Research map / Video / Video segmentation and tracking
AltFreezing for More General Video Face Forgery Detection
Read the paperCheck your idea against 108,967 papers
AI-extracted summary. Written by AI from the paper's title only (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 |
|---|---|---|---|
| Apply AltFreezing rather than end-to-end 3D R50 training (ablation) | 3D R50 · CelebDF v2 | video-level AUC (%): 81.8 -> 86.4 | Helped |
| Apply AltFreezing rather than end-to-end 3D R50 training (ablation) | 3D R50 · Temporal Set | video-level AUC (%): 76.5 -> 80.6 | Helped |
| Apply AltFreezing rather than end-to-end 3D R50 training (ablation) | 3D R50 · Spatial Set | video-level AUC (%): 71.5 -> 84.5 | Helped |
| Freeze spatial kernels throughout training instead of alternating freezes (ablation) | 3D R50 · CelebDF v2 | video-level AUC (%): 81.8 -> 76.8 | Hurt |
| Freeze temporal kernels throughout training instead of alternating freezes (ablation) | 3D R50 · CelebDF v2 | video-level AUC (%): 81.8 -> 80.6 | Hurt |
| Set spatial-to-temporal freezing ratio to 20:1 rather than end-to-end training (ablation) | 3D R50 · CelebDF v2 | video-level AUC (%): 81.8 -> 86.4 | Helped |
| Use video-level fake clip augmentation instead of no proposed augmentation (ablation) | AltFreezing · CelebDF v2 | video-level AUC (%): 86.4 -> 89.5 | Helped |
| Use image-level blending rather than video-level augmentation (ablation) | AltFreezing · CelebDF v2 | video-level AUC (%): 89.5 -> 78.6 | Hurt |
| Add temporal dropout and repeat augmentations without clip-level blending (ablation) | AltFreezing · DFD | video-level AUC (%): 97.6 -> 98.8 | Helped |
| Add temporal dropout and repeat augmentations without clip-level blending (ablation) | AltFreezing · CelebDF v2 | video-level AUC (%): 86.4 -> 84.5 | Hurt |
| Add clip-level blending to temporal augmentations (ablation) | AltFreezing · CelebDF v2 | video-level AUC (%): 84.5 -> 89.5 | Helped |
| Use 3D R101 instead of 3D R50 backbone (ablation) | AltFreezing · CelebDF v2 | video-level AUC (%): 89.5 -> 90.4 | Helped |
Closest papers by meaning
- Generalizing Deepfake Video Detection with Plug-and-Play: Video-Level Blending and Spatiotemporal Adapter Tuning (CVPR 2025 · similarity 0.90)
- Face Forgery Video Detection via Temporal Forgery Cue Unraveling (CVPR 2025 · similarity 0.89)
- Identity-Aware Vision-Language Model for Explainable Face Forgery Detection (AAAI 2026 · similarity 0.88)
- Forensics Adapter: Adapting CLIP for Generalizable Face Forgery Detection (CVPR 2025 · similarity 0.88)
- FakeDiffer: Distributional Disparity Learning on Differentiated Reconstruction for Face Forgery Detection (AAAI 2025 · similarity 0.87)
- On Learning Multi-Modal Forgery Representation for Diffusion Generated Video Detection (NeurIPS 2024 · similarity 0.87)
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.