AI Paper Atlas

Research map / Detection and segmentation / CNN features

Generalizing Face Forgery Detection With High-Frequency Features

Yuchen Luo, Yong Zhang, Junchi Yan, Wei Liu

CVPR 2021 · Detection and segmentation › CNN features · approach: Forgery detection · 563 citations (OpenAlex)

Check 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 the two-stream RGB and high-frequency model instead of RGB-only detection (ablation)Xception face forgery detector · FF++ F2F training → FSAUC: 0.762 -> 0.922Improved
Replace RGB input with SRM noise residuals (ablation)Xception face forgery detector · FF++ F2F training → NTAUC: 0.696 -> 0.858Improved
Add residual-guided spatial attention (ablation)Two-stream Fusion · FF++ F2F training → FSAUC: 0.922 -> 0.966Improved
Add dual cross-modality attention (ablation)Fusion + RSA · FF++ F2F training → FSAUC: 0.966 -> 0.975Improved
Add multi-scale high-frequency feature extraction (ablation)Fusion + RSA + DCMA · FF++ F2F training → FSAUC: 0.975 -> 0.987Improved
Use the proposed model instead of Face X-rayFace forgery detection · FF++ HQ → DFDAUC: 0.856 -> 0.919Improved
Use the proposed model instead of Face X-rayFace forgery detection · FF++ HQ → DFDCAUC: 0.700 -> 0.797Improved
Use the proposed model instead of Face X-rayFace forgery detection · FF++ HQ → CelebDFAUC: 0.742 -> 0.794Improved
Use the proposed model instead of Face X-rayFace forgery detection · FF++ HQ → DF1.0AUC: 0.723 -> 0.738Improved
Use the proposed model instead of F3NetVideo forgery detection · FF++ LQ, FSaccuracy: 0.965 -> 0.929Hurt
Use the proposed model instead of F3NetVideo forgery detection · FF++ LQ, DFaccuracy: 0.980 -> 0.986Improved
Use the proposed model instead of ForensicTransF2F-trained face forgery detector · F2F(HQ) → FS(HQ)accuracy: 0.726 -> 0.867Improved

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