-
Notifications
You must be signed in to change notification settings - Fork 15
/
merge.py
245 lines (197 loc) · 8.03 KB
/
merge.py
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
43
44
45
46
47
48
49
50
51
52
53
54
55
56
57
58
59
60
61
62
63
64
65
66
67
68
69
70
71
72
73
74
75
76
77
78
79
80
81
82
83
84
85
86
87
88
89
90
91
92
93
94
95
96
97
98
99
100
101
102
103
104
105
106
107
108
109
110
111
112
113
114
115
116
117
118
119
120
121
122
123
124
125
126
127
128
129
130
131
132
133
134
135
136
137
138
139
140
141
142
143
144
145
146
147
148
149
150
151
152
153
154
155
156
157
158
159
160
161
162
163
164
165
166
167
168
169
170
171
172
173
174
175
176
177
178
179
180
181
182
183
184
185
186
187
188
189
190
191
192
193
194
195
196
197
198
199
200
201
202
203
204
205
206
207
208
209
210
211
212
213
214
215
216
217
218
219
220
221
222
223
224
225
226
227
228
229
230
231
232
233
234
235
236
237
238
239
240
241
242
243
244
245
import os
import glob
import pickle
import argparse
from xml.dom import minidom
from xml.etree import ElementTree as ET
from termcolor import cprint
import re
import cv2
import numpy as np
import libs.merge_utility as ut
import libs.GTElement as gc
def col_merging(_table, ocr_data):
for word in ocr_data:
txt = re.sub('[^a-zA-Z0-9]', '', str(word[1]))
if len(txt) == 0:
continue
for col in _table.gtCols:
if (word[2] < col.x1) and (word[4] > col.x2):
merging_col = gc.Row(word[2], word[3], word[4])
_table.gtSpans.append(merging_col)
_table.evaluateCells()
def row_merging(_table, ocr_data):
for word in ocr_data:
txt = re.sub('[^a-zA-Z0-9]', '', str(word[1]))
if len(txt) == 0:
continue
for row in _table.gtRows:
if (word[3] < row.y1) and (word[5] > row.y2):
merging_row = gc.Column(word[2], word[3], word[5])
_table.gtSpans.append(merging_row)
_table.evaluateCells()
def remove_redundant_seperators(_table, ocr_data, check_threshold=0.4):
_table.populate_ocr(ocr_data)
remove_rows = [True for i in range(len(_table.gtRows))]
remove_cols = [True for i in range(len(_table.gtCols))]
row_threshold_count = [0 for i in range(len(_table.gtRows))]
col_threshold_count = [0 for i in range(len(_table.gtCols))]
for i in range(1, len(_table.gtCells)):
for j in range(len(_table.gtCells[i])):
if (
_table.gtCells[i][j].dontCare == False
and len(_table.gtCells[i][j].words) > 0
):
if len(_table.gtCells[0]) < 4:
current_thresh = 1.0
else:
row_threshold_count[i - 1] += 1
current_thresh = row_threshold_count[i -
1] / len(_table.gtCells[i])
if current_thresh >= check_threshold:
remove_rows[i - 1] = False
break
for j in range(1, len(_table.gtCells[0])):
for i in range(len(_table.gtCells)):
if (
_table.gtCells[i][j].dontCare == False
and len(_table.gtCells[i][j].words) > 0
):
if len(_table.gtCells) < 5:
current_thresh = 1.0
else:
col_threshold_count[j - 1] += 1
current_thresh = col_threshold_count[j -
1] / len(_table.gtCells)
if current_thresh >= check_threshold:
remove_cols[j - 1] = False
break
removed_rows = [
_table.gtRows[i].y1 for i in range(len(_table.gtRows)) if remove_rows[i]
]
removed_cols = [
_table.gtCols[i].x1 for i in range(len(_table.gtCols)) if remove_cols[i]
]
_table.gtRows = [
_table.gtRows[i] for i in range(len(_table.gtRows)) if not remove_rows[i]
]
_table.gtCols = [
_table.gtCols[i] for i in range(len(_table.gtCols)) if not remove_cols[i]
]
_table.evaluateCells()
return removed_rows, removed_cols
def get_table(_list_rows, _list_col, h, w):
_list_row_obj = []
_list_col_obj = []
_table = gc.Table(0, 0, w, h)
for row in _list_rows:
_table.gtRows.append(gc.Row(0, row, w))
for col in _list_col:
_table.gtCols.append(gc.Column(col, 0, h))
_table.evaluateCells()
return _table
def get_grid_structure(xml_filename):
