SWHL/RapidOCR
Updated • 31
Need help to make the dataset viewer work? Make sure to review how to configure the dataset viewer, and open a discussion for direct support.
可用于自定义数据集下的模型验证和性能评估等。
from datasets import load_dataset
dataset = load_dataset("SWHL/text_det_test_dataset")
test_data = dataset['test']
print(test_data)
数据来源于网络,如侵删。
数据集使用labelme标注,json格式差别在于imagePath字段改写为了file_name字段。其余相同。
数据集标注为json格式,示例如下:
{
"version": "4.5.6",
"flags": {},
"shapes": [
{
"label": "",
"points": [
[
486.0,
751.0
],
[
588.0,
751.0
],
[
588.0,
779.0
],
[
486.0,
779.0
]
],
"group_id": null,
"shape_type": "polygon",
"flags": {}
},
{
"label": "",
"points": [
[
385.1382113821138,
686.6178861788618
],
[
592.4552845528456,
688.2439024390244
],
[
589.2032520325204,
739.4634146341464
],
[
386.7642276422764,
737.0243902439024
]
],
"group_id": null,
"shape_type": "polygon",
"flags": {}
}
],
"file_name": "19.png",
"imageData": "xxxxxx",
"imageHeight": 784,
"imageWidth": 613
}