BlazeFace: Sub-millisecond Neural Face Detection on Mobile GPUs
By Valentin Bazarevsky, Yury Kartynnik, Andrey Vakunov, Karthik Raveendran, Matthias Grundmann. arXiv.org, 2019
378 citations recorded by the source corpus.
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.
Subfields: Computer Vision and Pattern Recognition
View Paper: https://www.semanticscholar.org/paper/fd8e358f306388eac6a9d9231e1ad9b5e9bc2d81
How researchers see this paper
The researchers whose trusted networks rank it highest, strongest advocate first. Every researcher gets a different order.
In Computer Vision and Pattern Recognition
- Fan Zhang
- Valentin Bazarevsky
- Deressa Wodajo
- Solomon Atnafu
- Ivan Grishchenko
- Andrey Vakunov
- Matthias Grundmann
- Karthik Raveendran
Citations in the corpus
- MediaPipe Hands: On-device Real-time Hand Tracking (2020, methodology): “To detect initial hand locations, we employ a single-shot detector model optimized for mobile real-time application similar to BlazeFace[1], which is also available in MediaPipe[12].”
- BlazePose: On-device Real-time Body Pose tracking (2020, methodology): “As its 33 keypoint topology is consistent with BlazeFace[2] and BlazePalm[3], it can be a backbone for subsequent hand pose[3] and facial geometry estimation[6] models.”
- Real-time Facial Surface Geometry from Monocular Video on Mobile GPUs (2019, neutral): “The whole frame from the camera input gets processed by a very lightweight face detector [2] the eye centers with the horizontal axis of the rectangle.”
- Deepfake Video Detection Using Convolutional Vision Transformer (2021, methodology): “BlazeFace neural face detector [5], MTCNN [55] and face recognition [17] DL libraries are used to extract the faces.”