Skip to main content

Detection Settings

Detection settings are configured per region. The editor uses tabs so you can switch between general sizing, object type, natural language, face recognition, and similarity workflows without leaving the selected region.

The screenshot below shows the detection settings tabs for a region.

EyesOnIt Live Search and Results

Motion detection​

Each video region includes a Motion Detection page with:

  • Enabled
  • Detection Threshold
  • Regular Check Frame Interval
  • No Motion Object Check Interval

Detection Threshold is controlled with a slider, and the UI draws a size hint on the preview so you can tune the threshold against the approximate size of the objects you expect to see.

The screenshot below shows the motion detection settings and the threshold hint overlay.

EyesOnIt Live Search and Results

Detection modes​

The Detection page is organized into these tabs:

  • General: shared object-size control for the active detection mode
  • Object Type: optional class selection before refining the query
  • Natural Language: text prompts that further describe what should be detected
  • Face Recognition: person or group matching
  • Similarity: image-based matching

Object type and size​

If Use Object Type Detection is enabled, you can select one of the supported classes:

  • Person
  • Vehicle
  • Animal
  • Bag
  • Something Else

The object-size slider is always important because it helps EyesOnIt focus on the expected scale of the target.

Natural language​

If Use Natural Language is enabled, you can add one or more descriptions. Each description supports:

  • description text
  • This is a Background Description
  • Alert for video workflows
  • Alert Threshold for video workflows

Background descriptions stay in the model context but do not generate alerts directly.

Face recognition​

If Use Face Recognition is enabled, configure:

  • Match Type: Person or Group
  • a person or group name from the autocomplete suggestions
  • Match Threshold

The UI queries the server for matching names while you type, so the safest workflow is to select from the suggestions instead of relying on free-form text.

The screenshot below shows either the face recognition tab in the region editor.

EyesOnIt Live Search and Results

Similarity detection​

If Use Similarity Detection is enabled, configure:

  • an uploaded reference image
  • Match Threshold

The UI stores the uploaded image in the region configuration so it can be reused when the stream is saved.