Home > Servers > PowerEdge Components > White Papers > Developing and Deploying Vision AI with Dell and NVIDIA Metropolis > NVIDIA DeepStream and Triton integration
The integration of DeepStream and Triton enables developers to quickly create prototypes for end-to-end intelligent video-analytic solutions and deploy models from multiple open-source frameworks such as TensorFlow, TensorRT, PyTorch, ONNX-Runtime. It also enables creating custom framework back ends on both NVIDIA T4 and Jetson platforms. DeepStream 6.0 integrates Triton Server directly from the application.
The following figure shows the inference service that is natively integrated with DeepStream through the plug-in Gst-nvinferserver through API.