Mario TFJS - ¡Play Super Mario Bros with your camera!

This is a project carried for the Machine Learning Engineering subject of the Master's Degree in IT, Data and Cloud at the University of Seville.

Authors

Andrés Martínez & Lorenzo Rondán

Objetives

The aim of this project is to realise an image classifier using 2 methods, pretrained MobileNet model and a custom on fly trained model classifier.

The pre-trained MobileNet detects the gestures that we indicate using the camera, the detections made will serve as an input device for a video game. In this case, we have selected the classic Super Mario Bros. level 1-1. The available actions are the following:

• Forward
• Backwards
• Jump

Each of these actions are the possible classes of the image classifier.

Deployment

The project is deployed using Firebase. It can be tested both on PC and mobile through this link: https://mario-tfjs.web.app/