Abstract

We present BlazeFace, a lightweight and well-performing face detectortailored for mobile GPU inference. It runs at a speed of 200-1000+ FPS onflagship devices. This super-realtime performance enables it to be applied toany augmented reality pipeline that requires an accurate facial region ofinterest as an input for task-specific models, such as 2D/3D facial keypoint orgeometry estimation, facial features or expression classification, and faceregion segmentation. Our contributions include a lightweight feature extractionnetwork inspired by, but distinct from MobileNetV1/V2, a GPU-friendly anchorscheme modified from Single Shot MultiBox Detector (SSD), and an improved tieresolution strategy alternative to non-maximum suppression.

 

Quick Read (beta)