Datasets:

Languages:
Indonesian
ArXiv:
License:
holylovenia commited on
Commit
cabfa8d
1 Parent(s): 0279a9c

Upload cub_bahasa.py with huggingface_hub

Browse files
Files changed (1) hide show
  1. cub_bahasa.py +386 -0
cub_bahasa.py ADDED
@@ -0,0 +1,386 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+ import json
2
+ from pathlib import Path
3
+ from typing import Dict, List, Tuple
4
+
5
+ import datasets
6
+ import pandas as pd
7
+
8
+ from seacrowd.utils import schemas
9
+ from seacrowd.utils.configs import SEACrowdConfig
10
+ from seacrowd.utils.constants import Tasks, Licenses
11
+
12
+ _CITATION = """\
13
+ @article{mahadi2023indonesian,
14
+ author = {Made Raharja Surya Mahadi and Nugraha Priya Utama},
15
+ title = {Indonesian Text-to-Image Synthesis with Sentence-BERT and FastGAN},
16
+ journal = {arXiv preprint arXiv:2303.14517},
17
+ year = {2023},
18
+ url = {https://arxiv.org/abs/2303.14517},
19
+ }
20
+ """
21
+
22
+ _DATASETNAME = "cub_bahasa"
23
+ _DESCRIPTION = """\
24
+ Semi-translated dataset of CUB-200-2011 into Indonesian. This dataset contains thousands
25
+ of image-text annotation pairs of 200 subcategories belonging to birds. The natural
26
+ language descriptions are collected through the Amazon Mechanical Turk (AMT) platform and
27
+ are required at least 10 words, without any information on subcategories and actions.
28
+ """
29
+
30
+ _LOCAL=False
31
+ _LANGUAGES = ["ind"] # We follow ISO639-3 language code (https://iso639-3.sil.org/code_tables/639/data)
32
+
33
+ _HOMEPAGE = "https://github.com/share424/Indonesian-Text-to-Image-synthesis-with-Sentence-BERT-and-FastGAN"
34
+ _LICENSE = Licenses.UNKNOWN.value
35
+ _URLS = {
36
+ "text": "https://raw.githubusercontent.com/share424/Indonesian-Text-to-Image-synthesis-with-Sentence-BERT-and-FastGAN/master/dataset/indo_cub_200_2011_captions.json",
37
+ "image": "https://data.caltech.edu/records/65de6-vp158/files/CUB_200_2011.tgz"
38
+ }
39
+
40
+ _SUPPORTED_TASKS = [Tasks.IMAGE_CAPTIONING]
41
+ _SOURCE_VERSION = "1.0.0"
42
+ _SEACROWD_VERSION = "2024.06.20"
43
+
44
+
45
+ class CubBahasaDataset(datasets.GeneratorBasedBuilder):
46
+ """CUB-200-2011 image-text dataset in Indonesian language for bird domain."""
