Provides quick access and useful resources for NVIDIA Jetson.

Quick Access

Before Purchasing

Check the specs of different Jetson Modules.

In general, AGX Orin > Orin NX > AGX Xavier > Xavier NX > TX2 > Nano.

For AGX Xavier, you may also want to purchase a compatible wireless network card.

For AGX Orin, you may want to purchase an active DisplayPort to HDMI adapter/cable.

Common Notes

If you encountered issues that are near impossible to solve, first check the required Jetpack version of the software you wish to install. In most case, re-flashing the board with a compatible Jetpack version (along with required SDKs) will solve the issue.

Quick Start

For older boards, you can use the raw image here and install it on a SD card. Plug the SD card on the Jetson board after installation and boot it up.

For AGX Xavier (newer boards), follow the steps:

  1. Use NVIDIA SDK Manager following the User Guide of your board, in the Jetpack section. (SDK Manager does not support Ubuntu 20.04, Use Ubuntu 18.04 instead)

    You may want to install Jetpack 4.4.1 for better compatibility. DeepStream is optional, depending on your usages.

  2. (Manual Install) Set Jetson board to Recovery Mode (Hold Recovery button, hold Power button, then release them both) and refresh the SDK Manager, your board module should pop up. If you have trouble recognizing the board interface or buttons, see the bottom of the board. It has tiny texts showing Power, Force Recovery, Reset, etc.
  3. Start the installation in SDK Manager.
  4. After the OS is installed on the board, a prompt will pop up on your PC.
  5. Now set up the OS on the Jetson board (may need to plug in a keyboard, mouse, and monitor). Set up the network (Wifi/Ethernet) on your Jetson board and use ifconfig to find your Jetson local IP.
  6. Type in the Jetson IP, username, and password on the prompt on your PC. The SDK Manager will now install the SDKs and tools on the Jetson baord.
  7. Install the SD card on your Jetson board (not required for AGX Xavier) and boot it up.

Detailed steps can be found in the SDK Manager Documentation.

Follow the tutorials for further instructions.

Show Jetpack Version

Use jetsonhacks/jetsonUtilities:

git clone
cd jetsonUtilities

Environment Setup

You may want to VNC or SSH into the board without plugging in keyboard & mouse & monitor.

Disable Screen Lock

gsettings set org.gnome.desktop.screensaver lock-enabled false


Follow this post to enable auto-login (for automatically launching VNC server). Specifically:

cd /etc/gdm3
sudo vim custom.conf
# Uncomment the following lines and enter your username:
    #AutomaticLogin = <Your_User_Name>

Execute the following commands:

sudo apt update
sudo apt install vino
# Inform the system which graphic display to interact with
export DISPLAY=:0
# Configure autostart
mkdir -p ~/.config/autostart
cp /usr/share/applications/vino-server.desktop ~/.config/autostart
# Enable the VNC server
gsettings set org.gnome.Vino prompt-enabled false
gsettings set org.gnome.Vino require-encryption false
# Reboot the system so that the settings take effect
sudo reboot

If VNC server doesn’t seem to work after reboot, follow the official tutorial.

Solely following the official tutorial above doesn’t seem to work on AGX Orin. Therefore, the commands provided above are required.

You can then connect to Jetson with any VNC client (e.g., RealVNC) from your PC.

Old Method (2021): Follow this thread and this thread to install VinoVNC.

The VNC screen resolution can be changed by xrandr and xorg.conf, following this guide.


Install openssh-server.

sudo apt install openssh-server


If you encountered issues when installing OpenCV, see this thead and this thread.

It can also be used in a virtual environment as mentioned in this post.

In general, you don’t need to compile OpenCV by yourself.


You need to install the wheels provided by NVIDIA instead of using pip.

Follow this tutorial to install PyTorch and TorchVision (expand the Installation instructions).

If you have trouble installing torchvision, see this thread.

If you encountered unknown issues, you may install the wrong version of PyTorch in most cases. Please keep in mind 3 versions: (1) Jetpack version, (2) PyTorch version, and (3) TorchVision version.


TensorRT comes with JetPack 5.0.2.

You can view your JetPack version by:

sudo apt-cache show nvidia-jetpack

as mentioned in this thread.

If TensorRT doesn’t exist on your system, please re-flash it (or install it) with NVIDIA SDK Manager.



If you have questions, ask on the NVIDIA Jetson Forum