CGI-DM: Digital Copyright Authentication for Diffusion Models via Contrasting Gradient Inversion
Published in CVPR 2024, 2024
CGI-DM removes part of an input image and recovers it by maximizing the conceptual difference between pretrained and fine-tuned diffusion models. Similarity between the original and recovered images provides an interpretable signal for potential infringement.
Recommended citation: Wu, X., Hua, Y., Liang, C., et al. (2024). "CGI-DM: Digital Copyright Authentication for Diffusion Models via Contrasting Gradient Inversion." CVPR 2024.
Download Paper