Research map / Image generation / Editing and style transfer
Face X-Ray for More General Face Forgery Detection
Check your idea against 108,967 papers
AI-extracted summary. Written by AI from the paper's title only (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 |
|---|---|---|---|
| Detect blending boundaries with face X-ray instead of classifying manipulation-specific artifacts | Face X-ray detector · FF++ unseen manipulations | AUC: Xception trained on DF: F2F 75.05, FS 49.13, NT 80.39; Face X-ray trained on DF: F2F 94.14, FS 75.34, NT 93.85 | Helped |
| Add blended-image training samples composited from real images | Face X-ray detector · FF++ unseen manipulations | AUC: DF training: F2F 94.14 -> 97.64; FS 75.34 -> 98.00; NT 93.85 -> 97.77 | Helped |
| Train with BI only, without manipulation-generated fake images | Face X-ray detector · DF, F2F, FS, NT | AUC: 99.17%, 98.57%, 98.21%, 98.13% | Helped |
| Add FF++ training data alongside BI | Face X-ray detector · DFD, DFDC, Celeb-DF | AUC: DFD 93.47 -> 95.40; DFDC 71.15 -> 80.92; Celeb-DF 74.76 -> 80.58 | Helped |
| Add mask deformation when generating blended training samples (ablation) | Face X-ray detector · FF++, DFD | AUC: FF++ 93.92 -> 98.52; DFD 85.89 -> 93.47 | Helped |
| Add color correction when generating blended training samples (ablation) | Face X-ray detector · FF++, DFD | AUC: FF++ 96.21 -> 98.52; DFD 89.91 -> 93.47 | Helped |
| Add face X-ray prediction loss (λ=1 rather than λ=0) (ablation) | Face X-ray detector · FF++ | AUC: 94.5% -> 97.9% | Helped |
| Vary λ among 10, 100, and 1000 (ablation) | Face X-ray detector · FF++ | AUC: 98.4%, 98.5%, 98.6% | No clear effect |
| Use alpha-blending training data and evaluate on Poisson blending | Face X-ray detector · Poisson-blended test data | AUC: Alpha blending 99.46; Possion blending 94.62 | Hurt |
| Evaluate on heavily compressed LQ rather than lightly compressed HQ images | Face X-ray detector · FF++ | AUC: 87.35% -> 61.6% | Hurt |
Closest papers by meaning
- FakeDiffer: Distributional Disparity Learning on Differentiated Reconstruction for Face Forgery Detection (AAAI 2025 · similarity 0.85)
- Forensics Adapter: Adapting CLIP for Generalizable Face Forgery Detection (CVPR 2025 · similarity 0.84)
- DevFD : Developmental Face Forgery Detection by Learning Shared and Orthogonal LoRA Subspaces (NeurIPS 2025 · similarity 0.84)
- From Specificity to Generality: Revisiting Generalizable Artifacts in Detecting Face Deepfakes (NeurIPS 2025 · similarity 0.84)
- Critical Forgetting-Based Multi-Scale Disentanglement for Deepfake Detection (AAAI 2025 · similarity 0.83)
- Stacking Brick by Brick: Aligned Feature Isolation for Incremental Face Forgery Detection (CVPR 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.