FitMate

Inspiration:

Problem: Uncertainty and guesswork in the gym, especially for newcomers. Solution: FitMate, a precise form analysis tool that acts as a personal trainer, guiding users through exercises with accuracy. What it does:

Functionality: Analyzes exercise form using uploaded videos, focusing on metrics like shoulder alignment and squat depth. Benefits: Provides personalized feedback to improve workouts and reduce injury risk, making fitness goals more achievable. How we built it:

Frontend: Used React Native and Expo for cross-platform development, enabling rapid prototyping and compatibility. Backend: Employed Python and Flask for efficient RESTful API endpoints and compatibility with machine learning models. Containerization: Utilized Docker for server backend containerization, enhancing scalability and maintainability. Computer Vision: Integrated YOLO for real-time object detection and pose estimation, enhancing user experience. Challenges:

Docker Configuration: Initial setup complexities with Docker container. React Native/Expo: Learning curve for mobile development, transitioning from web development. File Transmission: Difficulty in transmitting large video files as multipart data, solved through compression and base64 encoding. Accomplishments:

Successfully built an application highlighting key metrics in squatting form. Demonstrated robustness in identifying good and bad form, providing tips for injury reduction and performance improvement. Potential for scaling to support more exercises across weightlifting. What we learned:

Enhanced mobile development skills using React Native/Expo. Gained insights into YOLO and pose estimation techniques. Learned about Docker configuration and large file streaming management. What's next:

Adding support for more exercises like deadlift, bench press, and yoga. Introducing profiles for tracking fitness progress and connecting with gym friends. Continuing development beyond LAHacks 2024 to enhance FitMate's capabilities.

Built With

• java