BlazeFace: Sub-millisecond Neural Face Detection on Mobile GPUs
Valentin Bazarevsky, Yury Kartynnik, Andrey Vakunov, Karthik Raveendran, Matthias Grundmann
Code Available — Be the first to reproduce this paper.
ReproduceCode
- github.com/PaddlePaddle/PaddleDetectionpaddle★ 14,132
- github.com/CycloneBoy/PPDetectionPytorchpytorch★ 26
- github.com/longnguyen2/blazefacepytorch★ 4
- github.com/gouthamvgk/facemesh_coreml_tftf★ 0
- github.com/hollance/BlazeFace-PyTorchpytorch★ 0
- github.com/kentaroy47/BlazeFace_Person.pytorchpytorch★ 0
- github.com/dl-maxwang/blazeface-tensorflowtf★ 0
- github.com/minus31/BlazeFacetf★ 0
- github.com/MaximilianHollis/Maskifynone★ 0
- github.com/zineos/blazefacepytorch★ 0
Abstract
We present BlazeFace, a lightweight and well-performing face detector tailored for mobile GPU inference. It runs at a speed of 200-1000+ FPS on flagship devices. This super-realtime performance enables it to be applied to any augmented reality pipeline that requires an accurate facial region of interest as an input for task-specific models, such as 2D/3D facial keypoint or geometry estimation, facial features or expression classification, and face region segmentation. Our contributions include a lightweight feature extraction network inspired by, but distinct from MobileNetV1/V2, a GPU-friendly anchor scheme modified from Single Shot MultiBox Detector (SSD), and an improved tie resolution strategy alternative to non-maximum suppression.