发表机构
Deakin University; Charles Sturt University（迪肯大学; 查尔斯·斯特尔特大学）
机构由 AI 辅助整理，请以论文原文为准。AI 中文总结
针对人工智能生成图像检测中泛化难的问题，提出FAIR方法，通过引入场景构图结构（SCS）特征增强隐式正则化，训练时约束模型优化轨迹，推理时丢弃结构先验，提升跨生成器泛化能力，在多基准测试中提高准确率。
AI 中文摘要
泛化仍然是人工智能生成图像检测中的一个关键瓶颈。由于许多现代生成器是专有的或经过对抗性修改的，现有检测器会过度拟合可访问训练数据的低级纹理模式，导致在未见领域出现严重失败。传统正则化技术（如\(L_1\)/\(L_2\)范数、随机失活）应用不加区分的参数约束，无法提供跨生成器鲁棒性所需的域不变结构。为解决此问题，我们提出特征增强隐式正则化（FAIR）。FAIR在训练期间引入正交的宏观结构先验，即场景构图结构（SCS），以几何方式约束模型的优化轨迹。通过用域不变的SCS特征增强主特征空间，FAIR明确惩罚基于纹理的捷径学习。关键的是，这种结构先验在推理时完全被丢弃，产生一个平滑的、广义的决策边界，且没有架构或计算开销。在五个大规模基准上的广泛评估表明，将FAIR集成到现有检测器中可显著提高跨生成器泛化能力，在零样本转移场景中准确率提高高达8.04%，并建立了新的最先进的鲁棒性。
英文摘要
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 unseen domains. Conventional regularization techniques (e.g., $L_1$/$L_2$ norms, Dropout) apply indiscriminate parametric constraints and fail to provide the domain-invariant structure necessary for cross-generator robustness. To address this, we propose Feature-Augmented Implicit Regularization (FAIR). FAIR introduces an orthogonal, macro-structural prior, specifically, Scene Composition Structure (SCS), during training to geometrically constrain the model's optimization trajectory. By augmenting the primary feature space with domain-invariant SCS features, FAIR explicitly penalizes texture-biased shortcut learning. Crucially, this structural prior is entirely discarded at inference, yielding a smoothed, generalized decision boundary with zero architectural or computational overhead. Extensive evaluations across five massive benchmarks demonstrate that integrating FAIR into state-of-the-art detectors significantly improves cross-generator generalization, boosting accuracy by up to 8.04% and establishing new state-of-the-art robustness in zero-shot transfer scenarios.
CommentsAccepted to ECCV 2026
