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
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.
- Used a webcam to capture real-time images.
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.
- Used a breadboard-based setup with MAX30102 for heart rate & SpO2 and a temperature sensor for body temperature monitoring.
Real-Time Feedback System
- Displayed posture corrections and health insights on a Tkinter-based GUI running on another laptop.
- 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.
Log in or sign up for Devpost to join the conversation.