Design and Implementation of Hybrid CNN-Transformer Architecture for Face Anti-Spoofing
This paper introduces HybridLivenessNet, a novel deep neural architecture that combines MobileNetV3-Small as a computationally efficient convolutional feature extractor with a multi-head Transformer encoder, thus capturing diverse and complementary representations at the local texture and global spatial levels.