About the Project

Inspiration

Yoga is a powerful practice for both physical and mental well-being, but improper posture can lead to inefficiency or even injuries. Traditional yoga training often requires expert supervision, which may not always be accessible. This inspired me to develop an AI-powered system that helps users correct their yoga postures in real time while also monitoring their health metrics like heart rate, SpO2, and temperature.

What I Learned

Throughout this project, I gained hands-on experience in computer vision, AI-based pose estimation, and IoT integration. I explored different pose estimation models like MediaPipe Pose and OpenPose and learned how to process real-time sensor data for biometric monitoring. Additionally, I enhanced my understanding of interfacing hardware components with ESP32 and streaming data to a GUI for visualization.

How I Built It

  1. Computer Vision for Posture Correction

    • Used a webcam to capture real-time images.
    • Implemented MediaPipe Pose to track body landmarks and compare poses with ideal yoga postures.
  2. Health Monitoring System

    • Used a breadboard-based setup with MAX30102 for heart rate & SpO2 and a temperature sensor for body temperature monitoring.
    • Connected sensors to an ESP32, which transmitted data for visualization.
  3. Real-Time Feedback System

    • Displayed posture corrections and health insights on a Tkinter-based GUI running on another laptop.

Challenges Faced

  • Pose Estimation Accuracy: Ensuring the AI model correctly detects and evaluates yoga poses under different lighting conditions.
  • Real-Time Data Processing: Synchronizing sensor data with pose estimation for seamless feedback.
  • Hardware Limitations: Using a breadboard setup instead of a wearable device made mobility limited, but it helped in rapid prototyping.

This project serves as a foundation for AI-driven personalized yoga training, and future improvements will include wireless wearables and deep learning-based pose correction models for enhanced accuracy. 🚀

Challenges We Ran Into

  • Pose Estimation Accuracy: Ensuring the AI model correctly detects and evaluates yoga poses under different lighting conditions and camera angles.
  • Real-Time Data Processing: Synchronizing sensor data from the ESP32 with pose estimation output to provide seamless feedback.
  • Hardware Limitations: Using a breadboard setup instead of a wearable device restricted mobility, but it was useful for rapid prototyping.
  • GUI Integration: Developing a Tkinter-based interface for real-time feedback and ensuring smooth data transmission between devices.

What We Learned

  • AI & Computer Vision: Gained experience in MediaPipe Pose, OpenPose, and real-time pose correction techniques.
  • IoT & Sensor Integration: Learned how to interface MAX30102 (heart rate & SpO2) and a temperature sensor with ESP32 and transmit data wirelessly.
  • Data Synchronization: Explored ways to efficiently combine biometric data with AI-driven posture detection for an enhanced user experience.
  • User Experience Design: Understood the importance of providing clear, actionable feedback to users for effective yoga practice.

What's Next for AI-Enabled Posture Correction and Health Monitoring for Yoga

  • Wearable Device Integration: Replace the breadboard setup with a compact wristband or smart wearable for better usability.
  • Advanced AI Models: Improve accuracy with deep learning-based pose correction instead of rule-based comparisons.
  • Web & Mobile App: Expand the system to a web or mobile platform for greater accessibility.
  • Voice & Haptic Feedback: Implement audio guidance and haptic alerts to assist users without needing a screen.
  • Personalized Yoga Plans: Use AI to analyze long-term health data and recommend customized yoga routines based on trends.

Built With

Share this project:

Updates