SynthChartNet

# SynthChartNet

<div style="display: flex; justify-content: center; align-items: center;">
<img src="https://cdn-uploads.huggingface.co/production/uploads/663e1254887b6f5645a0399f/Kgt6S5S_zPGGQ3IlmyRVB.png" alt="Chart Example" style="width: 800px; height: auto">
</div>

**SynthChartNet** is a multimodal dataset designed for training the **SmolDocling** model on chart-based document understanding tasks. It consists of **1,981,157** synthetically generated samples, where each image depicts a chart (e.g., line chart, bar chart, pie chart, stacked bar chart), and the associated ground truth is given in **OTSL** format.

Charts were rendered at 120 DPI using a diverse set of visualization libraries: **Matplotlib**, **Seaborn**, and **Pyecharts**, enabling visual variability in layout, style, and color schemes.

---

## Dataset Statistics

* **Total samples**: 1,981,157

* **Training set**: 1,981,157

* **Modalities**: Image, Text (OTSL format)

* **Chart Types**: Line, Bar, Pie, Stacked Bar

* **Rendering Engines**: Matplotlib, Seaborn, Pyecharts

---

## Data Format

Each dataset entry is structured as follows:

```json
{
"images": [PIL Image],
"texts": [
{
"assistant": "<loc_x0><loc_y0><loc_x1><loc_y1><_Chart_>OTSL_REPRESENTATION</chart>",
"source": "SynthChartNet",
"user": "<chart>"
}
]
}
```

---

## Intended Use

* Training multimodal models for **chart understanding**, specifically:

* Chart parsing and transcription to structured formats (OTSL)

---

## Citation

If you use SynthChartNet, please cite:

```bibtex
@article{nassar2025smoldocling,
title={SmolDocling: An ultra-compact vision-language model for end-to-end multi-modal document conversion},
author={Nassar, Ahmed and Marafioti, Andres and Omenetti, Matteo and Lysak, Maksym and Livathinos, Nikolaos and Auer, Christoph and Morin, Lucas and de Lima, Rafael Teixeira and Kim, Yusik and Gurbuz, A Said and others},
journal={arXiv preprint arXiv:2503.11576},
year={2025}
}
@inproceedings{lysak2023optimized,
title={Optimized table tokenization for table structure recognition},
author={Lysak, Maksym and Nassar, Ahmed and Livathinos, Nikolaos and Auer, Christoph and Staar, Peter},
booktitle={International Conference on Document Analysis and Recognition},
pages={37--50},
year={2023},
organization={Springer}
}
```

# SynthChartNet

<div style="display: flex; justify-content: center; align-items: center;">
<img src="https://cdn-uploads.huggingface.co/production/uploads/663e1254887b6f5645a0399f/Kgt6S5S_zPGGQ3IlmyRVB.png" alt="Chart Example" style="width: 800px; height: auto">
</div>

**SynthChartNet** 是一款专为基于图表的文档理解任务训练**SmolDocling**模型而打造的多模态数据集。该数据集共包含**1,981,157**个合成生成样本，每个样本的图像均为各类图表（例如折线图、柱状图、饼图、堆叠柱状图），配套的标注真值采用**OTSL**格式给出。

图表以120 DPI渲染，使用了多样化的可视化库：**Matplotlib**、**Seaborn**与**Pyecharts**，这使得生成的图表在布局、样式与配色方案上具备丰富的视觉多样性。

---

## 数据集统计信息
* **总样本数**：1,981,157
* **训练集**：1,981,157
* **模态类型**：图像、文本（OTSL格式）
* **图表类型**：折线图、柱状图、饼图、堆叠柱状图
* **渲染引擎**：Matplotlib、Seaborn、Pyecharts

---

## 数据格式
每条数据集条目结构如下：
json
{
"images": [PIL图像],
"texts": [
{
"assistant": "<loc_x0><loc_y0><loc_x1><loc_y1><_Chart_>OTSL_REPRESENTATION</chart>",
"source": "SynthChartNet",
"user": "<chart>"
}
]
}

---

## 预期用途
* 用于训练面向**图表理解**的多模态模型，具体包括：
* 图表解析与转录为结构化格式（OTSL）

---

## 引用声明
若使用SynthChartNet，请引用以下文献：
bibtex
@article{nassar2025smoldocling,
title={SmolDocling: An ultra-compact vision-language model for end-to-end multi-modal document conversion},
author={Nassar, Ahmed and Marafioti, Andres and Omenetti, Matteo and Lysak, Maksym and Livathinos, Nikolaos and Auer, Christoph and Morin, Lucas and de Lima, Rafael Teixeira and Kim, Yusik and Gurbuz, A Said and others},
journal={arXiv preprint arXiv:2503.11576},
year={2025}
}
@inproceedings{lysak2023optimized,
title={Optimized table tokenization for table structure recognition},
author={Lysak, Maksym and Nassar, Ahmed and Livathinos, Nikolaos and Auer, Christoph and Staar, Peter},
booktitle={International Conference on Document Analysis and Recognition},
pages={37--50},
year={2023},
organization={Springer}
}