tree = ET.parse(xml_filename)
root = tree.getroot()
tables = root.findall(".//Table")
assert len(tables) == 1
w = int(tables[0].attrib["x1"])
h = int(tables[0].attrib["y1"])
columns = list(map(lambda x: int(x.attrib["x0"]), tables[0].findall(".//Column")))
rows = list(map(lambda x: int(x.attrib["y0"]), tables[0].findall(".//Row")))
return rows, columns, h, w
def execute_pipeline(xml_filename, ocr_data, org_image):
_list_rows, _list_col, h, w = get_grid_structure(xml_filename)
if org_image is not None:
org_image[_list_rows, :] = (0, 255, 0)
org_image[:, _list_col] = (0, 255, 0)
_table = get_table(_list_rows, _list_col, h, w)
col_merging(_table, ocr_data)
row_merging(_table, ocr_data)
removed_rows, removed_cols = remove_redundant_seperators(
_table, ocr_data, 0.2)
removed_rows_iter_2, removed_cols_iter_2 = remove_redundant_seperators(
_table, ocr_data, 0.4)
removed_rows.extend(removed_rows_iter_2)
removed_cols.extend(removed_cols_iter_2)
if org_image is not None:
org_image[removed_rows, :] = (0, 0, 255)
org_image[:, removed_cols] = (0, 0, 255)
_table.merge_header_v2(ocr_data)
return _table
def visualize_merging(_table, ocr_data, img, image_name):
for item in _table.gtSpans:
cv2.line(img, (item.x1, item.y1),
(item.x2 - 1, item.y2 - 1), (0, 0, 255), 2)
for item in ocr_data:
cv2.rectangle(img, (item[2], item[3]),
(item[4], item[5]), (245, 66, 176), 1)
cv2.imwrite(image_name, img)
def data_pipeline(xml_in_path, out_path, images_path, ocr_path):
filenames = map(os.path.basename, glob.glob(os.path.join(xml_in_path, "*.xml")))
for item in filenames:
try:
filename = item.rsplit(".", 1)[0]
xml_filename = os.path.join(xml_in_path, filename + ".xml")
if images_path:
org_img = cv2.imread(os.path.join(images_path, filename + ".png"))
else:
org_img = None
with (open(os.path.join(ocr_path, filename + ".pkl"), "rb")) as f:
ocr_data = pickle.load(f)
ocr_data = ut.clean_ocr_data(ocr_data)
ocr_data = ut.make_sentences(ocr_data)
final_table = execute_pipeline(xml_filename, ocr_data, org_img)
out_root = ET.Element("GroundTruth")
out_root.attrib["InputFile"] = filename + ".png"
out_tables = ET.SubElement(out_root, "Tables")
table_xml = final_table.get_xml_object()
out_tables.append(table_xml)
out_data = minidom.parseString(
ET.tostring(out_root)).toprettyxml(indent=" ")
with open(os.path.join(out_path, "xmls", filename + ".xml"), "w") as _file:
_file.write("\n".join(out_data.split("\n")))
if org_img is not None:
visualize_merging(
final_table,
ocr_data,
org_img,
os.path.join(out_path, "visualization", filename + ".png"),
)
cprint("Processed: ", "green", attrs=["bold"], end="")
print(filename)
except Exception as e:
cprint("Error: ", "red", attrs=["bold"], end="")
print(filename)
print(e)
if __name__ == "__main__":
parser = argparse.ArgumentParser()
parser.add_argument(
"-i",
"--input_xml_dir",
help="Path to folder containing XML files predicted by infer.py",
required=True,
)
parser.add_argument(
"-o",
"--output_dir",
help="Path to folder for writing output XML files and visualization (optional) of merge heuristics.",
required=True,
)
parser.add_argument(
"-ocr",
"--ocr_dir",
help="Path to folder containing table-level OCR files generated by prepare_data.py",
required=True,
)
parser.add_argument(
"-img",
"--images_dir",
help="Path to table-level images generated by prepare_data.py (Optional. If not provided merge visualization will not be written).",
)
args = parser.parse_args()
os.makedirs(args.output_dir, exist_ok=True)
os.makedirs(os.path.join(args.output_dir, "xmls"), exist_ok=True)
if args.images_dir:
os.makedirs(os.path.join(args.output_dir, "visualization"), exist_ok=True)
data_pipeline(args.input_xml_dir, args.output_dir, args.images_dir, args.ocr_dir)