add testing for merge

merge.py

@@ -100,7 +100,7 @@ def rm_compos_in_text(compos):
 
 
 def incorporate(img_path, compo_path, text_path, output_root, params,
-                resize_by_height=None, show=False):
+                resize_by_height=None, show=False, wait_key=0):
     org = cv2.imread(img_path)
 
     compos = []
@@ -118,29 +118,30 @@ def incorporate(img_path, compo_path, text_path, output_root, params,
         texts.append(element)
 
     org_resize = resize_img_by_height(org, resize_by_height)
-    draw_bounding_box_class(org_resize, compos, show=show, name='ip')
-    draw_bounding_box(org_resize, texts, show=show, name='ocr')
+    draw_bounding_box_class(org_resize, compos, show=show, name='ip', wait_key=wait_key)
+    draw_bounding_box(org_resize, texts, show=show, name='ocr', wait_key=wait_key)
 
     compos_merged = reclassify_text_by_ocr(org_resize, compos, texts)
     # compos_merged = merge_redundant_corner(org_resize, compos_merged)
-    draw_bounding_box_class(org_resize, compos_merged, name='text', show=show)
+    # draw_bounding_box_class(org_resize, compos_merged, name='text', show=show, wait_key=wait_key)
 
     # merge words as line
     compos_merged = merge_intersected_compos(org_resize, compos_merged, max_gap=(params['max-word-inline-gap'], 0), merge_class='Text')
-    draw_bounding_box_class(org_resize, compos_merged, name='merged line', show=show)
+    draw_bounding_box_class(org_resize, compos_merged, name='merged line', show=show, wait_key=wait_key)
     # merge lines as paragraph
     compos_merged = merge_intersected_compos(org_resize, compos_merged, max_gap=(0, params['max-line-gap']), merge_class='Text')
-    draw_bounding_box_class(org_resize, compos_merged, name='merged paragraph', show=show)
+    # draw_bounding_box_class(org_resize, compos_merged, name='merged paragraph', show=show)
     # clean compos intersected with paragraphs
     compos_merged = rm_compos_in_text(compos_merged)
-    board = draw_bounding_box_class(org_resize, compos_merged, name='merged paragraph', is_return=True, show=show)
+    board = draw_bounding_box_class(org_resize, compos_merged, name='merged paragraph', is_return=True, show=show, wait_key=wait_key)
 
-    draw_bounding_box_non_text(org_resize, compos_merged, org_shape=org.shape, show=show)
+    # draw_bounding_box_non_text(org_resize, compos_merged, org_shape=org.shape, show=show)
     compos_json = save_corners_json(output_root, background, compos_merged, org_resize.shape)
     dissemble_clip_img_fill(pjoin(output_root, 'clips'), org_resize, compos_json)
     cv2.imwrite(pjoin(output_root, 'result.jpg'), board)
 
     print('Merge Complete and Save to', pjoin(output_root, 'result.jpg'))
     print(time.ctime(), '\n')
-    if show:
-        cv2.destroyAllWindows()
+    # if show:
+    #     cv2.destroyAllWindows()
+

run_single.py

@@ -31,19 +31,19 @@ def resize_height_by_longest_edge(img_path, resize_length=800):
         mobile: {'min-grad':4, 'ffl-block':5, 'min-ele-area':50, 'max-word-inline-gap':6, 'max-line-gap':1}
         web   : {'min-grad':3, 'ffl-block':5, 'min-ele-area':25, 'max-word-inline-gap':4, 'max-line-gap':4}
     '''
-    key_params = {'min-grad':10, 'ffl-block':5, 'min-ele-area':50, 'merge-contained-ele':True,
+    key_params = {'min-grad':4, 'ffl-block':5, 'min-ele-area':50, 'merge-contained-ele':True,
                   'max-word-inline-gap':4, 'max-line-gap':4}
 
     # set input image path
-    input_path_img = 'data/input/30800.jpg'
+    input_path_img = 'data/input/472.jpg'
     output_root = 'data/output'
 
     resized_height = resize_height_by_longest_edge(input_path_img)
 
-    is_ip = True
+    is_ip = False
     is_clf = False
     is_ocr = False
-    is_merge = False
+    is_merge = True
 
     if is_ocr:
         import detect_text_east.ocr_east as ocr

run_testing(Used for Adjusting).py

@@ -12,6 +12,10 @@ def resize_height_by_longest_edge(img_path, resize_length=800):
         return int(resize_length * (height / width))
 
 
+def nothing(x):
+    pass
+
+
 if __name__ == '__main__':
 
     '''
@@ -35,16 +39,12 @@ def resize_height_by_longest_edge(img_path, resize_length=800):
                   'max-word-inline-gap':4, 'max-line-gap':4}
 
