Skip to navigation
DeepStream Getting Started

Installation

View as Markdown

NVIDIA ® DeepStream Software Development Kit (SDK) is an accelerated AI framework to build intelligent video analytics (IVA) pipelines. DeepStream runs on NVIDIA ® T4, NVIDIA® Hopper, NVIDIA ® Blackwell, NVIDIA ® Ampere, NVIDIA ® ADA, NVIDIA® DGX Spark, and platforms such as NVIDIA ® Jetson AGX Thor, NVIDIA ® Jetson Orin and NVIDIA ® IGX Thor. For dGPU platforms Enterprise GPUs are highly recommended for deployments that are expected to run 24x7. Gaming GPUs are not designed to perform in such type of environments.

Install from the GitHub Mono-repo (Build from Source)

The DeepStream mono-repo at https://github.com/NVIDIA/DeepStream is the recommended starting point for DeepStream 9.2. It contains the open-source components (GStreamer plugins, utilities, sample, reference and TAO applications, and Service Maker) together with build/build.sh, a single script that downloads the prebuilt proprietary runtime from the GitHub release, installs the open-source dependencies, builds the source components, and installs everything to /opt/nvidia/deepstream/deepstream-9.2/.

Choose an installation method

MethodUse when
Mono-repo build (build/build.sh)You want the sources, or need to modify, rebuild or extend DeepStream components.
Release packages (.deb / .tbz2)You want a binary installation without building. See the platform sections below.
ContainerYou want a ready-made environment with the compute stack installed. See Docker Containers.

Prerequisites

  • The NVIDIA compute stack (driver, CUDA, cuDNN and TensorRT) for your platform. Install it by following the Jetson, dGPU or dGPU on ARM sections below, and match the versions listed in Platform and OS Compatibility.
  • git, git-lfs, cmake and wget.
  • SBSA / DGX Spark: bare-metal installation is not supported. Build inside the DeepStream SBSA container instead. See FAQ for DeepStream GitHub Mono-repo.

For the complete list of system packages per platform, see build/BUILD.md in the mono-repo.

Build and install

# Install Git LFS (required for LFS-tracked sample media)
sudo apt-get install git-lfs && git lfs install
# Clone the repo with submodules
git clone --recurse-submodules https://github.com/NVIDIA/DeepStream.git && cd DeepStream
# Build and install. The script prompts for sudo only when writing to system paths.
bash build/build.sh

build/build.sh detects the platform and runs these stages in order:

  1. artifacts: downloads the prebuilt runtime libraries and sample data from the DeepStream GitHub release and installs them (x86 and Jetson).
  2. deps: builds and installs the open-source dependencies.
  3. Source stages: gst-utils, utils, gst-plugins, sample_apps, yolo, tao_apps, reference_apps and service-maker.
  4. install.sh: registers the binaries and installs the pyservicemaker Python wheel.

A transcript of each run is written to build/build.log.

Commonly used options:

OptionDescription
--only=STAGE[,STAGE]Build only the named component stage(s).
--resumeSkip stages already completed, after an interrupted build.
--skip-deps, --skip-artifactsSkip the dependency or artifact install when it is already done.
--install-method=<deb or tar>Install the release assets from Debian packages (deb, the default) or tarballs (tar).
--keep-assetsKeep the downloaded release assets in artifacts/ after a successful build.
-j NNumber of parallel build jobs.
--verboseShow sub-make errors.

The defaults are CUDA_VER=13.2 (13.0 on IGX) and NVDS_VERSION=9.2.0. Run bash build/build.sh --help for all options.

Verify the installation

After the first install, clear the GStreamer plugin cache, then run the reference application with a sample configuration:

rm -rf ~/.cache/gstreamer-1.0/
cd /opt/nvidia/deepstream/deepstream-9.2/samples/configs/deepstream-app
deepstream-app -c source30_1080p_dec_infer-resnet_tiled_display.txt

Notes

  • build/build.sh does not build gst-nvdsudp, gst-dsexample-cuda or gst-nvdstextembedder. Build these separately by following the README in each component directory.
  • Some sample applications are platform-specific and are skipped automatically on platforms where they are not supported.
  • On Jetson Orin, use external storage: the downloaded release assets can exhaust internal storage.
  • On SBSA / DGX Spark, to modify a single open-source component, run make in that component’s directory inside the container. See FAQ for DeepStream GitHub Mono-repo.

