Diffusion Art or Digital Forgery? Investigating Data Replication in Diffusion Models
This addresses concerns about data replication in diffusion models for users in commercial art and graphic design, highlighting potential copyright and originality issues.
The study investigated whether diffusion models replicate content from their training data by developing image retrieval frameworks to compare generated images with training samples, finding cases where models like Stable Diffusion blatantly copy from datasets such as Oxford flowers, Celeb-A, ImageNet, and LAION.
Cutting-edge diffusion models produce images with high quality and customizability, enabling them to be used for commercial art and graphic design purposes. But do diffusion models create unique works of art, or are they replicating content directly from their training sets? In this work, we study image retrieval frameworks that enable us to compare generated images with training samples and detect when content has been replicated. Applying our frameworks to diffusion models trained on multiple datasets including Oxford flowers, Celeb-A, ImageNet, and LAION, we discuss how factors such as training set size impact rates of content replication. We also identify cases where diffusion models, including the popular Stable Diffusion model, blatantly copy from their training data.