Search papers, labs, and topics across Lattice.
Affiliation:
3
0
3
Foundation models exploit subtle low-to-mid frequency cues to distinguish real images from diffusion-generated ones, challenging assumptions about semantic reliance in detection.
Color statistics expose a critical vulnerability in AI-generated images, enabling a 93.27% accuracy in distinguishing them from real ones.
Detecting AI-generated images in the wild gets a boost: by considering the quality of near-duplicate versions circulating online, detection accuracy improves by 8% over single-image methods.
