CrowdIQ: two cameras side by side, one in alert
Back to Apar Drishti
Apar DrishtiCrowds & public spacesWorking prototype

CrowdIQ

A crowd can build in minutes.CrowdIQ counts it as it grows.

How many people are here, right now? CrowdIQ, part of Apar Drishti, finds each person in the frame, even the small ones far from the camera, and follows them as they move. Each camera has its own limit, and your control room knows the moment a crowd goes over it.

Per camera
its own limit
3 ways
screen, serial line, buzzer
Logged
every alert, with its frame
On-prem
the footage never leaves
CrowdIQ · on your own server
CrowdIQ: a station platform over its limit, every person boxed, the camera in red

Watch it work

CrowdIQ on its real screens: people counted, the alert at the limit, the control room, and the vision model’s read.

CrowdIQ: people counted, an alert at your limit

The challenge

Your cameras see the crowd. Nobody can count it.

A crowd builds fast

A crowd can build in minutes, while the control room watches other screens.

Too many screens

Nobody can count heads on every camera, every minute.

An alarm that flickers

An alert that switches on and off with every dip is soon ignored.

Afterwards, no record

Without the picture and the count, the review runs on memory.

The idea

Every person counted. An alert at your limit.

1server at your site
RTSP streamsVideo filesAn uploadMultiple cameras, one screen
4jobs, one alert
Counts every personHolds the alert steadyTells the control roomReads the scene
Your control room knows at once: which camera, and how many people. Then the vision model reads the scene, rates the risk, and says what to do.

How it works

From a count to the control room

Real screens from the working prototype, on public footage.

01 · Step by step

Counted, checked, sent, explained, kept

Every camera goes through the same steps, every second.

  • The count is checked against that camera’s own limit
  • Over the limit, the camera turns red and the alert holds
  • The control room knows first; the vision model’s read follows
  1. 1The camera sees the crowd
  2. 2Each person found, even far from the camera
  3. 3The count, second by second, against the limit
  4. 4Over the limit: the camera turns red, the alert holds
  5. 5The control room knows: which camera, how many
  6. 6The vision model reads the scene and rates the risk

Every alert is kept, with what the operator saw and the frame the AI judged.

02 · Count

Every person in the frame

It finds each person, even the small ones far from the camera, and follows them as they move.

  • Each person boxed on the picture
  • People in frame, the peak, how many it is following, and the trend
  • A green banner while the crowd is under the limit
CrowdIQ · one camera
CrowdIQ: a plaza from above, every person boxed, with the count, the peak, the people followed and the trend
03 · Multiple cameras

Multiple cameras, one screen

Each camera has its own limit, and turns red when the crowd goes over it.

  • A limit for each camera: a platform and a plaza hold different crowds
  • The banner names the camera and the count
  • Pick a camera for its numbers and its chart
CrowdIQ · all cameras
CrowdIQ: two cameras side by side, the station platform in red over its limit
04 · Alert out

The control room knows at once

Which camera, and how many people. On a screen, over a serial line, or with a buzzer and a light.

  • Once it crosses, the alert holds through short dips, so it does not flicker
  • A control room screen, over your own network
  • A serial line (RS-232), or a buzzer and a light
On siteThe network alert is shown working here. The serial line, buzzer and light are options we wire to your hardware when we set it up.
CrowdIQ · control room
CrowdIQ alert dashboard in the control room: the alert count and each alert with its camera, count and limit
05 · It advises

A vision model reads the scene

On your own server: it reads the scene, rates the risk, and says what to do.

  • The scene, in plain words
  • The probable cause
  • Recommended actions for the team
On siteThe model runs on your own server. The footage never leaves your site.
CrowdIQ · AI safety analysis
CrowdIQ AI safety analysis: risk medium, the scene, the probable cause and the recommended actions
06 · The proof

Every alert is kept

With what the operator saw, and the frame the AI judged.

  • Both pictures, side by side
  • The vision model’s note, kept with each alert
  • Export the log for your records
CrowdIQ · alert history
CrowdIQ alert history: what the operator saw, what the AI looked at, and its note

How we engineered it

Six ideas behind an alert you can trust

What keeps CrowdIQ quick to warn, and steady once it has.

The small ones count too

It finds each person, even the small ones far from the camera, and follows them as they move.

A limit per camera

A platform and a plaza hold different crowds, so each camera has its own limit.

An alert that does not flicker

Once it crosses, the alert holds through short dips, so the control room sees one alert, not a light switching on and off.

The count alerts, the AI explains

The count raises the alert. The vision model then reads the scene, rates the risk and says what to do.

Built to plug into your site

Your cameras over RTSP, video files or an upload. The alert out to a screen on your network, a serial line, or a buzzer and a light.

Your own server

Counting, the alert and the vision model run on your own server. The footage never leaves your site.

What the build proves

A working prototype, built to your site

Per camera
limit
3 ways
the alert goes out
On-prem
nothing leaves the site
CSV
the alert log
  • Know how many people are here, right now, on every camera
  • An alert at the limit you set: which camera, and how many
  • The vision model’s read of the scene: the risk, and what to do
  • A record of every alert, for the review afterwards

More screens

Every screen, from the working prototype

CrowdIQ: a platform over its limit

A station platform over its limit, every person boxed.

1 / 6

Built to your site

On your own cameras?

CrowdIQ is part of Apar Drishti, a working prototype built to your site. Bring us a clip from one of your cameras, and we will show you the count.