VisionAI · Features
Eight ways VisionAI keeps your operations covered.
Every category runs on a shared set of computer-vision models — built once, applied across compliance, safety, loss prevention, and customer experience.
Each camera runs the detections chosen for it at setup. Items marked On request sit outside the standard F&B setup.
- AI detection models
- 117
- specialised model families
- 10
- object classes
- 87
F.01
SOP & PPE compliance monitoring
Flags missing PPE — masks, gloves, hairnets, safety gear — and SOP breaks such as phone use on shift, smoking in restricted zones, or staff asleep on duty. Reported by shift and by station.
Used in: kitchens, central kitchens, factories
What it detects
- PPE detection
- Phone usage
- Smoking detection On request
- Sleeping on duty
F.02
Queue & customer experience analytics
Counts people, tracks dwell time, and reads queue length and wait time in real time, so you can see where service is slowing down before customers walk out. Anonymous age-band and gender breakdowns show who's coming in — no personal data captured. Signals feed straight into operations dashboards.
Used in: retail, restaurants, venues
What it detects
- People counting
- Footfall
- Dwell time
- Queue length
- Demographics
F.03
Staff behaviour & productivity monitoring
Tracks movement and task time across cameras using skeleton-based action recognition, and flags unsafe behaviour, idle time, or off-task phone use. Surfaced as coaching data for supervisors, not surveillance.
Used in: production areas, service counters
What it detects
- Action recognition
- Theft monitoring
- Idle-time tracking
- Station absence
F.04
Theft & loss prevention (POS / self-checkout)
Watches self-checkout and register activity for suspicious scan behaviour, unpaid items, and mismatches between what's scanned and what's in hand.
Used in: retail chains, supermarkets, QSR counters
What it detects
- Theft detection
- Payment behaviour
- Register status
F.05
Fire, smoke & safety hazard detection
Watches for fire and smoke around the clock, and extends to water safety — spotting a swimmer in distress before it becomes an emergency.
Used in: kitchens, fuel stations, warehouses, pools and leisure sites
What it detects
- Fire detection
- Smoke detection
- Drowning detection
F.06
Hygiene & food safety monitoring
Checks hand-washing and station hygiene at the prep line, spots pests in food areas, and checks freshness, packaging, labelling, and portion consistency.
Used in: restaurant kitchens, central kitchens, hospital kitchens, food manufacturers
What it detects
- Hand-wash checks
- Pest detection
- Freshness
- Packaging & labelling
- Portion consistency
F.07
Violent behaviour & security detection
Recognises assault, weapon presentation, vandalism, and crowd unrest as they happen, so security responds in real time instead of reviewing footage afterwards.
Used in: malls, venues, convention centres, transport hubs
What it detects
- Assault detection On request
- Weapon presentation On request
- Vandalism On request
- Crowd unrest On request
F.08
People, vehicle & object recognition
The detection layer behind every other category — people, vehicles, and dozens of everyday object classes, recognised and tracked across cameras without identifying anyone.
Used in: car parks, forecourts, warehouses, secured areas
What it detects
- People detection
- Vehicle detection
- Object detection
Every plan
Built into every deployment.
-
Works with your existing CCTV
IP cameras of 2MP or more streaming RTSP or ONVIF, up to 8 per AI Box. No vendor lock-in, no rip-and-replace.
-
Runs on site, even offline
The AI Box analyses video on your network. If the internet drops, detection carries on and events sync when it reconnects.
-
Mobile push alerts and task generation
Every detection can become a task for the right person.
-
Compliance dashboard
One live view per camera, per outlet.
-
Calibrated to your site
Models are tuned to your lighting, layout, and policies before going live.
-
Connected to the Codemax stack
Events stream into the AI-Kitchen Command Center and can open incidents in RMS.
Anonymous by default.
VisionAI works on anonymised signals — no face recognition, no stored biometrics, no continuous recording. The AI Box analyses video on site, and only alerts and statistics leave it.