FireIQ — a control room watching a fire, illustration created with AI
Back to Apar Drishti
Apar DrishtiFire & smokeWorking prototype

FireIQ

A fire starts small.FireIQ sees it while it is.

A little smoke, in a corner of a room, where nobody is looking. Your camera can see it, but nobody can watch every screen, every minute of every shift. FireIQ, part of Apar Drishti, watches that camera for you. It runs on one computer at your site, and nothing leaves it.

1–5
how serious, from the picture
3 ways
screen, fire panel, buzzer
On-prem
nothing leaves the site
Logged
every alarm, with its picture
FireIQ · on your own server
FireIQ: a fire boxed on the camera feed, its severity, and the vision model's note on the scene

The challenge

Your cameras can see the smoke. Nobody is watching.

A fire starts small

A little smoke, in a corner where nobody is looking.

Too many screens

Nobody can watch every screen, every minute of every shift.

An alarm says little

Where is it, how bad is it, and what should the team do first?

Afterwards, no record

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

The idea

The camera you already have, watching for you.

1computer at your site
Cameras you already haveRTSP streamsVideo filesNothing leaves the site
4jobs, one alarm
Sees smoke and flameRates how seriousSends the alarmKeeps the proof
The alarm never waits for the AI. The vision model only advises: it can raise the severity, never lower it, and it never switches an alarm off.

How it works

From the first smoke to the day's log

Real screens from the working prototype, running on a demo clip.

01 · Step by step

Seen, rated, sent, explained, kept

Every alarm follows the same steps, so your team knows while the fire is still small.

  • When the smoke stays for half a second, the alarm goes off
  • When a flame shows, it turns to fire
  • The alarm goes out first; the vision model's note follows
  1. 1The camera sees a little smoke
  2. 2Smoke alarm: it stayed for half a second
  3. 3A flame shows: it turns to fire, rated 1 to 5
  4. 4The alarm goes out: screen, fire panel, buzzer
  5. 5The vision model reads the picture and advises
  6. 6Your team acts, while the fire is still small

Every step is logged with its picture and time.

02 · Detect

Smoke and flame, many times a second

It looks for smoke and flame, and marks what it finds, with how sure it is.

  • Each find is boxed on the picture, with its score
  • Once raised, the alarm holds through a missed frame or two, so it does not flicker
  • An optional second model double-checks smoke-only alarms, so steam or glare is less likely to set one off
FireIQ · camera feed
FireIQ: a flame boxed on the camera feed with its score, under a FIRE banner with the counts and SEV 3/5
03 · Severity

How serious, from one to five

It judges how serious it is by how much of the picture is burning.

  • The banner on the picture shows the status, the counts and the severity
  • Smoke and fire, and whether it is growing or steady
  • The vision model's own score sits beside the detector's
By designThe vision model can raise the severity, never lower it.
FireIQ · severity
FireIQ: the detector's severity and the vision model's severity side by side, with the scene description
04 · Alarm out

The alarm goes out at once

To a control room screen, to a fire panel, or to a light and a buzzer on the floor.

  • A control room screen, over your own network
  • A fire panel, over a serial line (RS-232)
  • A light and a buzzer on the floor
On siteThe demo shows them as options. We wire them to your panel and hardware when we set it up.
FireIQ · alert outputs
FireIQ alert outputs: console, network webhook, RS-232 serial, and a light and buzzer
05 · It advises

A vision model reads the picture

On the same computer: what it sees, the likely cause, and what to do.

  • The alarm never waits for it: it only advises
  • It can raise the severity, never lower it
  • It never switches an alarm off
On siteThe model runs on your own server. The picture never leaves the site.
FireIQ · AI fire-safety analysis
FireIQ: the vision model's scene description and probable cause beside the camera feed
06 · The proof

Every alarm keeps its proof

The picture, the time, and a line in the day's log, from the first smoke on.

  • Download the day's log as a CSV
  • The vision model's note is kept with each alarm: the scene, the cause, the actions
  • Review what happened after the moment has passed
FireIQ · alarm history
FireIQ alarm history: each alarm with its time, severity, the likely cause and the recommended actions

How we engineered it

Six ideas behind an alarm you can trust

What keeps FireIQ quick, steady and honest about what it knows.

One computer at your site

Detection, the alarm and the vision model run on one computer at your site. The footage never leaves it.

An alarm that does not flicker

Once raised, the alarm holds through a missed frame or two, so the control room sees one alarm, not a light switching on and off.

A second look at smoke

An optional second model double-checks smoke-only alarms, so steam, dust or glare is less likely to set one off.

The AI advises, never decides

The vision model can raise the severity, never lower it, and never switches an alarm off. The alarm never waits for it.

Built to plug into your site

Your cameras over RTSP, or video files. The alarm out to a screen on your network, a fire panel, or a light and a buzzer.

Proof from the first smoke

Every alarm keeps its picture, its time and a line in the day's log, with the vision model's note beside it.

What the build proves

A working prototype, built to your site

1–5
severity
3 ways
the alarm goes out
On-prem
nothing leaves the site
CSV
the day's log
  • Your team knows while the fire is still small, and has time to act
  • One alarm, sent at once to the screen, panel or buzzer you choose
  • The vision model's read of the scene: what it sees, the likely cause, what to do
  • A record of every alarm, for the review afterwards

More screens

Every screen, from the working prototype

FireIQ: a flame boxed on the camera feed

A flame boxed on the camera feed, with its score, the counts and the severity.

1 / 5

Built to your site

Want it watching your floor?

FireIQ 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 what it sees.