Installing DeepStream and CUDA for the Jetson

NVIDIA® DeepStream Software Development Kit (SDK) is an accelerated AI framework to build intelligent video analytics (IVA) pipelines. DeepStream runs on NVIDIA® Jetson NX™, NVIDIA® Jetson Orin™ NX, NVIDIA® Jetson Orin™ Nano in conjunction with the EchoPilot AI.

The CUDA Toolkit provides a development environment for creating high-performance GPU-accelerated applications.

The instructions below show how to install both. These instructions were developed using a Jetson Orin NX with a 256 GB SSD, running L4T 35.4.1. In most cases, you will not be able to install this software without significant available storage space. E.g., an Xavier NX with only a 16 GB eMMC will not have enough storage space. We recommend adding a NVMe SSD before proceeding.

Note

The instructions below assume that the EchoPilot AI has internet access and you are logged in to the console. Since EchoPilotAI Jetson hardware is provided with a static IP address, it is often simpler to enable DHCP and let the Jetson get internet through your LAN's router. To do this, you can use the nmcli commands to change the static-eth0 connection profile to auto/hdcp.

sudo nmcli con mod static-eth0 ipv4.method auto
sudo nmcli con down static-eth0
sudo nmcli con up static-eth0

When you are done and would like to return the IP config to static, use:

sudo nmcli con mod static-eth0 ipv4.addresses "10.223.44.55/16"   # For example
sudo nmcli con down static-eth0
sudo nmcli con up static-eth0

Remember if you change the network during a ssh session, you will lose connection. It is recommendced to make network system changes when on a USB Console connection.

Set system time

Assuming you have an internet connection at this point, you will want to ensure your system time is set correctly or package updates may fail.

sudo timedatectl set-ntp true
sudo systemctl restart systemd-timesyncd.service

now verify your system time is correct

sudo timedatectl

Install DeepStream

Start by doing an apt update.

sudo apt-get update

Install dependencies

sudo apt install \
build-essential \
libssl1.1 \
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

Install librdkafka

1. Clone the repo

   git clone https://github.com/edenhill/librdkafka.git
   

2. Configure, build and install

   cd librdkafka
   git reset --hard 7101c2310341ab3f4675fc565f64f0967e135a6a
   ./configure
   make
   sudo make install
   

3. Copy the generated files to the deepstream directory.

   sudo mkdir -p /opt/nvidia/deepstream/deepstream-6.3/lib
   sudo cp /usr/local/lib/librdkafka* /opt/nvidia/deepstream/deepstream-6.3/lib
   

Get and install the Deepstream SDK

wget --content-disposition 'https://api.ngc.nvidia.com/v2/resources/org/nvidia/deepstream/6.3/files?redirect=true&path=deepstream-6.3_6.3.0-1_arm64.deb' -O deepstream-6.3_6.3.0-1_arm64.deb
sudo apt-get -y install ./deepstream-6.3_6.3.0-1_arm64.deb

Install CUDA

wget https://developer.download.nvidia.com/compute/cuda/repos/ubuntu2004/sbsa/cuda-keyring_1.1-1_all.deb
sudo dpkg -i cuda-keyring_1.1-1_all.deb
sudo apt-get update
sudo apt-get -y install cuda-toolkit-12-3

Install CUDA driver

sudo apt-get install -y cuda-drivers

Boost the clocks

sudo nvpmodel -m 0 
# for Xavier NX, use nvpmodel -m 8
sudo jetson_clocks

Now Reboot the system.

Verify functionality

At this point, the deepstream-app should run without errors:

deepstream-app --help

Explore Sample Code

Browse and run precompiled sample applications in sources/apps/sample_apps.
Follow the directory’s README file to run the application.