MarkusDressel commited on
Commit
d3e509e
1 Parent(s): 55128e0

Upload cord.py

Browse files
Files changed (1) hide show
  1. cord.py +13 -30
cord.py CHANGED
@@ -5,8 +5,7 @@ import os
5
 
6
  import datasets
7
  import gdown
8
- from PIL import Image
9
- import numpy as np
10
  logger = datasets.logging.get_logger(__name__)
11
 
12
 
@@ -24,21 +23,6 @@ https://github.com/clovaai/cord
24
  """
25
  _URL = "https://drive.google.com/uc?id=1MqhTbcj-AHXOqYoeoh12aRUwIprzTJYI"
26
 
27
- def load_image(image_path):
28
- image = Image.open(image_path)
29
- image = image.convert("RGB")
30
- w, h = image.size
31
- return image, (w, h)
32
-
33
-
34
- def normalize_bbox(bbox, size):
35
- return [
36
- int(1000 * bbox[0] / size[0]),
37
- int(1000 * bbox[1] / size[1]),
38
- int(1000 * bbox[2] / size[0]),
39
- int(1000 * bbox[3] / size[1]),
40
- ]
41
-
42
 
43
  def gdrive_downloader(url, path):
44
  gdown.download(url, path, quiet=False)
@@ -71,6 +55,7 @@ class Cord(datasets.GeneratorBasedBuilder):
71
  "id": datasets.Value("string"),
72
  "tokens": datasets.Sequence(datasets.Value("string")),
73
  "bboxes": datasets.Sequence(datasets.Sequence(datasets.Value("int64"))),
 
74
  "ner_tags": datasets.Sequence(
75
  datasets.features.ClassLabel(
76
  names=['menu.cnt',
@@ -163,8 +148,16 @@ class Cord(datasets.GeneratorBasedBuilder):
163
  image_path = image_path.replace(
164
  filepath, filepaths[other_dir_idx])
165
 
166
- image, size = load_image(image_path)
167
- image = np.array(image)
 
 
 
 
 
 
 
 
168
 
169
  for item in data["valid_line"]:
170
  for word in item['words']:
@@ -172,30 +165,20 @@ class Cord(datasets.GeneratorBasedBuilder):
172
  txt = word['text']
173
 
174
  # get bounding box
175
- # important: each bounding box should be in (upper left, lower right) format
176
- # it took me some time to understand the upper left is (x1, y3)
177
- # and the lower right is (x3, y1)
178
  x1 = word['quad']['x1']
179
  y1 = word['quad']['y1']
180
  x3 = word['quad']['x3']
181
  y3 = word['quad']['y3']
182
 
183
  box = [x1, y1, x3, y3]
184
- box = normalize_bbox(box, size)
185
 
186
  # ADDED
187
  # skip empty word
188
  if len(txt) < 1:
189
  continue
190
- if min(box) < 0 or max(box) > 1000: # another bug in which a box had -4
191
- continue
192
- # another bug in which a box difference was -12
193
- if ((box[3] - box[1]) < 0) or ((box[2] - box[0]) < 0):
194
- continue
195
- # ADDED
196
 
197
  tokens.append(txt)
198
  bboxes.append(box)
199
  ner_tags.append(item['category'])
200
 
201
- yield guid, {"id": str(guid), "tokens": tokens, "bboxes": bboxes, "ner_tags": ner_tags, "image_path": image_path}
 
5
 
6
  import datasets
7
  import gdown
8
+
 
9
  logger = datasets.logging.get_logger(__name__)
10
 
11
 
 
23
  """
24
  _URL = "https://drive.google.com/uc?id=1MqhTbcj-AHXOqYoeoh12aRUwIprzTJYI"
25
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
26
 
27
  def gdrive_downloader(url, path):
28
  gdown.download(url, path, quiet=False)
 
55
  "id": datasets.Value("string"),
56
  "tokens": datasets.Sequence(datasets.Value("string")),
57
  "bboxes": datasets.Sequence(datasets.Sequence(datasets.Value("int64"))),
58
+ "roi": datasets.Sequence(datasets.Sequence(datasets.Value("int64"))),
59
  "ner_tags": datasets.Sequence(
60
  datasets.features.ClassLabel(
61
  names=['menu.cnt',
 
148
  image_path = image_path.replace(
149
  filepath, filepaths[other_dir_idx])
150
 
151
+ roi = data["roi"]
152
+ if roi:
153
+ top_left = [roi["x1"], roi["y1"]]
154
+ bottom_right = [roi["x3"], roi["y3"]]
155
+ bottom_left = [roi["x4"], roi["y4"]]
156
+ top_right = [roi["x2"], roi["y2"]]
157
+ roi = [top_left, top_right, bottom_right, bottom_left]
158
+ else:
159
+ roi = []
160
+
161
 
162
  for item in data["valid_line"]:
163
  for word in item['words']:
 
165
  txt = word['text']
166
 
167
  # get bounding box
 
 
 
168
  x1 = word['quad']['x1']
169
  y1 = word['quad']['y1']
170
  x3 = word['quad']['x3']
171
  y3 = word['quad']['y3']
172
 
173
  box = [x1, y1, x3, y3]
 
174
 
175
  # ADDED
176
  # skip empty word
177
  if len(txt) < 1:
178
  continue
 
 
 
 
 
 
179
 
180
  tokens.append(txt)
181
  bboxes.append(box)
182
  ner_tags.append(item['category'])
183
 
184
+ yield guid, {"id": str(guid), "tokens": tokens, "bboxes": bboxes, "ner_tags": ner_tags, "image_path": image_path, "roi":roi}