FUSE: Hybrid Approach for AI-Generated Image Detection

This paper introduces FUSE, a novel approach to detect AI-generated images by combining spectral and semantic features. The method's strength lies in its ability to generalize across different generative models, as demonstrated by strong performance on various datasets, including the challenging Chameleon benchmark. The integration of spectral and semantic information offers a more robust solution compared to existing methods that often struggle with high-fidelity images.

• •FUSE combines spectral (Fast Fourier Transform) and semantic (CLIP Vision encoder) features.
• •The method is trained in two stages.
• •Demonstrates strong generalization across multiple AI image generators.
• •Achieves state-of-the-art results on the Chameleon benchmark.