Research map / Detection and segmentation / CNN features
DiffusionFF: A Diffusion-based Framework for Joint Face Forgery Detection and Fine-Grained Artifact Localization
Siran Peng, Haoyuan Zhang, Li Gao, Tianshuo Zhang, Xiangyu Zhu, Bao Li, Weisong Zhao, Zhen Lei
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 DiffusionFF to EfficientNet-B4 detector | EfficientNet-B4 · DFDC (cross-dataset) | AUC (%): 72.42 -> 82.19 | Improved |
| add DiffusionFF to ConvNeXt-B detector | ConvNeXt-B · CDF2 (cross-dataset) | AUC (%): 96.08 -> 97.24 | Improved |
| add DiffusionFF to Swin-B detector | Swin-B · DFDC (cross-dataset) | AUC (%): 82.81 -> 84.54 | Improved |
| use diffusion instead of direct regression for DSSIM estimation (ablation) | DiffusionFF · FF++ DSSIM estimation | FID: 172.213 -> 43.093 | Improved |
| run diffusion at pixel level instead of latent space (ablation) | DiffusionFF · FF++ DSSIM estimation | FID: 56.384 -> 43.093 | Improved |
| multi-stage conditioning instead of final-stage only (ablation) | DiffusionFF · FF++ DSSIM estimation | FID: 43.812 -> 43.093 | Improved |
| encoder-based conditioning instead of decoder conditioning (ablation) | DiffusionFF · FF++ DSSIM estimation | FID: 46.329 -> 43.093 | Improved |
| gating fusion instead of addition (ablation) | DiffusionFF · Cross-dataset average AUC | AUC (%): 90.10 -> 90.85 | Improved |
| gating fusion instead of cross-attention (ablation) | DiffusionFF · Cross-dataset average AUC | AUC (%): 90.71 -> 90.85 | Improved |
| DiffusionFF vs existing DSSIM regression methods for map quality | LiSiam / LRL · CDF2 DSSIM estimation | FID: 256.201 -> 98.982 | Improved |
Earlier methods this paper compares with
Named baselines from earlier accepted papers in this paper's results tables: better than 1 of 1 on most shared results.
- AUNet — AUNet: Learning Relations Between Action Units for Face Forgery Detection (CVPR 2023) · better on 1 of 1 shared result
Matched by method name in results tables read by AI. Authors' own tables tend to favour the newer method, and a comparison can be missed when a method is renamed.
Closest papers by meaning
- Rethinking Vision-Language Model in Face Forensics: Multi-Modal Interpretable Forged Face Detector (CVPR 2025 · similarity 0.90)
- DiffusionFake: Enhancing Generalization in Deepfake Detection via Guided Stable Diffusion (NeurIPS 2024 · similarity 0.90)
- DiffForensics: Leveraging Diffusion Prior to Image Forgery Detection and Localization (CVPR 2024 · similarity 0.89)
- FakeShield: Explainable Image Forgery Detection and Localization via Multi-modal Large Language Models (ICLR 2025 · similarity 0.89)
- DF40: Toward Next-Generation Deepfake Detection (NeurIPS 2024 · similarity 0.88)
- VLForgery Face Triad: Detection, Localization and Attribution via Multimodal Large Language Models (NeurIPS 2025 · similarity 0.88)
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.
