# Camera planning guide for face recognition

The measurements to take, in the order to take them. This costs a morning per
site and prevents the most expensive category of mistake in face recognition
deployment.

This guide deliberately publishes no pixel threshold. A number quoted without
the algorithm, the lighting, the pose and the matching mode would be wrong for
some deployments and misleadingly reassuring for others. Measure yours and have
a representative image assessed.

Full guidance: https://facerec.ayonix.com/resources/camera-planning-guide

## The survey

- [ ] Stand where a person stands and capture a still from the existing camera — at the exact position where the decision has to be made, not a nearby one and not the centre of the frame.
- [ ] Measure the pixels across the eyes on that still. This is the number that bounds everything downstream. Record it per capture point.
- [ ] Repeat at the worst hour of the day: winter morning, summer afternoon, and after dark if the site operates then.
- [ ] Note the light source relative to the face — front, side or behind. Behind is the problem, and it is usually structural.
- [ ] Measure the downward angle from the camera to the capture point. The shallower the better.
- [ ] Check the stream profile the analytics will actually receive: resolution, frame rate and compression, not what the camera is capable of.
- [ ] Observe where people actually look for ten minutes at a busy time. People look at doors, screens and each other.
- [ ] Note walking speed at the capture point, and whether there is a natural pause nearby.
- [ ] Record the height range that must be served, including children and wheelchair users where in scope.
- [ ] Have a representative thirty-second clip assessed before buying anything.

## The six variables, in order of how often each is the problem

- [ ] Pixels across the eyes at the capture point. Fix: move the camera closer, narrow the lens, or move the capture point. A higher-resolution camera in the same position usually helps less than expected because the field of view grows with it.
- [ ] Lighting direction and dynamic range. A glass entrance backlights everyone. Fix: front fill light, reposition away from the bright background, or move the capture point deeper inside.
- [ ] Angle between camera and face. A camera high enough to be out of reach often sees the top of a head. Fix: lower the camera, or move the capture point further from it so the angle flattens.
- [ ] Stream compression and frame rate. Compression removes the fine detail matching depends on. Fix: give the analytics a higher-quality second stream profile where one is available.
- [ ] Subject movement and exposure. Motion blur is a function of shutter speed, which is a function of available light. Fix: more light, or choose a capture point where people naturally slow.
- [ ] Where people look. Fix: put the camera where attention already goes — a screen, a signal light, or the door itself.

## What good camera placement cannot fix

- [ ] An enrolment process carried out under conditions the site never reproduces.
- [ ] A threshold set without reference to what each error costs at that location.
- [ ] A gallery that has grown past the point where the 1:N error rate is acceptable.
- [ ] A deployment purpose that was never narrow enough to test.
- [ ] A missing fallback path for people the system will not recognise.
- [ ] Demographic performance differences, which must be measured at the deployment and acted on separately.

---

Published by Ayonix, facerec.ayonix.com. Ayonix sells face recognition and
therefore has an interest in how these checklists are used — they are written
so they can be applied to Ayonix as readily as to any other supplier, and if
following one leads you to a different supplier or a different category of
system, that is a legitimate outcome.

No accuracy percentage appears in any Ayonix material, for Ayonix or any other
vendor, because a figure without its threshold, dataset, gallery size and
demographic breakdown cannot be reproduced.

Corrections: infojp@ayonix.com. Last reviewed 11 September 2026.