Jetson Nano Setup #jetson-nano

jetson_nano_setup.md

Jetson Nano

NVIDIA Jetson Linux Developer Guide!comprehensive and useful resources

Jetson Zoo

NVIDIA Jetson Nano Setup

JetPack SDK

JetPack SDK Introduction

JetPack SDK includes:

• L4T
  • Ubuntu 18.04(Linux operating system)
  • bootloader
  • Linux kernel 4.9
  • necessary firmwares
  • NVIDIA drivers
• CUDA accelerated libraries
  • TensorRT
  • cuDNN
  • CUDA
  • Multimedia API
  • Computer Vision
  • Developer Tools
  • DeepStream

How to Upgrade JetPack

1. Upgrade L4T
2. Upgrade JetPack

Over-The-Air Updates

Note: upgrade L4T on a system where Jetpack is not installed.

Setup

Setup and First Boot

• Set power mode(5W or 10W)

  # To set Jetson Nano to 10W performance mode
  sudo nvpmodel -m 0
  sudo jetson_clocks

• Set CUDA toolkits in environment path

  in ~/.bashrc,

  export PATH=/usr/local/cuda/bin:${PATH}
  export LD_LIBRARY_PATH=/usr/local/cuda/lib64:${LD_LIBRARY_PATH}

  use script,

  #!/bin/bash
  set -e
  
  if ! grep 'cuda/bin' ${HOME}/.bashrc > /dev/null ; then
    echo "** Add CUDA stuffs into ~/.bashrc"
    echo >> ${HOME}/.bashrc
    echo "export PATH=/usr/local/cuda/bin:\${PATH}" >> ${HOME}/.bashrc
    echo "export LD_LIBRARY_PATH=/usr/local/cuda/lib64:\${LD_LIBRARY_PATH}" >> ${HOME}/.bashrc
  fi

Connect to Jetson

1. Ethernet/WiFi It needs Jetson to connect to the same network as your desktop pc via Ethernet or WiFi.

   # `jetson-desktop` is hostname of Jetson 
   ssh [email protected]

2. USB device mode Jetson Nano Micro-USB Login

   • serial console

     screen /dev/cu.usbmode* 115200 -L

   • ssh(over virtual network)

     ssh [email protected]

     the Jetson establishes a virtual network where it works as a network device and desktop PC works as a host. When Jetson is connected from PC by micro USB, PC will ask for a IP address of 192.168.55.100 via DHCP from Jetson and Jetson will 192.168.55.1 in default. This service is located at: /opt/nvidia/l4t-usb-device-mode

     IP address 192.68.55.1 in Jetson a bit of explanaion on SSH through micro USB USB device mode USB Gadget API for Linux

   • J40 header pins, Jetson Nano USB Login

Install Softwares

Docker

• Docker setting

  # setup docker without sudo
  sudo usermod -aG docker $USER
  # upgrade docker
  sudo apt-get --only-upgrade install docker.io

• Docker Compose

  Install Docker Compose

TensorFlow

Pytorch

JupyterLab

1. install NodeJS through installing nvm

2. install JupyterLab

3. configure JupyterLab

   Running a notebook server

   # configure /home/$USER/jupyter_notebook_config.py
   ...
   c.NotebookApp.ip = '*'
   c.NotebookApp.port = 8888
   c.NotebookApp.open_browser = False
   ...
   
   # configure password
   $ jupyter notebook password

   Run jupyterlab as ubuntu service

   # 1. create script
   $ vim /home/$USER/jupyterlab.sh
   
   #!/bin/sh
   /home/jetson/.local/bin/jupyter lab --ip=0.0.0.0 --port=8888 --no-browser
   
   $ sudo chmod +x /home/$USER/jupyterlab.sh
   
   # 2. create service
   $ sudo vim /etc/systemd/system/jupyterlab.service
   # replace $USER
   
   [Unit]
   Description=Jupyterlab
   After=syslog.target network.target
   
   [Service]
   User=jetson
   ExecStart=/bin/bash /home/$USER/jupyterlab.sh
   WorkingDirectory=/home/$USER/
   
   [Install]
   WantedBy=multi-user.target
   
   # 3. enable service
   $ sudo systemctl enable jupyterlab.service
   $ sudo service jupyterlab start

4. install JupyterLab extensions

   # install ipywidgets and enable
   jupyter labextension install @jupyter-widgets/jupyterlab-manager

   Language Server Protocol integration for Jupyter(Lab)

