Research map / Label-efficient and robust learning / Domain adaptation and generalization
Self-Supervised Domain Adaptation for Forgery Localization of JPEG Compressed Images
Yuan Rao, Jiangqun Ni
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 ICRF-Att attention module (ablation) | backbone network · Nim.16, DSO-1, Wild Web | MCC/F1 | Improved |
| add domain adaptation with JPEG-agent images (DA-Ia) (ablation) | backbone with ICRF-Att · Nim.16 | MCC: 0.351 -> 0.391 | Improved |
| add domain adaptation with JPEG-agent images (DA-Ia) (ablation) | backbone with ICRF-Att · DSO-1 | MCC: 0.805 -> 0.822 | Improved |
| use JPEG-agent images instead of real JPEG as DA target (ablation) | DA strategy · Nim.16, DSO-1, Wild Web | MCC: 3.2%, 2.7% and 1.4% | Improved |
| add real JPEG images as DA target (DA-Ic) (ablation) | backbone with ICRF-Att · Nim.16 | MCC: 0.351 -> 0.359 | Improved |
| mix JPEG-agent images into training without DA (ablation) | backbone with ICRF-Att · Nim.16 | MCC: 0.351 -> 0.333 | Hurt |
| mix JPEG-agent images instead of real JPEG images into training without DA (ablation) | backbone with ICRF-Att · Nim.16 | MCC: 0.256 -> 0.333 | Improved |
| mix JPEG-agent images into training without DA (ablation) | backbone with ICRF-Att · DSO-1 | MCC: 0.805 -> 0.774 | Hurt |
| use L2 domain loss rather than L1 or MMD (ablation) | DA strategy · Nim.16, DSO-1, Wild Web | MCC/F1 | Improved |
| use alpha = 1 for domain loss weight (ablation) | joint loss · IFC training data | localization | Improved |
| method vs second best Noiseprint | Proposed · 7 datasets average | MCC, F1, AP: 5.3%, 5.6% and 8.6% | Improved |
| lower JPEG quality QF=70 to QF=50 | Proposed · Nim.16 | F1: 0.317 -> 0.280 | Hurt |
Closest papers by meaning
- FakeShield: Explainable Image Forgery Detection and Localization via Multi-modal Large Language Models (ICLR 2025 · similarity 0.84)
- MUN: Image Forgery Localization Based on M³ Encoder and UN Decoder (AAAI 2025 · similarity 0.84)
- M²RL-Net: Multi-View and Multi-Level Relation Learning Network for Weakly-Supervised Image Forgery Detection (AAAI 2025 · similarity 0.84)
- Learning Discriminative Noise Guidance for Image Forgery Detection and Localization (AAAI 2024 · similarity 0.84)
- CatmullRom Splines-Based Regression for Image Forgery Localization (AAAI 2024 · similarity 0.83)
- Reinforced Multi-teacher Knowledge Distillation for Efficient General Image Forgery Detection and Localization (AAAI 2025 · similarity 0.83)
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.
