AI-Generated Images

15 different generative models. Specifically, BigGAN \cite{BigGAN}, ProGAN \cite{progan}, StyleGAN \cite{karras2}, and WhichFaceIsReal (WFIR) \cite{DavidWFIR} are contributed by CNNSpot \cite{wang24}; StarGAN \cite{choi1}, HiSD \cite{li3}, LDM \cite{LDM}, and DDIM \cite{DDIM} are contributed by HiFi-IFDL \cite{guo22}; DALL-E 2 \cite{ramesh10}, VQDM \cite{gu4}, ADM \cite{ADM}, Midjourney \cite{David9}, Stable Diffusion v1.4 and v1.5 \cite{SD} are provided by GenImage \cite{zhu2024genimage}. Note that GLIDE \cite{nichol5} during training is provided by HiFi-IFDL \cite{guo22} and during testing the data is provided by GenImage \cite{zhu2024genimage}.