Inspiration
After long exam study sessions and day-to-day homework, we noticed how easy it was to fall into poor posture habits slouching shoulders, craned necks, and overall discomfort. We wanted to create something that helps people stay mindful of their posture in real time, improving both physical wellness and daily performance.
What it does
Posture Power is a real-time posture coaching system that monitors your posture through your webcam and provides gentle reminders to sit upright. Using Google MediaPipe Pose, the program identifies key body landmarks like the shoulders, hips, and ears to detect slouching or misalignment. When poor posture is detected, the backend classifies it and the frontend displays a message enhanced with Gemini AI for personalized feedback to encourage better ergonomic alignment.
How we built it
We built our backend using Python, FastAPI, and MediaPipe Pose to handle image processing and geometric posture analysis. By computing angles between key body landmarks, the program detects when the user deviates from their calibrated “good” posture. On the frontend, we used Next.js for real-time interaction, Figma and CSS for design, and the Webcam API to capture live video. The frontend sends frames to the backend, receives posture assessments, and displays them to the user with a clean, responsive interface.
Challenges we ran into
Data sensitivity: MediaPipe’s landmark tracking can fluctuate with lighting and camera quality, requiring careful smoothing and calibration logic. Frontend-backend synchronization: Managing real-time frame capture without overloading the API required optimization and throttling.
Accomplishments that we're proud of
We’re proud to have achieved accurate, real-time posture classification using pure geometric analysis rather than deep learning models. We also built a fully functional full-stack prototype that connects the frontend and backend seamlessly, delivering live posture feedback through an intuitive user interface.
What we learned
We deepened our understanding of computer vision, real-time web systems, and human–computer interaction design. We also learned how small calibration changes can greatly improve usability and accuracy for health-focused applications.
What's next for Posture Pal
We plan to:
- Add personalized posture analytics over time for long-term wellness tracking.
- Collaborate with ergonomics and health professionals to refine feedback accuracy and recommendations.
Log in or sign up for Devpost to join the conversation.