umuthopeyildirim/svgen-500k

SVGen Vector Images Dataset

Overview

SVGen is a comprehensive dataset containing 300,000 SVG vector codes from a diverse set of sources including SVG-Repo, Noto Emoji, and InstructSVG. The dataset aims to provide a wide range of SVG files suitable for various applications including web development, design, and machine learning research.

Data Fields

• input: The name or label of the SVG item
• output: SVG code containing the vector representation
• description: Brief description of the SVG item
• source: The original source or collection of the SVG
• license: Licensing terms for using the SVG

Data Sources

• SVG-Repo
• Noto Emoji
• InstructSVG

Usage

The dataset is particularly useful for tasks such as icon classification, style transfer, image-to-vector translation, and much more. It serves as a rich resource for machine learning models that require high-quality SVG data.

Help Wanted

I wanted to use BILP to generate description's for each SVG, but It's not working well. If you have any ideas, please let me know. Here is the Github and it also contains Colab notebook links.

License

The dataset incorporates SVG files with varying licenses. Users are advised to consult the license field of each record for specific usage rights.

Contribution Guidelines

Contributions are welcome! If you find any issues or would like to add more SVGs to the dataset, please submit a pull request or open an issue in the repository.

Acknowledgements

A huge thanks to SVGRepo, Noto Emoji, and InstructSVG for providing the SVG files that make up this dataset.

For more details and to download the dataset, visit the project repository.