CCTV Planning Tool · 07 of 10

Face Recognition Feasibility Checker

Most "face recognition" deployments fail not because the AI is weak — they fail because the camera doesn't deliver enough pixels on the face, the angle's too steep, or the lighting kills the signal. This tool checks all four conditions at once and tells you straight: will this work? If not, exactly what to change.

The four things that decide if face rec will work

Face recognition isn't a single threshold like "4 MP camera = done". It's the simultaneous interaction of pixel density on the face, the angle from camera to subject, the lighting, and the frame rate. Get any one badly wrong and the AI fails — even with a top-tier engine.

1 · Pixels between the eyes

The canonical metric. Cooperative kiosk: ~20 px between eyes. Access control: ~32 px. Watchlist alert: ~45 px. Forensic ID: 60+ px. That translates to PPM thresholds of 300 – 1000+ at the face plane — well above general-surveillance PPM.

2 · Angle (pitch and yaw)

Past 15° pitch (camera looking down too steeply) or 30° yaw (subject's face turned away), accuracy drops sharply. A 3 m ceiling mount over a person 1.5 m horizontally gives 35° pitch — too steep for reliable matching, no matter how many pixels you have.

3 · Lighting on the face

Target: 200+ lux, fairly even, frontal. Backlit doorways are the silent killer — the AI sees a silhouette. IR at night works but loses colour features some engines need.

4 · Frame rate

For walking subjects: ≥15 fps to catch the frontal moment. For fast subjects (running, dense crowd): 25 fps. Irrelevant at kiosks where the subject stops.

👁️
Pixel density on the face
PPM at the face plane — driven by resolution, sensor width, focal length, and distance.
📐
Pitch & yaw angle
Mount height, distance to subject, and subject pose set the effective angle to the face.
💡
Lighting
Even, frontal lighting at 200+ lux. Backlit conditions and harsh shadows are deal-breakers.
🎬
Frame rate
Enough frames to catch the rare frontal moment for a moving subject.
⚠ The hidden trap
A camera that's perfect for "general surveillance" (250 PPM) is roughly 1/3 the pixel density needed for access-control face rec. Reusing existing cameras almost always means re-aiming, swapping lenses, or adding cameras at face level.

Step 1 · Pick your face-rec use case

Each tier demands different pixel density, angle tolerance, lighting, and frame rate. Picking the right tier is half the battle.

Step 2 · Start from a scenario preset (optional)

Real-world layouts pre-filled. Pick one to load camera + geometry + lighting, then tweak in Step 3.

Step 3 · Verify your setup

Set the camera, geometry, behaviour, and lighting. The verdict updates live as you change inputs.
Camera spec
Range: 2.8 mm (very wide) to 50 mm (narrow). Higher focal = more pixels on face but smaller field of view.
15+ fps recommended for moving subjects. 25+ for fast-walking or crowd flow.
Geometry & placement
Camera height from floor. Face-rec sweet spot is 2.2–2.6 m for face-level cameras, 2.8–3.2 m for ceiling.
Floor distance from camera base to where the subject's face will be.
Subject behaviour
Affects required frame rate and tolerance for yaw.
How much subjects naturally turn their face away from the lens.
Lighting
Lighting on the face — not the room average. Backlit doorways are the #1 cause of face-rec failure.
FEASIBLE
Setup will deliver reliable face recognition
All four factors clear the threshold for the chosen use case.
87/ 100
Pixels between eyes
Actual sampling at your face plane vs the use-case requirement.
Your sampling Target eye-px
Camera-to-face geometry
Side view — mount height, horizontal distance, and the resulting pitch angle.
Camera Subject (1.7 m) Pitch angle

Findings & specific fixes

Optimisation limits

Max distance at this focal
— m
Beyond this, the face won't have enough pixels for your selected use case. Stay closer than this in deployment.
Focal needed for current distance
— mm
The lens focal length that would put your current distance exactly at the use-case pixel target.

Refine your plan with IndoAi's companion tools

Face recognition feasibility is one piece of the puzzle. The other planning tools nail the rest — pixel density, lens choice, coverage, total camera count, and storage / bandwidth budgeting.

📐
PPM Calculator
Pixel density at any distance — the foundation behind face-rec math.
Check PPM →
🔍
Lens Selector
Pick the right focal length for your scene and the use-case target.
Pick lenses →
📷
Coverage Calculator
How many cameras at this focal length to cover a given area.
Plan coverage →
🧠
AI Camera Estimator
Total camera count and license tier for a full site rollout.
Estimate site →

Get this report as a PDF

A branded report with the feasibility verdict, four sub-scores, computed PPM and eye-pixel sampling, side-view geometry, issues with specific fixes, and the optimisation limits (max distance, required focal). Useful for customer quotes, installer briefs, or internal sales.