Using Pose Net from Tensorflow, we were able to take a video feed and compute the location of a human body in the shot. With this, we are able to compare live feed from a webcam with a video of a professional (or amazing amatuer) dancing. Written in JavaSCript with a NodeJS backend and a ReactJS frontend.

• Node

For package requirements you must run npm install in the Wiggl root dir and the client dir.

Then run in the root dir run node on the server.js file:

node server.js

Then in another terminal, enter the client dir and run via npm

cd client
npm run start

The app will automatically open locally in your web browser. It is located at localhost:3000.

Must be an mp4 video linked from imgur.com .

In the Room.js file located at Wiggl/client/src/routes/Room.js, change the src variable to equal the link (make sure the link ends in .mp4).

...
const v3 = (
    <video
      autoPlay
      loop
      id="dance_hidden"
      src="proxy.php?url=https://i.imgur.com/MCB77Uw.mp4" //HERE
      style={{ display: "none" }}
      crossOrigin="Anonymous"
    />
  );
...

• Add larger multiplayer functionality
• Optimze processing of poses
• Improve mobile layout