For the full build reference, including uninstall, Service Maker local compilation and packaging, see build/BUILD.md in the mono-repo.

Jetson Setup

This section explains how to prepare a Jetson device before installing the DeepStream SDK.

Steps to install DeepStream SDK locally, assume that ~/.local/bin/ has been added to the ~/.bashrc or ~/.profile

Install Jetson SDK components

Download NVIDIA SDK Manager from https://developer.nvidia.com/embedded/jetpack. You will use this to install JetPack 7.2.1 GA (corresponding to L4T 39.2 release)

  • NVIDIA SDK Manager is a graphical application which flashes and installs the JetPack packages.
  • The flashing procedure takes approximately 10-30 minutes, depending on the host system.

Install Dependencies

Install prerequisite packages

Enter the following commands to install the prerequisite packages:

$ sudo apt install \
libssl3 \
libssl-dev \
libcurl4-openssl-dev \
libgstreamer1.0-0 \
gstreamer1.0-tools \
gstreamer1.0-plugins-good \
gstreamer1.0-plugins-bad \
gstreamer1.0-plugins-ugly \
gstreamer1.0-libav \
libgstreamer-plugins-base1.0-dev \
libgstrtspserver-1.0-0 \
libjansson4 \
libyaml-cpp-dev \
libmosquitto1

Sometimes with RTSP streams the application gets stuck on reaching EOS. This is because of an issue in rtpjitterbuffer component. To fix this issue, On baremetal, run a script “update_rtpmanager.sh” at /opt/nvidia/deepstream/deepstream/. The script should be executed once user installs packages mentioned under “Install Dependencies” section of Quickstart Guide. On docker, users need to run user_additional_install.sh script to use the prebuilt binaries.

Install OpenTelemetry

Enter the following commands to install OpenTelemetry:

$ cd ~
$ mkdir otel_src && cd otel_src && git clone --branch v1.23.0 --recurse-submodules https://github.com/open-telemetry/opentelemetry-cpp
$ cd opentelemetry-cpp
$ mkdir build && cd build
$ sudo apt install cmake
$ cmake .. -DBUILD_SHARED_LIBS=ON -DWITH_OTLP_GRPC=ON -DWITH_OTLP_HTTP=ON -DBUILD_TESTING=OFF
$ cmake --build . --parallel $(nproc)
$ ctest -V (Optional: run tests for verification)
$ sudo cmake --install .

For more information, refer to the OpenTelemetry C++ installation guide: https://github.com/open-telemetry/opentelemetry-cpp/blob/v1.23.0/INSTALL.md

Install librdkafka (to enable Kafka protocol adaptor for message broker)

  1. Install confluent-platform:
sudo mkdir -p /etc/apt/keyrings
wget -qO - https://packages.confluent.io/deb/8.0/archive.key | gpg \
--dearmor | sudo tee /etc/apt/keyrings/confluent.gpg > /dev/null
CP_DIST=$(lsb_release -cs)
echo "Types: deb
URIs: https://packages.confluent.io/deb/8.0
Suites: stable
Components: main
Architectures: $(dpkg --print-architecture)
Signed-by: /etc/apt/keyrings/confluent.gpg
Types: deb
URIs: https://packages.confluent.io/clients/deb/
Suites: ${CP_DIST}
Components: main
Architectures: $(dpkg --print-architecture)
Signed-By: /etc/apt/keyrings/confluent.gpg" | sudo tee /etc/apt/sources.list.d/confluent-platform.sources > /dev/null
sudo apt-get update && sudo apt-get install confluent-platform
  1. Install librdkafka:
sudo apt-get install librdkafka-dev

Install latest NVIDIA BSP packages

Installation of JetPack 7.2.1 GA will ensure that latest NVIDIA BSP packages are installed.