47
+
48
+ SOURCE_VERSION = datasets.Version(_SOURCE_VERSION)
49
+ SEACROWD_VERSION = datasets.Version(_SEACROWD_VERSION)
50
+
51
+ SEACROWD_SCHEMA_NAME = "imtext"
52
+ IMAGE_CLASS = {
53
+ 1: '001.Black_footed_Albatross',
54
+ 2: '002.Laysan_Albatross',
55
+ 3: '003.Sooty_Albatross',
56
+ 4: '004.Groove_billed_Ani',
57
+ 5: '005.Crested_Auklet',
58
+ 6: '006.Least_Auklet',
59
+ 7: '007.Parakeet_Auklet',
60
+ 8: '008.Rhinoceros_Auklet',
61
+ 9: '009.Brewer_Blackbird',
62
+ 10: '010.Red_winged_Blackbird',
63
+ 11: '011.Rusty_Blackbird',
64
+ 12: '012.Yellow_headed_Blackbird',
65
+ 13: '013.Bobolink',
66
+ 14: '014.Indigo_Bunting',
67
+ 15: '015.Lazuli_Bunting',
68
+ 16: '016.Painted_Bunting',
69
+ 17: '017.Cardinal',
70
+ 18: '018.Spotted_Catbird',
71
+ 19: '019.Gray_Catbird',
72
+ 20: '020.Yellow_breasted_Chat',
73
+ 21: '021.Eastern_Towhee',
74
+ 22: '022.Chuck_will_Widow',
75
+ 23: '023.Brandt_Cormorant',
76
+ 24: '024.Red_faced_Cormorant',
77
+ 25: '025.Pelagic_Cormorant',
78
+ 26: '026.Bronzed_Cowbird',
79
+ 27: '027.Shiny_Cowbird',
80
+ 28: '028.Brown_Creeper',
81
+ 29: '029.American_Crow',
82
+ 30: '030.Fish_Crow',
83
+ 31: '031.Black_billed_Cuckoo',
84
+ 32: '032.Mangrove_Cuckoo',
85
+ 33: '033.Yellow_billed_Cuckoo',
86
+ 34: '034.Gray_crowned_Rosy_Finch',
87
+ 35: '035.Purple_Finch',
88
+ 36: '036.Northern_Flicker',
89
+ 37: '037.Acadian_Flycatcher',
90
+ 38: '038.Great_Crested_Flycatcher',
91
+ 39: '039.Least_Flycatcher',
92
+ 40: '040.Olive_sided_Flycatcher',
93
+ 41: '041.Scissor_tailed_Flycatcher',
94
+ 42: '042.Vermilion_Flycatcher',
95
+ 43: '043.Yellow_bellied_Flycatcher',
96
+ 44: '044.Frigatebird',
97
+ 45: '045.Northern_Fulmar',
98
+ 46: '046.Gadwall',
99
+ 47: '047.American_Goldfinch',
100
+ 48: '048.European_Goldfinch',
101
+ 49: '049.Boat_tailed_Grackle',
102
+ 50: '050.Eared_Grebe',
103
+ 51: '051.Horned_Grebe',
104
+ 52: '052.Pied_billed_Grebe',
105
+ 53: '053.Western_Grebe',
106
+ 54: '054.Blue_Grosbeak',
107
+ 55: '055.Evening_Grosbeak',
108
+ 56: '056.Pine_Grosbeak',
109
+ 57: '057.Rose_breasted_Grosbeak',
110
+ 58: '058.Pigeon_Guillemot',
111
+ 59: '059.California_Gull',
112
+ 60: '060.Glaucous_winged_Gull',
113
+ 61: '061.Heermann_Gull',
114
+ 62: '062.Herring_Gull',
115
+ 63: '063.Ivory_Gull',
116
+ 64: '064.Ring_billed_Gull',
117
+ 65: '065.Slaty_backed_Gull',
118
+ 66: '066.Western_Gull',
119
+ 67: '067.Anna_Hummingbird',
120
+ 68: '068.Ruby_throated_Hummingbird',
121
+ 69: '069.Rufous_Hummingbird',
122
+ 70: '070.Green_Violetear',
123
+ 71: '071.Long_tailed_Jaeger',
124
+ 72: '072.Pomarine_Jaeger',
125
+ 73: '073.Blue_Jay',
126
+ 74: '074.Florida_Jay',
127
+ 75: '075.Green_Jay',
128
+ 76: '076.Dark_eyed_Junco',
129
+ 77: '077.Tropical_Kingbird',
130
+ 78: '078.Gray_Kingbird',
131
+ 79: '079.Belted_Kingfisher',
132
+ 80: '080.Green_Kingfisher',
133
+ 81: '081.Pied_Kingfisher',
