---
title: SIBR
canonical_url: "https://www.modelscope.cn/datasets/iic/SIBR"
md_url: "https://www.modelscope.cn/datasets/iic/SIBR.md"
repository: iic/SIBR
chinese_name: "读光-OCR-自然场景信息抽取数据集-中英"
last_updated: 2023-03-14
license: "Apache License 2.0"
storage_size: "563 MB"
domain:
  - image
tasks:
  - image-to-text
language:
  - ch
  - en
type:
  - ner
downloads: 8845
stars: 5
---

# SIBR

> SIBR - iic 在 ModelScope 开源的数据集。SIBR是面向自然场景视觉信息抽取的数据集。相对已有FUNSD、XFUND、CORD等数据集，SIBR的图片具有噪音、打印偏移、重叠、模糊、褶皱等特点，对文字检测、识别、信息抽取都具有挑战。

iic/SIBR 是 ModelScope 魔搭社区上的image-to-text数据集，涉及 image 领域，存储大小 563 MB，采用 Apache License 2.0 许可。

- **Repository**: iic/SIBR
- **License**: Apache License 2.0
- **Tasks**: image-to-text
- **Domain**: image
- **Storage size**: 563 MB
- **Downloads**: 8845
- **Stars**: 5
- **Last updated**: 2023-03-14

Source: https://www.modelscope.cn/datasets/iic/SIBR

---

# SIBR自然场景信息抽取数据集(SIBR Dataset for VIE in the Wild)

## 数据集概述
SIBR是面向自然场景视觉信息抽取的数据集。

### 数据集简介
```
1）SIBR总的有1000张图片，400张测试，600张训练，包括中文、英文两种语言。
2）包含images.zip、label.zip、train.txt、test.txt四个文件，images.zip、label.zip中包含所有图片和标签，通过train.txt和test.txt区分训练和测试。
3）标注规则与FUNSD、XFUND一致，实体类别包括header、question、answer、other四类，但与之不同的是除了提供link id之外，还标注了link type。link type包括inter以及intra两种，inter表示实体间的link，也即是一对kv对之间的link；而intra表示同一个实体内部segment之间的link。通过intra将多行文字组成一个实体，通过inter将实体组成kv对。
```

### 标注格式
标注格式与FUNSD、XFUND一致，标签示例如下：
```
{
    "form":[
        {
            "box":[[1074.7,104.8],[2577.1,107.7],[2580.6,235.9],[1071.8,238.8]],
            "text": "取药窗口",
            "label": "question"
            "id": 2,
            "linking": [
                {
                    "type": "inter",
                    "id": [2,3]
                }
            ]
        },
        ……
    ]
}
```

### 引用方式
```bib
@inproceedings{yang2023modeling,
  title={Modeling Entities as Semantic Points for Visual Information Extraction in the Wild},
  author={Yang, Zhibo and Long, Rujiao and Wang, Pengfei and Song, Sibo and Zhong, Humen and Cheng, Wenqing and Bai, Xiang and Yao, Cong},
  booktitle={Proceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition},
  year={2023}
}
```

### Clone with HTTP
```bash
git clone https://www.modelscope.cn/datasets/damo/SIBR.git
```
