CDLA
CDLA copied to clipboard
托管到Hugging Face
感谢作者工作! 由于最近也在做相关方面工作,顺手将该数据集托管到了Hugging Face,便于大家后续使用吧
详情参见: https://huggingface.co/datasets/SWHL/CDLA
使用方式为:
from datasets import load_dataset
dataset = load_dataset("SWHL/CDLA")
train_data = dataset["train"]
print(train_data[0])
val_data = dataset["validation"]
print(val_data[0])
# {'image': <PIL.JpegImagePlugin.JpegImageFile image mode=RGB size=1240x1754 at 0x12FEE3DF0>,
# 'version': '4.5.6', 'flags': {},
# 'shapes': [
# {'label': 'Header', 'points': [[118.0, 135.66666666666669]], 'group_id': None, 'shape_type': 'polygon', 'flags': {}}
# ],
# 'imagePath': 'train_0001.jpg', 'imageData': None, 'imageHeight': 1754, 'imageWidth': 1240}