Moonlight Analytica Request demo
moonlight / janus / floor-01 / live Janusphysical-space intelligence Caposeosearch & AI-search Launch Sentinelproduct-market 42ms inf LIVE · 00:00:00
about this workspace · janusPNL-01
Physical-space intelligence

Cameras become a live operating model of your floor.

Janus turns ordinary cameras into zones, dwell data, queue depth, and role-aware people tracking — without surveillance creep. Generic CV tells you what it sees; Janus learns what your location means.

+53%Promo dwell lift
4.4sAvg wait tracked
94%Model confidence
42msInference latency
● workspace healthyuptime 99.98%janus-core v2.4.1
janus / convenience — live · 17 tracked · 4 staff separated
Janus dashboard — convenience store live view with 17 tracked people and 4 staff separated
● stream stablecam-array 12/12conf 0.94lat 42ms
floor-01 · live tracking · conveniencePNL-02 · CAM 04+07
ROLE RECOGNITION · STAFF SEPARATED
● 0 active tracksbusiest: —painter-sort ondpr≤260fps
mechanism · pipelinePNL-03
Signal gap detected

What happens on your floor is invisible to analytics.

You can see clicks, carts, and conversions — but the physical floor is a black box. Do shoppers pause at the promo display? Are staff crowding the areas customers need? Which zones bottleneck before a complaint surfaces? That gap kills margin in silence.

Janus closes this gap
01
Cameras streamYour existing CCTV or IP cameras — no new hardware required.
RUNNING
02
Role recognition separates who is whoCustomers, staff, security, and vendors classified at 94% confidence, so every metric is staff-excluded by default.
RUNNING
03
Zones, dwell & queues become live dataEvery named zone streams occupancy, dwell time, and queue depth at 42ms latency.
RUNNING
04
Operators act on the intelligence layerStaffing, promo placement, and layout decisions driven by what the floor actually shows — not gut feel.
RUNNING
● pipeline nominal4 stagesthroughput 30fps × 12 cams
deployments · 3 verticalsPNL-04
Three verticals · one intelligence layer

From retail floors to transit systems.

The same Janus stack — role recognition, zone intelligence, dwell tracking — adapts to any space where people move and operators need to act.

Janus — Retail deployment feed / retail · live
+53%Retail — Promo dwell — shoppers only, staff auto-excluded
Janus — Restaurant deployment feed / restaurant · live
+22%Restaurant — Guest table dwell — staff movement excluded
Janus — Transit deployment feed / transit · live
+19%Transit — Crowd flow — congestion zones flagged early
● 3 feeds attachedsame stack · zero retraining
dashboards · live capability viewsPNL-05 · 4 WIDGETS
Inside the dashboard

Live capability views.

The same Janus intelligence layer, the way operators actually read it — live occupancy, dwell heatmaps, campaign attribution, and predictive forecasting.

live occupancy2,340 visitors/hr
In store now 247 62% of capacity
Entries +318/hr Exits 291/hr Queue 4 lanes

Real-Time People Tracking

Count every person entering, exiting, and moving through your space with computer-vision powered tracking — live occupancy by zone, directional flow, and queue formation, no wearables or beacons required.

dwell heatmap3.2 min avg dwell
Avg dwell3.2min
LOWHIGH
⚠ bottleneck: Electronics · 4.1m

Dwell Time & Heatmaps

Animated heatmap overlays on your actual floorplan reveal engagement hotspots, dead zones, and bottleneck corridors — broken down by hour, day, or campaign period.

campaign attribution+23% conversion lift
In-store impact+23%
Visits340 Revenue$12.4K Cost/visit$0.42 ROAS4.2x

Campaign ROI Attribution

Janus traces digital ad impressions through to physical store visits and in-store conversions — true campaign ROI with zone-level attribution and A/B testing.

predictive forecast94% accuracy
Forecast accuracy94%
▲ PEAK 2:30PM

Predictive Forecasting

ML models trained on your historical traffic predict peak hours up to 7 days out — optimize staffing, inventory, and promo timing with 94% forecast accuracy across every zone.

● 4/4 widgets streamingrefresh 1.6ssource: floor-01
modules · intelligence layerPNL-06
Capabilities

Generic CV sees. Janus understands.

LOADEDRR

Role Recognition

Classifies every detected person as customer, employee, security, or vendor. All downstream metrics are staff-excluded by default.

LOADEDUC

Uniform Classifier

Learns your specific uniform colors and patterns to separate staff from guests across lighting conditions and camera angles.

LOADEDBT

Camera Blind-Spot Tuning

Maps every camera's coverage geometry; signals from low-coverage angles are down-weighted so no dead zone inflates your numbers.

LOADEDPR

Problem-Region Detection

Identifies zones that repeatedly generate queue buildup, low dwell, or staffing gaps — before it becomes a complaint or a lost sale.

LOADEDTR

Track Recovery

Re-identifies individuals after occlusion or camera handoff, maintaining continuous track IDs across your full floor.

LOADEDCS

Confidence & Coverage Scoring

Every metric ships with a coverage score and confidence band — you know exactly how reliable each number is.

● 6/6 modules loadedmem 412MB · gpu 31%
proof · field resultsPNL-07
By the numbers

The numbers Janus actually produced.

Promo dwell lift+53%Convenience store promo aisle. Shoppers (staff-excluded) dwelled 53% longer at instrumented displays.
Table dwell+22%Restaurant deployment. Guest-only table-dwell increase — staff movement excluded from the signal.
Congestion reduction+19%Transit deployment. Crowd-flow improvement after Janus flagged bottleneck zones.
Tracked at once17People tracked in the convenience deployment. Staff (4) automatically separated from customers.
● measured in productionnot projections
deploy · janus on your floorPNL-08
Get started

See your floor become data.

We instrument your existing cameras and deliver a live Janus intelligence layer — no new hardware, no long procurement. Most deployments go live in under two weeks. Send your email and we'll set up a walkthrough on your own space.

No surveillance data leaves your premises · 94% model confidence · 42ms inference

● provisioning readygo-live < 2 weeks