Install the DeepStream SDK

Recommended for new users: use Method 4 (Docker container) for the quickest, dependency-free setup. For native installs, the Debian (Method 3) and tar (Method 2) packages are published as GitHub release assets at https://github.com/NVIDIA/DeepStream/releases/tag/v9.2.0.

  • Method 1: Using SDK Manager

    Select DeepStreamSDK from the Additional SDKs section along with JP 7.2.1 GA software components for installation.

  • Method 2: Using the DeepStream tar package: https://github.com/NVIDIA/DeepStream/releases/tag/v9.2.0

    1. Download the DeepStream 9.2 Jetson tar package deepstream_sdk_v9.2.0_jetson.tbz2 to the Jetson device.

    2. Enter the following commands to extract and install the DeepStream SDK:

$  sudo tar -xvf deepstream_sdk_v9.2.0_jetson.tbz2 -C /
$ cd /opt/nvidia/deepstream/deepstream-9.2
$ sudo ./install.sh
$ sudo ldconfig
$ sudo apt-get install ./deepstream-9.2_9.2.0-1_arm64.deb

For scripted or headless downloads of the release assets using curl or wget, see FAQ for DeepStream GitHub Mono-repo.

Verification: Once DeepStream SDK installation is successful, refer to Expected output (deepstream-app) for the expected output.

dGPU Setup for Ubuntu

This section explains how to prepare an Ubuntu x86_64 system with NVIDIA dGPU devices before installing the DeepStream SDK.

Steps to install DeepStream SDK locally, assume that ~/.local/bin/ has been added to the ~/.bashrc or ~/.profile

This document uses the term dGPU (“discrete GPU”) to refer to NVIDIA GPU expansion card products such as NVIDIA Tesla ® T4, NVIDIA ® Hopper, NVIDIA ® Blackwell, NVIDIA ® Ampere, NVIDIA ® ADA, NVIDIA GeForce ® RTX 2080, NVIDIA GeForce ® RTX 3080, NVIDIA GeForce ® RTX 4080, NVIDIA GeForce ® RTX pro 4500 and GeForce ®/NVIDIA RTX/QUADRO. This version of DeepStream SDK runs on GPUs supported by NVIDIA driver 595.58.03 and NVIDIA TensorRT™ 10.16.0.72 and later versions.

Prerequisites

You must install the following components:

  • Ubuntu 24.04
  • GStreamer 1.24.2
  • NVIDIA driver 595.58.03
  • CUDA 13.2
  • TensorRT 10.16.0.72

Remove all previous DeepStream installations

Enter the following commands to remove all previous DeepStream 3.0 or prior installations:

$ sudo rm -rf /usr/local/deepstream /usr/lib/x86_64-linux-gnu/gstreamer-1.0/libgstnv* /usr/bin/deepstream* /usr/lib/x86_64-linux-gnu/gstreamer-1.0/libnvdsgst*
/usr/lib/x86_64-linux-gnu/gstreamer-1.0/deepstream*
/opt/nvidia/deepstream/deepstream*
$ sudo rm -rf /usr/lib/x86_64-linux-gnu/libv41/plugins/libcuvidv4l2_plugin.so

To remove DeepStream 4.0 or later installations:

  1. Open the uninstall.sh file in /opt/nvidia/deepstream/deepstream/
  2. Run the following script as sudo ./uninstall.sh

Install Dependencies

Install prerequisite packages

Enter the following commands to install the necessary packages before installing the DeepStream SDK:

$ sudo apt install \
libssl3 \
libssl-dev \
libcurl4-openssl-dev \
libgles2-mesa-dev \
libgstreamer1.0-0 \
gstreamer1.0-tools \
gstreamer1.0-plugins-good \
gstreamer1.0-plugins-bad \
gstreamer1.0-plugins-ugly \
gstreamer1.0-libav \
libgstreamer-plugins-base1.0-dev \
libgstrtspserver-1.0-0 \
libjansson4 \
libyaml-cpp-dev \
libjsoncpp-dev \
protobuf-compiler \
libmosquitto1 \
gcc \
make \
git \
python3

