DevMesh

MixPose Live Yoga Stream

Peter Ma

Peter Ma

San Francisco

4 0
  • 0 Collaborators

MixPose is personalized AI fitness with real-time live stream yoga, yoga for beginners friendly and weekly free yoga classes available. ...learn more

Project status: Published/In Market

oneAPI, Mobile, RealSense™, Artificial Intelligence, Graphics and Media

Groups
Movidius™ Neural Compute Group

Intel Technologies
DevCloud, oneAPI, Intel Opt ML/DL Framework, Intel Python, OpenVINO

Code Samples [1]Links [12]

Overview / Usage

Wellness has always been an important issue for us. COVID19 has resulted more and more people staying home, and gyms/yoga studios closing down. We want to better the experience for live streaming exercising as well as tackling the growing problem of loneliness.

MixPose is a live streaming platform designed for yoga and fitness classes, but what sets us apart is that we use real-time AI pose tracking on the mobile device itself. The information is then sent to our fitness instructors, which creates a much more immersive and interactive experience.

Methodology / Approach

Static fitness videos only share information in one direction. But with this app, teachers can customize their lessons and directly engage with their student’s needs.

We utilize our pose-tracking function to detect each user’s movements and positions, which are then classified using Artificial Intelligence. Then, live sensors and feedback systems inform users of their alignment. Added output to video features, such as Chromecast, easily connects to larger screens for more immersive viewing.

Technologies Used

We focus on the idea of training on the server and deploying to the cloud. We have used part of intel DevCloud to train the model on the server, as well as leveraging OpenVINO to test with NVIDIA's GPU. We are planning to implement RealSense as part of the client experience

Android/iOS will be the distribution platform for students, as they are mostly widely available distribution platform. Android app can be downloaded at https://play.google.com/store/apps/details?id=com.mixpose.shiva.android and iOS in alpha testing at https://testflight.apple.com/join/QCkmOttG

We are doing double inferencing on from pose tracking to pose classification. Leveraging oneAPI for training for pose classification.

Agora.io helps apps increase user engagement and retention with a developer platform for live, interactive audio, video, and messaging features at a planetary scale. With a dedicated network and device-optimized SDKs, Agora is the only real-time engagement platform designed to cross borders and reach users on low-bandwidth networks and low-powered devices.

Repository

https://github.com/mixpose/

Comments (0)