Parkinto Detection

How Parkinto detects free and occupied parking spaces

In short

Parkinto Detection reports each marked parking space as free or occupied from an existing static IP camera, every 30 seconds, for $69 per camera per month. AI finds the vehicles in each image with 99.7% accuracy in everyday operation.

Start free trial Book a 20-min demo

14-day free trial · no credit card required

Camera view of a business park in Sulgen, Switzerland, after detection — green is free, red is occupied — Business park in Sulgen, Switzerland, after detection — green is free, red is occupied. Spaces are polygons you draw once; painted lines are not needed.

From camera image to free space, in four steps

The camera sends an image

Every 30 seconds by default, over common camera streams. No extra hardware is installed on site.

AI detects the vehicles

Cars, vans, trucks, buses, motorcycles and bikes. Each gets a type and a confidence score.

Matched to your spaces

You draw each space once as a polygon on the camera image. A space is occupied when a vehicle’s centre lies inside it. Painted lines are not needed.

Delivered where you need it

Dashboard, public map, LED signs and your API.

See all outputs ↓

One camera on Velké náměstí, Kroměříž

Raw camera image of Velké náměstí in Kroměříž with parked cars — Camera imageWhat the camera sends — an ordinary image, nothing on the ground.

The same image with detected vehicles boxed and each space shaded green for free or red for occupied — DetectionThe same image after detection: vehicles found, every space free or occupied.

Parking guidance sign in Kroměříž with LED counts of free spaces for Velké náměstí and the parking house — On the streetFree-space counts on the city’s guidance signs, fed by the API.

Public Parkinto map of Velké náměstí showing 16 free spaces out of 80, with each space marked green or red on an aerial photo — Public mapThe same spaces on the public map for drivers, with directions to the car park.

Step by step through a real installation: how installation works

The whole topic in one guide: parking space occupancy detection explained

What Parkinto recognises

Every object gets a type and a confidence score, so the data tells you not only that a space is taken, but by what.

Vehicles

Car, van, truck, bus, motorcycle and bike. The type is reported through the API, so you can count free truck bays separately or spot a car standing on a truck space.

Licence plates

Included in the licence and read by an ordinary camera. Reading needs a distance of 4–8 m, so one camera covers 3–5 spaces. Each plate is available through the API, and a webhook tells your system when the car arrives or leaves — Parkinto does not store the plate.

Space types

Give each marked space a type — regular, disabled, electric vehicle or private. Occupancy of the car park is reported for regular spaces, and the Statistics add-on breaks usage down by type.

RegularDisabledEVPrivate

STEP 1
Camera 4–8 m from the space
One camera reads plates on 3–5 marked spaces.
STEP 2
Plate read on the marked space
The plate is matched to the space it stands on.
STEP 3
Sent to your system
Webhook: PARK 1NT0 arrived at A12 · later: A12 freed
STEP 4
Use it
Match plates to permits, guests or employees in your own system — e.g. hotel or company car parks.

Parkinto does not store plates — they are only passed on through the API. Occupancy detection works on every camera; licence plates only on cameras placed 4–8 m from the spaces.

For time limits alone you don’t need licence plates — stay length works on every camera. Stay length ↓

Detecting boats in a marina? See Parkinto Marina

Stay length and time limits — no licence plates needed

Parkinto shows how long each space has been occupied. Set a maximum stay for the car park — e.g. 2 hours at a shopping centre — and spaces over the limit are highlighted in the portal. Works with every camera, without reading licence plates or storing personal data. The REST API also returns when each space became occupied.

Shopping centre · level 1Max. stay 2 h

A12occupied 38 min
A13free
A14occupied 2 h 47 minover limit

How accurate is Parkinto?

99.7% is Parkinto’s own continuous measurement on live installations: the share of reported space states that match a person checking the same camera images. An independent audit by the Czech Technical University in Prague (Faculty of Transportation Sciences) measured an earlier version at 99.4%. If a person can clearly see from the camera image whether a space is taken, Parkinto usually can too. Accuracy keeps improving over the first weeks on a particular car park, as the model settles into that view. See where parking space detection fails and how to test it before you buy any system.

99.7%accuracy in everyday operationMeasured across live installations in city centres, truck parks and marinas.Share of reported space states that match a person checking the same camera images.How we measure

Works in real conditions

Car parks are not studio scenes. These are the conditions customers ask about most.

