Introduction

Au-Zone has three different models of Maivin AI vision system.  This guide will identity the Maivin based on quick, physical appearance, and contain links to lead you to hardware specific guides.

As well, this guide will run through some basic tasks that you should be to run on all Maivin units.

Identifying your Maivin

Maivin 1 -- "Maivin AI Vision Starter Kit - Standard 

The initial Maivin vision system can be identified by the long, rectangular form factor and the smaller, e-ConSystems camera with its 5MP AR0521 camera sensor.

Below are several links to help you work with your Maivin.

• Maivin AI Quick Start Guide
• Maivin AI User's Guide
• AI Application Zoo for Maivin

Maivin 1 Omni -- "Maivin AI Vision Starter Kit w/OmniVision 8MP Sensor"

The Maivin 1 Omni is an upgraded Maivin 1 which has the larger, OmniVision 8MP camera and OS08A20 camera sensor in the Maivin 1 form factor.  Despite this change, it is still a Maivin 1, and the above links should still work.

The Maivin 1 Omni, on left, beside the Maivin 1.

Maivin 2

This is the next generation of Maivin with a waterproof, ruggedize chassis and back-face mounting points suited for outdoor and marine deployment.  It also comes with the OmniVision 8MP camera and OS08A20 camera sensor, but contains several, under-the-hood improvements, including IMU and GPS sensors.

Below are several links to help you work with your Maivin.

• Maivin 2 User Guide

First Boot

Finding the Device Number

Regardless of the Maivin, each device should have an 8-digit number on its exterior, usually on the back or underside of the device.  It may be part of a "verdin-imx8mp-????????" label.  This is the device number, and is used to uniquely identify the Maivin.  It is also part of its hostname, "verdin-imx8mp-????????", though you may need to also add ".local" to the end depending on your network configuration.

Viewing the Web Application

Some Maivins come with a pre-installed VisionPack application already running.  To view the demo, just put the the hostname into your browser.  You may get a "Your connection is not private" warning -- it's okay! -- just click the "Advanced" and continue on to the Maivin.

In that picture, the device number is 15141027.  If your Maivin doesn't have an VisionPack application running, no worries!

SSH Into the Maivin

You can use SSH to access the Maivin.  All Maivin vision systems come with the user "torizon" and should have a default password of "1" or "torizon".  Running the following command:

ssh torizon@verdin-imx8mp-????????

Should provide you with a Linux BASH prompt to your device.  As above, you may need to add ".local" to the end of your hostname.

Our SSH into the Maivin page has more details regarding SSH connections.

Starting a VisionPack Web Application with systemd

The Maivin vision systems use systemd to manage the preinstalled VisionPack applications found in our AI Application Zoo for Maivin, though previous versions used Docker.  As such, starting one of the six applications is a simple as running the following command in your SSH session:

sudo systemctl start detectionui

Stopping a running application as done by:

sudo systemctl stop detectionui

The six application names are:

• detectionui -- an object detection application running a people-detection model
• facedetectui -- the same application running a face-detection model
• faceblurui -- an practical application of the face-detection model, using it to blur detected faces to anonymize people recorded
• segmentationui -- an application running a people-based segmentation model
• bodyposeui -- an application showing the location of arms, legs, and torso
• headposeui -- an application showing the three-dimensional vector of where a head is looking.

When one of the above applications are started, the application as described in the above "View the Web Application" section should appear.

To determine which application is running, the following command can be used:

systemctl show -P Description -P ActiveState detectionui facedetectui faceblurui segmentationui bodyposeui headposeui

Which will provide output as follows:

People Detection UI Demo
active

Face Detection UI Demo
inactive

Face Blur UI Demo
inactive

People Segmentation UI Demo
inactive

Body Pose UI Demo
inactive

Head Pose UI Demo
inactive

As well, we have a new recorder application that can take detections and images (and IMU data and GPS coordinates from the Maivin 2) and save them in MCAP format, which can be started with:

sudo systemctl start recorder

More information about this application is available on the Maivin Dataset Recording page.