AI Paper Atlas

Research map / Image generation / Editing and style transfer

FaceTracer: Unveiling Source Identities From Swapped Face Images and Videos for Fraud Prevention

Zhongyi Zhang, Jie Zhang, Wenbo Zhou, Xinghui Zhou, Qing Guo, Weiming Zhang, Tianwei Zhang, Nenghai Yu

TPAMI 2025 · Image generation › Editing and style transfer · approach: Identity-preserving face manipulation · 3 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
use FaceTracer source-identity extractor instead of ArcFace baselineArcFace ResNet18 · HiRes swapped faces (S1)Top-1 accuracy (%): 0.04 -> 98.80Improved
add reference images during training (S2 vs S3)FaceTracer · HiRes swapped facesTop-1 accuracy (%): 93.34 -> 99.27Improved
add reference images during training (S2 vs S3)FaceTracer · InfoSwap swapped facesTop-1 accuracy (%): 91.83 -> 98.19Improved
add reference images at inference (S1 vs S2)FaceTracer · HiRes swapped facesTop-1 accuracy (%): 99.27 -> 98.80No clear effect
use ViT-S backbone instead of ResNet18 (ablation)FaceTracer · HiRes swapped faces (S1)Top-1 accuracy (%): 78.71 -> 98.80Improved
use ViT-S backbone instead of ResNet18 (ablation)FaceTracer · FaceShifter swapped faces (S1)Top-1 accuracy (%): 99.25 -> 99.40No clear effect
use explicit disentanglement instead of cross-attention (ablation)FaceTracer · HiRes swapped faces (S1)Top-1 accuracy (%): 97.56 -> 98.80No clear effect
add AAM-softmax identity enhancement instead of plain softmax (ablation)FaceTracer · HiRes swapped faces (S1)Top-1 accuracy (%): 89.51 -> 98.80Improved
add AAM-softmax identity enhancement instead of plain softmax (ablation)FaceTracer · InfoSwap swapped faces (S1)Top-1 accuracy (%): 96.64 -> 97.97Improved
add reference images in training for cross-method transfer (S1/S2 vs S3)FaceTracer trained on HiRes · InfoSwap test imagesTop-1 accuracy (%): 87.84 -> 98.06Improved
train on mixture of four swap methods instead of oneFaceTracer · HiRes swapped faces (S1)Top-1 accuracy (%): 98.80 -> 99.23No clear effect
swap faces twice with two targets (continuous swapping)FaceTracer · HiRes swapped faces (S3)Top-1 accuracy (%): 93.34 -> 84.73Hurt

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