Installation

The preferred tool for installing VPI is the SDK Manager installer, which automates the installation and setup process on both the host and the target system.

For a full description of the installer, see the SDK Manager User Guide.

Note

When installing VPI via the SDK Manager installer, it's advisable to upgrade VPI to the most recent version to take advantage of the bug fixes made after VPI v2.0 was released. To do so, please execute the following commands in a terminal:

sudo apt-get update
sudo apt-get install libnvvpi2

Alternatively, you can use apt to install the VPI host components on a linux-x86_64 host.

1. Install the public key for the VPI repository.

   sudo apt install gnupg
   sudo apt-key adv --fetch-key https://repo.download.nvidia.com/jetson/jetson-ota-public.asc

2. Install some packages needed to add a new apt repository to the system.

   sudo apt install software-properties-common

3. Add the public repository server to the apt configuration.
   • For Ubuntu 18.04:

     sudo add-apt-repository 'deb https://repo.download.nvidia.com/jetson/x86_64/bionic r34.1 main'

   • For Ubuntu 20.04:

     sudo add-apt-repository 'deb https://repo.download.nvidia.com/jetson/x86_64/focal r34.1 main'

4. Update the local repository package list.

   sudo apt update

5. Install the VPI package and its dependencies on the host:

   sudo apt install libnvvpi2 vpi2-dev vpi2-samples

6. If there's a display connected, also install the VPI demo package.

   sudo apt install vpi2-demos

   Then follow the instructions in Demo Applications.
7. For python bindings, do the following
   • Python 3.8 (Ubuntu 18.04 and Ubuntu 20.04)

     sudo apt install python3.8-vpi2

   • Python 3.9 (Ubuntu 20.04 only)

     sudo apt install python3.9-vpi2

     In both cases, numpy module still needs to be installed, either via pip, conda, apt or another method.
8. (optional) Development packages for cross-compilation targeting aarch64-l4t architecture.

   sudo apt install vpi2-cross-aarch64-l4t

The installer copies VPI files to the directory /opt/nvidia/vpi2.

Inside this directory are the following sub-directories:

- bin	Demo applications and other executables
- include	C headers
- lib64	Shared libraries for host platform
- samples/*	Complete sample applications
- targets/aarch64-l4t/lib64	Shared libraries needed for cross-compilation

After a successful installation, read basic concepts to learn some basic VPI concepts, then go write your first application with VPI.