We present BlazePose, a lightweight convolutional neural network architecture for human pose estimation that is tailored for real-time inference on mobile devices. During inference, the network produces 33 body keypoints for a single person and runs at over 30 frames per second on a Pixel 2 phone. This makes it particularly suited to real-time use cases like fitness tracking and sign language recognition. Our main contributions include a novel body pose tracking solution and a lightweight body pose estimation neural network that uses both heatmaps and regression to keypoint coordinates.
@article{arxiv.2006.10204,
title = {BlazePose: On-device Real-time Body Pose tracking},
author = {Valentin Bazarevsky and Ivan Grishchenko and Karthik Raveendran and Tyler Zhu and Fan Zhang and Matthias Grundmann},
journal= {arXiv preprint arXiv:2006.10204},
year = {2020}
}
Comments
4 pages, 6 figures; CVPR Workshop on Computer Vision for Augmented and Virtual Reality, Seattle, WA, USA, 2020