animation demonstrating the form, notes on what it is looking for.
shows red when form breaks.
judges the set afterwards and gives constructive tips to help

Inspiration

we wanted to use computer vision to help us with our form in the gym. FormCoach was designed to give objective feedback on your workout without the need for a real life coach.

What it does

waits until the user lines up with a set of dashed guidelines to ensure proper starting position. After their position is confirmed, the users outline turns green and a countdown occurs on screen. As the user performs their set of pushups the computer watches to see if you break form, where the skeleton then flashes red. after the set is complete, FormCoach tells you how you broke form and suggestions on fix it for the future.

How we built it

We used open source tools like MediaPipe Tasks Vision for the computer vision , Vite and Node.js for the dev server, Canvas and WebAssembly for the visual overlays, and typescript with HTML and CSS for the rep counting and data analysis. We used codex to make us a Github repository where we all worked independently to build the structure, then worked together for bug fixing and patches to the code.

Challenges we ran into

there were many challenges creating a front view angle of the project so we opted to scrap it and develop on the side angle.

Accomplishments that we're proud of

We were proud that the bot works affectively using the open source tools. and after many iterations and bug fixes, we were able to create exactly what we had imagined.

What we learned

we learned how to affectively work in teams and how to use different ai agents simultaneously to code our project without the code failing or overwriting.

What's next for FormCoach

we want to add more exercises in later versions and allow the user to film their set in different angles. making the app compatible with mobile devices is also a priority so that people dont have to bring their laptops into the gym.