Sometimes with RTSP streams the application gets stuck on reaching EOS. This is because of an issue in rtpjitterbuffer component. To fix this issue, On baremetal, run a script “update_rtpmanager.sh” at /opt/nvidia/deepstream/deepstream/. The script should be executed once user installs packages mentioned under “Install Dependencies” section of Quickstart Guide. On docker, users need to run user_additional_install.sh script to use the prebuilt binaries.

For OpenTelemetry installation, follow the steps in Install OpenTelemetry under Jetson Setup.

Install CUDA Toolkit 13.2

Run the following commands (reference, https://developer.nvidia.com/cuda-downloads):

$ sudo apt-key adv --fetch-keys https://developer.download.nvidia.com/compute/cuda/repos/ubuntu2404/x86_64/3bf863cc.pub
$ sudo add-apt-repository "deb https://developer.download.nvidia.com/compute/cuda/repos/ubuntu2404/x86_64/ /"
$ sudo apt-get update
$ sudo apt-get install cuda-toolkit-13-2

If you observe following errors while CUDA installation, refer to https://developer.nvidia.com/blog/updating-the-cuda-linux-gpg-repository-key/.

W: GPG error: https://developer.download.nvidia.com/compute/cuda/repos/ubuntu2404/x86_64 InRelease: The following signatures couldn't be verified because the public key is not available: NO_PUBKEY A4B469963BF863CC
E: The repository 'https://developer.download.nvidia.com/compute/cuda/repos/ubuntu2404/x86_64 InRelease' is no longer signed.

Install NVIDIA driver 595.58.03

$ chmod 755 NVIDIA-Linux-x86_64-595.58.03.run
$ sudo ./NVIDIA-Linux-x86_64-595.58.03.run --no-cc-version-check

Ensure gdm, lightdm or Xorg service is stopped while installing nvidia driver

Use command : sudo service gdm stop sudo service lightdm stop sudo pkill -9 Xorg

Install TensorRT 10.16.0.72

Run the following command to install TensorRT 10.16.0.72:

version="10.16.0.72-1+cuda13.2"
sudo apt-get install libnvinfer-dev=${version} libnvinfer-dispatch-dev=${version} \
libnvinfer-dispatch10=${version} libnvinfer-headers-dev=${version} libnvinfer-headers-plugin-dev=${version} \
libnvinfer-safe-headers-dev=${version} libnvinfer-lean-dev=${version} libnvinfer-lean10=${version} \
libnvinfer-plugin-dev=${version} libnvinfer-plugin10=${version} libnvinfer-vc-plugin-dev=${version} \
libnvinfer-vc-plugin10=${version} libnvinfer10=${version} libnvonnxparsers-dev=${version} \
libnvonnxparsers10=${version} tensorrt-dev=${version} \
libnvinfer-headers-python-plugin-dev=${version} libnvinfer-win-builder-resource10=${version}

Install librdkafka (to enable Kafka protocol adaptor for message broker)

  1. Install confluent-platform:
sudo mkdir -p /etc/apt/keyrings
wget -qO - https://packages.confluent.io/deb/8.0/archive.key | gpg \
--dearmor | sudo tee /etc/apt/keyrings/confluent.gpg > /dev/null
CP_DIST=$(lsb_release -cs)
echo "Types: deb
URIs: https://packages.confluent.io/deb/8.0
Suites: stable
Components: main
Architectures: $(dpkg --print-architecture)
Signed-by: /etc/apt/keyrings/confluent.gpg
Types: deb
URIs: https://packages.confluent.io/clients/deb/
Suites: ${CP_DIST}
Components: main
Architectures: $(dpkg --print-architecture)
Signed-By: /etc/apt/keyrings/confluent.gpg" | sudo tee /etc/apt/sources.list.d/confluent-platform.sources > /dev/null
sudo apt-get update && sudo apt-get install confluent-platform
  1. Install librdkafka:
sudo apt-get install librdkafka-dev

Install the DeepStream SDK

Recommended for new users: use Method 3 (Docker container) for the quickest, dependency-free setup. For native installs, the Debian (Method 1) and tar (Method 2) packages are published as GitHub release assets at https://github.com/NVIDIA/DeepStream/releases/tag/v9.2.0.

