pdf-layout-chinese / README.md
Doraemon-AI's picture
up
293fc85 verified
metadata
license: afl-3.0
task_categories:
  - feature-extraction
language:
  - en
  - zh
size_categories:
  - 100M<n<1B

pdf-layout-chinese: A Chinese document layout PDF dataset

介绍

pdf-layout-chinese是一个中文文档版面分析数据集,面向中文文献类(论文)场景。包含以下10个label:

正文 标题 图片 图片标题 表格 表格标题 页眉 页脚 注释 公式
Text Title Figure Figure caption Table Table caption Header Footer Reference Equation

共包含5000张训练集和1000张验证集,分别在train和val目录下。每张图片对应一个同名的标注文件(.json)。

样例展示:

1 2

标注格式

使用的标注工具是labelme,所以标注格式和labelme格式一致。这里说明一下比较重要的字段。

"shapes": shapes字段是一个list,里面有多个dict,每个dict代表一个标注实例。

"labels": 类别。

"points": 实例标注。因为我们的标注是Polygon形式,所以points里的坐标数量可能大于4。

"shape_type": "polygon"

"imagePath": 图片路径/名

"imageHeight": 高

"imageWidth": 宽

展示一个完整的标注样例:

{
  "version":"4.5.6",
  "flags":{},
  "shapes":[
    {
      "label":"Title",
      "points":[
        [
          553.1111111111111,
          166.59259259259258
        ],
        [
          553.1111111111111,
          198.59259259259258
        ],
        [
          686.1111111111111,
          198.59259259259258
        ],
        [
          686.1111111111111,
          166.59259259259258
        ]
      ],
      "group_id":null,
      "shape_type":"polygon",
      "flags":{}
    },
    {
      "label":"Text",
      "points":[
        [
          250.5925925925925,
          298.0740740740741
        ],
        [
          250.5925925925925,
          345.0740740740741
        ],
        [
          188.5925925925925,
          345.0740740740741
        ],
        [
          188.5925925925925,
          410.0740740740741
        ],
        [
          188.5925925925925,
          456.0740740740741
        ],
        [
          324.5925925925925,
          456.0740740740741
        ],
        [
          324.5925925925925,
          410.0740740740741
        ],
        [
          1051.5925925925926,
          410.0740740740741
        ],
        [
          1051.5925925925926,
          345.0740740740741
        ],
        [
          1052.5925925925926,
          345.0740740740741
        ],
        [
          1052.5925925925926,
          298.0740740740741
        ]
      ],
      "group_id":null,
      "shape_type":"polygon",
      "flags":{}
    },
    {
      "label":"Footer",
      "points":[
        [
          1033.7407407407406,
          1634.5185185185185
        ],
        [
          1033.7407407407406,
          1646.5185185185185
        ],
        [
          1052.7407407407406,
          1646.5185185185185
        ],
        [
          1052.7407407407406,
          1634.5185185185185
        ]
      ],
      "group_id":null,
      "shape_type":"polygon",
      "flags":{}
    }
  ],
  "imagePath":"val_0031.jpg",
  "imageData":null,
  "imageHeight":1754,
  "imageWidth":1240
}

转coco格式

执行命令:

# train
python3 labelme2coco.py train train_save_path  --labels labels.txt

# val
python3 labelme2coco.py val val_save_path  --labels labels.txt

转换结果保存在train_save_path/val_save_path目录下。

labelme2coco.py取自labelme,更多信息请参考labelme官方项目