AiGAP Vision — new module

AI vision for what your
cameras already see.

A computer-vision layer that snaps frames from any camera stream, processes them asynchronously, and runs four ready-to-go analysis modules — counting people, recognising staff, tracking inventory on shelves, and reading the room's emotional temperature.

Cam-01 · EntranceCounter
27
People now
+12% vs 1h
Cam-02 · FloorStaff
AY
Olivia Reedentrance
On post
EL
Emma Carterfloor
On post
YT
Yuki Tanakaregister
On post
Cam-03 · AisleStock
B3 low · espresso beans
Cam-04 · LobbyMood
+0.62
Happy index · today
Ai
4 cameras · 4 modules · live | Async pipeline · 180ms avg
01 · Counter

People counting per frame.

Real-time headcounts, dwell time, queue length, and hourly/daily trends. Anonymised — no identities stored.

02 · Staff

Employee recognition.

Face matching against your team roster. Attendance, presence at posts, late arrivals, shift handovers.

03 · Stock

Inventory by sight.

Watches your shelves and bins. Detects low stock, mis-shelving, and replenishment events — by object class, not SKU.

04 · Mood

Emotion analytics.

Aggregated, anonymous facial-expression signal across the visit. Trends by hour, by zone, by shift.

Module 01 · People counter

Know exactly how many people are in every zone, every minute.

A continuous, anonymous headcount per zone. The counter handles overlapping detections, re-entry, and queue dynamics — and feeds charts you can act on without a data team.

Real-time per-zone counts

Define zones in the camera frame (entrance, queue, fitting room). The counter tracks live occupancy and queue length, exposing both as API and live chart.

Dwell time + flow direction

How long people stay, and which way they move. Heat-map ready out of the box.

Anonymous by design

The counter never persists identities. Faces are reduced to detection tokens that expire when the visitor leaves the frame.

📈

Hourly / daily / weekly trends

Generated charts and digests — emailed, posted to Slack, or exposed via the AiGAP analytics package.

Zone 01 · Entrance Live
27
people in zone · now
+12% vs 1h ago
Module 02 · Staff recognition

Your team, recognised on sight — only your team.

Match faces against a roster you maintain. Get attendance, presence at posts, late arrivals, and shift handovers — with the strict guarantee that nobody outside your roster is identified.

Active shift · 09:00 → 17:00 5/6 on post
AY
Olivia Reed08:54 · entrance
On post
EL
Emma Carter09:02 · floor
On post
YT
Yuki Tanaka09:12 · register
On post
MO
Marcus Okonkwo09:24 · stockroom
Late · 24m
KA
Alex Morgannot detected
Absent

Roster-only matching

The model only matches enrolled team members. Anyone outside the roster stays anonymous — the system never builds a face database of strangers.

Attendance + post-presence

First seen / last seen, time at assigned post, lunch break detection, late arrivals. Daily and weekly logs.

Shift handovers

Detects when the outgoing shift hasn't been replaced by the incoming one — alerts before the gap costs you a sale.

🔐

On-device encoding

Face embeddings are computed on-device. The raw images leave the camera as anonymous tokens; only matches against your roster surface to dashboards.

Module 03 · Inventory by sight

Your shelves, read by the camera.

Object-class inventory tracking. The camera watches what's where, how full it is, and what's missing — without barcode scans, weight sensors, or SKU mapping.

Object-class shelf reading

Trained on your product categories (or trained live by you, on-screen, by drawing boxes). The shelf shows up as a tiled grid with per-cell fullness.

Low-stock alerts

Configurable thresholds. When a cell drops below the line, an alert goes to the right phone, the back room screen, or the right Slack channel.

Mis-shelving detection

If a class shows up where it shouldn't, the camera flags it. Cuts the daily walk-the-floor audit to a glance.

📅

Replenishment history

Every restock event is timestamped. See how long shelves stay full, who restocks what, and when patterns drift.

Shelf · Aisle 03 · Coffee Last frame · 4s ago
A1100%
A280%
A360%
A440%
B1100%
B260%
B320%
B40%
B3 low · espresso beans (12 units left)
B4 empty · oat milk · last restock 14h ago
Module 04 · Emotion analytics

The temperature of the room, charted by the hour.

Aggregate, anonymous facial-expression signal — not as a per-person judgement, but as a zone-level mood index over time. Pair with hours and staffing to see what actually moves the dial.

Mood by hour · Cam-01 Today · 09:00 → 17:00
HappyNeutralFrustratedEngaged
+0.62happy
0.24neutral
-0.18frustration
0.71engagement

Aggregate-only, never per-person

The chart is a rolling zone-level mean. The system can't show you "person X looked happy" — that's a deliberate boundary, not a missing feature.

Hour, zone, and shift breakdowns

Slice the mood signal by hour-of-day, by camera zone, or by the shift on duty. Spot where service tilts the room one way or the other.

Pair with revenue + staffing

Cross-reference with your POS, your staffing roster, your queue data. The emotion signal isn't a verdict — it's a clue.

Privacy-first by construction

Expressions are encoded into anonymous mood vectors on-device. Raw frames are discarded after the snapshot pipeline finishes. No face library, no recording.

Under the hood · pipeline

Async camera-to-insight in five stages.

Each camera feeds an asynchronous snapshot processor; the snapshots are dispatched to whichever modules you've enabled. Results stream back to dashboards and the AiGAP core in seconds — without ever blocking the camera.

01 · Camera
Stream in

RTSP, MJPEG, ONVIF, IP cams, USB webcams, or phone cameras. AiGAP adapts to the source.

02 · Snapshot
Async frames

The processor pulls frames at a rate the analysis can keep up with — no dropped frames, no blocked camera.

03 · Dispatch
Module fan-out

The same snapshot is handed to every enabled module in parallel — count, staff, inventory, mood.

04 · Encode
On-device encoding

Faces and objects become anonymous vectors at the edge. Raw frames are discarded after the snapshot expires.

05 · Surface
Charts + alerts

Live dashboards, hourly digests, low-stock alerts, late-arrival pings — wherever your team works.

Where AiGAP Vision is in use

From retail floors to hospital wings.

A few sectors where camera-driven AiGAP is already paying its way.

Retail

Specialty café — Lisbon

Live queue length on the till tablet. Coffee bean shelf alerts. Mood by hour cross-referenced with the espresso machine's repair log.

Hospitality

Boutique hotel — Istanbul

Lobby presence tracking for staffing. Restaurant mood as a service-quality signal. Late-arrival nudges for night-shift.

Healthcare

Outpatient clinic — Berlin

Waiting-room queue analytics, no-show detection. Anonymised mood signal for waiting experience — never per patient.

Manufacturing

Workshop — Izmir

Workbench inventory tracking. Staff presence at safety-critical posts. Mood as a fatigue signal at end-of-shift.

Want AiGAP Vision running on your cameras next week?

We onboard pilot customers in a single day — bring your camera stream and your roster, we'll ship the dashboard and the alerts. Pricing is per-camera, paid monthly.