We were inspired by the difficulties those trying to commit to a healthy life-style face. Through the power of yoga, we strive to remedy anxiety, stress, and mood issues.

Our intelligent yoga web-app uses machine learning to recognize yoga poses and allows users to simultaneously learn a new skill and improve their overall health from the comfort of their own home.

Form correction: We believe that the best way to learn is by doing. Through your desktop camera, our model analyses your body position in real-time as you follow the steps on how to perform a yoga pose of your choice. Our web app provides visual cues, so you can see how accurate your form is and look at the proper steps to perform the pose.

Yoga has been proven to...

• prevents injuries (improves flexibility)
• promotes proper physical exercise (correct form)
• makes physical exercise easy and accessible (more likely to do daily)
• improves mental well-being (calms the mind, brings self-confidence, etc.)
• Tree “reward” system:

Every correct/corrected pose users gives them the chance to grow their very own virtual plant companion. These virtual plants all contribute to your online forest, a compile of the hard work and progress made. The more you practice, the more your tree grows, along with how YOU grow as a person. The more you stay consistent with yoga (everyday), the more your forest flourishes. this... makes it fun! (further promotes yoga for mental health and exercise) = encouragement/motivation fosters commitment through a sense of reward (may encourage to apply this learning mindset to other things) encourages mental health improvement

We used Google's TeachableMachines to train the model and exported the model with Tensorflow.js. - We made the choice to use vanilla HTML, CSS and JavaScript for this project due to some difficulties we ran into, such as performance overheads when experimenting with using React in conjunction with a client hosted machine learning model. We gathered our own data and took snapshots of ourselves doing yoga poses (someone's hip did NOT survive this)!

Difficulty integrating a front-end and back-end We designed and wanted to render 3D models to make a more engaging UX, but it turned out to take longer than we thought. We decided to reduce the scope of the project and pivoted to a simpler design instead Utilizing version control (we are SO BAD at GitHub)

Creating a fully functional app that helps people become healthier, and actually finishing our project on time yippee!

How to use GitHub and Version control effectively (It was really useful once we got the hang of it :D) How to use machine learning and tensorflow.js to track poses The design process and revision How to cache data on the browser

FUTURE ADD-ONS:

• More data, more poses to master!
• A database so you can compete with your friends and keep yourself accountable!
• Using generative/agentic AI to send motivational messages from time-to-time to encourage users; or real time coaching through speech synthesis
• Allowing a wider variety of activities beyond static poses, which may further encourage a bigger population to participate, and grow along (ex. dancing)
• A mascot we can put on top of the webcam to make it cuter, or maybe add some hardware to spice things up!