qurator-spk/eynollah
1
0
Fork 0
mirror of https://github.com/qurator-spk/eynollah.git synced 2026-03-02 05:11:57 +01:00
Code Issues Projects Releases Wiki Activity

musicregion is added to pagexml to label

Browse source
This commit is contained in:
vahidrezanezhad 2026-02-24 15:46:15 +01:00
parent 76ac4c5b71
commit d8667f46d7
2 changed files with 71 additions and 5 deletions
Download patch file Download diff file Expand all files Collapse all files

4
src/eynollah/training/generate_gt_for_training.py
View file

@ -483,9 +483,9 @@ def visualize_layout_segmentation(xml_file, dir_xml, dir_out, dir_imgs):
img_file_name_with_format = find_format_of_given_filename_in_dir(dir_imgs, f_name) img_file_name_with_format = find_format_of_given_filename_in_dir(dir_imgs, f_name)
img = cv2.imread(os.path.join(dir_imgs, img_file_name_with_format)) img = cv2.imread(os.path.join(dir_imgs, img_file_name_with_format))
co_text, co_graphic, co_sep, co_img, co_table, co_map, co_noise, y_len, x_len = get_layout_contours_for_visualization(xml_file) co_text, co_graphic, co_sep, co_img, co_table, co_map, co_music, co_noise, y_len, x_len = get_layout_contours_for_visualization(xml_file)
added_image = visualize_image_from_contours_layout(co_text['paragraph'], co_text['header']+co_text['heading'], co_text['drop-capital'], co_sep, co_img, co_text['marginalia'], co_table, co_map, img) added_image = visualize_image_from_contours_layout(co_text['paragraph'], co_text['header']+co_text['heading'], co_text['drop-capital'], co_sep, co_img, co_text['marginalia'], co_table, co_map, co_music, img)
cv2.imwrite(os.path.join(dir_out, f_name+'.png'), added_image) cv2.imwrite(os.path.join(dir_out, f_name+'.png'), added_image)
except: except:

