Md Redwanul Haque

FAIR: Feature-Augmented Implicit Regularization for AI-generated Fake Image Detection

Md Redwanul Haque, Manzur Murshed, Manoranjan Paul, Tsz-Kwan Lee

Generalization remains a critical bottleneck in AI-generated image detection. Because many modern generators are proprietary or adversarially modified, existing detectors overfit to the low-level textural patterns of accessible training data, resulting in severe failures on unsee…Read more