Current methods of working out with friends were bland and we felt they needed augmentation to be more fun.
What it does
Allows two people to workout together and do fun challenges at the same time.
How we built it
We connected two people via peer to peer video call powered by webrtc build on a project called cuckoo, while also at the same time feeding the video into a tensorflow posenet machine learning model to human pose estimation an in turn feeding that into a mahcine learning algorith to detect the position of the rep inorder to know when a rep has been completed. This facilites the rep based challenges which are run in the react frontend
Challenges we ran into
Lots of them. React challenges due to our inexperience in using react. Machine learning challenges. Backend challenges (how do you get node.js to work python with very low latency?)
Accomplishments that we're proud of
Strong react and typescript based frontend. Integration of video call with posenet. Custom machine learning algorithm to determine state of rep
What we learned
Lots more about react, machine learning and time management.
What's next for GymBattles
Finish it! Improved frontend. Better ML, more features.