Skip to main content
Version: V12

How to Set Up Face Recognition

Face recognition has two halves that have to meet in the middle: the faces you enroll in the Object Library, and the cameras you tell to look for them. Enroll a face but never turn recognition on for a camera, and nothing happens; turn recognition on but enroll no faces, and there is nothing to match. This guide walks the whole flow end to end, so a camera ends up naming the people on your watchlist as they appear.

Prerequisites​

  • You belong to a group (or hold a CAL) with the Object Library feature enabled, and you can add and edit cameras.
  • You have at least one camera added in AI Live Insight. See Add a Camera.
  • You have clear, front-facing reference photos of the people you want recognized, as JPEG, PNG, or WebP.

Note: A camera can only recognize a face it first detects, so face detection must be enabled on that camera before recognition can run (Step 2). See Understanding Face Recognition.

The Flow at a Glance​

StepWhereWhat you do
1. Enroll subjectsObject LibraryAdd each person as a Face subject with reference photos.
2. Detect facesCamera → Camera Alert tabAdd Face as a Detection Class.
3. Recognize facesCamera → Camera Alert tab → RecognitionAdd a Face recognition row; set the match threshold and recording.
4. Watch it workLive player, event feed, recordingsRecognized names appear on the video and in the feed.

Step 1: Enroll Faces in the Object Library​

This is the watchlist, the faces every camera with recognition on will be matched against.

  1. Open the Object Library.
  2. Select Add to Library.
  3. Choose the Face type, enter the person's Name, and set a Watchlist Priority.
  4. Upload one or more clear reference photos and select Enroll Subject.

Enroll several photos per person where you can, different angles and lighting make recognition more reliable. For the full enrollment, editing, and search workflow, see How to Manage the Object Library.

Note: Enrolled faces are matched only against your own portal's cameras. They are never compared against another tenant's cameras.

Step 2: Enable Face Detection on the Camera​

Recognition only runs on faces the camera detects, so first tell the camera to watch for faces.

  1. Open the camera and edit it, or add a new one, and go to the Camera Alert tab.
  2. Under the detection rows, add a row with Detection Class set to Face.
  3. Set the Confidence Threshold for detection (35 is the recommended starting point).

For the full camera setup, see Add a Camera.

Step 3: Turn On Recognition for the Camera​

Once the camera detects faces, a Recognition section appears on the same Camera Alert tab. This section only shows up after a recognition-capable class (Face) is added as a detection class, because recognition builds on detection.

  1. In the Recognition section, select Add Recognition Class.
  2. Set Recognition Class to Face.
  3. Optionally set the Match Threshold (%), the minimum match score for a face to count as a recognition (allowed range 30–99). Leave it blank to use the system default. A higher value means fewer, more certain recognitions; a lower value catches more, with a higher chance of a wrong name.
  4. Turn on Record this alert if you want a recording captured each time this camera recognizes a watchlisted face. This is off by default, and is independent of plain detection recording.
  5. Save the camera.

Note: The Match Threshold (%) here governs recognition (how confident the face match must be), which is separate from the Confidence Threshold in Step 2 that governs detection (how confident the AI is that there is a face at all).

Step 4: See Recognition Working​

With faces enrolled and recognition on, the camera starts naming people as they appear:

  • Live player: the recognized person's name appears on the bounding box around their face.
  • Event feed: recognitions are listed as their own entries, separate from plain detections.
  • Recordings: if you turned on Record this alert, a clip is captured around each recognition. See How to Review Surveillance Recordings.
  • Object Library: each recognition is logged against the subject, building its activity history.

Where Each Setting Lives​

  • Portal-wide detection defaults are set in the Surveillance app settings, the starting point new cameras inherit. See How to Configure VIDIZMO Surveillance App Settings.
  • Per-camera detection and recognition (classes, thresholds, recording) live on the camera's Camera Alert tab.
  • The faces themselves live in the Object Library, shared across every camera in the portal.

See Also​

Did you find this article helpful?