Create README.md

Load and check meta data:

from datasets import load_dataset 

dataset = load_dataset("yizhilll/arxiv_image_cap_mini")

# %%
dataset["train"][0]["caption"]
# %%
dataset["train"][0]["image"]

for split in ["train", "validation", "test"]:
    print(dataset[split])

Output:

Dataset({
    features: ['image', 'caption', 'captionBoundary', 'figType', 'imageText', 'name', 'page', 'regionBoundary', 'referred_text'],
    num_rows: 30328
})
Dataset({
    features: ['image', 'caption', 'captionBoundary', 'figType', 'imageText', 'name', 'page', 'regionBoundary', 'referred_text'],
    num_rows: 3803
})
Dataset({
    features: ['image', 'caption', 'captionBoundary', 'figType', 'imageText', 'name', 'page', 'regionBoundary', 'referred_text'],
    num_rows: 3776
})