Research map / Continual learning / Online and replay-based
Continual Face Forgery Detection via Historical Distribution Preserving
Ke Sun, Shen Chen, Taiping Yao, Xiaoshuai Sun, Shouhong Ding, Rongrong Ji
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 pseudo-forged entropy training with UAP samples to the baseline (ablation) | SFT · CFFD Protocol 1 | PRE ACC: 67.29 -> 74.33 | Improved |
| add pseudo-forged feature distillation on top of pseudo-forged training (ablation) | HDP · CFFD Protocol 1 | PRE ACC: 74.33 -> 77.39 | Improved |
| add real feature distillation alone to the baseline (ablation) | SFT · CFFD Protocol 1 | PRE ACC: 67.29 -> 81.19 | Improved |
| add real feature distillation to pseudo-forged training (ablation) | HDP · CFFD Protocol 1 | PRE ACC: 74.33 -> 90.71 | Improved |
| add pseudo-forged feature distillation on top of the other two terms (ablation) | HDP · CFFD Protocol 1 | PRE ACC: 90.71 -> 94.94 | Improved |
| lower UAP attack rate from 0.8 to 0.6 (ablation) | HDP · CFFD Protocol 1 | PRE ACC: 94.94 -> 90.33 | Hurt |
| raise UAP attack rate from 0.8 to 1.0 (ablation) | HDP · CFFD Protocol 1 | PRE ACC: 94.94 -> 93.67 | Hurt |
| add HDP to sequence fine-tuning | SFT · CFFD Protocol 1 | AVG ACC: 75.07 -> 93.32 | Improved |
| add a 50-image per-stage buffer to HDP | HDP · CFFD Protocol 2 | PRE ACC: 86.99 -> 88.57 | Improved |
| add a 50-image per-stage buffer to HDP | HDP · CFFD Protocol 3 | AVG ACC: 77.17 -> 80.15 | Improved |
| add HDP to MAT detector | MAT · CFFD Protocol 1 | AVG ACC: 76.81 -> 93.13 | Improved |
| add HDP to F3-Net detector | F3-Net · CFFD Protocol 1 | AVG ACC: 77.18 -> 91.49 | 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.
- iCaRL — iCaRL: Incremental Classifier and Representation Learning (CVPR 2017) · 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
- Stacking Brick by Brick: Aligned Feature Isolation for Incremental Face Forgery Detection (CVPR 2025 · similarity 0.89)
- DevFD : Developmental Face Forgery Detection by Learning Shared and Orthogonal LoRA Subspaces (NeurIPS 2025 · similarity 0.88)
- ResProto-FD: Visual-Language Residual Prototype Sets for Generalized Face Forgery Detection (AAAI 2026 · similarity 0.88)
- Unifying Locality of KANs and Feature Drift Compensation Projection for Data-Free Replay Based Continual Face Forgery Detection (AAAI 2026 · similarity 0.88)
- Face Forgery Video Detection via Temporal Forgery Cue Unraveling (CVPR 2025 · similarity 0.87)
- Identity-Aware Vision-Language Model for Explainable Face Forgery Detection (AAAI 2026 · similarity 0.86)
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.
