Kosuke-Yamada commited on
Commit
da9f614
1 Parent(s): 2e63281

commit for the first time 🎉

Browse files
Files changed (1) hide show
  1. camera.py +147 -0
camera.py ADDED
@@ -0,0 +1,147 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+ import ast
2
+
3
+ import datasets as ds
4
+ import pandas as pd
5
+
6
+ _DESCRIPTION = """\
7
+ CAMERA (CyberAgent Multimodal Evaluation for Ad Text GeneRAtion) is the Japanese ad text generation dataset.
8
+ """
9
+
10
+ _CITATION = """\
11
+ @misc{mita2024striking,
12
+ title={Striking Gold in Advertising: Standardization and Exploration of Ad Text Generation},
13
+ author={Masato Mita and Soichiro Murakami and Akihiko Kato and Peinan Zhang},
14
+ year={2024},
15
+ eprint={2309.12030},
16
+ archivePrefix={arXiv},
17
+ primaryClass={id='cs.CL' full_name='Computation and Language' is_active=True alt_name='cmp-lg' in_archive='cs' is_general=False description='Covers natural language processing. Roughly includes material in ACM Subject Class I.2.7. Note that work on artificial languages (programming languages, logics, formal systems) that does not explicitly address natural-language issues broadly construed (natural-language processing, computational linguistics, speech, text retrieval, etc.) is not appropriate for this area.'}
18
+ }
19
+ """
20
+
21
+ _HOMEPAGE = "https://github.com/CyberAgentAILab/camera"
22
+
23
+ _LICENSE = """\
24
+ This work is licensed under a Creative Commons Attribution-NonCommercial-ShareAlike 4.0 International License.
25
+ """
26
+
27
+ _URLS = {
28
+ "without-lp-images": "https://storage.googleapis.com/camera-public/camera-v2.2-minimal.tar.gz",
29
+ "with-lp-images": "https://storage.googleapis.com/camera-public/camera-v2.2.tar.gz",
30
+ }
31
+
32
+ _DESCRIPTION = {
33
+ "without-lp-images": "The CAMERA dataset w/o LP images (ver.2.2.0)",
34
+ "with-lp-images": "The CAMERA dataset w/ LP images (ver.2.2.0)",
35
+ }
36
+
37
+ _VERSION = ds.Version("2.2.0", "")
38
+
39
+
40
+ class CameraConfig(ds.BuilderConfig):
41
+ def __init__(self, name: str, version: ds.Version = _VERSION, **kwargs):
42
+ super().__init__(
43
+ name=name,
44
+ description=_DESCRIPTION[name],
45
+ version=version,
46
+ **kwargs,
47
+ )
48
+
49
+
50
+ class CameraDataset(ds.GeneratorBasedBuilder):
51
+ BUILDER_CONFIGS = [CameraConfig(name="without-lp-images")]
52
+
53
+ DEFAULT_CONFIG_NAME = "without-lp-images"
54
+
55
+ def _info(self) -> ds.DatasetInfo:
56
+ features = ds.Features(
57
+ {
58
+ "asset_id": ds.Value("int64"),
59
+ "kw": ds.Value("string"),
60
+ "lp_meta_description": ds.Value("string"),
61
+ "title_org": ds.Value("string"),
62
+ "title_ne1": ds.Value("string"),
63
+ "title_ne2": ds.Value("string"),
64
+ "title_ne3": ds.Value("string"),
65
+ "domain": ds.Value("string"),
66
+ "parsed_full_text_annotation": ds.Sequence(
67
+ {
68
+ "text": ds.Value("string"),
69
+ "xmax": ds.Value("int64"),
70
+ "xmin": ds.Value("int64"),
71
+ "ymax": ds.Value("int64"),
72
+ "ymin": ds.Value("int64"),
73
+ }
74
+ ),
75
+ }
76
+ )
77
+
78
+ if self.config.name == "with-lp-images":
79
+ features["lp_image"] = ds.Image()
80
+
81
+ return ds.DatasetInfo(
82
+ description=_DESCRIPTION,
83
+ citation=_CITATION,
84
+ homepage=_HOMEPAGE,
85
+ license=_LICENSE,
86
+ features=features,
87
+ )
88
+
89
+ def _split_generators(self, dl_manager: ds.DownloadManager):
90
+ base_dir = dl_manager.download_and_extract(_URLS[self.config.name])
91
+ lp_image_dir: str | None = None
92
+
93
+ if self.config.name == "without-lp-images":
94
+ data_dir = f"{base_dir}/camera-v2.2-minimal"
95
+ elif self.config.name == "with-lp-images":
96
+ data_dir = f"{base_dir}/camera-v2.2"
97
+ lp_image_dir = f"{data_dir}/lp-screenshot"
98
+ else:
99
+ raise ValueError(f"Invalid config name: {self.config.name}")
100
+
101
+ return [
102
+ ds.SplitGenerator(
103
+ name=ds.Split.TRAIN,
104
+ gen_kwargs={
105
+ "file": f"{data_dir}/train.csv",
106
+ "lp_image_dir": lp_image_dir,
107
+ },
108
+ ),
109
+ ds.SplitGenerator(
110
+ name=ds.Split.VALIDATION,
111
+ gen_kwargs={
112
+ "file": f"{data_dir}/dev.csv",
113
+ "lp_image_dir": lp_image_dir,
114
+ },
115
+ ),
116
+ ds.SplitGenerator(
117
+ name=ds.Split.TEST,
118
+ gen_kwargs={
119
+ "file_name": f"{data_dir}/test.csv",
120
+ "lp_image_dir": lp_image_dir,
121
+ },
122
+ ),
123
+ ]
124
+
125
+ def _generate_examples(self, file: str, lp_image_dir: str | None = None):
126
+ df = pd.read_csv(file)
127
+ for i, data_dict in enumerate(df.to_dict("records")):
128
+ asset_id = data_dict["asset_id"]
129
+ example_dict = {
130
+ "asset_id": asset_id,
131
+ "kw": data_dict["kw"],
132
+ "lp_meta_description": data_dict["lp_meta_description"],
133
+ "title_org": data_dict["title_org"],
134
+ "title_ne1": data_dict.get("title_ne1", ""),
135
+ "title_ne2": data_dict.get("title_ne2", ""),
136
+ "title_ne3": data_dict.get("title_ne3", ""),
137
+ "domain": data_dict.get("domain", ""),
138
+ "parsed_full_text_annotation": ast.literal_eval(
139
+ data_dict["parsed_full_text_annotation"]
140
+ ),
141
+ }
142
+
143
+ if self.config.name == "with-lp-images" and lp_image_dir is not None:
144
+ file_name = f"screen-1200-{asset_id}.png"
145
+ example_dict["lp_image"] = f"{lp_image_dir}/{file_name}"
146
+
147
+ yield i, example_dict