atasoglu commited on
Commit
4343508
1 Parent(s): 867d57f

Upload flickr8k-dataset.py

Browse files
Files changed (1) hide show
  1. flickr8k-dataset.py +112 -0
flickr8k-dataset.py ADDED
@@ -0,0 +1,112 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+ import datasets
2
+ import re
3
+ from pathlib import Path
4
+
5
+
6
+ class Flickr8kDataset(datasets.GeneratorBasedBuilder):
7
+ VERSION = datasets.Version("0.0.0")
8
+ BUILDER_CONFIGS = [
9
+ datasets.BuilderConfig(
10
+ name="flickr8k_dataset",
11
+ version=VERSION,
12
+ ),
13
+ ]
14
+
15
+ def _info(self):
16
+ feature_dict = {
17
+ "image_id": datasets.Value("string"),
18
+ "image_path": datasets.Value("string"),
19
+ "captions": datasets.Sequence(datasets.Value("string")),
20
+ }
21
+ return datasets.DatasetInfo(
22
+ description="Flickr8k dataset",
23
+ features=datasets.Features(feature_dict),
24
+ )
25
+
26
+ def _split_generators(self, dl_manager):
27
+ data_dir = self.config.data_dir
28
+ if not data_dir:
29
+ raise ValueError("'data_dir' argument is required.")
30
+
31
+ archive_path = dl_manager.extract(
32
+ {
33
+ "images": Path(data_dir).joinpath("Flickr8k_Dataset.zip"),
34
+ "texts": Path(data_dir).joinpath("Flickr8k_text.zip"),
35
+ }
36
+ )
37
+
38
+ splits = [
39
+ datasets.SplitGenerator(
40
+ name=datasets.Split.TRAIN,
41
+ gen_kwargs={
42
+ "image_ids": Path(archive_path["texts"]).joinpath(
43
+ "Flickr_8k.trainImages.txt"
44
+ ),
45
+ },
46
+ ),
47
+ datasets.SplitGenerator(
48
+ name=datasets.Split.TEST,
49
+ gen_kwargs={
50
+ "image_ids": Path(archive_path["texts"]).joinpath(
51
+ "Flickr_8k.testImages.txt"
52
+ ),
53
+ },
54
+ ),
55
+ datasets.SplitGenerator(
56
+ name=datasets.Split.VALIDATION,
57
+ gen_kwargs={
58
+ "image_ids": Path(archive_path["texts"]).joinpath(
59
+ "Flickr_8k.devImages.txt"
60
+ ),
61
+ },
62
+ ),
63
+ ]
64
+ for split in splits:
65
+ split.gen_kwargs.update(
66
+ {
67
+ "image_dir": Path(archive_path["images"]).joinpath(
68
+ "Flicker8k_Dataset"
69
+ ),
70
+ "captions_txt": Path(archive_path["texts"]).joinpath(
71
+ "Flickr8k.token.txt"
72
+ ),
73
+ }
74
+ )
75
+
76
+ return splits
77
+
78
+ def _generate_examples(
79
+ self,
80
+ image_ids,
81
+ image_dir,
82
+ captions_txt,
83
+ ):
84
+ def get_image_ids_as_list(fpath):
85
+ ids = []
86
+ with open(fpath, "r") as f:
87
+ for image in f.read().split("\n"):
88
+ if image != "":
89
+ ids.append(image.split(".")[0])
90
+ return ids
91
+
92
+ def get_captions_by_id(captions_txt, id):
93
+ captions = []
94
+ with open(captions_txt, "r") as f:
95
+ for line in f.read().split("\n"):
96
+ if line != "":
97
+ m = re.match(id + r".jpg#\d+(.*)", line)
98
+ if m != None:
99
+ captions.append(m.group(1)[1:])
100
+ return captions
101
+
102
+ ids = get_image_ids_as_list(image_ids)
103
+
104
+ for id in ids:
105
+ img_path = image_dir.joinpath(id + ".jpg")
106
+ captions = get_captions_by_id(captions_txt, id)
107
+ example = {
108
+ "image_id": id,
109
+ "image_path": str(img_path),
110
+ "captions": captions,
111
+ }
112
+ yield id, example