Deepfake Detection API

Production-ready deepfake detection service using EfficientNetB4 for Digital Identity Wallet integration.

🚀 Features

• EfficientNetB4 Model: State-of-the-art deepfake detection
• Fast Inference: Optimized for production use
• Grad-CAM Visualization: Visual explanations of predictions
• RESTful API: Easy integration with mobile/web apps
• Docker Support: Containerized deployment
• Health Checks: Built-in monitoring endpoints

📋 API Endpoints

1. Health Check

GET /health

2. Deepfake Detection (Recommended)

POST /api/v1/detect/deepfake
Content-Type: multipart/form-data

Parameters:
  - file: image file (JPEG, PNG)
  - include_visualization: boolean (default: true)

Response:
{
  "is_authentic": true,
  "confidence": 0.9234,
  "label": "Real",
  "probability_fake": 0.0766,
  "probability_real": 0.9234,
  "model": "EfficientNetB4",
  "processing_time_ms": 245.67,
  "visualizations": {
    "original": "base64_encoded_image",
    "gradcam_heatmap": "base64_encoded_image",
    "overlay": "base64_encoded_image"
  }
}

3. Legacy Endpoint (Backward Compatible)

POST /predict
Content-Type: multipart/form-data

Parameters:
  - file: image file

4. Liveness Detection

POST /api/v1/verify/liveness
Content-Type: multipart/form-data

Parameters:
  - file: image file

🛠️ Installation

Local Development

1. Install dependencies:

pip install -r requirements.txt

2. Place your model:

# Ensure your model is at: model/effb4_ai_detector.pt

3. Run the server:

uvicorn app:app --host 0.0.0.0 --port 8080 --reload

4. Test the API:

curl -X POST "http://localhost:8080/api/v1/detect/deepfake" \
  -F "file=@test_image.jpg"

Docker Deployment

1. Build the image:

docker build -t deepfake-detector .

2. Run the container:

docker run -p 8080:8080 deepfake-detector

3. Test:

curl http://localhost:8080/health

🧪 Testing

# Test with sample image
curl -X POST "http://localhost:8080/api/v1/detect/deepfake" \
  -F "file=@sample_real.jpg" \
  -F "include_visualization=true"

# Health check
curl http://localhost:8080/health

📝 Model Information

• Architecture: EfficientNetB4
• Input Size: 380x380 pixels
• Parameters: ~19M
• Model Size: ~70MB
• Training: Fine-tuned on deepfake dataset

---

Built for Digital Identity Wallet Integration 🔐