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,
|
||||
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(
|
||||
image,
|
||||
dir_in,
|
||||
|
|
@ -70,7 +64,6 @@ def linegt_cli(
|
|||
pref_of_dataset,
|
||||
do_not_mask_with_textline_contour,
|
||||
exclude_vertical_lines,
|
||||
page_alto,
|
||||
):
|
||||
assert bool(dir_in) ^ bool(image), "Set --dir-in or --image-filename, not both"
|
||||
if dir_in:
|
||||
|
|
@ -86,84 +79,6 @@ def linegt_cli(
|
|||
dir_xml = os.path.join(dir_xmls, file_name + '.xml')
|
||||
img = cv2.imread(dir_img)
|
||||
|
||||
if page_alto:
|
||||
h, w = img.shape[:2]
|
||||
|
||||
tree = ET.parse(dir_xml)
|
||||
root = tree.getroot()
|
||||
|
||||
NS = {'alto': root.tag.split('}')[0].strip('{')}#{"alto": "http://www.loc.gov/standards/alto/ns-v4#"}
|
||||
|
||||
results = []
|
||||
|
||||
indexer_textlines = 0
|
||||
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)
|
||||
if polygon_el is None:
|
||||
continue
|
||||
|
||||
points = polygon_el.attrib["POINTS"].split()
|
||||
coords = [
|
||||
(int(points[i]), int(points[i + 1]))
|
||||
for i in range(0, len(points), 2)
|
||||
]
|
||||
|
||||
coords = np.array(coords, dtype=np.int32)
|
||||
x, y, w, h = cv2.boundingRect(coords)
|
||||
|
||||
|
||||
if exclude_vertical_lines and h > 1.4 * w:
|
||||
img_crop = None
|
||||
continue
|
||||
|
||||
img_poly_on_img = np.copy(img)
|
||||
|
||||
mask_poly = np.zeros(img.shape)
|
||||
mask_poly = cv2.fillPoly(mask_poly, pts=[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 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"))
|
||||
|
|
|
|||
|
|
@ -73,14 +73,8 @@ def main():
|
|||
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.",
|
||||
)
|
||||
@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:
|
||||
with open(config) as 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")
|
||||
config_params = None
|
||||
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()
|
||||
@click.option(
|
||||
|
|
|
|||
|
|
@ -686,7 +686,7 @@ def get_layout_contours_for_visualization(xml_file):
|
|||
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
|
||||
|
||||
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.
|
||||
"""
|
||||
|
|
@ -699,19 +699,6 @@ def get_images_of_ground_truth(gt_list, dir_in, output_dir, output_type, config_
|
|||
print(gt_list[index])
|
||||
|
||||
try:
|
||||
if page_alto:
|
||||
tree = ET.parse(dir_in+'/'+gt_list[index])
|
||||
root = tree.getroot()
|
||||
|
||||
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()]
|
||||
|
|
@ -797,24 +784,6 @@ def get_images_of_ground_truth(gt_list, dir_in, output_dir, output_type, config_
|
|||
|
||||
textline_rgb_color = (255, 0, 0)
|
||||
|
||||
if page_alto:
|
||||
co_use_case = []
|
||||
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)
|
||||
if polygon_el is None:
|
||||
continue
|
||||
|
||||
points = polygon_el.attrib["POINTS"].split()
|
||||
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':
|
||||
region_tags = np.unique([x for x in alltags if x.endswith('TextLine')])
|
||||
elif config_params['use_case']=='word':
|
||||
|
|
|
|||
Loading…
Add table
Add a link
Reference in a new issue