Skip to content
Branch: master
Find file History
Fetching latest commit…
Cannot retrieve the latest commit at this time.
Permalink
Type Name Latest commit message Commit time
..
Failed to load latest commit information.
README.md
index.html
tsbodytracking.gif

README.md

Tensorflow PoseNet body tracking for virtual reality

VRDemo

The Project [Work-in-Progress]

This is a demo created for use in my Master's Thesis. The thesis focuses on analyzing the social VR landscape to better understand how to make co-located VR experiences more collaborative and social.

Demo Goals

This purpose of this demo is to provide a way for mobile VR users to engage in VR experiences together simply by placing their phones in a low-cost VR device like Google cardboard. By using phone camera-based body tracking, VR users do not have to instrument an environment to get realistic body representation in the virtual environment.

We will use this demo as a part of a user study to discover if body representation in co-located social VR experiences encourages collaboration and increases feelings of togetherness. Scenarios will be developed to study multiple co-located VR participants and co-located VR + non-VR participants.

Additional work to be done

  • Use additional sensor or marker to identify z coordinate.
  • Animate a 3D model with poses.
  • Add facial expression tracking of both VR and non-VR participants.
  • Provide multi-user support.
  • Prototype will be configurable to support multiple study activities.

Demo

Built With

  • Tensorflow.js - Javascript library for training & deploying ML models in the browser and on Node.js.
  • Tensorflow.js PoseNet - In browser human body pose estimation model.
  • A-Frame - The web framework used web-based VR prototypes.

Authors

Acknowledgments

Thanks to my advisor, Professor Michael Nebeling, and the University of Michigan Information Interaction Lab for supporting the work in my master's thesis.

You can’t perform that action at this time.