$ sudo apt-get install ./deepstream-9.2_9.2.0-1_amd64.deb
$ sudo tar -xvf deepstream_sdk_v9.2.0_x86_64.tbz2 -C /
$ cd /opt/nvidia/deepstream/deepstream-9.2/
$ sudo ./install.sh
$ sudo ldconfig

For scripted or headless downloads of the release assets using curl or wget, see FAQ for DeepStream GitHub Mono-repo.

Verification: Once DeepStream SDK installation is successful, refer to Expected output (deepstream-app) for the expected output.

dGPU on ARM Setup for Ubuntu

This section explains how to prepare an Ubuntu aarch64 system with NVIDIA dGPU devices before installing the DeepStream SDK.

You must install the following components:

  • NVIDIA driver 595.58.03
  • Docker
  • Nvidia Container Toolkit

Install NVIDIA driver 595.58.03

$ chmod 755 NVIDIA-Linux-aarch64-595.58.03.run
$ sudo ./NVIDIA-Linux-aarch64-595.58.03.run --no-cc-version-check

Ensure gdm, lightdm or Xorg service is stopped while installing Nvidia driver

Use command : sudo service gdm stop sudo service lightdm stop sudo pkill -9 Xorg

Run dGPU on ARM Docker (SBSA)

A compatible NVIDIA driver must be installed on the host for SBSA. The driver is not included in the Docker image; the NVIDIA Container Toolkit makes the host driver available to the container.

  1. Pull the DeepStream Triton Inference Server docker
docker pull ghcr.io/nvidia/deepstream:9.2-triton-sbsa-dgx-spark
  1. Start the docker
sudo docker run -it --rm --runtime=nvidia --network=host -e NVIDIA_DRIVER_CAPABILITIES=compute,utility,video,graphics --gpus all -e DISPLAY=:0 -v /tmp/.X11-unix:/tmp/.X11-unix -v /etc/X11:/etc/X11 ghcr.io/nvidia/deepstream:9.2-triton-sbsa-dgx-spark

--privileged is omitted above; add it only if the container needs access to all host devices or low-level driver operations. It grants near-root host access, so on shared or production systems prefer --device / --cap-add instead.

DGX Spark Setup for Ubuntu

This section explains how to prepare a NVIDIA DGX Spark system before installing DeepStream SDK. DGX Spark is an aarch64-based desktop AI system running Ubuntu with an integrated NVIDIA Grace Blackwell GPU.

Native DeepStream SDK installation on DGX Spark is not supported. Use the Docker container method described below.

A compatible NVIDIA driver must be installed on the host for DGX Spark. The driver is not included in the Docker image; the NVIDIA Container Toolkit makes the host driver available to the container.

Refer NVIDIA DGX Spark Release Notes for stack details.

Run DGX Spark Docker

  1. Pull the DeepStream Triton Inference Server docker
docker pull ghcr.io/nvidia/deepstream:9.2-triton-sbsa-dgx-spark
  1. Start the docker
sudo docker run -it --rm --runtime=nvidia --network=host \
    -e NVIDIA_DRIVER_CAPABILITIES=compute,utility,video,graphics \
    --gpus all \
    -e DISPLAY=:0 \
    -v /tmp/.X11-unix:/tmp/.X11-unix \
    -v /etc/X11:/etc/X11 \
    ghcr.io/nvidia/deepstream:9.2-triton-sbsa-dgx-spark
  • --privileged is omitted above; add it only if the container needs access to all host devices or low-level driver operations. It grants near-root host access, so on shared or production systems prefer --device / --cap-add instead.
  • The triton docker for DGX Spark is based on the SBSA Tritonserver image and has Ubuntu 24.04.
  • When the triton docker is launched for the first time, it might take a few minutes to start since it has to generate compute cache.
  • DGX Spark docker currently supports only nv3dsink for video display. nveglglessink is not supported.
  • VIC is not supported on DGX Spark. Set tiler compute-hw=1 (GPU) in your application config.

dGPU Setup for RedHat Enterprise Linux (RHEL)

DeepStream for RHEL is not supported in this release.

