YogaAI is a real-time AI-powered web application that detects yoga poses using a webcam or uploaded video and provides instant posture feedback based on a trained machine learning model.
This repository contains the web application (frontend + backend integration).
The ML training code and model development are maintained in a separate repository:
👉 ML Repository: github.com/aayush-12321/YogAI_Project
- 🎥 Real-time pose detection using webcam
- 🤖 XGBoost multi-class classification model
- 🧍 Pose confidence score
⚠️ Incorrect posture feedback- 📊 Evaluation using confusion matrix, feature importance, learning curve, and 5-fold cross-validation
- 🔊 Voice guidance (Text-to-Speech)
- 🧾 Session-based feedback tracking
- Mountain Pose (Tadasana – Arms Down)
- Warrior II
- Plank
Webcam → MediaPipe Pose → Angle Feature Extraction →
Trained XGBoost Model → Pose Classification →
Real-time Feedback → Frontend Display
Frontend: HTML, CSS, JavaScript
Backend: Django , OpenCV, MediaPipe
Machine Learning: XGBoost, NumPy, Pandas
| Parameter | Value |
|---|---|
| Algorithm | XGBoost (Multi-class classifier) |
| Train/Test Split | 80/20 |
| Validation | 5-Fold Cross-Validation |
| Features | Joint angle–based biomechanical features |
For the full training pipeline, feature engineering, and experiments, see the ML repository:
👉 github.com/aayush-12321/YogAI_Project
1. Clone the repository
git clone https://github.com/aayush-12321/yoga-ai-webapp.git
cd yoga-ai-webapp2. Install dependencies
pip install -r Requirements.txt3. Run the backend server
python manage.py runserver4. Open in browser
http://127.0.0.1:8000
Allow webcam access when prompted.