AI Paper Atlas

Research map / Label-efficient and robust learning / Domain adaptation and generalization

Group-wise Scaling and Orthogonal Decomposition for Domain-Invariant Feature Extraction in Face Anti-Spoofing

Seungjin Jung, Kanghee Lee, Yonghyun Jeong, Haeun Noh, Jungmin Lee, Jongwon Choi

ICCV 2025 · Label-efficient and robust learning › Domain adaptation and generalization · approach: Augmentation and mixture-of-experts generalization · 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
add GS-RM and FOD together to the base model (ablation)GD-FAS · OCI->MHTER: 2.92 -> 0.42Improved
add GS-RM alone (ablation)GD-FAS base · OCI->MHTER: 2.92 -> 2.50Improved
add FOD alone (ablation)GD-FAS base · OCI->MHTER: 2.92 -> 2.08Improved
add FOD alone to reduce calibration error (ablation)GD-FAS base · OCI->MECE: 5.60 -> 1.60Improved
add GS-RM on top of FOD (ablation)GD-FAS (FOD only) · OCI->MECE: 1.60 -> 3.39Hurt
add GS-RM on top of FOD (ablation)GD-FAS (FOD only) · OCI->MHTER: 2.08 -> 0.42Improved
add GS-RM alone (ablation)GD-FAS base · OMI->CHTER: 2.22 -> 2.22No clear effect
predict with image-text embedding similarity instead of classifier output (ablation)GD-FAS · OCI->MHTER: 2.50 -> 0.42Improved
replace text embeddings with random vectors (ablation)GD-FAS · OCM->IHTER: 3.33 -> 11.83Hurt
use batch size 16 instead of 3 (ablation)GD-FAS · OCI->MHTER: 5.00 -> 0.42Improved
increase batch size from 16 to 32 (ablation)GD-FAS · ICM->OHTER: 2.64 -> 3.01No clear effect
add extra CelebA-Spoof source dataGD-FAS · Average over 4 protocolsHTER: 1.83 -> 0.95Improved

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