Dataset Card for GlyphByT5Pretraining

Dataset Summary

This dataset contains data used in the glyph-alignment pretraining stage introduced in Glyph-ByT5. The dataset contains 100K typography data annotations that can be used to render visual text.

Dataset Sources

• Repository: [https://github.com/AIGText/Glyph-ByT5]
• Paper: [https://arxiv.org/abs/2403.09622]
• Project Page: [https://glyph-byt5.github.io/]

Dataset Structure

The dataset contains a json file that contains the detailed annotations. Each annotation is organized as follows:

{
    "texts": [
        "text string 1",
        "text string 2",
        ...
        "text string k"
    ],
    "styles": [
        {
            'color': 'color 1', # a hex code that represents color
            'font-family': 'font 1' # font code that represents font
        },
        {
            'color': 'color 2',
            'font-family': 'font 2'
        },
        ...
        {
            'color': 'color k',
            'font-family': 'font k'
        },
    ],
    "bbox": [
        [x1, y1, w1, h1, a1], # absolute coordinates in between [0, 1645],
        [x2, y2, w2, h2, a2],
        ...
        [xk, yk, wk, hk, ak]
    ]
}

Uses

Please check our paper and project page for more details.

Citation

If you find our work useful in your research, please consider citing:

@misc{liu2024glyphbyt5,
      title={Glyph-ByT5: A Customized Text Encoder for Accurate Visual Text Rendering}, 
      author={Zeyu Liu and Weicong Liang and Zhanhao Liang and Chong Luo and Ji Li and Gao Huang and Yuhui Yuan},
      year={2024},
      eprint={2403.09622},
      archivePrefix={arXiv},
      primaryClass={cs.CV}
}