Inspiration
The inspiration behind MyMuscle came from a shared frustration with poor workout form, lack of motivation, and inconsistent progress tracking during home workouts. With the rise of remote fitness and wearable technology, we wanted to create a solution that feels like having a personal trainer—without the cost or commitment. We envisioned a tool that empowers users to stay active, avoid injuries, and meet their goals using the power of AI and real-time feedback. Our goal was to make fitness more intelligent, accessible, and personalized for everyone.
What it does
MyMuscle is a mobile-responsive web application that helps users improve their workout form using real-time pose detection through their device's camera. It counts reps, gives instant technique feedback, and generates personalized workout plans based on user input like fitness goals, available equipment, and timeline. The app features a rich exercise library with detailed descriptions, a smart dashboard to track user progress, and even integrates motivational visuals and summaries to keep users engaged and on track.
How we built it
We built MyMuscle using React.js for the frontend and Tailwind CSS for responsive styling. We integrated TensorFlow.js to run models directly in the browser for pose detection and real-time analysis. State management is handled through Zustand, and React Router organizes the multi-page layout. We used Supabase for backend services including authentication, user data storage, and file management. React Query manages data fetching and caching efficiently. For visualizing user progress, we used Chart.js and Recharts. The platform is deployed using Vercel, with PWA capabilities to support mobile installation. Additionally, we use the Google Gemini API to generate dynamic, informative exercise descriptions.
Challenges we ran into
TensorFlow compatibility with Expo Real-time pose detection on the browser came with performance and accuracy tradeoffs—especially across different devices and lighting conditions.
Accomplishments that we're proud of
Successfully implementing real-time AI-powered pose detection and rep counting right in the browser.
Building a fully mobile-responsive Progressive Web App (PWA) that users can install and use like a native app.
Creating a personalized fitness experience with customized plans based on user input and goals.
What we learned
How to use TensorFlow.js for browser-based computer vision and real-time analysis.
Importance of user experience design when building fitness tools that involve movement and attention to form.
Deepened our skills in React, state management, and data visualization libraries.
What's next for MyMuscle
Expanding the exercise library with more routines and difficulty levels.
Adding form correction tips using AI-powered feedback and voice guidance.
Implementing wearable integration for heart rate and calorie estimation.
Introducing social features, like sharing progress, challenges, and community leaderboards.
Launching guided workout sessions with video/audio prompts for an even more immersive experience.
Continuing to optimize performance and detection accuracy, especially for low-end mobile devices.
Log in or sign up for Devpost to join the conversation.