Research map / Image generation / Autoregressive and latent generation
Pixel-Inconsistency Modeling for Image Manipulation Localization
Chenqi Kong, Anwei Luo, Shiqi Wang, Haoliang Li, Anderson De Rezende Rocha, Alex Chichung Kot
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 boundary decoder (ablation) | PIM backbone · 11 unseen datasets (trained on CASIAv2), average | avg F1: .211 -> .220 | Improved |
| add regular data augmentation (RDA) (ablation) | PIM backbone + boundary decoder · 11 unseen datasets (trained on CASIAv2), average | avg F1: .220 -> .233 | Improved |
| add PIDA instead of RDA (ablation) | PIM backbone + boundary decoder · 11 unseen datasets (trained on CASIAv2), average | avg F1: .233 -> .260 | Improved |
| add PIDA on top of RDA (ablation) | PIM backbone · 11 unseen datasets (trained on CASIAv2), average | avg F1: .260 -> .283 | Improved |
| add central difference convolution (CDC) (ablation) | PIM · 11 unseen datasets (trained on CASIAv2), average | avg F1: .283 -> .304 | Improved |
| add radial difference convolution (RDC) alongside CDC (ablation) | PIM · 11 unseen datasets (trained on CASIAv2), average | avg F1: .304 -> .312 | Improved |
| add Learning-to-Weight Module (ablation) | PIM · 11 unseen datasets (trained on CASIAv2), average | avg F1: .312 -> .317 | Improved |
| add compactness loss (ablation) | PIM · 11 unseen datasets (trained on CASIAv2), average | avg IoU: .271 -> .269 | No clear effect |
| add Global Pixel Dependency Encoder (masked attention) (ablation) | PIM · 11 unseen datasets (trained on CASIAv2), average | avg F1: .323 -> .330 | Improved |
| add image reconstruction loss (ablation) | PIM · 11 unseen datasets (trained on CASIAv2), average | avg F1: .330 -> .333 | Improved |
| multi-head instead of single-head attention in global encoder (ablation) | PIM · 11 unseen datasets (trained on CASIAv2), average | avg F1: .312 -> .333 | Improved |
| multi-head instead of single-head attention in local encoder (ablation) | PIM · 11 unseen datasets (trained on CASIAv2), average | avg F1: .318 -> .333 | Improved |
Closest papers by meaning
- DG-Force: Disentangling and Gathering Forensic Cues is Needed for Image Manipulation Localization (ECCV 2026 · similarity 0.91)
- Learning Discriminative Noise Guidance for Image Forgery Detection and Localization (AAAI 2024 · similarity 0.89)
- FakeShield: Explainable Image Forgery Detection and Localization via Multi-modal Large Language Models (ICLR 2025 · similarity 0.88)
- Towards Modern Image Manipulation Localization: A Large-Scale Dataset and Novel Methods (CVPR 2024 · similarity 0.88)
- ForensicHub: A Unified Benchmark & Codebase for All-Domain Fake Image Detection and Localization (NeurIPS 2025 · similarity 0.87)
- Can We Model the Artifacts Explicitly? Disentangle Artifacts via Pairwise Edit Relations for Image Manipulation Localization (NeurIPS 2026 · similarity 0.87)
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.