134
+ 82: '082.Ringed_Kingfisher',
135
+ 83: '083.White_breasted_Kingfisher',
136
+ 84: '084.Red_legged_Kittiwake',
137
+ 85: '085.Horned_Lark',
138
+ 86: '086.Pacific_Loon',
139
+ 87: '087.Mallard',
140
+ 88: '088.Western_Meadowlark',
141
+ 89: '089.Hooded_Merganser',
142
+ 90: '090.Red_breasted_Merganser',
143
+ 91: '091.Mockingbird',
144
+ 92: '092.Nighthawk',
145
+ 93: '093.Clark_Nutcracker',
146
+ 94: '094.White_breasted_Nuthatch',
147
+ 95: '095.Baltimore_Oriole',
148
+ 96: '096.Hooded_Oriole',
149
+ 97: '097.Orchard_Oriole',
150
+ 98: '098.Scott_Oriole',
151
+ 99: '099.Ovenbird',
152
+ 100: '100.Brown_Pelican',
153
+ 101: '101.White_Pelican',
154
+ 102: '102.Western_Wood_Pewee',
155
+ 103: '103.Sayornis',
156
+ 104: '104.American_Pipit',
157
+ 105: '105.Whip_poor_Will',
158
+ 106: '106.Horned_Puffin',
159
+ 107: '107.Common_Raven',
160
+ 108: '108.White_necked_Raven',
161
+ 109: '109.American_Redstart',
162
+ 110: '110.Geococcyx',
163
+ 111: '111.Loggerhead_Shrike',
164
+ 112: '112.Great_Grey_Shrike',
165
+ 113: '113.Baird_Sparrow',
166
+ 114: '114.Black_throated_Sparrow',
167
+ 115: '115.Brewer_Sparrow',
168
+ 116: '116.Chipping_Sparrow',
169
+ 117: '117.Clay_colored_Sparrow',
170
+ 118: '118.House_Sparrow',
171
+ 119: '119.Field_Sparrow',
172
+ 120: '120.Fox_Sparrow',
173
+ 121: '121.Grasshopper_Sparrow',
174
+ 122: '122.Harris_Sparrow',
175
+ 123: '123.Henslow_Sparrow',
176
+ 124: '124.Le_Conte_Sparrow',
177
+ 125: '125.Lincoln_Sparrow',
178
+ 126: '126.Nelson_Sharp_tailed_Sparrow',
179
+ 127: '127.Savannah_Sparrow',
180
+ 128: '128.Seaside_Sparrow',
181
+ 129: '129.Song_Sparrow',
182
+ 130: '130.Tree_Sparrow',
183
+ 131: '131.Vesper_Sparrow',
184
+ 132: '132.White_crowned_Sparrow',
185
+ 133: '133.White_throated_Sparrow',
186
+ 134: '134.Cape_Glossy_Starling',
187
+ 135: '135.Bank_Swallow',
188
+ 136: '136.Barn_Swallow',
189
+ 137: '137.Cliff_Swallow',
190
+ 138: '138.Tree_Swallow',
191
+ 139: '139.Scarlet_Tanager',
192
+ 140: '140.Summer_Tanager',
193
+ 141: '141.Artic_Tern',
194
+ 142: '142.Black_Tern',
195
+ 143: '143.Caspian_Tern',
196
+ 144: '144.Common_Tern',
197
+ 145: '145.Elegant_Tern',
198
+ 146: '146.Forsters_Tern',
199
+ 147: '147.Least_Tern',
200
+ 148: '148.Green_tailed_Towhee',
201
+ 149: '149.Brown_Thrasher',
202
+ 150: '150.Sage_Thrasher',
203
+ 151: '151.Black_capped_Vireo',
204
+ 152: '152.Blue_headed_Vireo',
205
+ 153: '153.Philadelphia_Vireo',
206
+ 154: '154.Red_eyed_Vireo',
207
+ 155: '155.Warbling_Vireo',
208
+ 156: '156.White_eyed_Vireo',
209
+ 157: '157.Yellow_throated_Vireo',
210
+ 158: '158.Bay_breasted_Warbler',
211
+ 159: '159.Black_and_white_Warbler',
212
+ 160: '160.Black_throated_Blue_Warbler',
213
+ 161: '161.Blue_winged_Warbler',
214
+ 162: '162.Canada_Warbler',
215
+ 163: '163.Cape_May_Warbler',
216
+ 164: '164.Cerulean_Warbler',
217
+ 165: '165.Chestnut_sided_Warbler',
