Aggregating Diverse Cue Experts for AI-Generated Image Detection

TL;DR

The paper tackles the generalization gap in AI-generated image detection by introducing MCAN, a unified framework that fuses three complementary cues—raw image content, high-frequency details, and Chromaticity Inconsistency (CI)—through a Mixture-of-Encoder Adapter. By leveraging a frozen CLIP ViT-B/16 backbone, applying discrete wavelet transforms for frequency cues, and employing chromaticity-based normalization to highlight noise, MCAN achieves state-of-the-art performance on GenImage, Chameleon, and UniversalFakeDetect benchmarks, with notable gains in cross-model settings. Key contributions include the CI representation, the MoEA-based multi-cue fusion, and comprehensive ablations and visualizations demonstrating improved generalization and robustness. The approach offers a practical impact by delivering more reliable detectors across diverse generative models and can inform future design of multi-cue, transformer-backed detection systems.

Abstract

The rapid emergence of image synthesis models poses challenges to the generalization of AI-generated image detectors. However, existing methods often rely on model-specific features, leading to overfitting and poor generalization. In this paper, we introduce the Multi-Cue Aggregation Network (MCAN), a novel framework that integrates different yet complementary cues in a unified network. MCAN employs a mixture-of-encoders adapter to dynamically process these cues, enabling more adaptive and robust feature representation. Our cues include the input image itself, which represents the overall content, and high-frequency components that emphasize edge details. Additionally, we introduce a Chromatic Inconsistency (CI) cue, which normalizes intensity values and captures noise information introduced during the image acquisition process in real images, making these noise patterns more distinguishable from those in AI-generated content. Unlike prior methods, MCAN's novelty lies in its unified multi-cue aggregation framework, which integrates spatial, frequency-domain, and chromaticity-based information for enhanced representation learning. These cues are intrinsically more indicative of real images, enhancing cross-model generalization. Extensive experiments on the GenImage, Chameleon, and UniversalFakeDetect benchmark validate the state-of-the-art performance of MCAN. In the GenImage dataset, MCAN outperforms the best state-of-the-art method by up to 7.4% in average ACC across eight different image generators.

Figures (5)

• Figure 1: (a) Motivation for MCAN: In the top panel, yellow, green, and red represent the three generated images; Using multiple cues: image, frequency, and chromaticity enhances AI-generated image detection by leveraging the cues' complementary strengths. (b) Motivation for Chromaticity Inconsistency: In the bottom panel, real images show chromaticity inconsistencies due to noise, while fake images appear smoother with uniform chromaticity.
• Figure 2: Overall architecture: MCAN combines image representation, high-frequency representation, and the novel chromaticity inconsistency as three distinct cues. To effectively integrate these cues, MCAN uses a mixture of encoder adapters that adapt efficiently to each cue's representation.
• Figure 3: Visualization of Chromaticity Inconsistency (CI): Real images, affected by noise, show weaker consistency in CI images compared to generated images.
• Figure 4: Performance of MCAN under structures: Optimal reaches when MoEA contains 4 experts in the last 4 blocks.
• Figure 5: Visualization of learned feature space of MCAN (a) and classification results for generated images under different cues (b). (a) While all cues face generalization challenges, MCAN improves generalization by leveraging complementary features across cues. (b) To demonstrate the complementary nature of CI to 'Img' and 'HF', we specifically select failure cases for 'Img' and 'HF'. The values shown indicate the confidence scores from each cue, representing the likelihood that the corresponding image is classified as real.