Research map / Detection and segmentation / CNN features
Patch Diffusion: A General Module for Face Manipulation Detection
Baogen Zhang, Sheng Li, Guorui Feng, Zhenxing Qian, Xinpeng Zhang
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 Patch Diffusion to Xception | Xception · FaceForensics++ LQ | AUC: 84.15 -> 89.99 | Improved |
| Add Patch Diffusion to WM-Wsdan-Effb | WM-Wsdan-Effb · FaceForensics++ HQ | AUC: 98.02 -> 98.51 | Improved |
| Add Patch Diffusion to Xception | Xception · FaceForensics++ cross-manipulation, trained on FS | AUC: 67.85 -> 83.49; 75.81 -> 93.27; 59.46 -> 82.12 | Improved |
| Add Patch Diffusion to WM-Wsdan-Effb | WM-Wsdan-Effb · DFD | AUC: 96.46 -> 97.38 | Improved |
| Add Patch Diffusion to WM-Wsdan-Effb | WM-Wsdan-Effb · Celeb-DFv1 | AUC: 75.15 -> 75.39 | Improved |
| Add Patch Diffusion to WM-Wsdan-Effb | WM-Wsdan-Effb · Celeb-DFv2 | AUC: 80.80 -> 82.25 | Improved |
| Add DPFL only, without AMP (ablation) | Xception · FaceForensics++ LQ | ACC: 84.41 -> 85.10 | Improved |
| Add AMP only, without DPFL (ablation) | Xception · FaceForensics++ LQ | ACC: 84.41 -> 83.42 | Hurt |
| Add both DPFL and AMP (ablation) | Xception · FaceForensics++ LQ | ACC: 84.41 -> 86.37 | Improved |
| Use an explicit attention-based adjacent matrix instead of a learned one (ablation) | Xception+PD · FaceForensics++ LQ | AUC: 88.94 -> 89.99 | Improved |
| Weight patches using the proposed patch-importance scores instead of equal scores (ablation) | Xception+PD · FaceForensics++ LQ | ACC: 85.53 -> 86.37 | Improved |
Closest papers by meaning
- FaceForensics++: Learning to Detect Manipulated Facial Images (ICCV 2019 · similarity 0.86)
- On the Detection of Digital Face Manipulation (CVPR 2020 · similarity 0.86)
- DeepFake Detection Based on Discrepancies Between Faces and Their Context (TPAMI 2021 · similarity 0.85)
- PatchNet: A Simple Face Anti-Spoofing Framework via Fine-Grained Patch Recognition (CVPR 2022 · similarity 0.84)
- Thinking in Frequency: Face Forgery Detection by Mining Frequency-aware Clues (ECCV 2020 · similarity 0.84)
- Towards Reliable Identification of Diffusion-based Image Manipulations (NeurIPS 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.
