BlazeFace: Sub-millisecond Neural Face Detection on Mobile GPUs

TL;DR

Outlines submission guidelines and formatting requirements for CVPR/IEEE Computer Society Press, detailing language, submission policies, page limits, ruler usage, numbering, blind review, and miscellaneous formatting rules. It prescribes final-copy formatting standards including margins, fonts, footnotes, references, figures, and color usage, as well as camera-ready copy procedures. The document serves as a comprehensive protocol to ensure consistent, fair review and publication, reducing formatting errors and expediting the CVPR submission process. Its practical impact lies in providing a clear, machine-readable suite of requirements for authors preparing manuscripts for CVPR submissions.

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.

Figures (2)

• Figure 1: Example of caption. It is set in Roman so that mathematics (always set in Roman: $B \sin A = A \sin B$) may be included without an ugly clash.
• Figure 2: Example of a short caption, which should be centered.