Inspiration

People who do not know proper workout form are more accident prone when they work out. They risk straining the wrong muscles and injuring themselves. This also might scare first-time gym-goers from going to the gym because they cannot afford or do not want a personal trainer.

What it does

FormFit AI is an AI analysis software that can help users improve their form in with real-time feedback while providing users with encouragement. Users can choose from a selection of exercises sorted by the muscle group it is exercising. Then, users receive vocal cues to start moving and any adjustments that they should make to their form. Users have a video reference of what the exercise looks like to get started. There is also a rep counter that tracks all the reps the user has made during the workout across all exercises.

How we built it

We built this using the Google AI's MediaPipe Pose Landmarker Detection model that detects the different parts of the body, then we compared the coordinates and angles from the user's livestream to standardized coordinates from some reference material of the exercises. This was then integrated into an HTML front end with features from React to make the program more intuitive.

Challenges we ran into

Initially we found it hard to calibrate the dimensions and the movement parameters which we later figured out using much trial and error. Also when we tried to implement verbal feedback, the audio and the textual feedback were out of sync. This took us a few hours of debugging and changing to fix the issue.

Accomplishments that we're proud of

What we learned

We learned about the benefits that ML models can bring to solutions for issues that people might struggle with in their daily lives. We learned the ways in which we should implement models like the MediaPipe Pose Landmark Model. We also learned the best methods to create a front end that is polished and professional

What's next for FormFit AI

We hope to expand to mobile phones for more convenience, add more exercises to our database, improve our analysis accuracy, include accommodated exercise options, and allow for workout tracking and summaries.

Built With

Share this project:

Updates