Setting Up an Edge Device for LEIP Recipes

Device Preparation

NVIDIA AGX or NX

It is recommended that you start with a clean environment on the target device. The system should be wiped to remove any packages that were installed previously, and a fresh JetPack should be installed. Instructions for installing a fresh JetPack image can be found in the Jetson AGX Developer Manual.

Raspberry Pi

Your system must be a Raspberry Pi 4B or a Raspberry Pi 400. It is recommended that you start with a clean environment on the target device. The system should be wiped to remove any packages that were installed previously, and a fresh 64-bit operating system should be installed. We use the Raspberry Pi 64-bit operating system for test and verification.

Other Devices

If you have requirements for specific devices and/or operating systems, please contact Latent AI for further assistance.

Installing Base Dependencies

To use the Latent AI Object Runner (LOR) or Python example applications, there are a number of dependencies that you will need to install. These dependencies will support the LOR and the example pre- and post- processing used. To facilitate the installation of these dependencies, we provide installation scripts in the setup_scripts directory of the example application git repo.

CODE

# Download the repo referred to by the link above
git clone <example application repo>
cd example_applications/setup_scripts

# Find the directory containing the scripts for your device type
ls 
agx_nx    rpi

Installing and Running the LOR

Once the base dependencies are installed, you can install the LOR using the following steps:

Copy the LOR wheel from the LEIP SDK to the edge device
Install the base dependencies using the appropriate install-lor-depsscript for your device
Install the LOR on the edge device
Launch the LOR on the edge device

In the LEIP SDK Container:

CODE

# The actual version number in your SDK may differ from the example below
scp /latentai/packages/latentai_lor-1.4.5-py3-none-any.whl <remote user/device>

On the Remote Device:

CODE

# Install the base dependencies
# For example on an agx/nx with jetpack 4.6, use the following script:
bash agx_nx/install-lor-deps-jp46.sh

# The actual version number in your SDK may differ from the example below
pip3 install latentai_lor-1.4.5-py3-none-any.whl

# lor_server will be installed in ~/.local/bin  Make sure to add that to your $PATH
# You may wish to launch lor_server using nohup or similar to stay running after you
# close your current session
lor_server

Dependencies for C++ Examples

In addition to the base dependencies, there are some additional dependencies that need to be installed to run the C++ examples. These dependencies include the C++ LEIP runtime and associated header files. To install these dependencies, use the install-for-examples.sh script in the appropriate device directory under the setup_scripts path. If you are only downloading that script from the repo, note that it depends on another script in the setup_scripts directory to add the Latent AI apt server to your apt sources.