Research map / Video / Video segmentation and tracking
Delving into Sequential Patches for Deepfake 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 |
|---|---|---|---|
| use LTTD's local temporal patch modeling and global cross-patch learning | deepfake detection trained on FF++ · Cross-dataset average | AUC (%): 89.6 -> 91.9 | Helped |
| use LTTD's local temporal patch modeling and global cross-patch learning | deepfake detection · FF++ perturbed videos, average | AUC (%): 92.6 -> 95.0 | Helped |
| add the Local Sequence Transformer instead of standard patch embedding and Transformer blocks (ablation) | LTTD · Cross-Avg | AUC (%): 88.48 -> 92.80 | Helped |
| add the Cross-Patch Inconsistency loss (ablation) | LTTD · Cross-Avg | AUC (%): 91.19 -> 92.80 | Helped |
| add Cross-Patch Aggregation instead of average pooling and a fully connected classifier (ablation) | LTTD · Cross-Avg | AUC (%): 88.87 -> 92.80 | Helped |
| reduce clip length from 16 to 8 (ablation) | LTTD · Average | AUC (%): 92.80 -> 91.11 | Hurt |
| increase clip length from 16 to 32 (ablation) | LTTD · Average | AUC (%): 92.80 -> 92.93 | No clear effect |
| increase clip length from 16 to 64 (ablation) | LTTD · Average | AUC (%): 92.80 -> 92.37 | No clear effect |
| increase frame sampling space from 1 to 2 (ablation) | LTTD · Average | AUC (%): 92.80 -> 90.88 | Hurt |
| increase frame sampling space from 1 to 4 (ablation) | LTTD · Average | AUC (%): 92.80 -> 87.62 | Hurt |
| apply LTTD to high-level gaussian noise perturbation | LTTD · FF++ perturbed videos | AUC (%): 82.6 | Mixed |
Closest papers by meaning
- Generalizing Deepfake Video Detection with Plug-and-Play: Video-Level Blending and Spatiotemporal Adapter Tuning (CVPR 2025 · similarity 0.91)
- Exposing the Deception: Uncovering More Forgery Clues for Deepfake Detection (AAAI 2024 · similarity 0.90)
- Towards More General Video-based Deepfake Detection through Facial Component Guided Adaptation for Foundation Model (CVPR 2025 · similarity 0.89)
- Exploiting Style Latent Flows for Generalizing Deepfake Video Detection (CVPR 2024 · similarity 0.88)
- DF40: Toward Next-Generation Deepfake Detection (NeurIPS 2024 · similarity 0.88)
- Rethinking Vision-Language Model in Face Forensics: Multi-Modal Interpretable Forged Face Detector (CVPR 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.