Skip to content
// signal 04 · crowds · in the pipeline

Count the room.
Read the flow.

Crowd intelligence is the in-the-pipeline signal that turns lobbies, parking decks, and festival gates into live occupancy readouts. Flow direction, dwell-time anomalies, and density thresholds, designed to be computed on-device, aggregate-only, with no individual ever identified.

// signal 04 · crowds
1,284
occupancy · aggregate only
no identifiers off the box
  • in the pipeline
  • in the pipeline
  • in the pipeline
  • in the pipeline
  • in the pipeline
  • in the pipeline
// crowds · in one paragraph

Crowds are not faces. They are counts, directions, and dwell-time signals.

Atlas is designed to aggregate these on-device, expose them through a simple webhook, and never identify the individuals being counted.

// read the room

Drag the crowd. Watch the threshold.

ATLAS · ATRIUM OCCUPANCYlive
// live headcount · aggregate-only
94in zone
capacity 240 · 39% of limit
// density
Calmnominal
0capacity 240
empty39% of limit
status — 94 present — calm.

For illustrative purposes only.

// capabilities

What Atlas does
for crowds.

01
Live occupancy counts

Per-zone headcount with smoothing, for lobbies, gyms, clubhouse capacity, and parking decks.

02
Flow direction

Net inbound vs outbound across any boundary. Catches reverse-flow at one-way ramps and after-hours exits.

03
Dwell-time anomalies

Surfaces a group that has been stationary in the wrong place for too long, without naming anybody.

04
Density thresholds

Fire-safety occupancy ceilings raised as alerts to the duty desk before a crowd becomes a violation.

05
Festival-mode tuning

Per-event profiles for Diwali, Ganpati, and New Year, suppressing normal-traffic alerts and sharpening density thresholds.

06
BMS / fire-panel webhooks

Designed to plug occupancy and density events into the building-management system, fire panel, or duty-app stack.

// what the pilot proves

Built for the conditions, measured in the pilot.

The pilot protocol →
By design · aggregate-only
A count, never a person.

The crowd signal is designed to emit counts and flow vectors, never a track tied to a person. That keeps occupancy intelligence on the right side of the DPDP Act while still answering 'how full is the atrium right now?'.

aggregateno individual tracking, by design
In the pipeline · per room
Accuracy depends on the room.

Crowd accuracy turns on sightlines, mounting height, and density. Rather than fake a generic web demo, Atlas will tune this signal on each site when crowds ships, which is why this is an in-the-pipeline conversation, not a canned number.

on-sitecalibrated per geometry
// see it run

In the pipeline,
tuned per site.

Crowd analytics is in the pipeline: accuracy will depend on each room's geometry, so it's tuned per site when it ships.