Research map / Label-efficient and robust learning / Self-supervised and contrastive
Editprint: General Digital Image Forensics via Editing Fingerprint with Self-Augmentation Training
Haiwei Wu, Kemou Li, Yuanman Li, Jiantao 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 |
|---|---|---|---|
| train on simulated in- and out-camera editing chains with language-guided contrastive learning, instead of labeled camera data (ExifMeta) | self-supervised forensic features · SID open-set (mean over 10 generators) | AUC: .8539 -> .9625 | Improved |
| replace ExifMeta with Editprint for social network provenance | ExifMeta · SNP open-set, SDR (10 OSNs) | AUC: .5092 -> .8120 | Improved |
| replace ExifMeta with Editprint for generation attribution | ExifMeta · SID generation attribution (10 generators) | average F1: 0.8608 -> 0.9206 | Improved |
| Editprint trained without annotations versus supervised UniDet | Editprint · SID open-set (mean over 10 generators) | AUC: .9625 vs .9700 | No clear effect |
| Editprint (self-supervised) versus supervised SNP methods | Editprint · SNP open-set (mean AUC) | AUC: .8853 vs .8952 (Seq2Seq), .8986 (MultiClue) | No clear effect |
| Editprint versus ExifMeta for camera source identification | ExifMeta · CSI open-set (mean of 4 test sets) | AUC: .9211 -> .9199 | No clear effect |
| supervised SID methods versus Editprint on generation attribution | Editprint · SID generation attribution (10 generators) | F1: 0.9206 vs 0.9213 to 0.9529 (supervised) | No clear effect |
| Editprint on diffusion-model detection (SD and MJ) is weaker than on GANs | Editprint · SID open-set | AUC: SD .8821, MJ .8454 vs StarGAN .9999 | Mixed |
Closest papers by meaning
- Forensic Self-Descriptions Are All You Need for Zero-Shot Detection, Open-Set Source Attribution, and Clustering of AI-generated Images (CVPR 2025 · similarity 0.87)
- ForensicHub: A Unified Benchmark & Codebase for All-Domain Fake Image Detection and Localization (NeurIPS 2025 · similarity 0.85)
- Revisiting Tampered Scene Text Detection in the Era of Generative AI (AAAI 2025 · similarity 0.85)
- FakeShield: Explainable Image Forgery Detection and Localization via Multi-modal Large Language Models (ICLR 2025 · similarity 0.84)
- Forensics Adapter: Adapting CLIP for Generalizable Face Forgery Detection (CVPR 2025 · similarity 0.84)
- A Lottery Ticket Hypothesis Approach with Sparse Fine-tuning and MAE for Image Forgery Detection and Localization (AAAI 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.
