Detection & AI

Vistralio supports object detection, motion detection, ALPR, and face matching.

Per-device detection settings

For each device you can configure:

• detection enabled
• detector model
• object list
• detection FPS
• minimum confidence
• hardware mode
• per-object overrides

Object selection

The detector label list is loaded dynamically from the selected detector.

Vistralio keeps support for labels such as:

• object labels from the model
• motion
• face
• plate

Zones vs masks

Detection behavior now separates zones from masks.

Zones

Used for event context.

• named alert areas
• optional note per zone
• optional threshold override
• optional dwell time override

Masks

Used for ignore areas.

• separate from zones
• can be titled and documented
• suppress object/motion pickup inside the masked area

No zones

If a label has no zones configured, Vistralio scans the entire frame.

Inferred scene details

Vistralio can add readable inferred details to event metadata without using a dedicated attribute model.

Examples:

• person upper/lower clothing colour
• likely hi-vis / hard-hat / jeans cues
• vehicle colour
• animal colour

These are shown in the Events UI and per-device event views.

ALPR

Plate recognition is per-device. Plate events include:

• plate text
• plate crop
• event metadata

Known plates are managed through the plates API/UI section.

Face recognition

Face recognition is per-device. Face events include:

• matched face label
• event metadata

Known faces are managed through the faces API/UI section.

Related endpoints

• GET /api/system/detectors/{detector}/labels
• GET /api/streams/{camera_id}/detections
• GET /api/events
• GET /api/faces
• POST /api/faces
• GET /api/plates
• POST /api/plates