Much information provided is directly from https://developer.nvidia.com/embedded/learn/get-started-jetson-nano-devkit .
UPDATE (01-12-2021) : You now must install JetPack 4.4 with the newest installers provided.
UPDATE (24-7-2021) : when this article was originally written there was only the 4GB model. The 2GB Model is not recommended to use with Shinobi and will not be supported.
First you need to prepare the operating system on the Jetson Nano. We will use their provided image that is based on Ubuntu. Make sure the Jetson Nano is off until step 4.
"To prepare your microSD card, you’ll need a computer with Internet connection and the ability to read and write SD cards, either via a built-in SD card slot or adapter." - NVIDIA Documentation
sudo su
apt install curl wget nano net-tools -y
bash <(curl -s https://gitlab.com/Shinobi-Systems/Shinobi-Installer/raw/master/shinobi-install.sh)
Please note : You should avoid using the browser on the Jetson Nano to view the Shinobi dashboard. The Jetson Nano is already going to be working hard with it processing streams, the browser actions running at the same time is going to give it a bad day.
You must install JetPack 4.3 to use the TensorFlow plugin.
These steps can be followed on Raspberry Pi with exclusion of step 4. Other CPU types can exclude steps 3 and 4.
cd /home/Shinobi/plugins/tensorflow
sh INSTALL-jetson.sh
pm2 start shinobi-tensorflow.js && pm2 save
node shinobi-tensorflow.js
pm2 restart camera
You're probably wondering, why? Well the purpose is to make it run with less load. If you aren't using the desktop then you don't need it to take resources.
Generally this is the case when Shinobi is installed on a Jetson Nano because you want to keep as much resources available to Shinobi as possible.
Be aware that simply re-installing the desktop interface may not make it operate as expected. Do this if you really want to remove the desktop environment.
cd /home/Shinobi/INSTALL
bash <(curl -s https://gitlab.com/Shinobi-Systems/Shinobi/-/raw/master/INSTALL/jetson-nano-convert-to-headless.sh)