Beyond Vanilla Convolution: Random Pixel Difference Convolution for Face Perception

Face perception is an essential and significant problem in pattern recognition, concretely including Recognition (FR), Facial Expression (FER), Race Categorization (RC). Though handcrafted features perform well on face images, Deep Convolutional Neural Networks (DCNNs) have brought new vitality to this field recently. Vanilla DCNNs are powerful at learning high-level semantic features, but weak capturing low-level image characteristic changes illumination, intensity,and texture regarded as key traits facial processing feature extraction, which alternatively the strength of human-designed descriptors. To integrate best both worlds, we proposed novel Random Pixel Difference Convolution (RPDC) efficient alternatives vanilla convolutional layers standard CNNs can promote extract discriminative diverse features. By means searched RPDC high efficiency, build S-RaPiDiNet, achieve promising extensive experiment results FR (≈0.5% improvement), FER (over 1% growth), RC (0.25%-3% increase) than baseline network convolution, showing strong generalization RPDC.