Ghosts Beneath Textures: Texture-Relation Cues for Cross-Paradigm AI-Generated Image Detection

2026-07-04

AI Summary

Existing methods for detecting AI-generated images exhibit limited generalization across free-form and conditional generation paradigms. To address this, this work introduces the ConImageGen benchmark to expose the failure of cross-paradigm detection and proposes a novel, artifact-agnostic detection paradigm based on modeling semantic-irrelevant local-to-global texture relationships. By leveraging semantic interference suppression to extract intrinsic texture relational features, the authors develop the DTS-Det framework, which achieves 99.6% accuracy on ConImageGen—surpassing baseline methods by 10.5%. Furthermore, it demonstrates strong cross-dataset generalization, attaining 93.2% and 94.1% accuracy on PicoBanana and RAID, respectively, while maintaining over 88.1% detection rates under image reconstruction and black-box adversarial attacks.

Abstract

AI-generated images have proliferated rapidly, motivating extensive research. Most existing AI-generated image detectors are developed and evaluated under image-free generation paradigms, such as noise-based or text-guided generation. However, image-conditioned generation has become increasingly important in practical applications, as it enables more fine-grained control over generated content. Detecting AI-generated images across these two paradigms creates a critical cross-paradigm detection problem that has long been overlooked. To study this problem, we construct ConImageGen, a benchmark for cross-paradigm AI-generated image detection. Evaluations on ConImageGen show that existing detectors fail to generalize reliably across image-free and image-conditioned generation. To address this failure, this paper identifies a cross-paradigm forensic cue and provides a new perspective for generalized AI-generated image detection. Specifically, by suppressing semantic interference, we visualize, for the first time, semantics-irrelevant texture patterns across generation paradigms. These patterns exhibit structured local-global texture relations, indicating a generalizable form of forensic evidence. Motivated by this finding, we shift the focus from directly exploiting explicit artifacts to modeling texture relations and propose DTS-Det, a detection framework that captures and leverages such relations for generalized AI-generated image detection. Extensive experiments validate the effectiveness of our method. DTS-Det achieves state-of-the-art performance across diverse evaluation settings, reaching 99.6% ACC on ConImageGen with a 10.5% gain over the best baseline. It also achieves 93.2%/94.1% ACC in cross-dataset evaluation on PicoBanana/RAID and maintains detection rates of 95.2%/88.1% under reconstruction attacks and black-box adversarial attacks, respectively.

Problem

• AI-generated image detection
• cross-paradigm
• image-conditioned generation
• texture relations
• forensic cue

Innovation

• cross-paradigm detection
• texture-relation cues
• AI-generated image forensics
• generalizable detection
• DTS-Det