冻结的DINO无需定位器即可定位图像编辑
Frozen DINO Localizes Image Edits Without a Localizer
- St Paul’s School London（圣保罗学校伦敦分校）
- Imperial College London（帝国理工学院）
- FAU（弗里德里希-亚历山大-埃尔兰根-纽伦堡大学）
机构由 AI 辅助整理，请以论文原文为准。
中文总结 AI 辅助
本文提出无需训练的TRAIL方法，利用冻结DINO的补丁标记漂移定位图像编辑，在CocoGlide等数据集上取得优于多数基准的定位性能，验证了后期层定位信号的有效性。
中文摘要 AI 辅助
局部图像编辑可在保留照片大部分真实性的同时改变其含义，因此取证分析必须识别编辑发生的位置。本文提出一种无需训练的AI图像编辑定位方法TRAIL（基于补丁标记漂移的AI图像编辑定位），该方法应用一次全局Haar扰动并映射对应补丁标记间的余弦漂移。在80个源不相交的CocoGlide测试图像上，TRAIL的补丁级AUROC达0.903，而受掩码监督的Detective SAM为0.912；固定阈值Dice系数为0.619，Detective SAM为0.709，采用最优阈值可将TRAIL提升至0.790。将TRAIL直接迁移至泊松图像插值任务，其AUROC达0.855，对比基准为0.864，表明该线索在无生成器时仍存在。在16个DINO编码器中，最优块位于归一化深度0.80-0.94处；全局上下文至关重要，当采用画布内局部扰动时，AUROC从全局的0.903降至0.857，独立编码裁剪时则降至0.735。因此，冻结的DINO补丁标记包含强的后期层定位信号，其可见性取决于扰动和保留的上下文。代码：this https URL
英文摘要
Localized image edits can change a photograph's meaning while leaving most of it authentic, so forensic analysis must identify where an edit occurred. We show that patch-level perturbation responses from frozen DINO encoders are themselves localization maps. Training-free Localization of AI-image Edits from patch-token Drift (TRAIL) applies one global Haar perturbation and maps cosine drift between corresponding patch tokens. On 80 source-disjoint CocoGlide test images, TRAIL reaches .903 patch AUROC versus .912 for the mask-supervised Detective SAM; fixed-threshold Dice is .619 versus .709, while an oracle threshold raises TRAIL to .790. Transferred unchanged to Poisson image interpolation, TRAIL reaches .855 AUROC versus .864, showing that the cue persists without a generator. Across sixteen DINO encoders, the best block lies at normalized depth .80-.94. Global context matters: AUROC falls from .903 globally to .857 for local-in-canvas perturbations and .735 for independently encoded crops. Frozen DINO patch tokens therefore contain a strong late-layer localization signal whose visibility depends on the perturbation and preserved context. Code: https://github.com/VishalJ99/trail-image-edit-localization.
