Research map / Detection and segmentation / Open-set and few-shot detection
DeepfakeBench: A Comprehensive Benchmark of Deepfake Detection
Zhiyuan Yan, Yong Zhang, Xinhang Yuan, Siwei Lyu, 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 |
|---|---|---|---|
| replace first conv of ResNet with depthwise separable convolution (ablation) | ResNet · average over 6 test sets | AUC: 0.7456 -> 0.7877 | Improved |
| replace first conv of ResNet with depthwise separable convolution (ablation) | ResNet · DFDCP | AUC: 0.6170 -> 0.6766 | Improved |
| remove all data augmentations (ablation) | Xception and SPSL · FF++_c23 within-domain | AUC | Improved |
| apply Gaussian blur augmentation (ablation) | Xception and SPSL · FF++_c40 | AUC | Improved |
| apply compression and blur augmentation (ablation) | SPSL · DFD and DFDCP | AUC | Hurt |
| apply blur augmentation (ablation) | Xception · cross-domain test sets | AUC | Hurt |
| use Xception or EfficientNet-B4 backbone instead of ResNet34 (ablation) | CORE/SPSL/UCF/Face X-ray · within and cross-domain | AUC | Improved |
| use ImageNet pre-trained weights (ablation) | Xception, EfficientB4 · DFDCP / DFD | AUC: about 10% gain | Improved |
| use ImageNet pre-trained weights (ablation) | ResNet34 · cross-domain | AUC | No clear effect |
| increase ResNet depth from 34 to 50 (ablation) | ResNet · deepfake test sets | AUC | Improved |
| increase ResNet depth from 50 to 152 (ablation) | ResNet · deepfake test sets | AUC | No clear effect |
| use sophisticated spatial/frequency detector instead of naive Xception | Xception · FF-c23 within-domain | AUC: 0.9637 -> 0.9705 (UCF) | No clear effect |
Closest papers by meaning
- DF40: Toward Next-Generation Deepfake Detection (NeurIPS 2024 · similarity 0.90)
- D^3: Scaling Up Deepfake Detection by Learning from Discrepancy (CVPR 2025 · similarity 0.90)
- Standing on the Shoulders of Giants: Reprogramming Visual-Language Model for General Deepfake Detection (AAAI 2025 · similarity 0.89)
- Exposing the Deception: Uncovering More Forgery Clues for Deepfake Detection (AAAI 2024 · similarity 0.89)
- Exploring Unbiased Deepfake Detection via Token-Level Shuffling and Mixing (AAAI 2025 · similarity 0.89)
- ODDN: Addressing Unpaired Data Challenges in Open-World Deepfake Detection on Online Social Networks (AAAI 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.
