The problem FitNet solves

Today, the majority of people tend to stay in their homes playing video games rather than going out and keeping themselves fit. Making the best out of the worse, we offer an online platform for playing those video games with fitness in mind. Not only laziness but let us take into account our lockdown situation. People are always at home and physical movement is negligible. One can easily open their systems, turn on their webcams, and play games.

The Application in connection with the MI band offers heart rate values and calories burnt. Unlike other fitness Apps on can actually have fun while working out. Other gaming platforms and play stations usually cost thousands of dollars and also need high maintenance. Our Application removes these unnecessary expenditures and provides stability.

Challenges we ran into

1. Working with pose estimation AI on Unity Barracuda and deploying it using WebGL. WebGL does not support GPU for the implementation of the AI model.
2. Integrating TFlite Posenet on javascript with Unity WebGL game. To overcome problems with Barracuda we tried using the model of TensorFlow.js. The live sending of data from the webcam to the Unity engine for character mapping was a huge challenge.
3. Creating a companion Application for sending the MI band data to Firebase.
4. Connecting the MI band with the Application for data communication.