Public Domain 12M: A Highly Aesthetic Image-Text Dataset with Novel Governance Mechanisms

TL;DR

Public Domain 12M (PD12M) is the largest high-quality public domain image-text dataset with innovative community governance, enabling safer and more reproducible training of text-to-image models.

Contribution

The paper introduces PD12M, the largest public domain image-text dataset, along with novel community-driven governance mechanisms for dataset management.

Findings

Largest public domain image-text dataset to date

Community governance mechanisms reduce harm and enhance reproducibility

Supports training of foundation models with minimal copyright concerns

Abstract

We present Public Domain 12M (PD12M), a dataset of 12.4 million high-quality public domain and CC0-licensed images with synthetic captions, designed for training text-to-image models. PD12M is the largest public domain image-text dataset to date, with sufficient size to train foundation models while minimizing copyright concerns. Through the Source.Plus platform, we also introduce novel, community-driven dataset governance mechanisms that reduce harm and support reproducibility over time.