What inspired this project? I wanted to build a motivational tool that actually keeps me moving! The goal was to create a tracker that doesn't just watch you work out, but actively encourages you to stay consistent and improve.
What does MoveAI do? MoveAI is a real-time AI rehabilitation monitoring system. Using a standard webcam and MediaPipe pose estimation, it tracks human body movements to automatically count reps for exercises like squats, sit-ups, and jumping jacks. Beyond fitness, it’s designed for physical therapy, allowing users to deliver accurate movement data directly to their physicians.
How did you build it?
The system is built on a modular Python architecture:
Computer Vision: MediaPipe and OpenCV for landmark extraction.
Logic Engines: Custom modules for exercise detection and gesture control.
Intelligence: Gemini API for movement analysis and a memory system to track progress.
Audio/UX: ElevenLabs for a realistic voice assistant and a dedicated "CoachBot" for real-time feedback.
What challenges did you face? Honestly, everything! As a self-described "not-great coder," managing the logic was a steep mountain to climb. The biggest hurdle was calibration—getting the AI to distinguish between a proper rep and accidental "noise." Even now, rapid movements can sometimes trigger false counts, which is something I'm constantly refining.
What accomplishments are you most proud of? Mastering MediaPipe: Deep-diving into how pose estimation actually runs.
Technical Integration: Successfully managing different Python versions and dependencies to get all these complex APIs (Gemini, ElevenLabs, etc.) to talk to each other.
Persistence: Turning a difficult coding journey into a functional prototype.
What did you learn? I learned that the best way to understand a complex system is to build it from scratch. By struggling through the integration of different libraries, I gained a much deeper understanding of how data flows between an AI analyzer and a physical movement detector.
What’s next for MoveAI? I’m moving from macro-movements to micro-precision. I want to adapt the technology for musicians—specifically for flute players. The next version will monitor delicate posture adjustments and embouchure (mouth positioning) to help musicians maintain perfect form during long practice sessions.
Log in or sign up for Devpost to join the conversation.