Camera Selection & Placement

Camera choice and physical placement decide recognition quality before any software setting does. This guide covers the site-survey side; the device-selection criteria and compatible hardware live in Choosing a camera, and the two processing models in Cameras.

Placement rules of thumb

Mount cameras at roughly eye level, facing oncoming travelers, and keep head angles modest — roll within ±15°, pitch and yaw within ±30° at the recognition point. The working target is at least 50 pixels between the eyes where identification should happen; walk the corridor with the intended lens and mounting distance and verify that number where it matters, not at the camera. In corridors, plan several cameras along the passage so a traveler is seen from more than one angle — CIGS merges those detections into one identity.

Survey the light first

Lighting is the most common cause of poor results. Prefer consistent frontal light, keep light sources out of the camera's field of view, and avoid backlight (windows and glass facades behind travelers are the classic failure). Check the site at different times of day — a corridor that works at noon can fail at sunset. Leave Wide Dynamic Range off where liveness matters, and confirm the stream holds 15 fps or more in the darkest expected conditions.

Plan the infrastructure with the camera

Treat power and network as part of placement: PoE carries both on one cable and simplifies mounting-point choices, and each RTSP camera needs about 4 Mbps of stable bandwidth back to the node. If cabling or server capacity is the constraint, consider on-edge devices for the affected positions — placement rules stay identical, only the compute location changes.

Validate before you fix anything to a wall

Do a temporary rig test: mount the shortlisted camera at the intended height and angle, capture real walkthroughs, and check face size, angles, and match scores in VPP. An afternoon of validation routinely saves weeks of re-mounting and re-cabling.