Object Detector is an AI utility that runs on EdgeVis encoders and analyses video streams to detect different types of object.

The Object Detector can only be configured from the EdgeVis Client for Windows. All analytics configuration is done through the Settings->Change analytics menu item, which is available when an encoder has been selected and viewed.

Selecting the the Settings->Change analytics menu item displays the Change analytics section across the bottom of the view. The top-level controls for the analytics are displayed on the left of the section.

To configure Object Detector:

Clicking the CONFIGURE button with the Object Detector option selected opens the Options popup for the Object Detector.

The Sensitivity level is controlled using a dropdown menu. It sets how many detected objects are displayed and registered as events.

The Sensitivity setting can be set to Very Low, Low, Medium, High or Very High.

Choosing the right setting for your situation will depend on how important catching all of the positive identifications is to you, compared to the overhead of having to deal with lots of possible false positive identifications.

Selecting or deselecting each of the checkboxes configures the use of a differently biased AI detector, geared towards detecting the particular type of object.

There currently three types of object that can be detected:

The objects will be detected according to how long they have been present in the video and the area of detection. By default this is set to 0.0 secs, meaning that as soon as the object is present and recognized it will be identified and labeled as either a Person, Car or Truck.

The duration can be changed between the default of 0.0 secs up to 60.0 secs, using the range of values displayed in the dropdown list. When set to a value greater than 0.0, the object will need to have been in the zone for that length of time before being labelled. For example, in low frequented areas you could set the Minimum presence in zone to 60 seconds, meaning that a Person, Car or Truck will need to have been in the zone for a minute before being labelled as such. With the idea that the presence in the zone for that length of time is more significant than an object just passing through - focusing attention on the important events.

This option lets you select whether the encoder sends the analytics data generated by the Object Detector as a secondary ONVIF Profile M channel. This allows the data to be consumed by a VMS without the need to reference a corresponding video stream.

This is for a specific use-case and the secondary channel does have a slight overhead, so the option is turned off by default.

The Object Detector settings can be used as alarm rule triggers but the OD analysis must first be enabled on one of the encoder video inputs using the EdgeVis Client.

The following steps show how OD analysis can be set as the trigger for an alarm. This example uses a new alarm rule, which we've named Object Detector Alarm.

Clicking on the Add alarm trigger link displays the Select trigger controls. Here you select the Analytic option.

Doing this displays the Analytic type controls, where you can choose between Object Detector or SafeZone 2D. In our example only the Object Detector has been enabled on the encoder using the EdgeVis Client, so the SafeZone 2D option is greyed out.

Selecting Object Detector displays all the options available for OD. Something to note is that all the Category types are available, regardless as to whether each type has been individually enabled, when using EdgeVis Client to configure the Object Detector on the video input, or not.

You can only select one Category type for each trigger. In our example we've chosen Person, Detection start and the BuchananStreetCamera for our trigger.

Clicking Add stores the new trigger in the alarm rule being edited.