A face anti-spoofing method that uses ViT's 8th block for scoring and 11th block for an extra loss, plus two augmentations, outperforms prior methods on OULU-NPU and SiW.
Title resolution pending
1 Pith paper cite this work. Polarity classification is still indexing.
1
Pith paper citing it
fields
cs.CV 1years
2025 1verdicts
CONDITIONAL 1representative citing papers
citing papers explorer
-
Leveraging Intermediate Features of Vision Transformer for Face Anti-Spoofing
A face anti-spoofing method that uses ViT's 8th block for scoring and 11th block for an extra loss, plus two augmentations, outperforms prior methods on OULU-NPU and SiW.