qengineering Jetson Nano Ubuntu 20.04 Image

If you download the image in chunks (e.g., JetsonNanoUb20_3b.img.xz.001 through ...013), you can reconstruct the original .img.xz file using 7-Zip.

Place all chunks in one folder and run the extraction command on the first file. 7-Zip will automatically process the subsequent parts in order.

If 7z is not in your system PATH, you must provide the full path to the executable.

# Standard usage
7z x JetsonNanoUb20_3b.img.xz.001

# Windows example with full path
"C:\Program Files\7-Zip\7z.exe" x JetsonNanoUb20_3b.img.xz.001

The image includes both GCC/G++ version 8 and version 9. Since many CUDA-related software packages require version 8, you can switch between them using update-alternatives.

Use the following commands to select your preferred version:

$ sudo update-alternatives --config gcc
$ sudo update-alternatives --config g++

To enable VNC on Ubuntu 20.04, you must first configure the GNOME Vino server to allow connections without encryption.

1. Run the following command to disable required encryption: $ gsettings set org.gnome.Vino require-encryption false
2. On your VNC client, ensure you uncheck the 'Authenticate' boxes during connection.

Because the image contains a large amount of software, you should expand the partition to utilize the full capacity of your SD card (especially if using 64GB+).

1. Install GParted: $ sudo apt-get install gparted.
2. Use GParted to resize the partition to fill the remaining unallocated space on the SD card.

To reduce the OS footprint to approximately 420 MB, you can remove the desktop environment and associated GUI packages. This is useful for resource-constrained headless deployments.

Warning: This will remove the graphical user interface.

sudo chown root:root / /lib
sudo apt purge ubuntu-desktop -y && sudo apt autoremove -y && sudo apt autoclean
sudo apt-get remove nautilus nautilus-* gnome-power-manager gnome-screensaver gnome-termina* gnome-pane* 
sudo apt-get remove gnome-applet* gnome-bluetooth gnome-desktop* gnome-sessio* gnome-user* gnome-shell-common
sudo apt-get remove zeitgeist-core libzeitgeist* gnome-control-center gnome-screenshot && sudo apt-get autoremove
sudo apt-get remove --purge libreoffice* 
sudo apt-get remove libreoffice-core
sudo apt-get remove snapd lightdm cups chromium*

When using an aarch64 system (like the Jetson Nano), importing both OpenCV and TensorFlow (or TensorRT) in Python can trigger the error: cannot allocate memory in static TLS block. This is caused by OpenMP memory requirements.

To resolve this, use one of the following methods:

1. Recommended: Import Order Import cv2 (OpenCV) at the very beginning of your Python script, before importing TensorFlow or TensorRT.

2. Alternative: Disable OpenMP Rebuild OpenCV with the -DBUILD_OPENMP=OFF and -DWITH_OPENMP=OFF flags. This forces OpenCV to use pthread or TBB for parallelization instead. Note that this is not recommended as not all OpenCV algorithms will automatically switch to pthread.

import cv2
import tensorflow as tf
# or
import cv2
import tensorrt

If you encounter a boot warning stating that lavapipe is broken, you can resolve it by removing the Vulkan ICD directory:

$ sudo rm -rf /usr/share/vulkan/icd.d

The following software versions are pre-installed on the image:

• OpenCV: 4.8.0
• TensorFlow: 2.4.1 (Note: Versions 2.5+ require CUDA 11, which is not supported on Jetson Nano)
• PyTorch: 1.13.0
• TorchVision: 0.14.0
• TensorRT: 8.0.1.6
• TeamViewer aarch64: 15.24.5
• Jtop: 4.2.1