Can We Build a Monolithic Model for Fake Image Detection? SICA: Semantic-Induced Constrained Adaptation for Unified-Yet-Discriminative Artifact Feature Space Reconstruction
Official repository for the paper: Can We Build a Monolithic Model for Fake Image Detection? SICA: Semantic-Induced Constrained Adaptation for Unified-Yet-Discriminative Artifact Feature Space Reconstruction.
We propose the first Fake Image Detection (covering 4 image forensic domains) monolithic paradigm, termed SICA, and the first comprehensive dataset, OpenMMSec, designed for FID. The full code will be released upon paper acceptance. If you are interested in the dataset, please feel free to contact me at 2500074187@qq.com.