Sports Injury Prevention Platform (SIPP)

Inspiration

We observed that within many sports, particularly combat sports, athletes were sustaining injuries due to poor technique. We wanted to create a program that could detect bad form in practice so it's corrected before competitions.

What it does

The program takes a sample input of an athlete's technique (in this case, punching) and compares it to "proper" technique to determine if there's potential for injury.

How we built it

We used Gemini's ability to view images to pinpoint the knuckle, wrist, and elbow on the arm. We then checked if these points were in line to see if the punching technique was safe.

Challenges we ran into

The model has a tough time differentiating different parts of the arm, so we decided to simplify it to make sure the model was accurate enough with where it placed the points.

Accomplishments that we're proud of

In many of our pictures, we were able to accurately pinpoint the parts of the arm, which allowed us to detect alignment.

What we learned

It takes a lot more data to train models to look for very specific pieces. In the future, we'll be sure to make adequate samples before trying to have the model generate everything.