Headless Jetson Nano 4GB setup with with Micro-USB cable Without Using Monitor. USB Dongle is required for Wifi Connection. JupyterLab setup included. JupyterLab doesn't need docker. Torch, Torch2trt, ONNX, ONNXRuntime-GPU Torchvision and TensorFlow Installation Included. ROS2 Foxy Installation Included.
- Install Jetson Nano Image
- Boot Jetson Nano
- Wifi Connection Setup
- Download This Repository
- One Command Install All
- Enable I2C Permissions
- Install Jupyter Lab
- Setup Jetbot Stats
- Install ROS2 Foxy
- Install Torch
- Install Torch2trt and ONNX
- Install TensorFlow
- Issues
Download balenaEtcher and install it.
Download Jetson Nano Ubuntu 20.04, and write it on a microSD card with balenaEtcher.
Insert the microSD card to Jetson Nano.
Connect Jetson Nano with a PC/Laptop using Micro-USB cable.
Power On the Jetson Nano and wait for 1-2 minutes.
Download PuTTY.
Open Putty, select connection type serial.
The serial line value COM# can be found from following in Windows OS.
Device Manager > Ports (COM&LPT) > USB Serial Device (COM#)
Once connected, go through initial setup until Network Configuration.
Select dummy0: Unknown Interface
Once it fails to connect, select Do not configure the network at this time
Go through the rest of the steps with the default settings.
Once completed, wait couple of minutes and go to next step.
Open Putty and connect using host name 192.168.55.1 or using COM Port
Run following command and reboot Jetson Nano.
sudo systemctl restart network-manager.service
Log into Jetson Nano and run following command to detect availability of your wifi SSID.
nmcli device wifi list
Connect to wifi using following command.
sudo nmcli device wifi connect <SSID> password <Wifi_Password>
Run following command to get the IP address (2nd one).
hostname -I
Open Putty and connect using the IP address.
Once login completes, the Micro-USB cable can be disconnected.
sudo apt-get --purge remove opencv
git clone https://github.com/Agroecology-Lab/Headless-Jetson-Nano-Setup/
Installs Jupyter Lab, ROS2 Dashing, PyTorch, Torchvision, Torch2trt, ONNX, ONNXRuntime-GPU, Inputs and Adafruit_MotorHat.
Enables I2C Permissions.
Sets Jetbot Stats.
Takes aroud 20 minutes to 50 minutes.
cd ~/Headless-Jetson-Nano-Setup
chmod +x ./*.sh && ./install.sh
sudo usermod -aG i2c $USER
cd ~/Headless-Jetson-Nano-Setup
chmod +x ./jupyter.sh && ./jupyter.sh
Open Jupytar Lab in a browser with <IP Address>:8888 link. Default password is jetbot.
The interface should look like this (without watermark).
Setup Jetbot Stats Service to show Jebot status in OLED display.
cd ~/Headless-Jetson-Nano-Setup
chmod +x ./jetbot_stats.sh && ./jetbot_stats.sh
cd ~/Headless-Jetson-Nano-Setup
chmod +x ./foxy.sh && ./foxy.sh
Installs PyTorch and TorchVision.
cd ~/Headless-Jetson-Nano-Setup
chmod +x ./pytorch.sh && ./pytorch.sh
Installs Torch2trt, ONNX and ONNXRuntime-GPU.
cd ~/Headless-Jetson-Nano-Setup
chmod +x ./torch2trt_onnx.sh && ./torch2trt_onnx.sh
Installs TensorFLow.
cd ~/Headless-Jetson-Nano-Setup
chmod +x ./tensorflow.sh && ./tensorflow.sh
See here: https://github.com/mgonzs13/ros2_rover
See here: https://github.com/kerry-t-johnson/i2c_pwm
See here: https://github.com/Ar-Ray-code/ps_ros2_common/
See here: https://github.com/dheera/rosboard Maybe also here: http://robotwebtools.org/
Installing or modifying OpenCV will break Gstreamer Pipeline.
Therefore Albumation package should not be installed for Deep Learning.
Also Numpy version 1.19.4 is required to work Gstreamer Pipeline.
Updating Setuptools will generate deprication error during ROS2 workspace building.