     # set input image path
-    input_path_img = 'data/input/30800.jpg'
+    input_path_img = 'data/input/472.jpg'
     output_root = 'data/output'
 
     resized_height = resize_height_by_longest_edge(input_path_img)
-
-    is_ip = True
     is_clf = False
     is_ocr = False
-    is_merge = False
-
     if is_ocr:
         import detect_text_east.ocr_east as ocr
         import detect_text_east.lib_east.eval as eval
@@ -53,38 +53,44 @@ def resize_height_by_longest_edge(img_path, resize_length=800):
         ocr.east(input_path_img, output_root, models, key_params['max-word-inline-gap'],
                  resize_by_height=resized_height, show=False)
 
+    '''
+    ******** Testing with adjustable parameters ********
+    '''
+    testing_ip = False
+    testing_merge = True
 
-    def nothing(x):
-        pass
     cv2.namedWindow('parameters')
-    cv2.createTrackbar('min-grad', 'parameters', 4, 20, nothing)
-    cv2.createTrackbar('min-ele-area', 'parameters', 20, 200, nothing)
-
-    while(1):
-        key_params['min-grad'] = cv2.getTrackbarPos('min-grad', 'parameters')
-        key_params['min-ele-area'] = cv2.getTrackbarPos('min-ele-area', 'parameters')
-
-        if is_ip:
+    if testing_ip:
+        cv2.createTrackbar('min-grad', 'parameters', 4, 20, nothing)
+        cv2.createTrackbar('min-ele-area', 'parameters', 20, 200, nothing)
+        while(1):
+            key_params['min-grad'] = cv2.getTrackbarPos('min-grad', 'parameters')
+            key_params['min-ele-area'] = cv2.getTrackbarPos('min-ele-area', 'parameters')
             import detect_compo.ip_region_proposal as ip
             os.makedirs(pjoin(output_root, 'ip'), exist_ok=True)
             # switch of the classification func
             classifier = None
             if is_clf:
                 classifier = {}
                 from cnn.CNN import CNN
-
                 # classifier['Image'] = CNN('Image')
                 classifier['Elements'] = CNN('Elements')
                 # classifier['Noise'] = CNN('Noise')
             ip.compo_detection(input_path_img, output_root, key_params,
                                classifier=classifier, resize_by_height=resized_height, show=True, wai_key=None)
-
-        if is_merge:
+            cv2.waitKey(10)
+
+    if testing_merge:
+        cv2.createTrackbar('max-word-inline-gap', 'parameters', 4, 20, nothing)
+        cv2.createTrackbar('max-line-gap', 'parameters', 20, 200, nothing)
+        while(1):
+            key_params['max-word-inline-gap'] = cv2.getTrackbarPos('max-word-inline-gap', 'parameters')
+            key_params['max-line-gap'] = cv2.getTrackbarPos('max-line-gap', 'parameters')
             import merge
             name = input_path_img.split('/')[-1][:-4]
             compo_path = pjoin(output_root, 'ip', str(name) + '.json')
             ocr_path = pjoin(output_root, 'ocr', str(name) + '.json')
             merge.incorporate(input_path_img, compo_path, ocr_path, output_root, params=key_params,
-                              resize_by_height=resized_height, show=True)
+                              resize_by_height=resized_height, show=True, wait_key=None)
+            cv2.waitKey(10)
 
-        cv2.waitKey(10)

utils/util_merge.py

@@ -14,7 +14,8 @@
 C = Config()
 
 
-def draw_bounding_box_class(org, compos, color_map=C.COLOR, line=2, show=False, is_return=False, name='img'):
+def draw_bounding_box_class(org, compos, color_map=C.COLOR, line=2,
+                            show=False, is_return=False, name='img', wait_key=0):
     if not show and not is_return: return
     board = org.copy()
 
@@ -29,18 +30,20 @@ def draw_bounding_box_class(org, compos, color_map=C.COLOR, line=2, show=False,
         #                     cv2.FONT_HERSHEY_SIMPLEX, 0.5, class_colors[compo_class[i]], 2)
     if show:
         cv2.imshow(name, board)
-        cv2.waitKey(0)
+        if wait_key is not None:
+            cv2.waitKey(wait_key)
     return board
 
 
-def draw_bounding_box(org, compos,  color=(0, 255, 0), line=2, show=False, name='board'):
+def draw_bounding_box(org, compos,  color=(0, 255, 0), line=2, show=False, name='board', wait_key=0):
     board = org.copy()
     for compo in compos:
         corner = compo.put_bbox()
         board = cv2.rectangle(board, (corner[0], corner[1]), (corner[2], corner[3]), color, line)
     if show:
         cv2.imshow(name, board)
-        cv2.waitKey(0)
+        if wait_key is not None:
+            cv2.waitKey(wait_key)
     return board