Platform and OS Compatibility

The following table provides information about platform and operating system compatibility in the current and earlier versions of DeepStream.

Jetson model Platform and OS Compatibility

DS releaseDS 6.1.1DS 6.2DS 6.3DS 6.4DS 7.0DS 7.1DS 8.0DS 9.0DS 9.1DS 9.2
Jetson platformsAGX Xavier, Jetson NX, Jetson OrinXavier AGX , Xavier NX, AGX Orin, Orin NXXavier AGX , Xavier NX, AGX Orin, Orin NX, Orin nanoAGX Orin, Orin NX, Orin nanoAGX Orin, Orin NX, Orin nanoAGX Orin, Orin NX, Orin nanoAGX ThorAGX ThorAGX Thor, Jetson OrinAGX Thor, Jetson Orin, IGX Thor
OSL4T Ubuntu 20.04L4T Ubuntu 20.04L4T Ubuntu 20.04L4T Ubuntu 22.04L4T Ubuntu 22.04L4T Ubuntu 22.04L4T Ubuntu 24.04L4T Ubuntu 24.04L4T Ubuntu 24.04L4T Ubuntu 24.04
JetPack release5.0.2 GA Revision 15.1 GA5.1.2 GA6.0 DP6.0 GA6.1 GA7.0 GA7.1 GA7.2 GA7.2.1 GA
L4T release35.135.2.135.436.236.336.438.238.439.239.2
CUDA releaseCUDA 11.4CUDA 11.4CUDA 11.4CUDA 12.2CUDA 12.2CUDA 12.6CUDA 13.0CUDA 13.0CUDA 13.2CUDA 13.2
cuDNN releasecuDNN 8.4.1.50+cuDNN 8.6.0.166+cuDNN 8.6.0.166+cuDNN 8.9.4.25+cuDNN 8.9.4.25+cuDNN 9.3.0CuDNN 9.12.0CuDNN 9.12.0CuDNN 9.20.0.46CuDNN 9.20.0.48
TensorRT releaseTRT 8.4.1.5TRT 8.5.2.2TRT 8.5.2.2TRT 8.6.2.3TRT 8.6.2.3TRT 10.3.0.31TRT 10.13.2.6TRT 10.13.2.6TRT 10.16.1.7TRT 10.16.1.7
OpenCV releaseOpenCV 4.2.0OpenCV 4.2.0OpenCV 4.5.4OpenCV 4.8.0OpenCV 4.8.0OpenCV 4.8.0OpenCV 4.8.0OpenCV 4.8.0OpenCV 4.8.0OpenCV 4.8.0
VisionWorksNANANANANANANANANANA
GStreamerGStreamer 1.16.2GStreamer 1.16.3GStreamer 1.16.3GStreamer 1.20.3GStreamer 1.20.3GStreamer 1.20.3GStreamer 1.24.2GStreamer 1.24.2GStreamer 1.24.2GStreamer 1.24.2
Docker imagedeepstream-l4t:6.1.1deepstream-l4t:6.2deepstream-l4t:6.3deepstream:6.4deepstream:7.0deepstream:7.1deepstream:8.0deepstream:9.0deepstream:9.1deepstream:9.2

