Using computer vision analytics

Computer vision processes events generated by face and object detection.

FortiRecorder can detect important objects such as a vehicle or a weapon. Object detection in FortiRecorder is treated as a camera-related event, similar to motion detection.

For face recognition, FortiCentral installations can create models of the faces that your cameras see, and then store them in a database on FortiRecorder. Every FortiCentral that runs face detection can contribute to your face recognition database. This distributed processing scales well as your organization grows. FortiRecorder can then use computer vision to analyze digital images to identify important people via face recognition.

Alternatively, you can perform face modeling and face recognition at a centralized location, on FortiRecorder. Both methods store their analysis results on FortiRecorder. However, the results are stored in separate databases. On FortiRecorder: Analytics > Computer Vision: Face modeling by FortiCentral. Both FortiCentral and FortiRecorder can query this database for face recognition. For details on face recognition in FortiCentral, see the FortiCentral User Guide. Face Recognition > User Asset, etc.: Face modeling by FortiRecorder. Currently, only FortiRecorder can query this database for face recognition. FortiCentral cannot use it. For details, see Face recognition .

For information about viewing logs and managing the databases via REST API, see the REST API Reference.

To enable detection actions

1. Go to Analytics > Computer Vision > Computer Vision.

2. Enter the amount of time to use to decide on a recognized face or object in the Accept detection after field.

   More time allows the camera a chance to get a more complete view of the object or face so that FortiRecorder can avoid misinterpretations.

3. Enable Face detection action. Face detection is organized into categories:

   • Prohibited: Individuals who are prohibited from accessing the building.
   • VIP: Individuals who are very important and are not security threats.
   • Expired: The time is past the expiry date of the person recorded.
   • Unknown: Individuals who are detected but not matched to an individual in the database.
   • Regular Person: Individuals who are recognized as common visitors, such as guests, employees, or contractors.

4. Enable Object detection action. Object detection is organized into categories:

   • Person: An individual.
   • Motion: Something that has generated motion.
   • Weapon: A detected tool that can cause harm.
   • Vehicle: A transportation machine, such as a bike, a car, or a train.
   • Animal: A non-human creature, such as a bird or a dog.
   • Item: An object, such as a backpack or a suitcase.
   • Sport: Athletic equipment, such as skis or a skateboard.

   Each category can be further filtered by only enabling parts of the group in the analytics processor.

5. From the drop-down menu, select the action of FortiRecorder when it detects each category:

   • No action: Ignore events of this type.
   • Event: Accept events of this type, process them for notification, and display them on timelines and logs.
   • Event clip: Accept events of this type, process them for notification, display them on timelines and logs, and generate a video clip at the moment of detection.
6. Click Apply.