Book a demo

Counting and occupancy In pilot

Count people, cars and free spaces.

Draw on the camera image what you want to measure: a zone, a line people cross or each parking space. The system counts, decides when something is full or free, and every change can start an automation.

Book a demo
Demo footage
Recording of the points-of-interest editor: on a parking-lot snapshot the "Cocheras fila sur" rectangle is drawn, four spaces are drawn inside it, two are set to "car or vehicle" and it is saved. Recorded in the TeroVision panel, with simulated cameras on a demo database. The panel is in Spanish.

Measure every spot in your shop

Each point of interest is saved as a camera position: the checkout, the display, the entrance. The camera goes through them, takes a photo of each one every minute and the system counts the people inside the zone you marked.

CAMERA 3 • OVERVIEW

Position
Overview
Shows every position
In the zone
–
Detected in frame
–
The shop is an AI-generated video. The detected people and the counts are real: they come from a detection model running on every frame. At each position only people inside the zone are counted.

Crowding

In pilot

For the checkout, the entrance, the fitting rooms. You set how many people are too many and for how many minutes in a row. When that happens, the manager gets an alert with the marked photo.

  • People and minutes threshold, per position.
  • Someone walking past does not trigger it: the count has to hold.
  • The photo arrives with the zone and every person marked.

Engagement zones

In pilot

For displays, promotions and tasting points. It counts minute by minute how many people come close. At a stand, it is what measures how many stop.

  • People in the zone, minute by minute.
  • Compare campaigns, shops and hours with the same number.
  • One zone per position: one camera covers several spots.
  • It detects and counts people. It does not recognize faces or identify anyone. In pilot
  • Automatic tour between positions, one photo of each per minute. In pilot

Save up to 90% on installation

To count people you do not need an NVR or DVR recorder, a PC running an agent in the shop, network cabling or expensive cameras. The camera plugs in, joins the WiFi and sends its photos to the system.

Savings depend on the shop. We compare against an installation with a recorder, a PC, cabling and cameras with built-in analytics.

  • Traditional installation: NVR or DVR recorderWith TeroVision: Not needed: it records to the microSD and the system does the counting.
  • Traditional installation: PC running an agent in the shopWith TeroVision: Not needed: the camera connects on its own.
  • Traditional installation: Network cabling, conduits and building workWith TeroVision: A power outlet and the shop's WiFi.
  • Traditional installation: Expensive cameras with analyticsWith TeroVision: An affordable camera: the intelligence lives in the system.
  • Traditional installation: Installer and days of workWith TeroVision: Anyone on the team puts it up in five minutes.

Three ways to count

Free or taken spaces

Parking spaces, tables or beds drawn as rectangles or polygons. Each is taken by a person, a car or either.

  • Alerts when a space is taken, freed, when all are full and when there is room again.
  • It changes state only after several snapshots in a row confirm it: a passing car does not count.
  • A car parked across two spaces takes only one.
In pilot

Crowding within hours

N people or more for M minutes, in a zone and within opening hours. With a start and an end event.

  • The end event carries the peak and how long it lasted.
  • A count hovering around the limit does not cut it short.
  • Outside the hours it does not start; if it was ongoing, it ends.
In pilot

Live counting

On the live video, following each person: line crossings with direction, dwell time in a zone and unique people.

  • "4 in, 2 out" per line, with the daily total.
  • "Someone stayed more than 5 minutes", once per person.
  • Summaries at the end of every hour and every day.

Needs the live video processor, on a server with a GPU: up to 8 cameras per space.

In pilot

Which model is watching

Detection runs on our servers with D-FINE, an open-source detector under the Apache 2.0 licence, on a GPU. Live counting follows each person with ByteTrack, under the MIT licence. For counting we only use models licensed for commercial use.

  • It counts people and vehicles. It does not recognize faces or identify anyone.
  • What is stored are numbers and events. Counting samples are deleted after 60 days.
  • Fall alerts, in pilot, are separate: they look at body posture on the live video, only on the cameras where you turn them on.

In pilot

Every count is a trigger

"Spaces: taken, free or full", "Crowding started or ended", "A line was crossed", "Someone stayed more than X minutes" and "Counting summary" start flows. "Current count" asks, mid-flow, how many there are right now.

In pilot See the counting templates

What is new

Shall we try it with your cameras?

We show you the panel running and build the first automations for your place together. Whatever is in pilot gets tested with you before we promise it.

Book a demo Write to us