doolayer's picture
Update README.md
2b3f901 verified
|
raw
history blame
840 Bytes
---
dataset_info:
features:
- name: image
dtype: image
- name: ground_truth
dtype: string
splits:
- name: train
num_bytes: 714443184.0
num_examples: 777
- name: validation
num_bytes: 83073676.0
num_examples: 97
- name: test
num_bytes: 99014242.0
num_examples: 97
download_size: 883571440
dataset_size: 896531102.0
configs:
- config_name: default
data_files:
- split: train
path: data/train-*
- split: validation
path: data/validation-*
- split: test
path: data/test-*
---
This dataset was created by using images from the [SROIE](https://arxiv.org/abs/2103.10213) dataset.
Labels were newly generated by [HCX Vision 3B](https://huggingface.co/naver-hyperclovax/HyperCLOVAX-SEED-Vision-Instruct-3B) following the [CORD](naver-clova-ix/cord-v2)-style annotation format.