Project Overview & Goal
TrueForm is a real-time augmented reality (AR) platform leveraging AI pose estimation to revolutionize at-home fitness.
- Mission: Train Smarter – Not Harder
- Goal: Reduce injury risk and maximize workout efficacy by delivering immediate, precise guidance in an accessible environment.
- Approach: Overlay a dynamic skeletal guide onto the user’s video feed to instantly correct form deviations.
The Opportunity
Traditional video tutorials are passive, preventing users from observing themselves while learning. This leads to:
- Injury Risk: 48% increase in ER visits due to home exercise injuries (2019–2021)
- Retention Deficit: Users forget ~70% of new cues within 24 hours
- User Demand: Professional users want tools to monitor technique at home but lack access
The Solution
TrueForm bridges the gap between professional coaching and home workouts through a browser-based “Smart Mirror” experience:
- Real-Time Pose Analysis: Detects key body points and analyzes movement geometry
- Immediate Feedback: Skeleton changes color (green → red) to indicate form deviations
- Prevent Injuries: Corrects exercises before they lead to strain or injury
Target Audience
Primary users:
- Home Fitness Enthusiasts: Maintain effective and safe workouts at home
- Performance-Focused Users: Focus on strength, form, and measurable progress
- Independent Trainees: Seek personalized guidance digitally without a trainer
Design Process & Prototype
Design focused on user experience and intuitive interactions:
- Ideation: Brainstormed flows and interface sketches
- Wireframes: Low-fidelity layouts for camera setup, real-time feedback, and metrics
- Prototype: Clickable AR demo showing skeleton overlay and visual alerts
- Iteration: Collected feedback and refined skeleton colors, threshold sensitivity, and status bar
Key Features
Real-Time Form Correction:
Immediate visual feedback via skeleton color change to prevent injury.Repetition & Session Tracking:
Automatic rep counting and saving session metrics to Firestore for progress monitoring.Dynamic Status Feedback:
Text-based status bar displays “Waiting…”, “Form Correct”, or “Warning!” to complement visual cues.Accessibility & Privacy:
Browser-based solution with low latency ensures easy access without app installation and secure data storage.Technical Stack & Implementation
Built using a privacy-first, browser-based architecture with JavaScript and JSON:
- AI & Computer Vision: TensorFlow.js + MoveNet (Lightning) for sub-millisecond pose estimation
- Algorithmic Logic: Trigonometric calculations for vectors (Hip, Knee, Ankle)
- Thresholds:
- Squat Down: Knee Angle < 120°
- Form Error: Hip Angle < 170° → Red warning
- Web Technologies: JavaScript ES6 Modules, Firebase Firestore for data persistence and secure authentication
- Data Handling: Used JSON to structure session and user data
Future Roadmap & Impact
Short-Term: Add support for complex movements like Deadlifts and LungesMedium-Term: Implement audio coaching with Text-to-Speech to reduce screen dependencyLong-Term: Introduce gamification, Form Scores, and social leaderboards to enhance engagement and communityReflection & Learnings
Design Insights:
Developing TrueForm taught me how to create an intuitive AR experience that combines real-time AI feedback with clear visual guidance, improving user safety and engagement.Challenges Overcome:
Balancing technical complexity with usability was key. Implementing sub-millisecond pose estimation, accurate angle detection, and responsive visual feedback required careful calibration and testing.Future Considerations:
Adding audio coaching, more exercises, and gamified progress metrics could further enhance user experience and motivation while maintaining safety.
.jpg)
.jpg)
.jpg)
.jpg)
.jpg)
.jpg)
.jpg)
.jpg)
.jpg)
