Research map / Label-efficient and robust learning / Domain adaptation and generalization
SARIF: Segment Anything for Robust Image Forensics
Dong-Hyun Moon, Ju-Hyeon Nam, Sang-Chul Lee
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 LoRA adapters to the SAM image encoder (ablation) | vanilla SAM · CASIAv2-trained, unseen average | DSC: 23.3 -> 33.0 | Improved |
| add FSIE (dual-encoder residual cues) on top of adapters (ablation) | SAM + Adapter · CASIAv2-trained, unseen average | DSC: 33.0 -> 44.7 | Improved |
| add FGMD (feedback mask decoder) on top of adapters (ablation) | SAM + Adapter · CASIAv2-trained, unseen average | DSC: 33.0 -> 42.6 | Improved |
| add FGMD to adapter plus FSIE (ablation) | SAM + Adapter + FSIE · CASIAv2-trained, unseen average | DSC: 44.7 -> 57.9 | Improved |
| add FSIE to adapter plus FGMD (ablation) | SAM + Adapter + FGMD · CASIAv2-trained, unseen average | DSC: 42.6 -> 57.9 | Improved |
| add FGMD to adapter plus FSIE (ablation) | SAM + Adapter + FSIE · CASIAv2 seen domain | DSC: 62.4 -> 63.1 | No clear effect |
| add FSIE to adapter (ablation) | SAM + Adapter · FLOPs | FLOPs(G): 503.79 -> 994.88 | Hurt |
| progressive feedback refinement from initial mask to final mask (ablation) | SARIF · CASIAv2 seen | IoU/DSC: 53.3/61.9 -> 56.7/63.1 | Improved |
| progressive feedback refinement from initial mask to final mask (ablation) | SARIF · IMD2020 | IoU/DSC: 38.2/46.6 -> 40.4/48.4 | Improved |
| use SAM-based SARIF instead of M2SFormer | SARIF · CocoGlide | DSC: 12.2 -> 34.0 | Improved |
| use SARIF instead of SAFIRE | SARIF · Unseen datasets DSC | DSC: 65.4 vs 39.5 on CoMoFoD; 48.4 vs 53.1 on IMD2020 | Mixed |
| use SARIF instead of FBINet on authentic images | SARIF · CASIAv2 authentic | FPR: 3.0 -> 5.1 | Hurt |
Closest papers by meaning
- SAFIRE: Segment Any Forged Image Region (AAAI 2025 · similarity 0.85)
- A Lottery Ticket Hypothesis Approach with Sparse Fine-tuning and MAE for Image Forgery Detection and Localization (AAAI 2025 · similarity 0.85)
- ForensicHub: A Unified Benchmark & Codebase for All-Domain Fake Image Detection and Localization (NeurIPS 2025 · similarity 0.84)
- Learning Discriminative Noise Guidance for Image Forgery Detection and Localization (AAAI 2024 · similarity 0.84)
- SUMI-IFL: An Information-Theoretic Framework for Image Forgery Localization with Sufficiency and Minimality Constraints (AAAI 2025 · similarity 0.84)
- Forensics-Bench: A Comprehensive Forgery Detection Benchmark Suite for Large Vision Language Models (CVPR 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.
