Research map / Detection and segmentation / Open-set and few-shot detection
A Sanity Check for Multi-In-Domain Face Forgery Detection in the Real World
Jikang Cheng, Renye Yan, Zhiyuan Yan, Yaozhong Gan, Xueyi Zhang, Zhongyuan Wang, Wei Peng, Ling Liang
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 FFDev without fine-tuning (ablation) | Effnb4 · MID-FFD P1 | M-ACC: 0.7624 -> 0.7823; cross 0.6826 -> 0.5735 | Mixed |
| add fixed-dose fine-tuning to FFDev (ablation) | Effnb4 · MID-FFD P1 | M-ACC: 0.7823 -> 0.8526; cross 0.6826 -> 0.5851 | Mixed |
| add parallel adaptive-dose training (AdaD-P) (ablation) | Effnb4+FFDev · MID-FFD P1 | M-ACC: 0.7624 -> 0.8130; cross 0.6826 -> 0.6341 | Mixed |
| train FFDev and adaptive-dose detector sequentially (DAFT-S) (ablation) | Effnb4+FFDev · MID-FFD P1 | M-ACC: 0.8130 -> 0.8676; cross 0.6341 -> 0.6896 | Improved |
| add adaptive dose instead of fixed dose (ablation) | Effnb4+FFDev · Cross-dataset | mean ACC: 0.5851 -> 0.6896 | Improved |
| select HF+ER samples instead of HF-only (small set) (ablation) | FFDev training · MID-FFD P1 | Avg ACC: 0.7708 -> 0.8675 | Improved |
| select HF+ER instead of HF+HR (large set) (ablation) | FFDev training · MID-FFD P1 | Avg ACC: 0.8051 -> 0.8809 | Improved |
| select HF+ER instead of All samples (large set) (ablation) | FFDev training · MID-FFD P1 | Avg ACC: 0.8405 -> 0.8809 | Improved |
| increase FFDev sample set from small (5000) to large (20000) for HF+ER (ablation) | FFDev training · MID-FFD P1 | Avg ACC: 0.8675 -> 0.8809 | No clear effect |
| apply DevDet to Xception (ablation) | Xception · MID-FFD P1 / DF40 cross | ACC: FF++ 0.7764 -> 0.8783; CDF3 0.6996 -> 0.6739 | Mixed |
| apply DevDet to Effort | Effort · MID-FFD P1 FF++ | ACC: 0.8274 -> 0.8931 | Improved |
| apply DevDet to Effort | Effort · Cross-dataset DF40 | ACC: 0.8051 -> 0.7935 | No clear effect |
Closest papers by meaning
- DF40: Toward Next-Generation Deepfake Detection (NeurIPS 2024 · similarity 0.92)
- Rethinking Vision-Language Model in Face Forensics: Multi-Modal Interpretable Forged Face Detector (CVPR 2025 · similarity 0.91)
- FreqBlender: Enhancing DeepFake Detection by Blending Frequency Knowledge (NeurIPS 2024 · similarity 0.90)
- From Specificity to Generality: Revisiting Generalizable Artifacts in Detecting Face Deepfakes (NeurIPS 2025 · similarity 0.90)
- FakeDiffer: Distributional Disparity Learning on Differentiated Reconstruction for Face Forgery Detection (AAAI 2025 · similarity 0.89)
- Towards More General Video-based Deepfake Detection through Facial Component Guided Adaptation for Foundation Model (CVPR 2025 · similarity 0.89)
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.