dGPU model Platform and OS Compatibility

DS releaseDS 6.1.1DS 6.2DS 6.3DS 6.4DS 7.0DS 7.1DS 8.0DS 9.0DS 9.1DS 9.2
GPU platformsT4, V100, A2, A10, A30, A100, RTX Ampere (Ax000/RTX30x0)T4, V100, A2, A10, A30, A100, RTX Ampere (Ax000/RTX30x0), Hopper, ADAT4, V100, A2, A10, A30, A100, RTX Ampere (Ax000/RTX30x0), Hopper, ADAT4, V100, A2, A10, A30, A100, RTX Ampere (Ax000/RTX30x0), Hopper, ADAT4, V100, A2, A10, A30, A100, RTX Ampere (Ax000/RTX30x0), Hopper, ADAT4, A2, A10, A30, A100, RTX Ampere (Ax000/RTX30x0), Hopper, ADATuring, Ampere, Hopper, ADA, BlackwellTuring, Ampere, Hopper, ADA, BlackwellTuring, Ampere, Hopper, ADA, BlackwellTuring, Ampere, Hopper, ADA, Blackwell
OSUbuntu 20.04Ubuntu 20.04Ubuntu 20.04Ubuntu 22.04Ubuntu 22.04Ubuntu 22.04Ubuntu 24.04Ubuntu 24.04Ubuntu 24.04Ubuntu 24.04
GCCGCC 9.4.0GCC 9.4.0GCC 9.4.0GCC 11.4.0GCC 11.4.0GCC 11.4.0GCC 11.4.0GCC 11.4.0GCC 11.4.0GCC 11.4.0
CUDA releaseCUDA 11.7.1CUDA 11.8CUDA 12.1CUDA 12.2CUDA 12.2CUDA 12.6CUDA 12.8CUDA 13.1CUDA 13.2CUDA 13.2
cuDNN releasecuDNN 8.4.1.50+cuDNN 8.7.0.84-1+cuDNN 8.8.1.3-1+cuDNN 8.9.4.25-1+cuDNN 8.9.6.50-1+cuDNN 9.3.0CuDNN 9.7.1CuDNN 9.18.0CuDNN 9.20.0.48CuDNN 9.20.0.48
TRT releaseTRT 8.4.1.5TRT 8.5.2.2TRT 8.5.3.1TRT 8.6.1.6TRT 8.6.1.6TRT 10.3.0.26TRT 10.9.0.34TRT 10.14.1.48TRT 10.16.0.72TRT 10.16.0.72
Display DriverR515.65.01R525.85.12R525.125.06R535.104.12R535.161.08R535.183.06(Data Center GPUs), R560.35.03(RTX GPUs)R570.133.20R590.48.01R595.58.03R595.58.03
VideoSDK releaseSDK 9.1SDK 9.1SDK 9.1SDK 9.1SDK 9.1SDK 9.1SDK 9.1SDK 9.1SDK 9.1SDK 9.1
OFSDK release2.0.232.0.232.0.232.0.232.0.232.0.232.0.232.0.232.0.232.0.23
GStreamer releaseGStreamer 1.16.2GStreamer 1.16.3GStreamer 1.16.3GStreamer 1.20.3GStreamer 1.20.3GStreamer 1.20.3GStreamer 1.24.2GStreamer 1.24.2GStreamer 1.24.2GStreamer 1.24.2
OpenCV releaseOpenCV 4.2.0OpenCV 4.2.0OpenCV 4.2.0OpenCV 4.5.4OpenCV 4.5.4OpenCV 4.5.4OpenCV 4.6OpenCV 4.6OpenCV 4.6OpenCV 4.6
Docker imagedeepstream:6.1.1deepstream:6.2deepstream:6.3deepstream:6.4deepstream:7.0deepstream:7.1deepstream:8.0deepstream:9.0deepstream:9.1deepstream:9.2
NVAIE releaseNANVAIE-3.xNVAIE-3.xNVAIE-4.xNANANANANANA

By default, OpenCV has been deprecated. However, OpenCV can be enabled in plugins such as nvinfer (nvdsinfer) and dsexample (gst-dsexample) by setting WITH_OPENCV=1 in the Makefile of these components. Please refer component README for more instructions.

NA: “Not Applicable”

DeepStream 9.2 support is currently enabled by SBSA triton docker. SBSA category currently supports GH200, GB200, and DGX Spark.