nielsr HF staff commited on
Commit
f515662
1 Parent(s): 6ba49fe

First commit

Browse files
Files changed (1) hide show
  1. funsd.py +123 -0
funsd.py ADDED
@@ -0,0 +1,123 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+ # coding=utf-8
2
+ import json
3
+ import os
4
+
5
+ import datasets
6
+
7
+ from PIL import Image
8
+ import numpy as np
9
+
10
+ logger = datasets.logging.get_logger(__name__)
11
+
12
+
13
+ _CITATION = """\
14
+ @article{Jaume2019FUNSDAD,
15
+ title={FUNSD: A Dataset for Form Understanding in Noisy Scanned Documents},
16
+ author={Guillaume Jaume and H. K. Ekenel and J. Thiran},
17
+ journal={2019 International Conference on Document Analysis and Recognition Workshops (ICDARW)},
18
+ year={2019},
19
+ volume={2},
20
+ pages={1-6}
21
+ }
22
+ """
23
+ _DESCRIPTION = """\
24
+ https://guillaumejaume.github.io/FUNSD/
25
+ """
26
+
27
+ def load_image(image_path):
28
+ image = Image.open(image_path).convert("RGB")
29
+ w, h = image.size
30
+ return image, (w, h)
31
+
32
+ def normalize_bbox(bbox, size):
33
+ return [
34
+ int(1000 * bbox[0] / size[0]),
35
+ int(1000 * bbox[1] / size[1]),
36
+ int(1000 * bbox[2] / size[0]),
37
+ int(1000 * bbox[3] / size[1]),
38
+ ]
39
+
40
+ class FunsdConfig(datasets.BuilderConfig):
41
+ """BuilderConfig for FUNSD"""
42
+
43
+ def __init__(self, **kwargs):
44
+ """BuilderConfig for FUNSD.
45
+
46
+ Args:
47
+ **kwargs: keyword arguments forwarded to super.
48
+ """
49
+ super(FunsdConfig, self).__init__(**kwargs)
50
+
51
+ class Funsd(datasets.GeneratorBasedBuilder):
52
+ """FUNSD dataset."""
53
+
54
+ BUILDER_CONFIGS = [
55
+ FunsdConfig(name="funsd", version=datasets.Version("1.0.0"), description="FUNSD dataset"),
56
+ ]
57
+
58
+ def _info(self):
59
+ return datasets.DatasetInfo(
60
+ description=_DESCRIPTION,
61
+ features=datasets.Features(
62
+ {
63
+ "id": datasets.Value("string"),
64
+ "tokens": datasets.Sequence(datasets.Value("string")),
65
+ "bboxes": datasets.Sequence(datasets.Sequence(datasets.Value("int64"))),
66
+ "ner_tags": datasets.Sequence(
67
+ datasets.features.ClassLabel(
68
+ names=["O", "B-HEADER", "I-HEADER", "B-QUESTION", "I-QUESTION", "B-ANSWER", "I-ANSWER"]
69
+ )
70
+ ),
71
+ "image_path": datasets.Value("string"),
72
+ }
73
+ ),
74
+ supervised_keys=None,
75
+ homepage="https://guillaumejaume.github.io/FUNSD/",
76
+ citation=_CITATION,
77
+ )
78
+
79
+ def _split_generators(self, dl_manager):
80
+ """Returns SplitGenerators."""
81
+ downloaded_file = dl_manager.download_and_extract("https://guillaumejaume.github.io/FUNSD/dataset.zip")
82
+ return [
83
+ datasets.SplitGenerator(
84
+ name=datasets.Split.TRAIN, gen_kwargs={"filepath": f"{downloaded_file}/dataset/training_data/"}
85
+ ),
86
+ datasets.SplitGenerator(
87
+ name=datasets.Split.TEST, gen_kwargs={"filepath": f"{downloaded_file}/dataset/testing_data/"}
88
+ ),
89
+ ]
90
+
91
+ def _generate_examples(self, filepath):
92
+ logger.info("⏳ Generating examples from = %s", filepath)
93
+ ann_dir = os.path.join(filepath, "annotations")
94
+ img_dir = os.path.join(filepath, "images")
95
+ for guid, file in enumerate(sorted(os.listdir(ann_dir))):
96
+ tokens = []
97
+ bboxes = []
98
+ ner_tags = []
99
+ file_path = os.path.join(ann_dir, file)
100
+ with open(file_path, "r", encoding="utf8") as f:
101
+ data = json.load(f)
102
+ image_path = os.path.join(img_dir, file)
103
+ image_path = image_path.replace("json", "png")
104
+ image, size = load_image(image_path)
105
+ for item in data["form"]:
106
+ words, label = item["words"], item["label"]
107
+ words = [w for w in words if w["text"].strip() != ""]
108
+ if len(words) == 0:
109
+ continue
110
+ if label == "other":
111
+ for w in words:
112
+ tokens.append(w["text"])
113
+ ner_tags.append("O")
114
+ bboxes.append(normalize_bbox(w["box"], size))
115
+ else:
116
+ tokens.append(words[0]["text"])
117
+ ner_tags.append("B-" + label.upper())
118
+ bboxes.append(normalize_bbox(words[0]["box"], size))
119
+ for w in words[1:]:
120
+ tokens.append(w["text"])
121
+ ner_tags.append("I-" + label.upper())
122
+ bboxes.append(normalize_bbox(w["box"], size))
123
+ yield guid, {"id": str(guid), "tokens": tokens, "bboxes": bboxes, "ner_tags": ner_tags, "image_path": image_path}