218
+ 166: '166.Golden_winged_Warbler',
219
+ 167: '167.Hooded_Warbler',
220
+ 168: '168.Kentucky_Warbler',
221
+ 169: '169.Magnolia_Warbler',
222
+ 170: '170.Mourning_Warbler',
223
+ 171: '171.Myrtle_Warbler',
224
+ 172: '172.Nashville_Warbler',
225
+ 173: '173.Orange_crowned_Warbler',
226
+ 174: '174.Palm_Warbler',
227
+ 175: '175.Pine_Warbler',
228
+ 176: '176.Prairie_Warbler',
229
+ 177: '177.Prothonotary_Warbler',
230
+ 178: '178.Swainson_Warbler',
231
+ 179: '179.Tennessee_Warbler',
232
+ 180: '180.Wilson_Warbler',
233
+ 181: '181.Worm_eating_Warbler',
234
+ 182: '182.Yellow_Warbler',
235
+ 183: '183.Northern_Waterthrush',
236
+ 184: '184.Louisiana_Waterthrush',
237
+ 185: '185.Bohemian_Waxwing',
238
+ 186: '186.Cedar_Waxwing',
239
+ 187: '187.American_Three_toed_Woodpecker',
240
+ 188: '188.Pileated_Woodpecker',
241
+ 189: '189.Red_bellied_Woodpecker',
242
+ 190: '190.Red_cockaded_Woodpecker',
243
+ 191: '191.Red_headed_Woodpecker',
244
+ 192: '192.Downy_Woodpecker',
245
+ 193: '193.Bewick_Wren',
246
+ 194: '194.Cactus_Wren',
247
+ 195: '195.Carolina_Wren',
248
+ 196: '196.House_Wren',
249
+ 197: '197.Marsh_Wren',
250
+ 198: '198.Rock_Wren',
251
+ 199: '199.Winter_Wren',
252
+ 200: '200.Common_Yellowthroat'
253
+ }
254
+
255
+ BUILDER_CONFIGS = [
256
+ SEACrowdConfig(
257
+ name=f"{_DATASETNAME}_source",
258
+ version=SOURCE_VERSION,
259
+ description=f"{_DATASETNAME} source schema",
260
+ schema="source",
261
+ subset_id=f"{_DATASETNAME}",
262
+ ),
263
+ SEACrowdConfig(
264
+ name=f"{_DATASETNAME}_seacrowd_{SEACROWD_SCHEMA_NAME}",
265
+ version=SEACROWD_VERSION,
266
+ description=f"{_DATASETNAME} SEACrowd schema",
267
+ schema=f"seacrowd_{SEACROWD_SCHEMA_NAME}",
268
+ subset_id=f"{_DATASETNAME}",
269
+ ),
270
+ ]
271
+
272
+ DEFAULT_CONFIG_NAME = f"{_DATASETNAME}_source"
273
+
274
+ def _info(self) -> datasets.DatasetInfo:
275
+ if self.config.schema == "source":
276
+ features = datasets.Features(
277
+ {
278
+ "image_id": datasets.Value("int64"),
279
+ "class_id": datasets.Value("int64"),
280
+ "image_path": datasets.Value("string"),
281
+ "class_name": datasets.Value("string"),
282
+ "captions": [
283
+ {
284
+ "caption_eng": datasets.Value("string"),
285
+ "caption_ind": datasets.Value("string"),
286
+ }
287
+ ]
288
+ }
289
+ )
290
+ elif self.config.schema == f"seacrowd_{self.SEACROWD_SCHEMA_NAME}":
291
+ features = schemas.image_text_features(label_names=list(self.IMAGE_CLASS.values()))
292
+
293
+ return datasets.DatasetInfo(
294
+ description=_DESCRIPTION,
295
+ features=features,
296
+ homepage=_HOMEPAGE,
297
+ license=_LICENSE,
298
+ citation=_CITATION,
299
+ )
300
+
301
+ def _split_generators(self, dl_manager: datasets.DownloadManager) -> List[datasets.SplitGenerator]:
302
+ # expect several minutes to download image data ~1.2GB
303
+ data_path = dl_manager.download_and_extract(_URLS)
