Dogeball is a motion-controlled browser game built with the PoseNet model of TensorFlow JS and the Canvas API. Players need to use their head and hands to intercept avatars on screen and get rewarded or penalized according to the type of avatars they "catch".

The project is built with React JS, CSS/SASS, Bootstrap, Axios, TensorFlow JS, and HTML Canvas on the front end. On the backend, it uses PostgreSQL database and consumes JSON APIs using Ruby on Rails.

Front-end:

• Go into the client directory and run npm i or npm install

Back-end:

• The project dependencies require that Bundler 2 be installed, which requires at least Ruby v2.3.0 and RubyGems v2.5.0. To install Bundler 2, simply run "gem install bundler". For more information, visit https://bundler.io/guides/bundler_2_upgrade.html
• Please make sure you have Rails 5.2 or above installed.
• In the rails_backend directory, run bundle install

The database is currently configured for PG as the DBMS to be used

• In the rails_backedn directory, run rake db:reset and rake db:migrate

• In the rails_backedn directory, run rails s -p 3001 (for Vagrant: rails s -p 3001 -b 0.0.0.0)
• In the client directory, run npm start

The PoseNet model being used has several different configuration options that vary in accuracy vs. performance. The app comes with these settings preconfigured to favor performance over accuracy, but use the guide below to adjust (code located in NewCamera.js):

architecture: 'MobileNetV1',
      outputStride: 16, // 8, 16, 32 ---> smaller equals more accurate, but more taxing on performance
      inputResolution:  { width: 320, height: 240 }, // default 257, can be provided as an object as well, e.g. { width: 320, height: 240 }. Higher is more accurate, but more       taxing on performance.
      multiplier: .5 // 1.0, .75, .50 ---> higher is more accurate, but more taxing on performance

In order to adjust how many DogeBalls are rendered on screen, you can increase the stop condition for the iterator on the below loop (also located in NewCamera.js):

for (let i = 0; i < 9; i++) {
      projectileGenerator(projectileCoords, badProjectile);  
}