Research map / Detection and segmentation / CNN features
DG-Force: Disentangling and Gathering Forensic Cues is Needed for Image Manipulation Localization
Yong Yao, Zhenyu Cui, Lei Chen, Jiwen Lu, Jiahuan Zhou
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 PFD branch to SegFormer (ablation) | SegFormer · Ablation, avg over 4 datasets | AVG F1: 0.6479 -> 0.6632 | Improved |
| add EFD branch to SegFormer (ablation) | SegFormer · Ablation, avg over 4 datasets | AVG F1: 0.6479 -> 0.6619 | Improved |
| use both PFD and EFD with fixed equal weights (ablation) | SegFormer · Ablation, avg over 4 datasets | AVG F1: 0.6479 -> 0.6653 | No clear effect |
| add adaptive reweighting of patch and edge cues (ablation) | PFD+EFD · Ablation, avg over 4 datasets | AVG F1: 0.6653 -> 0.6882 | Improved |
| add Multi-scale Forensic Transfer (MFT) (ablation) | DG-Force · Ablation, avg over 4 datasets | AVG F1: 0.6882 -> 0.6973 | Improved |
| DG-Force vs strongest baseline Mesorch on CASIAv1 | DG-Force · CASIAv1 | pixel F1: 0.8398 -> 0.8063 | Hurt |
| DG-Force vs Mesorch on average localization | DG-Force · IML avg over 4 datasets | pixel F1: 0.6771 -> 0.6973 | Improved |
| DG-Force under degradations vs Mesorch | DG-Force · Robustness (rotation, noise, JPEG, scale, color) | pixel F1: rotation 0.5733 -> 0.6362 | Improved |
| DG-Force on unseen AI-generated tampering | DG-Force · AI tampering avg (CoCoGlide, AutoSplice) | F1* (optimal threshold): 0.7708 -> 0.8049 | Improved |
Closest papers by meaning
- ForensicHub: A Unified Benchmark & Codebase for All-Domain Fake Image Detection and Localization (NeurIPS 2025 · similarity 0.88)
- DiffForensics: Leveraging Diffusion Prior to Image Forgery Detection and Localization (CVPR 2024 · similarity 0.88)
- Training-Free Image Manipulation Localization Using Diffusion Models (AAAI 2025 · similarity 0.87)
- FakeShield: Explainable Image Forgery Detection and Localization via Multi-modal Large Language Models (ICLR 2025 · similarity 0.86)
- Learning Discriminative Noise Guidance for Image Forgery Detection and Localization (AAAI 2024 · similarity 0.86)
- Can We Model the Artifacts Explicitly? Disentangle Artifacts via Pairwise Edit Relations for Image Manipulation Localization (NeurIPS 2026 · similarity 0.86)
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.
