BlazePose: On-device Real-time Body Pose tracking
Published in
arXiv.org(2020)External Links:
TL
TL;DR
BlazePose is presented, a lightweight convolutional neural network architecture for human pose estimation that is tailored for real-time inference on mobile devices that uses both heatmaps and regression to keypoint coordinates.
Abstract
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.
References11 items
1
Real-time Facial Surface Geometry from Monocular Video on Mobile GPUs
2
BlazeFace: Sub-millisecond Neural Face Detection on Mobile GPUs
3
PifPaf: Composite Fields for Human Pose Estimation
4
Deep High-Resolution Representation Learning for Human Pose Estimation
5