MC_OCR2021 / my_dataset.py
tqhuyen's picture
Upload my_dataset.py
aa46b7b
raw
history blame
1.18 kB
import json
import datasets
datasets.Features(
{
"img_id": datasets.Image("path"),
"anno_polygons": datasets.Sequence(
{
[
{'category_id': datasets.Value("int32"),
'segmentation': datasets.Sequence(feature=list, ),
'area': datasets.Value("int64"),
'bbox': datasets.Sequence('list'),
'width': datasets.Value('int32'),
'height': datasets.Value('int32')
},
]
}
),
"anno_texts": datasets.Value("string"),
"anno_labels": datasets.Value("string"),
"anno_num": datasets.Value("int32"),
"anno_image_quality": datasets.Value("float64")
}
)
_URL = ""
def _split_generators(self, dl_manager):
return [
datasets.SplitGenerator(name=datasets.Split.TRAIN, gen_kwargs={"filepath": ".\mcocr_public_train_test_shared_data\mcocr_train_data"}),
datasets.SplitGenerator(name=datasets.Split.VALIDATION, gen_kwargs={"filepath": ".\mcocr_public_train_test_shared_data\mcocr_val_data"}),
]