SquatBuddy

Inspiration

Our inspiration for the project was as beginner lifters, we always felt shy and never had the confidence to ask people for help. Due to this, we started working out with terrible form, and luckily this only turned into us plateauing in our lifts, but for this others it meant serious injury. We built a solution to solve this, because no one should have to risk their body because they're not comfortable asking for help.

What it does

Our project teaches users the correct form for each of the compound lifts, including squat, bench, and deadlift. In addition, the project also has a built-in neural network that is trained on hundreds of powerlifters compound lifts, to ensure that your squat is perfect, or give you feedback to improve it.

How we built it

We built the frontend using NextJS, Tailwind CSS, and React Hooks which connects to a backend build with REST API. The model consists of a dataset on Kaggle, image analysis using OpenCV, and an LSTM model to train on the dataset.

Challenges we ran into

It was very difficult to get high accuracy with our model due to limited sample size. We also found it very difficult dividing and conquering, with so many different complex tasks going on with this project from the model to the frontend and backend.

Accomplishments that we're proud of

We were ale to build a fully function full-stack project with an integrated machine learning model built in house, all in our first hackathon. We also learned about 3D rendering, using Gemini API, and caught up with the more modern technologies.

What we learned

We learned how to build an ML model, integrate it into a full-stack site, and create a fully-functioning startup that could be used to help others.

What's next for SquatBuddy

We will integrate it into a mobile app for more easy access, and begin marketing towards beginner lifters so they can improve their lifts while staying safe.