Towards Cross-materials: Fingerprint Liveness Detection based on Style Transfer

While a number of CNN-based fingerprint liveness detection (FLD) methods have been proposed, they still provide limited detection performance on untrained materials that can be used for generating fake fingerprints. In this paper, we present a cross-material FLD method that improves generalization performance by transferring the material styles between finger-prints. As a result of evaluation using real-world datasets, our proposed method successfully provides improved generalization performance for cross-materials.