AI Paper Atlas

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

IJCV 2024 · Video › Video segmentation and tracking · approach: Deepfake video detection · 50 citations (OpenAlex)

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 changedWhereMeasuredEffect
add thumbnail layout (TALL) to Swin-B (ablation)Swin-B · FF++ -> Celeb-DFAUC: 83.13 -> 87.13Improved
add fixed-position mask to TALL (ablation)TALL on Swin-B · FF++ -> Celeb-DFAUC: 87.13 -> 90.79Improved
add graph reasoning block to TALL (ablation)TALL on Swin-B · FF++ -> Celeb-DFAUC: 90.79 -> 91.42Improved
add semantic consistency loss (ablation)TALL+GRB on Swin-B · FF++ -> Celeb-DFAUC: 91.42 -> 91.96Improved
add GRB to plain Swin-B without TALL (ablation)Swin-B · FF++ -> Celeb-DFAUC: 83.13 -> 83.61Improved
enlarge window from 7 to 14 in first three stages (ablation)TALL++ Swin-B · FF++ -> Celeb-DFAUC: 88.24 -> 91.96Improved
enlarge window further to 28 (ablation)TALL++ Swin-B · FF++ -> Celeb-DFAUC: 91.96 -> 89.26Hurt
use forward instead of random frame order (ablation)TALL++ · FF++ -> Celeb-DFAUC: 88.34 -> 91.96Improved
keep all four sub-frames instead of dropping the last (ablation)TALL++ · FF++ -> Celeb-DFAUC: 89.73 -> 91.96Improved
3x3 layout with 224 sub-frames (672 thumbnail) (ablation)TALL++ · FF++ -> Celeb-DFAUC: 91.96 -> 93.13Mixed
squeeze 3x3 layout into 224 thumbnail (ablation)TALL++ · FF++ -> Celeb-DFAUC: 91.96 -> 85.57Hurt
add TALL to ResNet50ResNet50 · FF++ -> Celeb-DFAUC: 76.38 -> 80.90Improved

Closest papers by meaning

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.

Open scime draft

Get the next scime Dispatch and a monthly field update

Free. We send a confirmation link first; every email has a one-click unsubscribe. Privacy