304
+
305
+ # working with image dataset
306
+ image_meta = Path(data_path["image"]) / "CUB_200_2011" / "images.txt"
307
+ df_image = pd.read_csv(image_meta, sep=" ", names=["image_id", "image_path"])
308
+ df_image['image_path'] = df_image['image_path'].apply(lambda x: Path(image_meta.parent, 'images', x))
309
+
310
+ label_meta = Path(data_path["image"]) / "CUB_200_2011" / "image_class_labels.txt"
311
+ df_label = pd.read_csv(label_meta, sep=" ", names=["image_id", "class_id"])
312
+
313
+ # working with text dataset
314
+ text_path = Path(data_path["text"])
315
+ with open(text_path, "r") as f:
316
+ text_data = json.load(f)
317
+
318
+ df_text = pd.DataFrame([
319
+ {
320
+ 'image_name': item['filename'],
321
+ 'en_caption': caption['english'],
322
+ 'id_caption': caption['indo']
323
+ } for item in text_data['dataset'] for caption in item['captions']
324
+ ])
325
+ grouped_text = df_text.groupby('image_name').agg(list).reset_index()
326
+
327
+ # working with split
328
+ split_dir = Path(data_path["image"]) / "CUB_200_2011" / "train_test_split.txt"
329
+ df_split = pd.read_csv(split_dir, sep=" ", names=["image_id", "is_train"])
330
+
331
+ # merge all data
332
+ df_image['image_name'] = df_image['image_path'].apply(lambda x: x.name)
333
+ df = pd.merge(df_image, grouped_text, on="image_name")
334
+ df.drop(columns=['image_name'], inplace=True)
335
+
336
+ df = pd.merge(df, df_label, on="image_id")
337
+ df = pd.merge(df, df_split, on="image_id")
338
+
339
+ return [
340
+ datasets.SplitGenerator(
341
+ name=datasets.Split.TRAIN,
342
+ gen_kwargs={
343
+ "data": df[df['is_train'] == 1],
344
+ "split": "train",
345
+ },
346
+ ),
347
+ datasets.SplitGenerator(
348
+ name=datasets.Split.TEST,
349
+ gen_kwargs={
350
+ "data": df[df['is_train'] == 0],
351
+ "split": "test",
352
+ },
353
+ ),
354
+ ]
355
+
356
+ def _generate_examples(self, data: pd.DataFrame, split: str) -> Tuple[int, Dict]:
357
+ if self.config.schema == "source":
358
+ for key, row in data.iterrows():
359
+ example = {
360
+ "image_id": row["image_id"],
361
+ "class_id": row["class_id"],
362
+ "image_path": row["image_path"],
363
+ "class_name": self.IMAGE_CLASS[row["class_id"]],
364
+ "captions": [
365
+ {
366
+ "caption_eng": row["en_caption"][i],
367
+ "caption_ind": row["id_caption"][i],
368
+ } for i in range(len(row["en_caption"]))
369
+ ]
370
+ }
371
+ yield key, example
372
+ elif self.config.schema == f"seacrowd_{self.SEACROWD_SCHEMA_NAME}":
373
+ key = 0
374
+ for _, row in data.iterrows():
375
+ for i in range(len(row["id_caption"])):
376
+ example = {
377
+ "id": str(key),
378
+ "image_paths": [row["image_path"]],
379
+ "texts": row["id_caption"][i],
380
+ "metadata": {
381
+ "context": row["en_caption"][i],
382
+ "labels": [self.IMAGE_CLASS[row["class_id"]]],
383
+ }
384
+ }
385
+ yield key, example
386
+ key += 1