CHAIN_APPROX_SIMPLE)
|
imgray = cv2.cvtColor(img_poly, cv2.COLOR_BGR2GRAY)
|
||||||
|
_, thresh = cv2.threshold(imgray, 0, 255, 0)
|
||||||
|
|
||||||
cnt_size = np.array([cv2.contourArea(contours[j]) for j in range(len(contours))])
|
contours, _ = cv2.findContours(thresh, cv2.RETR_TREE, cv2.CHAIN_APPROX_SIMPLE)
|
||||||
|
|
||||||
try:
|
cnt_size = np.array([cv2.contourArea(contours[j]) for j in range(len(contours))])
|
||||||
cnt = contours[np.argmax(cnt_size)]
|
|
||||||
x, y, w, h = cv2.boundingRect(cnt)
|
|
||||||
except:
|
|
||||||
x, y , w, h = 0, 0, x_len, y_len
|
|
||||||
|
|
||||||
bb_xywh = [x, y, w, h]
|
try:
|
||||||
|
cnt = contours[np.argmax(cnt_size)]
|
||||||
|
x, y, w, h = cv2.boundingRect(cnt)
|
||||||
|
except:
|
||||||
|
x, y , w, h = 0, 0, x_len, y_len
|
||||||
|
|
||||||
|
bb_xywh = [x, y, w, h]
|
||||||
|
|
||||||
|
|
||||||
if config_file and (config_params['use_case']=='textline' or config_params['use_case']=='word' or config_params['use_case']=='glyph' or config_params['use_case']=='printspace'):
|
if config_file and (config_params['use_case']=='textline' or config_params['use_case']=='word' or config_params['use_case']=='glyph' or config_params['use_case']=='printspace'):
|
||||||
|
|
@ -797,67 +784,49 @@ def get_images_of_ground_truth(gt_list, dir_in, output_dir, output_type, config_
|
||||||
|
|
||||||
textline_rgb_color = (255, 0, 0)
|
textline_rgb_color = (255, 0, 0)
|
||||||
|
|
||||||
if page_alto:
|
if config_params['use_case']=='textline':
|
||||||
co_use_case = []
|
region_tags = np.unique([x for x in alltags if x.endswith('TextLine')])
|
||||||
for line in root.findall(".//alto:TextLine", NS):
|
elif config_params['use_case']=='word':
|
||||||
string_el = line.find("alto:String", NS)
|
region_tags = np.unique([x for x in alltags if x.endswith('Word')])
|
||||||
textline_text = string_el.attrib["CONTENT"] if string_el is not None else None
|
elif config_params['use_case']=='glyph':
|
||||||
|
region_tags = np.unique([x for x in alltags if x.endswith('Glyph')])
|
||||||
|
elif config_params['use_case']=='printspace':
|
||||||
|
region_tags = np.unique([x for x in alltags if x.endswith('PrintSpace')])
|
||||||
|
|
||||||
polygon_el = line.find("alto:Shape/alto:Polygon", NS)
|
co_use_case = []
|
||||||
if polygon_el is None:
|
|
||||||
continue
|
|
||||||
|
|
||||||
points = polygon_el.attrib["POINTS"].split()
|
for tag in region_tags:
|
||||||
coords = [
|
|
||||||
(int(points[i]), int(points[i + 1]))
|
|
||||||
for i in range(0, len(points), 2)
|
|
||||||
]
|
|
||||||
|
|
||||||
co_use_case.append( np.array(coords, dtype=np.int32) )
|
|
||||||
else:
|
|
||||||
if config_params['use_case']=='textline':
|
if config_params['use_case']=='textline':
|
||||||
region_tags = np.unique([x for x in alltags if x.endswith('TextLine')])
|
tag_endings = ['}TextLine','}textline']
|
||||||
elif config_params['use_case']=='word':
|
elif config_params['use_case']=='word':
|
||||||
region_tags = np.unique([x for x in alltags if x.endswith('Word')])
|
tag_endings = ['}Word','}word']
|
||||||
elif config_params['use_case']=='glyph':
|
elif config_params['use_case']=='glyph':
|
||||||
region_tags = np.unique([x for x in alltags if x.endswith('Glyph')])
|
tag_endings = ['}Glyph','}glyph']
|
||||||
elif config_params['use_case']=='printspace':
|
elif config_params['use_case']=='printspace':
|
||||||
region_tags = np.unique([x for x in alltags if x.endswith('PrintSpace')])
|
tag_endings = ['}PrintSpace','}printspace']
|
||||||
|
|
||||||
co_use_case = []
|
if tag.endswith(tag_endings[0]) or tag.endswith(tag_endings[1]):
|
||||||
|
for nn in root1.iter(tag):
|
||||||
for tag in region_tags:
|
c_t_in = []
|
||||||
if config_params['use_case']=='textline':
|
sumi = 0
|
||||||
tag_endings = ['}TextLine','}textline']
|
for vv in nn.iter():
|
||||||
elif config_params['use_case']=='word':
|
# check the format of coords
|
||||||
tag_endings = ['}Word','}word']
|
if vv.tag == link + 'Coords':
|
||||||
elif config_params['use_case']=='glyph':
|
coords = bool(vv.attrib)
|
||||||
tag_endings = ['}Glyph','}glyph']
|
if coords:
|
||||||
elif config_params['use_case']=='printspace':
|
p_h = vv.attrib['points'].split(' ')
|
||||||
tag_endings = ['}PrintSpace','}printspace']
|
c_t_in.append(
|
||||||
|
np.array([[int(x.split(',')[0]), int(x.split(',')[1])] for x in p_h]))
|
||||||
if tag.endswith(tag_endings[0]) or tag.endswith(tag_endings[1]):
|
|
||||||
for nn in root1.iter(tag):
|
|
||||||
c_t_in = []
|
|
||||||
sumi = 0
|
|
||||||
for vv in nn.iter():
|
|
||||||
# check the format of coords
|
|
||||||
if vv.tag == link + 'Coords':
|
|
||||||
coords = bool(vv.attrib)
|
|
||||||
if coords:
|
|
||||||
p_h = vv.attrib['points'].split(' ')
|
|
||||||
c_t_in.append(
|
|
||||||
np.array([[int(x.split(',')[0]), int(x.split(',')[1])] for x in p_h]))
|
|
||||||
break
|
|
||||||
else:
|
|
||||||
pass
|
|
||||||
|
|
||||||
if vv.tag == link + 'Point':
|
|
||||||
c_t_in.append([int(float(vv.attrib['x'])), int(float(vv.attrib['y']))])
|
|
||||||
sumi += 1
|
|
||||||
elif vv.tag != link + 'Point' and sumi >= 1:
|
|
||||||
break
|
break
|
||||||
co_use_case.append(np.array(c_t_in))
|
else:
|
||||||
|
pass
|
||||||
|
|
||||||
|
if vv.tag == link + 'Point':
|
||||||
|
c_t_in.append([int(float(vv.attrib['x'])), int(float(vv.attrib['y']))])
|
||||||
|
sumi += 1
|
||||||
|
elif vv.tag != link + 'Point' and sumi >= 1:
|
||||||
|
break
|
||||||
|
co_use_case.append(np.array(c_t_in))
|
||||||
|
|
||||||
|
|
||||||
if "artificial_class_label" in keys:
|
if "artificial_class_label" in keys:
|
||||||
|
|
|
||||||
Loading…
Add table
Add a link
Reference in a new issue