OpenVPN

sudo apt-get install openvpn

# Then you can connect like this:
sudo openvpn --config /path/to/config.ovpn

Samba

sudo apt install samba samba-common-bin
sudo nano /etc/samba/smb.conf

# edit `/etc/samba/smb.conf` file
workgroup = your_workgroup_name
wins support = yes

[pihome]
   comment= Pi Home
   path=/home/pi
   browseable=Yes
   writeable=Yes
   only guest=no
   create mask=0777
   directory mask=0777
   public=no

# save and exit
sudo smbpasswd -a pi

Samba Installation Anonymous Read-Write Document Server

OpenCV(GPU)

Using GPU with OpenCV on Jetson Xavier - Jetson AGX Xavier - NVIDIA Developer Forums

GitHub - mdegans/nano_build_opencv: Build OpenCV on Nvidia Jetson Nano

tips: make the CMake version >3.20 or support ssl. Otherwise it will not support https download for xfeatures2d in opencv

software installation - How do I install the latest version of cmake from the command line? - Ask Ubuntu

Check OpenCV Build Information: getBuildInformation()

python3 -c 'import cv2; print(cv2.getBuildInformation())'

Some Tips

Connect to WiFi from terminal

nmcli device wifi rescan
nmcli device wifi list
nmcli device wifi connect <ssid_name> password <password>

Check L4T Version

dpkg-query --show nvidia-l4t-core

JETSON_L4T_STRING=$(head -n 1 /etc/nv_tegra_release)
JETSON_L4T_RELEASE=$(echo $JETSON_L4T_STRING | cut -f 2 -d ' ' | grep -Po '(?<=R)[^;]+')
JETSON_L4T_REVISION=$(echo $JETSON_L4T_STRING | cut -f 2 -d ',' | grep -Po '(?<=REVISION: )[^;]+')
JETSON_L4T_VERSION="$JETSON_L4T_RELEASE.$JETSON_L4T_REVISION"
echo $JETSON_L4T_VERSION

or

JETSON_L4T_STRING=$(dpkg-query --showformat='${Version}' --show nvidia-l4t-core)
JETSON_L4T_ARRAY=$(echo $JETSON_L4T_STRING | cut -f 1 -d '-')
JETSON_L4T_RELEASE=$(echo $JETSON_L4T_ARRAY | cut -f 1 -d '.')
JETSON_L4T_REVISION=${JETSON_L4T_ARRAY#"$JETSON_L4T_RELEASE."}
JETSON_L4T_VERSION="$JETSON_L4T_RELEASE.$JETSON_L4T_REVISION"
echo $JETSON_L4T_VERSION

Check JetPack Version

sudo apt-cache show nvidia-jetpack

Check Cudnn Version

cat /usr/include/cudnn_version.h | grep CUDNN_MAJOR -A 2

Install Pycuda

export CPATH=$CPATH:/usr/local/cuda/include
export LIBRARY_PATH=$LIBRARY_PATH:/usr/local/cuda/lib64

pip3 install pycuda --user

Disable GUI/Install lightdm instead of gdm

sudo systemctl set-default multi-user.target
sudo systemctl get-default

Disable GUI in Ubuntu 18.04. desktop? (SSH/Webmin headless setup) How to disable GUI from starting on boot?

Troubleshooting

Camera failed to create CaptureSession

# restart camera daemon
sudo systemctl restart nvargus-daemon

xRDP disconnects immediately after connection

# find reason from logs
sudo journalctl -S -3m

# add user xrdp into ssl-cert group
sudo adduser xrdp ssl-cert

# configure desktop session for xRDP
# rdp works better with XFCE
sudo apt-get install xfce4
echo xfce4-session > ~/.xsession
sudo sh -c 'cat /dev/null > /etc/xrdp/startwm.sh'

cat <<EOF > /tmp/startwm.sh
#!/bin/sh
if [ -r /etc/default/locale ]; then
  . /etc/default/locale
  export LANG LANGUAGE
fi
startxfce4

EOF

sudo mv /tmp/startwm.sh /etc/xrdp/startwm.sh
sudo systemctl restart xrdp.service

xRDP cannot read /etc/xrdp/key.pem. Permission denied error explained

How to Use Remote Desktop Connection in Ubuntu Linux: Complete Walkthrough

How to configure XRDP to start cinnamon as default desktop session

SSH Connection Hangs

Check this issue by opening ssh daemon log(/var/log/auth.log in ubuntu) to solve. In some cases, upgrade libs (sudo apt-get upgrade, systemd,...etc) or install missing libs(sudo apt-get install libpam-kwallet5)

ssh connection takes forever to initiate, stuck at “pledge: network”

AI-IOT

JetCam ! camera interface for NVIDIA Jetson

JetBot

Getting Started with AI on Jetson Nano

jetson-inference

Test JetPack On Jetson Nano

JetPack-4.5