yuxi-liu-wired's picture
Update README.md
9d6ca7d verified
metadata
license: cc0-1.0
dataset_info:
  features:
    - name: style
      dtype: string
    - name: content
      dtype: string
    - name: image
      dtype: image
    - name: id
      dtype: string
  splits:
    - name: train
      num_bytes: 3155812770
      num_examples: 1738
  download_size: 3155909210
  dataset_size: 3155812770
configs:
  - config_name: default
    data_files:
      - split: train
        path: data/train-*

Style Content Grid SDXL

Dataset Structure

The dataset contains 1738 images of resolution 1024x1024, generated by Stable Diffusion XL (sd_xl_base_1.0 with model hash 31e35c80fc). They were all generated in lllyasviel/stable-diffusion-webui-forge, with the following positive and negative prompts:

Positive prompt: <style> of a <content>, \n masterpiece, best quality, high quality, 

Negative prompt: (worst quality, low quality, normal quality), 

with the following generation details:

Steps: 20, Sampler: Euler a, CFG scale: 8, Size: 1024x1024, Model hash: 31e35c80fc, Model: sd_xl_base_1.0, Version: f0.0.17v1.8.0rc-previous-1-ga9e0c387

The generation process is automated by the adieyal/sd-dynamic-prompts plugin, with the following template:

{1$$pencil sketch by da Vinci| painting by Rembrandt| drawing by Picasso| glazed ceramic sculpture| Chinese ink wash painting| ukiyo-e painting by Tsukioka Yoshitoni| painting by van Gogh| photograph| painting by Frida Kahlo| colorful collage} of a {1$$horse|cat|dog|man|woman|library|mountains|fish|solar system|bedroom}, \nmasterpiece, best quality, high quality, 

Each row of the dataset contains:

  • image: The image itself, in PNG format.
  • id: Its original filename.
  • style: A string describing what the style is.
    • Can be one of 10 values: "pencil sketch by da Vinci| painting by Rembrandt| drawing by Picasso| glazed ceramic sculpture| Chinese ink wash painting| ukiyo-e painting by Tsukioka Yoshitoni| painting by van Gogh| photograph| painting by Frida Kahlo| colorful collage"
  • content: A string describing what the content is.
    • Can be one of 10 values: "horse|cat|dog|man|woman|library|mountains|fish|solar system|bedroom"

Example uses

To load the dataset:

from datasets import load_dataset

ds = load_dataset("yuxi-liu-wired/style-content-grid-SDXL")

TODO: load CSD.

Licensing

The dataset is released under Creative Commons Zero v1.0 Universal.

@misc{yuxi-liu-wired_2024_style-content-grid-SDXL,
  author = {Yuxi Liu},
  title = {Style Content Grid SDXL},
  year = {2024},
  howpublished = {Hugging Face Dataset},
  url = {https://huggingface.co/datasets/yuxi-liu-wired/style-content-grid-SDXL}
}