Frigate camera FOV & identification planner
A camera can be brilliant at detecting a person and useless at identifying one — the difference is pixels per foot at the distance that matters. Tell this planner your camera's resolution and lens, pick the distance to the gate, door or driveway spot you care about, and it works out the pixels per foot, the band that lands in — Detection, Observation, Recognition or Identification — and the maximum range for each level, using the DORI thresholds from IEC 62676-4.
Private by design. Everything on this page runs locally in your browser. No camera details are uploaded or sent anywhere.
Plan it: one camera, one target distance
Use the resolution of the stream Frigate runs detection on — usually a substream, not the headline sensor resolution — and the lens's horizontal field of view. Presets fill in the common values; every field can be overridden.
Lens focal lengths are approximations — the same 2.8mm lens gives a different field of view on different sensor sizes. If the camera's spec sheet lists a horizontal FOV, trust that number and type it in directly.
Pixels per foot at your target
—
DORI ranges for this camera & lens
| Level | Threshold (px/ft) | Max distance | Max distance | Your target |
|---|
What this means for Frigate
What pixels per foot actually means
Spread a camera's horizontal pixels across the width of the scene it can see at a given distance and you get pixels per foot (or per metre). Walk away from the camera and the scene gets wider while the pixel count stays fixed, so pixels per foot falls steadily — detail is a budget that distance spends. The maths is one line: the scene width at distance d is 2 × d × tan(HFOV ÷ 2), and pixels per foot is the stream width divided by that scene width.
Two consequences surprise people. First, a wider lens is not free coverage: doubling the field of view halves the pixels per foot everywhere, which is why one wide camera over a whole yard identifies nobody. Second, resolution only counts where it is used — Frigate detects on the stream you give the detect role, so a 4K camera detecting on a 640-pixel substream plans like a 640-pixel camera.
The DORI levels, in plain terms
DORI comes from the IEC 62676-4 video surveillance standard, which defines four detail levels by pixel density: Detection at 25 px/m (7.6 px/ft), Observation at 62.5 px/m (19 px/ft), Recognition at 125 px/m (38 px/ft) and Identification at 250 px/m (76 px/ft). Detection means you can say a person is there. Observation means you can follow their actions and describe them generally. Recognition means someone familiar can tell who it is. Identification is the evidential bar: enough facial detail to identify a stranger.
Frigate's own bar sits lower than most people assume. Object detection — the person/car/animal boxes that drive events and alerts — works well at detection-band densities. It is the layer on top (sub-labels, vehicle details, GenAI descriptions of a scene) that starves first as pixels per foot drops, well before basic detection does.
Choosing a lens: width versus detail
Every camera placement is a trade between covering ground and resolving detail, and the planner above makes the trade visible. A 2.8mm lens on a 4MP camera identifies out to roughly 12 feet — fine over a porch, hopeless across a driveway. The same camera behind a 6mm lens identifies past 30 feet but sees a corridor, not a yard. The usual answer is not a better camera, it is a choke point: aim a narrower lens at the one place everyone must pass — the gate, the front door, the gap between parked cars — and let a wide camera provide context around it.
Height matters too, though this planner keeps the geometry flat: a camera mounted high looks down at steep angles that shrink faces regardless of pixel density. For identification duty, closer to head height and closer to the choke point beats higher resolution almost every time.
Worked example
A 4MP camera (2560 px) with a standard 3.6mm lens (84°) watching a gate 25 feet away: the scene at the gate is about 45 feet wide, giving ≈57 px/ft — solidly in the Recognition band. The family will recognise visitors; a stranger's face will not quite hold up. Identification on this camera runs out at about 18.7 feet, so moving the same camera to cover the path six feet nearer the house — or swapping in a 6mm lens for the gate view — is what closes the gap. Run the numbers above with your own distances before drilling any holes.
Sources
- IEC 62676-4 — the DORI pixel densities (25 / 62.5 / 125 / 250 px per metre) this planner converts to px/ft.
- Frigate object detector docs — detect stream roles and resolution guidance.
- Frigate config validator & storage planner — the companion tool: SHM sizing, recording storage and 0.18 config checks.
- Frigate NVR with Home Assistant — the full setup this tool supports.
- Frigate zones & masks tuning — once the camera can see it, teach Frigate where to care.
Frequently asked questions
Why can my camera detect people but never identify them?
Because the two jobs need very different pixel densities. Frigate object detection works at roughly 7.6 px/ft (the DORI Detection level), while identifying a stranger needs about 76 px/ft — ten times as much. A wide camera at the far end of a yard usually sits between the two: plenty for alerts, far too little for faces.
Which resolution should I enter — the camera's or the detect stream's?
The detect stream's. Frigate runs detection on the stream assigned the detect role, which is usually a low-resolution substream. Enter that stream's width, or the planner will flatter the camera by a factor of four or more.
Is a narrower lens or a higher-resolution camera the better fix?
Usually the lens, aimed at a choke point. Doubling resolution doubles pixels per foot; swapping a 2.8mm lens for a 6mm roughly doubles it too — but the lens also removes wasted pixels spent on sky, walls and the neighbour's fence. The cheapest fix of all is moving the camera closer to the spot that matters.
Does this planner upload anything?
No. It runs entirely in your browser with JavaScript — your camera setup and distances never leave your device.
Related guides
- Frigate config validator & storage planner — SHM, storage and 0.18 config checks for the same cameras.
- Frigate NVR with Home Assistant — detectors, go2rtc streams and zones, set up end to end.
- Frigate zones & masks tuning — fewer false alerts once detection is running.
- NVR storage calculator — drive sizing for the footage these cameras produce.

