Research map / Video / Video segmentation and tracking
Learning Spatiotemporal Inconsistency via Thumbnail Layout for Face Deepfake Detection
Yuting Xu, Jian Liang, Lijun Sheng, Xiaoyu Zhang
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 thumbnail layout (TALL) to Swin-B (ablation) | Swin-B · FF++ -> Celeb-DF | AUC: 83.13 -> 87.13 | Improved |
| add fixed-position mask to TALL (ablation) | TALL on Swin-B · FF++ -> Celeb-DF | AUC: 87.13 -> 90.79 | Improved |
| add graph reasoning block to TALL (ablation) | TALL on Swin-B · FF++ -> Celeb-DF | AUC: 90.79 -> 91.42 | Improved |
| add semantic consistency loss (ablation) | TALL+GRB on Swin-B · FF++ -> Celeb-DF | AUC: 91.42 -> 91.96 | Improved |
| add GRB to plain Swin-B without TALL (ablation) | Swin-B · FF++ -> Celeb-DF | AUC: 83.13 -> 83.61 | Improved |
| enlarge window from 7 to 14 in first three stages (ablation) | TALL++ Swin-B · FF++ -> Celeb-DF | AUC: 88.24 -> 91.96 | Improved |
| enlarge window further to 28 (ablation) | TALL++ Swin-B · FF++ -> Celeb-DF | AUC: 91.96 -> 89.26 | Hurt |
| use forward instead of random frame order (ablation) | TALL++ · FF++ -> Celeb-DF | AUC: 88.34 -> 91.96 | Improved |
| keep all four sub-frames instead of dropping the last (ablation) | TALL++ · FF++ -> Celeb-DF | AUC: 89.73 -> 91.96 | Improved |
| 3x3 layout with 224 sub-frames (672 thumbnail) (ablation) | TALL++ · FF++ -> Celeb-DF | AUC: 91.96 -> 93.13 | Mixed |
| squeeze 3x3 layout into 224 thumbnail (ablation) | TALL++ · FF++ -> Celeb-DF | AUC: 91.96 -> 85.57 | Hurt |
| add TALL to ResNet50 | ResNet50 · FF++ -> Celeb-DF | AUC: 76.38 -> 80.90 | Improved |
Closest papers by meaning
- TALL: Thumbnail Layout for Deepfake Video Detection (ICCV 2023 · similarity 0.93)
- Towards More General Video-based Deepfake Detection through Facial Component Guided Adaptation for Foundation Model (CVPR 2025 · similarity 0.88)
- Generalizing Deepfake Video Detection with Plug-and-Play: Video-Level Blending and Spatiotemporal Adapter Tuning (CVPR 2025 · similarity 0.87)
- From Specificity to Generality: Revisiting Generalizable Artifacts in Detecting Face Deepfakes (NeurIPS 2025 · similarity 0.86)
- Standing on the Shoulders of Giants: Reprogramming Visual-Language Model for General Deepfake Detection (AAAI 2025 · similarity 0.86)
- Exploring Unbiased Deepfake Detection via Token-Level Shuffling and Mixing (AAAI 2025 · similarity 0.85)
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.
