---
title: TableVerse-5K
canonical_url: "https://www.modelscope.cn/datasets/pspdada/TableVerse-5K"
md_url: "https://www.modelscope.cn/datasets/pspdada/TableVerse-5K.md"
repository: pspdada/TableVerse-5K
chinese_name: TableVerse-5K
last_updated: 2026-09-30
license: other
storage_size: "806 MB"
downloads: 56226
stars: 1
---

# TableVerse-5K

> TableVerse-5K - pspdada 在 ModelScope 开源的数据集。面向 StrucTab 框架的表格解析评测基准

pspdada/TableVerse-5K 是 ModelScope 魔搭社区上的数据集，存储大小 806 MB，采用 other 许可。

- **Repository**: pspdada/TableVerse-5K
- **License**: other
- **Storage size**: 806 MB
- **Downloads**: 56226
- **Stars**: 1
- **Last updated**: 2026-09-30

Source: https://www.modelscope.cn/datasets/pspdada/TableVerse-5K

---

# TableVerse-5K <!-- omit in toc -->

**面向 StrucTab 框架的表格解析评测基准**

<p align="center">
  <a href="https://github.com/VirtualLUOUCAS/StrucTab">GitHub 仓库</a> •
  <a href="https://huggingface.co/datasets/psp-dada/TableVerse-5K">HuggingFace 数据集</a> •
  <a href="https://arxiv.org/abs/2606.29905">Arxiv 论文</a>
</p>

## 新闻 <!-- omit in toc -->

- [2026.06] 📖 代码与 TableVerse-5K 基准已发布！
- [2026.06] 🎉 我们的 StrucTab 已被 ECCV 2026 接收！

## 概览 <!-- omit in toc -->

**TableVerse-5K** 是 **StrucTab** 的评测基准。StrucTab 是一个面向 **表格解析 (table parsing)** 任务的结构化优化框架，目标是将表格图像转换为结构化的 HTML 表格。数据集中每条样本由一张表格图像、一段指令提示词与一份标准答案 HTML 表格组成，模型解析结果使用 TEDS / TEDS-S 指标进行评分。

<table align="center">
    <p align="center">
      <img src="/docs/figures/introduction.jpg" width="80%" />
    </p>
</table>

基准的数据集构建流程如下：

<table align="center">
    <p align="center">
      <img src="/docs/figures/benchmark_pipeline.jpg" width="80%" />
    </p>
</table>

## 目录 <!-- omit in toc -->

- [基准统计](#基准统计)
- [数据集结构](#数据集结构)
- [数据格式](#数据格式)
- [下载方式](#下载方式)
- [使用方法](#使用方法)
- [引用](#引用)
- [许可证](#许可证)

## 基准统计

| 项目     | 详情                         |
| -------- | ---------------------------- |
| 样本规模 | 约 5K 张表格图像             |
| 任务     | 表格解析（图像 → HTML 表格） |
| 语言     | 双语（表格内容含中文与英文） |
| 输出格式 | HTML（`<table>...</table>`） |
| 评分指标 | TEDS、TEDS-S                 |

## 数据集结构

```
data/
├── TableVerse_5K.jsonl     # 所有样本的标注信息
└── images/                 # 表格图片（*.jpg）
```

## 数据格式

`TableVerse_5K.jsonl` 每一行为一个 JSON 对象：

```json
{
  "image_path": "images/xxx.jpg",
  "question": "You are an AI specialized in recognizing and extracting table from images...",
  "ref_answer": "<table>...</table>"
}
```

| 字段         | 类型   | 说明                                              |
| ------------ | ------ | ------------------------------------------------- |
| `image_path` | string | 相对于 `data/` 的图片路径，同时作为样本的唯一主键 |
| `question`   | string | 与图像一同送入模型的指令 / 提示词                 |
| `ref_answer` | string | HTML 格式的标准答案表格（`<table>...</table>`）   |

## 下载方式

:modelscope-code[]{type="sdk"}
:modelscope-code[]{type="git"}

## 使用方法

请参考 [GitHub 仓库](https://github.com/VirtualLUOUCAS/StrucTab) 获取完整的推理与评测脚本。

```bash
# 1. 克隆代码仓库
git clone https://github.com/VirtualLUOUCAS/StrucTab
cd StrucTab/benchmark
pip install -r requirements.txt

# 2. 克隆本数据集，并将内容放入 benchmark/data/ 目录下，
#    使其结构为 benchmark/data/TableVerse_5K.jsonl 与 benchmark/data/images/

# 3. 推理
python infer.py --api_type openai_compat --model_name <模型名> --base_url <接口地址>

# 4. 评分（需先启动 TEDS 打分服务，详见仓库 README）
python judge.py
```

## 引用

如果本数据集对你有帮助，欢迎引用：

```bibtex
@inproceedings{StrucTab_2026,
  title     = {{StrucTab}: A Structured Optimization Framework for Table Parsing},
  author    = {Li, Gengluo and Peng, Shangpin and Zhang, Chengquan and Wu, Binghong and Feng, Hao and Wang, Weinong and Lyu, Pengyuan and Shen, Huawen and Wan, Xingyu and Tian, Zhuotao and others},
  booktitle = {Proceedings of the European Conference on Computer Vision},
  pages     = {672--690},
  year      = {2026}
}
```

## 许可证

本数据集仅供**学术研究使用**。
