Research map / Image generation / Autoregressive and latent generation
APT: Anchor-aligned Perturbations for Tamper Localization in Fully Regenerated Images
Suhyeon Ha, Woo Jae Kim, Joonsung Jeon, Sooel Son, Sung-eui Yoon
Read the paperCheck your idea against 137,217 papers
AI-extracted. The measured effects below were read by AI from the paper's full text (no summary yet). Numbers and claims can be wrong or missing context, so check the paper before you cite or build on it.
Measured effects reported in the paper
| What was changed | Where | Measured | Effect |
|---|---|---|---|
| add noisy perturbation branch loss (ablation) | APT · COCO FR | IoU: +0.24 | Improved |
| add noisy perturbation branch loss (ablation) | APT · COCO SP | IoU: +0.23 | Improved |
| add hard negative mining loss (ablation) | APT · COCO FR | IoU: +0.26 | Improved |
| add hard negative mining loss (ablation) | APT · COCO SP | IoU: +0.29 | Improved |
| add shallow mask decoder to training-free prediction | APT · COCO 500 images FR | IoU: 0.89 -> 0.91 | Improved |
| add shallow mask decoder under corruptions | APT · Corruption average FR | IoU: 0.80 -> 0.90 | Improved |
| use encoder layer 1 instead of layers 0, 2, 3 (ablation) | APT · FR center-mask SD-Painter | cosine similarity foreground/background separation: 0.124 vs 0.265 | Improved |
| reduce optimization steps from 150 to 50 (ablation) | APT* · COCO FR and SP | IoU: 42.52 s -> 14.34 s per image | No clear effect |
| tighten perturbation budget epsilon (ablation) | APT* · COCO | IoU: modest localization cost for better quality | Mixed |
| embed latent perturbation (APT) instead of pixel/frequency embedding | APT vs StableGuard/OmniGuard · COCO FR SD-Painter | AUC: 0.5 -> 0.97 | Improved |
| apply APT* to non-identical VAE and flow-matching inpainters | APT* · FR transfer | IoU: IoU >= 0.91 | No clear effect |
Earlier methods this paper compares with
Named baselines from earlier accepted papers in this paper's results tables: better than 2 of 2 on most shared results.
- OmniGuard — OmniGuard: Hybrid Manipulation Localization via Augmented Versatile Deep Image Watermarking (CVPR 2025) · better on 5 of 6 shared results
- StableGuard — StableGuard: Towards Unified Copyright Protection and Tamper Localization in Latent Diffusion Models (NeurIPS 2025) · better on 5 of 6 shared results
Matched by method name in results tables read by AI. Authors' own tables tend to favour the newer method, and a comparison can be missed when a method is renamed.
Closest papers by meaning
- Less Supervision, Better Generalization: Weakly Supervised Fake Region Localization in Diffusion-Edited Images (NeurIPS 2026 · similarity 0.86)
- Toward Real-world Text Image Forgery Localization: Structured and Interpretable Data Synthesis (NeurIPS 2025 · similarity 0.85)
- FakeShield: Explainable Image Forgery Detection and Localization via Multi-modal Large Language Models (ICLR 2025 · similarity 0.85)
- Training-Free Image Manipulation Localization Using Diffusion Models (AAAI 2025 · similarity 0.84)
- Learning Discriminative Noise Guidance for Image Forgery Detection and Localization (AAAI 2024 · similarity 0.84)
- ForensicHub: A Unified Benchmark & Codebase for All-Domain Fake Image Detection and Localization (NeurIPS 2025 · similarity 0.84)
Writing a paper in this area? scime draft is a free LaTeX editor with AI included, real-time co-authors and a check of your references and venue rules before you submit. It also imports OpenAI Prism projects.
