examples/inputs/SOURCE.json

{
  "dataset": "creative-graphic-design/PubLayNet",
  "dataset_url": "https://huggingface.co/datasets/creative-graphic-design/PubLayNet",
  "license": "CDLA-Permissive-1.0",
  "citation": "Zhong, X., Tang, J., & Yepes, A. J. (2019). PubLayNet: largest dataset ever for document layout analysis. arXiv:1908.07836",
  "note": "Pages sourced from PubMed Central open-access scientific articles. Selected for high layout-element density from a scan of the dataset train split (see fetch_example_images.py).",
  "images": [
    {
      "file_name": "PMC5883194_00003.jpg",
      "width": 596,
      "height": 794,
      "n_boxes": 31,
      "n_classes": 4
    },
    {
      "file_name": "PMC5883225_00001.jpg",
      "width": 596,
      "height": 794,
      "n_boxes": 28,
      "n_classes": 4
    },
    {
      "file_name": "PMC4413546_00014.jpg",
      "width": 596,
      "height": 794,
      "n_boxes": 21,
      "n_classes": 4
    },
    {
      "file_name": "PMC5942346_00002.jpg",
      "width": 596,
      "height": 791,
      "n_boxes": 21,
      "n_classes": 4
    }
  ]
}