Cyclically Disentangled Feature Translation for Face Anti-spoofing

TL;DR

This paper introduces CDFTN, a novel domain adaptation method for face anti-spoofing that disentangles features to generate pseudo-labeled data, improving cross-scenario generalization.

Contribution

The paper proposes a cyclically disentangled feature translation network that enhances domain adaptation by generating pseudo-labeled samples with invariant and specific features.

Findings

Significantly outperforms state-of-the-art methods on public datasets.

Effectively disentangles domain-invariant and domain-specific features.

Extends to multi-target domain adaptation with multiple unlabeled domains.

Abstract

Current domain adaptation methods for face anti-spoofing leverage labeled source domain data and unlabeled target domain data to obtain a promising generalizable decision boundary. However, it is usually difficult for these methods to achieve a perfect domain-invariant liveness feature disentanglement, which may degrade the final classification performance by domain differences in illumination, face category, spoof type, etc. In this work, we tackle cross-scenario face anti-spoofing by proposing a novel domain adaptation method called cyclically disentangled feature translation network (CDFTN). Specifically, CDFTN generates pseudo-labeled samples that possess: 1) source domain-invariant liveness features and 2) target domain-specific content features, which are disentangled through domain adversarial training. A robust classifier is trained based on the synthetic pseudo-labeled images under the supervision of source domain labels. We further extend CDFTN for multi-target domain adaptation by leveraging data from more unlabeled target domains. Extensive experiments on several public datasets demonstrate that our proposed approach significantly outperforms the state of the art.