def refine_text(org, corners_text, max_line_gap, min_word_length):
def refine(bin):
head = 0
rear = 0
gap = 0
get_word = False
for i in range(bin.shape[1]):
# find head
if not get_word and np.sum(bin[:, i]) != 0:
head = i
rear = i
get_word = True
continue
if get_word and np.sum(bin[:, i]) != 0:
rear = i
continue
if get_word and np.sum(bin[:, i]) == 0:
gap += 1
if gap > max_line_gap:
if (rear - head) > min_word_length:
corners_text_refine.append(
(head + col_min, row_min, rear + col_min, row_max))
gap = 0
get_word = False
if get_word and (rear - head) > min_word_length:
corners_text_refine.append(
(head + col_min, row_min, rear + col_min, row_max))
corners_text_refine = []
pad = 1
for corner in corners_text:
(col_min, row_min, col_max, row_max) = corner
col_min = max(col_min - pad, 0)
col_max = min(col_max + pad, org.shape[1])
row_min = max(row_min - pad, 0)
row_max = min(row_max + pad, org.shape[0])
if row_max <= row_min or col_max <= col_min:
continue
clip = org[row_min:row_max, col_min:col_max]
clip_bin = pre.binarization(clip)
refine(clip_bin)
return corners_text_refine
def compo_detection(input_img_path,
output_root,
uied_params,
resize_by_height=600,
classifier=None,
show=False,
wai_key=10):
start = time.clock()
name = input_img_path.split('/')[-1][:-4]
ip_root = file.build_directory(pjoin(output_root, "ip"))
# *** Step 1 *** pre-processing: read img -> get binary map
org, grey = pre.read_img(input_img_path, resize_by_height)
binary = pre.binarization(org,
grad_min=int(uied_params['min-grad']),
show=show,
wait_key=wai_key)
# *** Step 2 *** element detection
det.rm_line(binary, show=show, wait_key=wai_key)
# det.rm_line_v_h(binary, show=show)
uicompos = det.component_detection(binary,
min_obj_area=int(
uied_params['min-ele-area']))
# draw.draw_bounding_box(org, uicompos, show=show, name='components', wait_key=wai_key)
# *** Step 3 *** results refinement
uicompos = det.merge_intersected_corner(
uicompos,
org,
is_merge_contained_ele=uied_params['merge-contained-ele'],
max_gap=(0, 0),
max_ele_height=25)
Compo.compos_update(uicompos, org.shape)
Compo.compos_containment(uicompos)
# draw.draw_bounding_box(org, uicompos, show=show, name='merged', wait_key=wai_key)
# *** Step 4 ** nesting inspection: treat the big compos as block and check if they have nesting element
uicompos += nesting_inspection(org,
grey,
uicompos,
ffl_block=uied_params['ffl-block'])
uicompos = det.compo_filter(uicompos,
min_area=int(uied_params['min-ele-area']))
Compo.compos_update(uicompos, org.shape)
draw.draw_bounding_box(org,
uicompos,
show=show,
name='merged compo',
write_path=pjoin(ip_root, 'result.jpg'),
wait_key=wai_key)
# *** Step 6 *** element classification: all category classification
if classifier is not None:
classifier['Elements'].predict(seg.clipping(org, uicompos), uicompos)
draw.draw_bounding_box_class(org,
uicompos,
show=show,
name='cls',
write_path=pjoin(ip_root, 'result.jpg'))
draw.draw_bounding_box_class(org,
uicompos,
write_path=pjoin(output_root,
'result.jpg'))
seg.dissemble_clip_img_fill(pjoin(output_root, 'new'), org, uicompos)
if classifier is None:
Compo.compos_update(uicompos, org.shape)
file.save_corners_json(pjoin(ip_root, name + '.json'), uicompos)
file.save_corners_json(pjoin(output_root, 'compo.json'), uicompos)
else:
Compo.compos_update(uicompos, org.shape)
file.save_corners_json(pjoin(ip_root, "clf_" + name + '.json'),
uicompos)
file.save_corners_json(pjoin(output_root, 'clf_compo.json'), uicompos)
if show:
cv2.destroyAllWindows()
print("[Compo Detection Completed in %.3f s] %s" %
(time.clock() - start, input_img_path))
def compo_detection(input_img_path,
output_root,
uied_params,
resize_by_height=600,
batch=False,
classifier=None,
show=False):
start = time.time()
name = input_img_path.split('/')[-1][:-4]
ip_root = file.build_directory(pjoin(output_root, "ip"))
# *** Step 1 *** pre-processing: read img -> get binary map
org, grey = pre.read_img(input_img_path, resize_by_height)
binary = pre.binarization(org,
grad_min=int(uied_params['min-grad']),
show=show)
# *** Step 2 *** element detection
det.rm_line(binary, show=show)
# det.rm_line_v_h(binary, show=show)
uicompos = det.component_detection(binary,
min_obj_area=int(
uied_params['min-ele-area']))
file.save_corners_json(pjoin(ip_root, name + '_all.json'), uicompos)
draw.draw_bounding_box(org, uicompos, show=show, name='components')
# *** Step 3 *** results refinement
# uicompos = det.rm_top_or_bottom_corners(uicompos, org.shape)
# uicompos = det.merge_text(uicompos, org.shape)
uicompos = det.merge_intersected_corner(
uicompos,
org,
is_merge_contained_ele=uied_params['merge-contained-ele'],
max_gap=(0, 0),
max_ele_height=25)
Compo.compos_update(uicompos, org.shape)
Compo.compos_containment(uicompos)
draw.draw_bounding_box(org, uicompos, show=show, name='merged')
