Our new X account is live! Follow @wizwand_team for updates
WorkDL logo mark

Rethinking Domain Generalization for Face Anti-spoofing: Separability and Alignment

About

This work studies the generalization issue of face anti-spoofing (FAS) models on domain gaps, such as image resolution, blurriness and sensor variations. Most prior works regard domain-specific signals as a negative impact, and apply metric learning or adversarial losses to remove them from feature representation. Though learning a domain-invariant feature space is viable for the training data, we show that the feature shift still exists in an unseen test domain, which backfires on the generalizability of the classifier. In this work, instead of constructing a domain-invariant feature space, we encourage domain separability while aligning the live-to-spoof transition (i.e., the trajectory from live to spoof) to be the same for all domains. We formulate this FAS strategy of separability and alignment (SA-FAS) as a problem of invariant risk minimization (IRM), and learn domain-variant feature representation but domain-invariant classifier. We demonstrate the effectiveness of SA-FAS on challenging cross-domain FAS datasets and establish state-of-the-art performance.

Yiyou Sun, Yaojie Liu, Xiaoming Liu, Yixuan Li, Wen-Sheng Chu• 2023

Related benchmarks

TaskDatasetResultRank
Face Anti-SpoofingOULU-NPU ICM → O
HTER10
115
Face Anti-SpoofingIdiap Replay-Attack OCM → I
HTER6.58
96
Face Anti-SpoofingMSU-MFSD OCI → M
HTER5.95
85
Face Anti-SpoofingCASIA-FASD OMI → C
HTER8.78
41
Face Anti-SpoofingReplay-Attack I (test)
HTER6.58
33
Face Anti-SpoofingMSU-MFSD M (test)
HTER5.95
33
Face Anti-SpoofingCASIA-MFSD C (test)
HTER8.78
18
Face Anti-SpoofingOULU-NPU CASIA-MFSD Idiap MSU-MFSD Average (test)
HTER7.83
14
Face Anti-SpoofingOULU-NPU O (test)
HTER0.1
14
Face Anti-SpoofingMSU-MFSD OCI -> M Leave-one-out (test)
HTER0.1317
5
Showing 10 of 11 rows

Other info

Follow for update