Car park after a snowfall with detected vehicles and every space reported — After a snowfallSnow on the ground and on the cars, white cars included — vehicles found and every space reported.

Car park at night with detected vehicles — At nightA car park at 1 am under the camera’s own infrared — cars detected down to the far row.

Unmarked car park where Parkinto counts vehicles — Unmarked car parkNo painted spaces at this mountain trailhead — Parkinto only counts the cars. Against the car park’s capacity, the town’s signs then send drivers to the lot that most likely still has room.

Rain, snow and fog: Works while vehicles stay visible in the image. The camera needs a weatherproof IP66 housing.
Unmarked and gravel lots: Spaces are polygons on the image, not painted lines. Zones can count vehicles instead of bays.
Garages and parking houses: Works indoors. Low ceilings mean roughly 5–25 spaces per camera.
Big and complex sites: Several cameras can watch the same space; the one most confident about what it sees decides.
Occlusion: Works when at least half of each vehicle is visible from the camera.

Cameras: what you need

Any static IP or CCTV camera that provides a snapshot URL or a supported stream, new or already installed. Moving PTZ cameras are not suitable, because the marked spaces would no longer match the view. Why cameras rather than ground sensors: parking occupancy detection, camera vs sensor.

Two static cameras mounted high on a street light pole against a blue sky — Static cameras on a street light pole — high enough to see over the parked cars.

Resolution

HD 1280 × 720 is enough; Full HD 1920 × 1080 for spaces far from the camera. Even 640 × 480 will run — with fewer spaces per camera.

Position

As high as possible — a street light pole at 5–6 m is ideal. The higher the mount, the fewer vehicles hide behind each other.

Brands

Tested with Hikvision, Axis, Uniview, Reolink and TP-Link. A proven budget choice is the Hikvision DS-2CD2045FWD-I.

Spaces per camera

Outdoor, typical mounting25–35

Outdoor, high mounting and good angleup to ~100

Indoor, garage or parking house5–25

With licence plate reading3–5

Real coverage depends on the mounting height and the angle onto the spaces. Send us the camera view and we will tell you what it covers.

Not sure your camera fits? Send us the view.

Send a snapshot from the camera — we reply with how many spaces it can cover.

Check my camera setup Start free trial

Connection and cloud

Detection runs in the cloud, so nothing has to be installed at the car park beyond the camera itself.

Supported:HTTP(S)FTPRTSPRTSP.meMJPEGYouTube live

Reaching the camera

The camera or recorder must be reachable from the cloud — a public IP with a forwarded port, or a VPN. Cameras that push snapshots need no inbound access.

Connectivity

Ethernet, Wi-Fi or mobile data (LTE/4G/5G). Roughly 1–3 Mbit/s per camera.

Cloud in the EU

Redundant servers in Belgium and the Czech Republic. No server on your site. On-premise or a dedicated regional instance for large deployments.

Why not in the camera

Cloud detection works with any inexpensive camera, the models improve centrally for every customer, and adding a camera is configuration rather than a device purchase.

Comparing with in-ground sensors? See the technologies side by side

Data and privacy

Camera images are not archived

Each image is processed to determine occupancy. Images and video are not archived — only the latest frame is kept so you can see the camera view in your portal. Licence plates are not stored — they are only passed on through the API.

GDPR compliance (PDF)

✓Only the latest frame kept for your portal view
✓Licence plates passed on through the API, never stored
✓All processing inside the EU
✓Compliance confirmed in writing by ÚOOÚ, the Czech Office for Personal Data Protection

Integrations: getting the data out

Occupancy is only useful where people see it — in your dashboard, on a sign at the entrance or inside your own system.

REST API

Query the current state of any site or space with an HTTP GET; the answer is JSON.

Webhooks

A notification whenever the car park changes — a car arrives, a car leaves, space A12 is freed — sent as an HTTP POST to your endpoint.

Public map for drivers

A web page showing free spaces, embeddable in your own site as an iframe.

LED signs

Any sign can read free-space counts from the API; there is a ready integration with Tetronik.

Endpoints, payloads and webhook events: API documentation

Optional add-on+$89 per parking site / month

Statistics

Average and peak occupancy, stay length, turnover and arrivals — per site and per space, compared with the previous period and exportable to CSV, Excel or PDF, with a summary at the end of each month.