Accelerated Applications; Accelerated Application Package Selection - Xilinx Kria KV260 User Manual

Accelerated Applications

The Xilinx SOM platforms are enabled with a number of accelerated applications that can be
dynamically installed on the SOM platform. The SOM starter Linux image is application agnostic
and provides a set of utilities for pulling the hardware accelerated application examples from the
SOM Linux package feeds.
Accelerated applications are software controllable, application-specific reference designs for AI
developers, embedded developers, and system architects to customize and enhance the
functionality through software control or updating the AI models. The following table outlines
some featured accelerated applications available for the KV260 Starter Kit. Consider this list as a
starting point because the SOM
accelerated applications availability.
Table 5: KV260 Accelerated Applications
Name
Smart camera
AI box-ReID
Defect detection

Accelerated Application Package Selection

RECOMMENDED: Public Ethernet connectivity is necessary to dynamically pull down the latest
accelerated application designs.
1. If you have not already verified Internet connectivity do so before proceeding via ping test or
DNS lookup (e.g., nslookup).
2. The sudo xmutil getpkgs lists a series of package groups that apply to your platform.
The package group naming convention is: packagegroup-kit_name-
application_name. For example, the Smart Camera application for the KV260 platform
has the following package group name packagegroup-kv260-smartcamera. You can
install any number of matching accelerated applications to your platform by executing a DNF
install for the chosen application package group. For example, to install the smart camera
application, run sudo dnf install packagegroup-kv260-smartcamera.
Note: You should only install package-groups that are compatible with your particular starter kit
configuration.
UG1089 (v1.2) July 26, 2022
KV260 Starter Kit
Getting Started web page contains the most up-to-date
Configurable camera/sensor input options with hardware accelerated ML
inference with face detection and other models that can be dynamically loaded.
Inference information is outputted to a monitor or network via RTSP.
Multi-stream IP camera RTSP inputs with hardware accelerated ML inference
models supporting face detection and re-identification (ReID) models. Outputs
are displayed to monitor.
Hardware accelerated machine vision application as applied in defect detection.
Chapter 4: Software Getting Started
Description
www.xilinx.com
Send Feedback
15