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

2024/10/30 by Meyer, Jordan, Padgett, Nick, Miller, Cullen +1 · 7 citations
#Artificial Intelligence (cs.AI) #FOS: Computer and information sciences

paper · doi:10.48550/arxiv.2410.23144

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.

Cited by

• SCAdapter: Content-Style Disentanglement for Diffusion Style Transfer
• EditMGT: Unleashing Potentials of Masked Generative Transformers in Image Editing
• ParaUni: Enhance Generation in Unified Multimodal Model with Reinforcement-driven Hierarchical Parallel Information Interaction
• Cultural Alien Sampler: Open-ended art generation balancing originality and coherence
• HeadsUp! High-Fidelity Portrait Image Super-Resolution
• Factuality Matters: When Image Generation and Editing Meet Structured Visuals
• Kwai Keye-VL 1.5 Technical Report

Related

• From Pixels to Prose: A Large Dataset of Dense Image Captions
• RedCaps: web-curated image-text data created by the people, for the people
• The KL3M Data Project: Copyright-Clean Training Resources for Large Language Models
• LAION-5B: An open large-scale dataset for training next generation image-text models
• Alchemist: Turning Public Text-to-Image Data into Generative Gold