Virtual reality for fitness and health care applications require accurate and real-time pose estimation for interactive features. Yet, they suffer either a limited angle of view when using handset devices such as smartphones and VR gears for capturing human pose or a limited input interfaces when using distant imaging/computing devices such as Kinect. Our goal is to marry these two into an interactive metaverse system with human pose estimation. This paper describes the design and implementation of Yoroke, a distributed system designed specifically for human pose estimation for interactive metaverse system. It consists of a remote imaging device for estimating human pose, and a handset device for implementing a multi-user interactive metaverse system. We have implemented and deployed Yoroke on embedded platforms and evaluated its effectiveness in delivering accurate and real-time pose estimation for multi-user interactive metaverse platform.
- Ubuntu 18.04
- python-opencv 3.4.10
- Tensorflow 1.14.0
- Tensorflow-GPU
- Cuda 11.2
- Cudnn 8.1.0
- Unity
- v1: Basic network and spatial configuration using Photon2 and unitychan_dynamic_locomotion
- v2: pose estimation based on 2 people
- v3: Pose estimation of 3 or more people, lobby, and character selection possible
- v4 : UI upgrade
- v5: Change the way data is read for pose estimation without delay on Unity
- yoroke: Final deployment
- https://github.com/SrikanthVelpuri/tf-pose
- https://github.com/Jacob12138xieyuan/real-time-3d-pose-estimation-with-Unity3D
- https://openaccess.thecvf.com/content_cvpr_2017/papers/Chen_3D_Human_Pose_CVPR_2017_paper.pdf
- https://junsk1016.github.io/deeplearning/tf-pose-estimation-%EB%B9%8C%EB%93%9C/
- https://www.youtube.com/watch?v=mXPndbtKbTo