TOC

• TOC
• Overview
• How to Install
  • Install by RPM
  • Manually install
• reference

Overview

VVAS(Vitis Video Analytics SDK) contains plugin developed by xilinx based on gstreamer. It can use hardware IP to accelerate image processing and AI inference in gstreamer. The overall architecture is as shown in the figure below. The video source is opened by gstreamer plugin and passed to VVAS for image processing and AI computing. The corresponding accelerator application is linked to VVAS according to Json config.

How to Install

• Since BSP 1.2.0 which with Vitis AI 2.5 & VVAS 2.0, VVAS is default built-in with the system, user do not need to install again.
• Change the version of VVAS by table below, if using different vesion of Vitis AI.

Vitis AI version	VVAS version
1.4	1.0
2.0	1.1
2.5	2.0

Install by RPM

VVAS can be installed by using the prebuilt rpm package.

rpm -ivh --force vvas-0.1-1.aarch64.rpm

Manually install

If you are using customised BSP, there may have dependency issue. So we will suggest manually install rather than install by RPM(Red Hat Package Manager).

• Preparation Petalinux SDK including Vitis AI, opencv (over 4.4), jansson.

1. Download source code which match to your Vitis AI version on x86 host.

   ​​​​git clone -b VVAS_REL_v1.0 https://github.com/Xilinx/VVAS.git
   

2. Source petalinux SDK.

   ​​​​unset LD_LIBRARY_PATH
   ​​​​source <path-to-SDK>/environment-setup-aarch64-xilinx-linux
   

3. Build VVAS on x86 host.

   ​​​​chmod 755 build-ivas-essential.sh
   ​​​​./build-ivas-essential.sh Edge
   

   demo

4. Copy the result from x86 host into rootfs of k26.

   ​​​​scp ./install/ivas_installer.tar.gz petalinux@<IP>:/
   

5. Install the VVAS on k26.

   ​​​​cd /
   ​​​​sudo tar -xvf ivas_installer.tar.gz
   

reference

Xilinx VVAS
Xilinx VVAS github
Xilinx VVAS tutorial