# *** Step 4 ** nesting inspection: treat the big compos as block and check if they have nesting element
uicompos += nesting_inspection(org,
grey,
uicompos,
ffl_block=uied_params['ffl-block'])
uicompos = det.compo_filter(uicompos,
min_area=int(uied_params['min-ele-area']))
Compo.compos_update(uicompos, org.shape)
draw.draw_bounding_box(org,
uicompos,
show=show,
name='nesting compo',
write_path=pjoin(ip_root, 'result.jpg'))
draw.draw_bounding_box(org,
uicompos,
write_path=pjoin(output_root, 'result.jpg'))
# *** Step 5 *** Image Inspection: recognize image -> remove noise in image
# -> binarize with larger threshold and reverse -> rectangular compo detection
# if classifier is not None:
# classifier['Image'].predict(seg.clipping(org, uicompos), uicompos)
# draw.draw_bounding_box_class(org, uicompos, show=show)
# uicompos = det.rm_noise_in_large_img(uicompos, org)
# draw.draw_bounding_box_class(org, uicompos, show=show)
# det.detect_compos_in_img(uicompos, binary, org)
# draw.draw_bounding_box(org, uicompos, show=show)
# if classifier is not None:
# classifier['Noise'].predict(seg.clipping(org, uicompos), uicompos)
# draw.draw_bounding_box_class(org, uicompos, show=show)
# uicompos = det.rm_noise_compos(uicompos)
# *** Step 6 *** element classification: all category classification
if classifier is not None:
classifier['Elements'].predict(seg.clipping(org, uicompos), uicompos)
draw.draw_bounding_box_class(org,
uicompos,
show=show,
name='cls',
write_path=pjoin(ip_root, 'result.jpg'))
draw.draw_bounding_box_class(org,
uicompos,
write_path=pjoin(output_root,
'result.jpg'))
Compo.compos_update(uicompos, org.shape)
file.save_corners_json(pjoin(ip_root, name + '.json'), uicompos)
file.save_corners_json(pjoin(output_root, 'compo.json'), uicompos)
# seg.dissemble_clip_img_fill(pjoin(output_root, 'clips'), org, uicompos)
if not batch:
print("[Compo Detection Completed in %.3f s] %s" %
(time.time() - start, input_img_path))
if show:
cv2.destroyAllWindows()
def compo_detection(input_img_path,
output_root,
uied_params,
resize_by_height=800,
classifier=None,
show=False,
wai_key=0):
start = time.clock()
name = input_img_path.split(
'/')[-1][:-4] if '/' in input_img_path else input_img_path.split(
'\\')[-1][:-4]
ip_root = file.build_directory(pjoin(output_root, "ip"))
# *** Step 1 *** pre-processing: read img -> get binary map
org, grey = pre.read_img(input_img_path, resize_by_height)
binary = pre.binarization(org, grad_min=int(uied_params['min-grad']))
# *** Step 2 *** element detection
det.rm_line(binary, show=show, wait_key=wai_key)
uicompos = det.component_detection(binary,
min_obj_area=int(
uied_params['min-ele-area']))
# *** Step 3 *** results refinement
uicompos = det.compo_filter(uicompos,
min_area=int(uied_params['min-ele-area']),
img_shape=binary.shape)
uicompos = det.merge_intersected_compos(uicompos)
det.compo_block_recognition(binary, uicompos)
if uied_params['merge-contained-ele']:
uicompos = det.rm_contained_compos_not_in_block(uicompos)
Compo.compos_update(uicompos, org.shape)
Compo.compos_containment(uicompos)
# *** Step 4 ** nesting inspection: check if big compos have nesting element
uicompos += nesting_inspection(org,
grey,
uicompos,
ffl_block=uied_params['ffl-block'])
Compo.compos_update(uicompos, org.shape)
draw.draw_bounding_box(org,
uicompos,
show=show,
name='merged compo',
write_path=pjoin(ip_root, name + '.jpg'),
wait_key=wai_key)
# *** Step 5 *** image inspection: recognize image -> remove noise in image -> binarize with larger threshold and reverse -> rectangular compo detection
# if classifier is not None:
# classifier['Image'].predict(seg.clipping(org, uicompos), uicompos)
# draw.draw_bounding_box_class(org, uicompos, show=show)
# uicompos = det.rm_noise_in_large_img(uicompos, org)
# draw.draw_bounding_box_class(org, uicompos, show=show)
# det.detect_compos_in_img(uicompos, binary_org, org)
# draw.draw_bounding_box(org, uicompos, show=show)
# if classifier is not None:
# classifier['Noise'].predict(seg.clipping(org, uicompos), uicompos)
# draw.draw_bounding_box_class(org, uicompos, show=show)
# uicompos = det.rm_noise_compos(uicompos)
# *** Step 6 *** element classification: all category classification
# if classifier is not None:
# classifier['Elements'].predict([compo.compo_clipping(org) for compo in uicompos], uicompos)
# draw.draw_bounding_box_class(org, uicompos, show=show, name='cls', write_path=pjoin(ip_root, 'result.jpg'))
# draw.draw_bounding_box_class(org, uicompos, write_path=pjoin(output_root, 'result.jpg'))
# *** Step 7 *** save detection result
Compo.compos_update(uicompos, org.shape)
file.save_corners_json(pjoin(ip_root, name + '.json'), uicompos)
print(
"[Compo Detection Completed in %.3f s] Input: %s Output: %s" %
(time.clock() - start, input_img_path, pjoin(ip_root, name + '.json')))