Research map / Detection and segmentation / Open-set and few-shot detection
DeepFake Detection Based on Discrepancies Between Faces and Their Context
Yuval Nirkin, Lior Wolf, Yosi Keller, Tal Hassner
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 face-vs-context identity discrepancy to a binary Xception detector (ablation) | Binary XceptionNet · FaceSwap-3D unseen swap (fixed threshold) | total accuracy: 76.55 -> 81.65 | Improved |
| add face-vs-context identity discrepancy to a binary Xception detector (ablation) | Binary XceptionNet · FSGAN unseen swap (fixed threshold) | total accuracy: 59.74 -> 68.85 | Improved |
| keep identity networks frozen instead of end-to-end training (ablation) | Ours · FSGAN (fixed threshold) | total accuracy: 63.52 -> 68.85 | Improved |
| keep identity networks frozen instead of end-to-end training (ablation) | Ours · 3D-based swap (fixed threshold) | total accuracy: 76.22 -> 81.65 | Improved |
| keep identity networks frozen instead of end-to-end training (optimal threshold) (ablation) | Ours · FSGAN (optimal threshold) | total accuracy: 75.49 -> 75.19 | No clear effect |
| use entire region instead of face only for identification (ablation) | Xception identification · VGGFace2 validation | accuracy: 95.10 -> 96.98 | Improved |
| context alone for identification vs face (ablation) | Xception identification · VGGFace2 validation | accuracy: 95.10 -> 87.06 | Hurt |
| add face-context discrepancy network to XceptionNet-c23 on Celeb-DF-v2 | XceptionNet · Celeb-DF-v2 | frame-level AUC: 65.3 -> 66.0 | Improved |
| add face-context discrepancy network on FF-DF | XceptionNet-c23 · FF-DF | frame-level AUC: 99.7 -> 99.7 | No clear effect |
| add face-context discrepancy network on FaceForensics++ | Xception · FaceForensics++ image benchmark | total accuracy: 71.0 -> 75.0 | Improved |
Closest papers by meaning
- DF40: Toward Next-Generation Deepfake Detection (NeurIPS 2024 · similarity 0.89)
- From Specificity to Generality: Revisiting Generalizable Artifacts in Detecting Face Deepfakes (NeurIPS 2025 · similarity 0.89)
- DiffusionFake: Enhancing Generalization in Deepfake Detection via Guided Stable Diffusion (NeurIPS 2024 · similarity 0.89)
- Guard Me If You Know Me: Protecting Specific Face-Identity from Deepfakes (NeurIPS 2025 · similarity 0.89)
- Rethinking Vision-Language Model in Face Forensics: Multi-Modal Interpretable Forged Face Detector (CVPR 2025 · similarity 0.87)
- FreqBlender: Enhancing DeepFake Detection by Blending Frequency Knowledge (NeurIPS 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.
