Breeze Fitness

Inspiration

Fitness and lack of exercise is a huge problem in America and other developed nations. We noticed that for many, the biggest barrier to more rigorous and more frequent exercise was simply their motivation. This, combined with the often overwhelming amount of information that fitness newcomers face, can create a large barrier of entry. Thus, we decided to create a website that would tackle both problems. By helping to keep you motivated while also streamlining your fitness journey, we strove to reduce the barrier of entry to fitness as much as possible.

What it does

Our website has many features that make it extremely valuable to anyone interested in improving their health -- not just fitness newcomers. This includes dedicated meal planning, guides on proper exercise form to reduce the risk of injury, and a system of tracking your progress. Additionally, our application has two keystone features that we believe can significantly boost anyone's fitness journey. The first of which is our token pledging system. We found that many of the solutions to help motivate you to achieve your long-term goals were ineffective. Tools such as motivational quotes and mindset changes can undoubtedly help, but often provide less benefit than many expect. Our system of token pledging lets users make real commitments towards their goals. Users can pledge a certain amount of tokens, which correspond to real-world dollars, on their goals. If they achieve their goals, they get the tokens, and thus the money, back. If they fail, the money is donated to a charitable cause. The idea of staking towards your goals has been done before. However, we also allow the friends or relatives of those pledging to also pledge on their cause with them. This method fosters healthy encouragement from the close relatives or friends for the subject of the goals to achieve what they set out to do. But what if someone decides to lie to achieve their goal? That's where our verification system comes in. We use state-of-the-art pose estimation to actively verify that users are completing the fitness goals that they set out to do. Additionally, instead of machine learning to analyze the poses, our verification system uses pure vector math and trigonometry, which allows us to verify any form of repetition workout.

How we built it

We built the website using Next.js, Tailwind CSS, and Firebase for the backend. For our computer vision model, we used ONNXRuntime for Java in a fully asynchronous application. Our computer vision system also utilized the Java Native Interface, which allowed us to accelerate our computer vision models with NVIDIA's CUDA & TensorRT, Apple's CoreML, and AMD's ROCm

Challenges we ran into

Because we were using multiple different apis, it was challenging to manage them. Additionally, we ran into some issues with CMake when building ONNXRuntime, and some threading issues with the computer vision application.

Accomplishments that we're proud of

We are proud of finishing most of our planned product, even though the initial goal was very ambitious. We are also proud of the computer vision application and the amount we did with it.

What we learned

We learned more webdev, how to push forward with very little sleep, and how to improvise on-the-go. We also learned more about machine learning.

What's next for Breeze Fitness

We hope to improve the polish of the platform and provide more API integrations, thus giving more people access to fitness.

Built With

• coreml
• firebase
• java
• javascript
• nextjs
• onnxruntime
• openai
• opencv
• react
• tailwind
• whoop