Research map / Detection and segmentation / CNN features
M2SFormer: Multi-Spectral and Multi-Scale Attention with Edge-Aware Difficulty Guidance for Image Forgery Localization
Ju-Hyeon Nam, Dong-Hyun Moon, 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 multi-scale attention to single-spectral M2S block (ablation) | M2SFormer · Unseen datasets avg (CASIAv2 training) | DSC: 27.1 -> 33.6 | Improved |
| add multi-spectral attention to single-scale M2S block (ablation) | M2SFormer · Unseen datasets avg (CASIAv2 training) | DSC: 27.1 -> 36.1 | Improved |
| combine multi-spectral and multi-scale attention (full M2S) (ablation) | M2SFormer · Unseen datasets avg (CASIAv2 training) | DSC: 36.1 -> 43.0 | Improved |
| add multi-scale attention on seen domain (ablation) | M2SFormer (single-spectral) · CASIAv2 (seen) | DSC: 56.3 -> 55.5 | Hurt |
| add difficulty-guided attention with edge-aware calculator (EADC) (ablation) | M2SFormer · Unseen datasets avg (CASIAv2 training) | DSC: 32.3 -> 43.0 | Improved |
| add difficulty-guided attention with edge-aware calculator (EADC) (ablation) | M2SFormer · CASIAv2 (seen) | DSC: 55.5 -> 58.8 | Improved |
| use simple difficulty calculator instead of edge-aware one (ablation) | M2SFormer DGA · Unseen datasets avg (CASIAv2 training) | DSC: 43.0 -> 30.8 | Hurt |
| add simple difficulty calculator + DGA over no DGA (ablation) | M2SFormer · Unseen datasets avg (CASIAv2 training) | DSC: 32.3 -> 30.8 | Hurt |
| PVT-v2-B2 backbone instead of ResNet50 (ablation) | M2SFormer · Unseen datasets avg (CASIAv2 training) | DSC: 35.6 -> 43.0 | Improved |
| PVT-v2-B2 backbone instead of MiT-B2 (ablation) | M2SFormer · Unseen datasets avg (CASIAv2 training) | DSC: 41.2 -> 43.0 | Improved |
| M2SFormer versus PIMNet on CASIAv2 training, DSC on Columbia | M2SFormer · Columbia, trained on CASIAv2 | DSC: 32.5 -> 42.4 | Improved |
| M2SFormer versus EITLNet on DIS25k training, seen DIS25k | M2SFormer · DIS25k (seen), trained on DIS25k | DSC: 90.6 -> 87.7 | Hurt |
Closest papers by meaning
- Pixel-Inconsistency Modeling for Image Manipulation Localization (TPAMI 2025 · similarity 0.90)
- FakeShield: Explainable Image Forgery Detection and Localization via Multi-modal Large Language Models (ICLR 2025 · similarity 0.89)
- Learning Discriminative Noise Guidance for Image Forgery Detection and Localization (AAAI 2024 · similarity 0.88)
- A Lottery Ticket Hypothesis Approach with Sparse Fine-tuning and MAE for Image Forgery Detection and Localization (AAAI 2025 · similarity 0.87)
- DiffForensics: Leveraging Diffusion Prior to Image Forgery Detection and Localization (CVPR 2024 · similarity 0.87)
- MGQFormer: Mask-Guided Query-Based Transformer for Image Manipulation Localization (AAAI 2024 · 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.
