Face recognition and gallery
Recognizes registered people in the person gallery and writes their name to the event; the gallery is managed entirely within ProVife.
Face recognition compares faces seen by the camera with the records in the person gallery and writes the name of the matching person to the event. To add a person to the gallery, ProVife shows the faces in the camera’s current frame; you select the right face and enter the first and last name, or you can upload a photo. The face similarity threshold sets how certain a match must be; no match is made for faces that resemble two people in the gallery almost equally. For example, when Maria Smith walks through the door, the event is recorded with her name and face image. Gallery data is stored encrypted on your own server.
How it works
Open the person gallery and list the faces in the camera with Capture frame.
Select the right face and enter the first and last name, or upload the person’s photo.
Create a Face recognition scenario on the camera and draw the area where faces will be seen.
Set the face similarity threshold; a high value reduces false matches.
When a person in the gallery is seen with similarity above the threshold, an event is created with the person’s name and face image.
When the gallery is updated, the change is pushed automatically to all face recognition scenarios.
Settings
| Person gallery | People registered with first name, last name and face image; added by selecting from the camera or with a photo. |
|---|---|
| Face similarity threshold | How closely the face must resemble the person in the gallery. |
| Area | The zone where face recognition is performed. |
| Repeat interval | The minimum time between two events for the same recognized person. |
| Analytics hours | The days and hours recognition runs. |
What it produces
- Face recognition event (analytics.face.recognized): the person’s name, ID and similarity value.
- Event image with the face labeled with a red box and the person’s name.
- The recognized person’s name in live video, and a No match label for unrecognized faces.
- Every scene in which the person was seen, by typing their name in AI Playback search.
What you can do with the event
- Unlocking the door lock or turnstile via a smart relay for a recognized person.
- A WhatsApp message with an image to the manager when a specific person is seen.
- A welcome announcement over the speaker in areas such as the executive floor.
- Forwarding the event via MQTT to the access control system.
- Building the check-in and check-out list on the Face attendance screen.
Where it is used
Office
Registered people entering the executive floor are logged by name, and unregistered faces are flagged.
Retail
With the store team’s own gallery, staff movements are tracked by name.
Education
Registered parents picking up students are recognized at the school entrance.
Residential complexes
Residents of the complex are recognized and distinguished from visitors.
Camera placement
- Position the camera to see faces from the front, at an angle close to eye level.
- Selecting the face added to the gallery from the same camera and in similar light strengthens matching.
- Favor spots in front of turnstiles and doors where people pass one at a time.
- Make sure the face is clear and large enough in the image.