AI Paper Atlas

Research map / Video / Video segmentation and tracking

DeepShield: Fortifying Deepfake Video Detection with Local and Global Forgery Analysis

Yinqi Cai, Jichang Li, Zhaolun Li, Weikai Chen, Rushi Lan, Xi Xie, Xiaonan Luo, Guanbin Li

ICCV 2025 · Video › Video segmentation and tracking · approach: Deepfake video detection · 4 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 ST-Adapter fine-tuning to CLIP-ViT (ablation)CLIP-ViT · CDF (train FF++)video AUC (%): 60.0 -> 85.4Improved
add DeepShield training strategies (SAM, DFA, LPG, GFD) to adapted CLIP-ViT (ablation)CLIP-ViT (adapted) · CDF (train FF++)video AUC (%): 85.4 -> 92.2Improved
add DeepShield training strategies to adapted CLIP-ViT (ablation)CLIP-ViT (adapted) · DFDC (train FF++)video AUC (%): 78.4 -> 82.8Improved
add DeepShield training strategies to adapted CLIP-ViT (ablation)CLIP-ViT (adapted) · DFD (train FF++)video AUC (%): 95.3 -> 96.1Improved
add Temporal Artifact Generating (TAG) to SAM (ablation)DeepShield · Avg of CDF/DFDCP/DFDC/DFDvideo AUC (%): 88.6 -> 91.1Improved
add TAG to SAM (ablation)DeepShield · DFDvideo AUC (%): 96.0 -> 96.1No clear effect
add TAG to SAM (ablation)DeepShield · DFDCPvideo AUC (%): 90.0 -> 93.2Improved
lower patch scoring threshold theta from 150 to 10 (ablation)DeepShield · CDF/DFDCP avgvideo AUC (%): 92.2 -> 92.7Improved
vary theta between 10 and 100 (ablation)DeepShield · CDF/DFDCP avgvideo AUC (%): 92.4 -> 92.7No clear effect
increase omega from 0.5 to 1 (ablation)DeepShield · CDF/DFDC/DFDCP avgvideo AUC (%): 89.4 -> 88.8Hurt
increase omega from 0.5 to 5 (ablation)DeepShield · CDF/DFDC/DFDCP avgvideo AUC (%): 89.4 -> 88.3Hurt
increase upsilon from 0.5 to 1 (ablation)DeepShield · DFDCvideo AUC (%): 82.8 -> 83.1Mixed

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