README.md

---
language:
  - en
license: mit
pretty_name: Digital Signatures Dataset
tags:
  - digital-signatures
  - synthetic-data
  - image-classification
  - computer-vision
task_categories:
  - image-classification
  - zero-shot-image-classification
  - image-feature-extraction
size_categories:
  - 1K<n<10K
---

# Digital Signatures Dataset

This dataset contains unique synthetic digital signatures rendered in different fonts:
- 4,000 synthetic signatures in Rage font
  ![Rage font sample](https://huggingface.co/datasets/Benjy/digital_signatures/resolve/main/rage/0000.jpg)
- 4,000 synthetic signatures in Mistral font  
  ![Mistral font sample](https://huggingface.co/datasets/Benjy/digital_signatures/resolve/main/mistral/0000.jpg)
- 2,000 synthetic signatures in Arial Unicode font
  ![Arial Unicode font sample](https://huggingface.co/datasets/Benjy/digital_signatures/resolve/main/arial/0000.jpg)


## Purpose
For the development of models that can detect digital signatures in documentation using the publicly available Docusign® font styles.


## Structure  

The dataset is organized into three folders:
- `rage/` - Contains synthetic signatures rendered in Rage font
- `mistral/` - Contains synthetic signatures rendered in Mistral font
- `arial_unicode/` - Contains synthetic signatures rendered in Arial Unicode font

Each image follows the naming convention: `{font_name}_{index:04d}.jpg`
Examples:
- `rage_0000.jpg`
- `mistral_0001.jpg`
- `arial_unicode_0002.jpg`

Total dataset size: 10,000 signature images

## Usage

This dataset can be used for:
- Training signature verification models
- Testing font recognition systems
- Developing digital document processing pipelines

## License
This dataset is released under the MIT license.

## Citation

If you use this dataset in your research, please cite:
```
@dataset{digital_signatures,
  author = {Benjy},
  title = {Digital Signatures Dataset},
  year = {2024},
  publisher = {Hugging Face},
  url = {https://huggingface.co/datasets/Benjy/digital_signatures}
}
```

## Acknowledgments

This dataset was created to support research in digital signature verification and synthesis, while acknowledging the limitations of publicly available font alternatives to proprietary signature systems.

Credit to https://github.com/elifiner for the open source list of random names.