🐍 Python Backend - Iris Stress Detection API

Flask-based REST API for ML-powered stress detection from eye images.

🚀 Quick Start

1. Install Dependencies

pip install -r requirements.txt

2. Start the Server

python app.py

Server will start on http://localhost:5000

📡 API Endpoints

Health Check

GET /
GET /health

Predict Stress Level

POST /predict

Body (multipart/form-data):
  - image: File (required) - Eye image (JPG/PNG)
  - age: Integer (optional) - User age (default: 30)

Response:
{
  "success": true,
  "prediction": {
    "stress_level": "Normal" | "Stress",
    "stress_probability": 0.76,
    "stress_percentage": 76.0,
    "confidence": "High" | "Medium" | "Low"
  },
  "pupil_detection": {...},
  "iris_detection": {...},
  "measurements": {...}
}

🏗️ Architecture

Frontend (React)
    ↓
Flask API (Port 5000)
    ↓
├── Detection Pipeline
├── Measurement Pipeline
└── ML Model (TensorFlow)

📁 Project Structure

Python_Backend/
├── app.py                 # Flask application
├── config.py              # Configuration
├── requirements.txt       # Dependencies
├── detection/             # Eye detection modules
├── measurement/           # Pupil & iris measurements
├── pipeline/              # ML inference pipeline
├── layers/                # Custom TensorFlow layers
├── utils/                 # Preprocessing utilities
└── Model/                 # Trained model files

🔧 Configuration

Edit config.py to customize:

• Model path
• Detection thresholds
• Stress classification threshold
• Image preprocessing parameters

🐛 Troubleshooting

Model Not Loading

• Check Model/best_dual_stream_age_aware_model.keras exists
• Verify TensorFlow installation: python -c "import tensorflow; print(tensorflow.__version__)"

Detection Failing

• Ensure image clearly shows pupil and iris
• Try different lighting conditions
• Image should be at least 200x200 pixels

CORS Errors

• Check Flask CORS configuration in app.py
• Frontend must run on port 5173 or 5174

📊 Model Details

• Type: Dual-stream CNN (Pupil + Iris)
• Architecture: EfficientNetB0 + Weighted Feature Fusion
• Training Accuracy: 99.91% AUC-PR
• Input: Eye images (224x224x5 channels)
• Output: Stress probability (0-1)

⚡ Performance

• Average inference time: ~2-3 seconds
• Supports concurrent requests (Flask threaded mode)
• Model loaded once at startup

🔒 Security

• File validation (image types only)
• Size limits enforced
• Temporary files auto-deleted
• No sensitive data stored

📝 Notes

• This backend is independent from the Node.js server
• Both servers run simultaneously
• Node.js handles: Auth, MongoDB, file storage
• Flask handles: ML predictions only