MobileFace provides lightweight deep learning models for face-related computer vision tasks on mobile hardware. It includes face detection, face identification (99.653% LFW accuracy), facial landmark detection, and head pose estimation models. The models are optimized for edge devices with small model sizes (2-3MB for identification) and fast inference times (3ms on CPU for the latest model). The project uses MXNet/GluonCV for training and supports deployment on mobile devices.