Research map / Detection and segmentation / CNN features
UCF: Uncovering Common Features for Generalizable Deepfake Detection
Zhiyuan Yan, Yong Zhang, Yanbo Fan, Baoyuan Wu
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 disentanglement framework (content removal) to Xception (ablation) | Xception · CelebDF (train FF++) | AUC: 0.672 -> 0.785 | Improved |
| add multi-task learning (method-specific classification) to disentanglement (ablation) | UCF · CelebDF (train FF++) | AUC: 0.785 -> 0.804 | Improved |
| add contrastive regularization loss (ablation) | UCF · CelebDF (train FF++) | AUC: 0.804 -> 0.824 | Improved |
| use AdaIN conditional decoder instead of linear addition (ablation) | UCF · DFDC (train FF++) | AUC: 0.786 -> 0.805 | Improved |
| detect with common features only instead of whole forgery features (ablation) | UCF · CelebDF (train FF++) | AUC: 0.785 -> 0.824 | Improved |
| detect with method-specific features only (ablation) | UCF · CelebDF (train FF++) | AUC: 0.824 -> 0.681 | Hurt |
| apply framework to ConvNeXt backbone instead of Xception (ablation) | UCF · cross-dataset average (train FF++) | AUC: 0.808 -> 0.866 | Improved |
| disentanglement and multi-task components on within-dataset FF++ test (ablation) | UCF · FF++ (within-dataset) | AUC: 0.986 -> 0.996 | No clear effect |
Closest papers by meaning
- Exposing the Deception: Uncovering More Forgery Clues for Deepfake Detection (AAAI 2024 · similarity 0.93)
- Transcending Forgery Specificity with Latent Space Augmentation for Generalizable Deepfake Detection (CVPR 2024 · similarity 0.92)
- Critical Forgetting-Based Multi-Scale Disentanglement for Deepfake Detection (AAAI 2025 · similarity 0.91)
- Exploring Unbiased Deepfake Detection via Token-Level Shuffling and Mixing (AAAI 2025 · similarity 0.91)
- DF40: Toward Next-Generation Deepfake Detection (NeurIPS 2024 · similarity 0.91)
- Veritas: Generalizable Deepfake Detection via Pattern-Aware Reasoning (ICLR 2026 · similarity 0.90)
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.
