🚀 Triton Inference Server

This repository provides a Triton Inference Server setup for face detection and face recognition using ONNX models with dynamic batching.

📌 Models Used

• Face Detection: RetinaFace-MobileNetV2 (Dynamic Batch, ONNX)
• Face Recognition: MobileNetV2 (Dynamic Batch, ONNX)

---

🛠️ Build & Run the Triton Server

1️⃣ Build the Docker Image

docker build -t tritonserver .

2️⃣ Run the Container

docker run --gpus all --rm -p 8000:8000 -p 8001:8001 -p 8002:8002 tritonserver:latest

---

🐳 Using Docker Compose (Recommended)

Start the Server

docker compose up --build -d

Stop the Server

docker compose down

---

Run API server

to visualize and test the API endpoints.

Run the API Server

uvicorn api.face_api:app --host 0.0.0.0 --port 8000 --reload

Test the API Endpoints

Open your browser and navigate to http://localhost:8000/docs to access the FastAPI Swagger UI. Here, you can test the available endpoints for face detection and recognition.

---

📌 TODO: TensorRT Optimization

To improve inference performance using TensorRT, enable GPU acceleration by adding the following configuration:

execution_accelerators {
  gpu_execution_accelerator : [
    {
      name : "tensorrt"
      parameters { key: "precision_mode" value: "FP16" }  # Run inference in FP16 for better performance
    }
  ]
}

✅ Upcoming Enhancements:

• Integrate TensorRT optimizations
• Improve model inference speed

---

📢 Contributions & Feedback

Feel free to open an issue or submit a PR if you have improvements or suggestions! 🚀

---

📜 License

This project is open-source and available under the MIT License.