README.md

annotations_creators:

• machine-generated language_creators:
• machine-generated license:
• cc-by-4.0 multilinguality:
• multilingual pretty_name: laion-publicdomain size_categories:
• 100K<n<1M source_datasets: -laion/laion2B-en tags:
• laion task_categories:
• text-to-image

Dataset Card for laion-publicdomain

Table of Contents

• Table of Contents
• Dataset Description
  • Dataset Summary
• Dataset Structure
  • Data Fields
• Dataset Creation
  • Curation Rationale
  • Source Data
  • Annotations
  • Personal and Sensitive Information
• Considerations for Using the Data
  • Social Impact of Dataset
  • Discussion of Biases
  • Other Known Limitations
• Additional Information
  • Licensing Information

Dataset Description

• Homepage: https://huggingface.co/datasets/devourthemoon/laion-publicdomain
• Repository: https://huggingface.co/datasets/devourthemoon/laion-publicdomain
• Paper: do i look like a scientist to you
• Leaderboard:
• Point of Contact: @devourthemoon on twitter

Dataset Summary

This dataset contains metadata about images from the LAION2B-eb dataset curated to a reasonable best guess of 'ethically sourced' images.

Dataset Structure

Data Fields

See the laion2B release notes.

Dataset Creation

Curation Rationale

This dataset contains images whose URLs are either from archive.org or whose license is Creative Commons of some sort. This is a useful first pass at "public use" images, as the Creative Commons licenses are primarily voluntary and intended for public use, and archive.org is a website that archives public domain images.

Source Data

The source dataset is at laion/laion2B-en and is not affiliated with this project.

Annotations

Annotation process

Laion2B-en is assembled from Common Crawl data.

Personal and Sensitive Information

[More Information Needed]

Considerations for Using the Data

Social Impact of Dataset

Is this dataset as ethical as possible?

No. This dataset exists as a proof of concept. Further research could improve the sourcing of the dataset in a number of ways, particularly improving the attribution of files to their original authors.

Can I willingly submit my own images to be included in the dataset?

This is a long term goal of this project with the ideal being the generation of 'personalized' AI models for artists. Contact @devourthemoon on Twitter if this interests you.

Is this dataset as robust as e.g. LAION2B?

Absolutely not. About 0.17% of the images in the LAION2B dataset matched the filters, leading to just over 600k images in this dataset.

Discussion of Biases

[More Information Needed]

Other Known Limitations

[More Information Needed]

Additional Information

Licensing Information

When using images from this dataset, please acknowledge the combination of Creative Commons licenses. This dataset itself follows CC-BY-4.0