DeepStream SDK 6.0


  • Graph Composer. Assemble complex pipelines using an intuitive and easy-to-use UI and quickly deploy them with Container Builder.
  • Action Recognition. Create pipelines to easily identify action in your scene with the new pre-processing plugin. Specify regions of interest and the number of frames you want to process simultaneously.
  • Audio-video synchronization for applications such as broadcasting and web conferencing.
  • Automatic Speech Recognition (ASR) support. This new plugin performs automatic speech recognition on input audio data and outputs transcribed text.
  • Source code for the Python Bindings available on
  • Tracker Updates
    • Performance Improvements
    • Easy integration of custom trackers.
    • New DeepSORT tracker.
  • Full REDIS support, previously alpha in version 5.1
  • New and updated sample applications available on the SDK and NVIDIA IOT GitHub. Check DeepStream documentation for additional details:
    • ASR
    • Action Recognition
    • Body Pose Estimation
    • Emotion
    • Gestures
    • Many more.
Jetson Tesla GPUs(x86)
Operating System Ubuntu 18.04 Ubuntu 18.04
Dependencies CUDA: 10.2.460
cuDNN: 8.2+
TensorRT: 8.0.1
JetPack: 4.6
CUDA: 11.4
cuDNN: 8.2+
TensorRT: 8.0.1
Driver: R470.63.01+

Getting Started Resources

Product Advisory

DeepStream 6.0 represents a major release, including significant improvements over the earlier versions. These new features and capabilities take advantage of TensorRT 8, providing best-in-class inference optimization and performance.

To ensure compatibility with TensorRT 8, users planning to use models developed with TAO Toolkit (formerly TLT) 3.0-21.08 or earlier MUST re-generate the INT8 calibration cache before using them with DeepStream 6.0.

Those who are using models and the INT8 calibration cache from previous versions of TensorRT will also need to re-generate the cache.

You can find details regarding regenerating the cache in the Readme First section of the documentation.

For new DeepStream developers or those not reusing old models, this is NOT an issue.


I Agree To the Terms of the NVIDIA DeepStream SDK 6.0 Software License Agreement

DeepStream 5.x applications are fully compatible with DeepStream 6.0. Please read the migration guide for more information.

Python Sample Apps & Bindings

The Python bindings source code and pre-built wheels are now available on GitHub. The Bindings build is no longer included with the DeepStream SDK.

Visit the DeepStream Python Apps Github page for documentation and sample apps.

Check out the DeepStream SDK technical FAQ for questions commonly asked.


  • Check out the frequently asked questions on DeepStream SDK in the technical FAQ

Documentation & Forums

Reference Implementations

Success Stories

Blogs & Tutorials

Beginner Friendly Free Self-Paced DLI Online Courses

  • Learn how to build end-to-end intelligent video analytics pipelines using DeepStream and Jetson Nano >> Enroll now
  • Learn how to get started with AI using Jetson Nano >> Enroll now


GTC 2021 November

Additional Resources

Ethical AI

NVIDIA’s platforms and application frameworks enable developers to build a wide array of AI applications. Consider potential algorithmic bias when choosing or creating the models being deployed. Work with the model’s developer to ensure that it meets the requirements for the relevant industry and use case; that the necessary instruction and documentation are provided to understand error rates, confidence intervals, and results; and that the model is being used under the conditions and in the manner intended.