72
src/eynollah/training/gt_gen_utils.py
View file

@ -15,7 +15,7 @@ with warnings.catch_warnings():
warnings.simplefilter("ignore") warnings.simplefilter("ignore")
def visualize_image_from_contours_layout(co_par, co_header, co_drop, co_sep, co_image, co_marginal, co_table, co_map, img): def visualize_image_from_contours_layout(co_par, co_header, co_drop, co_sep, co_image, co_marginal, co_table, co_map, co_music, img):
alpha = 0.5 alpha = 0.5
blank_image = np.ones( (img.shape[:]), dtype=np.uint8) * 255 blank_image = np.ones( (img.shape[:]), dtype=np.uint8) * 255
@ -29,6 +29,7 @@ def visualize_image_from_contours_layout(co_par, co_header, co_drop, co_sep, co_
col_marginal = (106, 90, 205) col_marginal = (106, 90, 205)
col_table = (0, 90, 205) col_table = (0, 90, 205)
col_map = (90, 90, 205) col_map = (90, 90, 205)
col_music = (90, 90, 0)
if len(co_image)>0: if len(co_image)>0:
cv2.drawContours(blank_image, co_image, -1, col_image, thickness=cv2.FILLED) # Fill the contour cv2.drawContours(blank_image, co_image, -1, col_image, thickness=cv2.FILLED) # Fill the contour
@ -57,6 +58,9 @@ def visualize_image_from_contours_layout(co_par, co_header, co_drop, co_sep, co_
if len(co_map)>0: if len(co_map)>0:
cv2.drawContours(blank_image, co_map, -1, col_map, thickness=cv2.FILLED) # Fill the contour cv2.drawContours(blank_image, co_map, -1, col_map, thickness=cv2.FILLED) # Fill the contour
if len(co_music)>0:
cv2.drawContours(blank_image, co_music, -1, col_music, thickness=cv2.FILLED) # Fill the contour
img_final =cv2.cvtColor(blank_image, cv2.COLOR_BGR2RGB) img_final =cv2.cvtColor(blank_image, cv2.COLOR_BGR2RGB)
added_image = cv2.addWeighted(img,alpha,img_final,1- alpha,0) added_image = cv2.addWeighted(img,alpha,img_final,1- alpha,0)
@ -387,6 +391,7 @@ def get_layout_contours_for_visualization(xml_file):
co_img=[] co_img=[]
co_table=[] co_table=[]
co_map=[] co_map=[]
co_music=[]
co_noise=[] co_noise=[]
types_text = [] types_text = []
@ -629,6 +634,31 @@ def get_layout_contours_for_visualization(xml_file):
break break
co_map.append(np.array(c_t_in)) co_map.append(np.array(c_t_in))
if tag.endswith('}MusicRegion') or tag.endswith('}musicregion'):
#print('sth')
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
#print(vv.tag,'in')
elif vv.tag!=link+'Point' and sumi>=1:
break
co_music.append(np.array(c_t_in))
if tag.endswith('}NoiseRegion') or tag.endswith('}noiseregion'): if tag.endswith('}NoiseRegion') or tag.endswith('}noiseregion'):
#print('sth') #print('sth')
@ -654,7 +684,7 @@ def get_layout_contours_for_visualization(xml_file):
elif vv.tag!=link+'Point' and sumi>=1: elif vv.tag!=link+'Point' and sumi>=1:
break break
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_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): def get_images_of_ground_truth(gt_list, dir_in, output_dir, output_type, config_file, config_params, printspace, dir_images, dir_out_images):
""" """
@ -873,7 +903,7 @@ def get_images_of_ground_truth(gt_list, dir_in, output_dir, output_type, config_
types_graphic_label = list(types_graphic_dict.values()) types_graphic_label = list(types_graphic_dict.values())
labels_rgb_color = [ (0,0,0), (255,0,0), (255,125,0), (255,0,125), (125,255,125), (125,125,0), (0,125,255), (0,125,0), (125,125,125), (255,0,255), (125,0,125), (0,255,0),(0,0,255), (0,255,255), (255,125,125), (0,125,125), (0,255,125), (255,125,255), (125,255,0), (125,255,255)] labels_rgb_color = [ (0,0,0), (255,0,0), (255,125,0), (255,0,125), (125,255,125), (125,125,0), (0,125,255), (0,125,0), (125,125,125), (255,0,255), (125,0,125), (0,255,0),(0,0,255), (0,255,255), (255,125,125), (0,125,125), (0,255,125), (255,125,255), (125,255,0), (125,255,255), (125,125,255)]
region_tags=np.unique([x for x in alltags if x.endswith('Region')]) region_tags=np.unique([x for x in alltags if x.endswith('Region')])
@ -885,6 +915,7 @@ def get_images_of_ground_truth(gt_list, dir_in, output_dir, output_type, config_
co_img=[] co_img=[]
co_table=[] co_table=[]
co_map=[] co_map=[]
co_music=[]
co_noise=[] co_noise=[]
for tag in region_tags: for tag in region_tags:
@ -1127,6 +1158,32 @@ def get_images_of_ground_truth(gt_list, dir_in, output_dir, output_type, config_
break break
co_map.append(np.array(c_t_in)) co_map.append(np.array(c_t_in))
if 'musicregion' in keys:
if tag.endswith('}MusicRegion') or tag.endswith('}musicregion'):
#print('sth')
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
#print(vv.tag,'in')
elif vv.tag!=link+'Point' and sumi>=1:
break
co_music.append(np.array(c_t_in))
if 'noiseregion' in keys: if 'noiseregion' in keys:
if tag.endswith('}NoiseRegion') or tag.endswith('}noiseregion'): if tag.endswith('}NoiseRegion') or tag.endswith('}noiseregion'):
#print('sth') #print('sth')
@ -1203,6 +1260,10 @@ def get_images_of_ground_truth(gt_list, dir_in, output_dir, output_type, config_
erosion_rate = 0#2 erosion_rate = 0#2
dilation_rate = 3#4 dilation_rate = 3#4
co_map, img_boundary = update_region_contours(co_map, img_boundary, erosion_rate, dilation_rate, y_len, x_len ) co_map, img_boundary = update_region_contours(co_map, img_boundary, erosion_rate, dilation_rate, y_len, x_len )
if "musicregion" in elements_with_artificial_class:
erosion_rate = 0#2
dilation_rate = 3#4
co_music, img_boundary = update_region_contours(co_music, img_boundary, erosion_rate, dilation_rate, y_len, x_len )
@ -1230,6 +1291,8 @@ def get_images_of_ground_truth(gt_list, dir_in, output_dir, output_type, config_
img_poly=cv2.fillPoly(img, pts =co_table, color=labels_rgb_color[ config_params['tableregion']]) img_poly=cv2.fillPoly(img, pts =co_table, color=labels_rgb_color[ config_params['tableregion']])
if 'mapregion' in keys: if 'mapregion' in keys:
img_poly=cv2.fillPoly(img, pts =co_map, color=labels_rgb_color[ config_params['mapregion']]) img_poly=cv2.fillPoly(img, pts =co_map, color=labels_rgb_color[ config_params['mapregion']])
if 'musicregion' in keys:
img_poly=cv2.fillPoly(img, pts =co_music, color=labels_rgb_color[ config_params['musicregion']])
if 'noiseregion' in keys: if 'noiseregion' in keys:
img_poly=cv2.fillPoly(img, pts =co_noise, color=labels_rgb_color[ config_params['noiseregion']]) img_poly=cv2.fillPoly(img, pts =co_noise, color=labels_rgb_color[ config_params['noiseregion']])
@ -1293,6 +1356,9 @@ def get_images_of_ground_truth(gt_list, dir_in, output_dir, output_type, config_
if 'mapregion' in keys: if 'mapregion' in keys:
color_label = config_params['mapregion'] color_label = config_params['mapregion']
img_poly=cv2.fillPoly(img, pts =co_map, color=(color_label,color_label,color_label)) img_poly=cv2.fillPoly(img, pts =co_map, color=(color_label,color_label,color_label))
if 'musicregion' in keys:
color_label = config_params['musicregion']
img_poly=cv2.fillPoly(img, pts =co_music, color=(color_label,color_label,color_label))
if 'noiseregion' in keys: if 'noiseregion' in keys:
color_label = config_params['noiseregion'] color_label = config_params['noiseregion']
img_poly=cv2.fillPoly(img, pts =co_noise, color=(color_label,color_label,color_label)) img_poly=cv2.fillPoly(img, pts =co_noise, color=(color_label,color_label,color_label))