DeepStream SDK on Jetson

Rapidly build intelligent video analytics applications at the edge.

NVIDIA DeepStream SDK on Jetson makes it easy for developers to create and deploy AI-based intelligent video analytics (IVA) capabilities in their products. Application developers can use this SDK to rapidly prototype and build products ranging from intelligent cameras to appliances for applications in smart cities, robotics, and industrial automation.

With Jetson’s advanced AI capabilities and rich set of imaging and I/Os, Jetson enables developers to build highly integrated systems that can be deployed at the edge in small form factors for as little as 7.5W.

DeepStream SDK on Jetson uses Jetpack, which includes L4T, Multimedia APIs, CUDA, and TensorRT. The SDK offers a rich collection of plug-ins and libraries, built using the Gstreamer framework to enable developers to build flexible applications for transforming video into valuable insights. DeepStream also comes with sample applications including source code and an application adaptation guide to help developers jumpstart their builds.

Developing IVA applications for complete scene understanding entails creating multimedia processing graphs and orchestrating the dataflow of multiple DNNs working in sync. This can be a complex and time consuming process. This SDK is designed to remove the complexity, so you can focus on application development. By leveraging the Gstreamer plug-in architecture, DeepStream makes it easier than ever to combine image capture, encoding, decoding, scaling, and inference using TensorRT. The architecture lets you utilize all of Jetson’s hardware capabilities to maximize throughput and performance. By leveraging Jetson’s unified memory architecture, DeepStream SDK enables you to reduce memory management overhead and deliver low latency solutions.


  • Modular and scalable framework for building AI-based video analytics application based on GStreamer plugins
  • Application configuration using a simple key file format
  • Optimized memory management like zero buffer copies to maximize throughput and performance
  • Samples include end-to-end application combining multiple deep learning networks to transform pixels into rich metadata
  • Pre-trained example DNNs to classify objects such as cars and pedestrians and to understand their attributes (e.g. car make, car color, and car type) based on ResNet and GoogLeNet
  • Multi-stream input from sources like USB and CSI cameras, RTSP stream inputs, or recorded files from disk
  • Multiple output sinks included rendering to display, metadata logging to file, and saving to disk
  • Deploy neural networks in full (FP32) or optimized precision (FP16)
  • Optimized for NVIDIA Jetson


  • Jetpack 3.2 which includes L4T R28.2, CUDA 9.0, TensorRT 3.0 GA, cuDNN 7.0.5, VisionWorks 1.6