Ladle hook engagement
Confirms the crane hook is correctly seated before a lift begins. The camera watches the hook, lid and bucket region and the PLC gets a verdict, so no lift proceeds on an unverified engagement.
We build the device, and the intelligence that runs on it.
Purview customises AI cameras and edge devices end to end, from enclosure and board to model and dashboard. The inference happens on site, so a decision does not wait for a network round trip and raw footage never has to leave the building.
One platform. Any technology. Complete control.
Every deployment below sits on the same four layers. Signals come from whatever the site already has, edge hubs collect and normalise them, analytics turn them into prediction and anomaly detection, and a digital twin makes the result something an operator can act on.
04
Data visualisation
Dashboards, digital twin, monitoring, ticketing, reports, alerts and workflows.
twinalertsreports
03
Analytics with intelligence
Validated streams become prediction, optimisation and anomaly insight.
predictdetectoptimise
02
Collection & edge hubs
Gateways, nodes, edge devices and protocols collect inputs across the facility.
gatewayedgeMQTT
01
Data generation sources
RFID, NFC, cameras, sensors, tickets, user feedback, and whatever gets added next year.
RFIDNFCcamerasensor
Technology-agnostic onboarding
Loose integration
AI-ready insight
Digital twin actions
01 Smart Industries
Vision AI that makes the call on the plant floor
Cameras that reach a safety or quality verdict in the moment and hand a clean OK / not-OK signal straight to the control system. The model runs at the edge, next to the machine, because a safety-critical lift cannot wait on a network round trip.
01
Trigger
A sensor detects arrival and wakes the camera at the right moment.
02
Capture
Camera and integrated light capture the region of interest (hook, lid, bucket, belt).
03
Infer
The model runs locally, on-premises, for low and predictable latency.
04
Signal
The PLC or control system receives a simple OK / not-OK result.
05
Evidence
HMI, alarm and dashboard show the event with the frame that caused it.
Why the integrated light matters. A hook area can be backlit, smoky or partly obscured. Controlled illumination is part of the camera design rather than an afterthought, because it keeps the model consistent shift to shift and cuts false rejections, which is what stops operators from switching the system off.
Conveyor belt damage inspection
Continuous inspection of a running belt for tears, edge damage, splice failure and foreign objects, so a belt is pulled on evidence rather than on a schedule or a rupture.
Seafood processing inspection
Grading, foreign-object detection and count or weight verification in a wet, cold, high-hygiene line where the hardware has to survive washdown as well as the model has to be accurate.
Manufacturing quality inspection
Dimension, defect and assembly verification at line rate, with every reject stored alongside the frame that triggered it for later review and model retraining.
Mining haul road & site safety
Vehicle proximity, restricted-zone entry and PPE compliance on haul roads and around plant, running on cameras placed where fixed infrastructure is expensive to bring in.
Logistics yard & gate
Plate reading at the gate, trailer and dock occupancy, and load verification, turning a yard that runs on radio calls into one that runs on a live record.
02 Smart Living
Care that notices, without a video feed leaving the building
Safety and privacy are usually presented as a trade. They are not. The camera understands activity locally and sends alerts, event summaries and analytics, never a raw stream. Families, residents and regulators all get a better answer from that arrangement than from a monitored feed.
- Fall and accident detection. Immediate alerts for falls and high-risk events, in rooms and corridors where a delay is the whole problem.
- Inactivity and distress alerts. Detects abnormal inactivity, distress or unsafe movement, rather than waiting on a call button being pressed.
- Automatic escalation. Critical events route to two or three named contacts, so a night shift is never the only line of defence.
- Accountability insights. Event summaries and routine analytics for clinical or care review, without anyone watching a feed.
- All processing on the device. Video is analysed on the edge box in the building, and only structured events travel.
Care homes & elder care
Healthcare ICU
Assisted living
Corrections
Secure transport
Restricted-zone facilities
03 Smart Enterprises
Retail, logistics and warehousing
Environments where throughput is the metric that matters. Queues managed before they form, stock that stays counted between audits, and yards that know what came through the gate, all decided on the device so the answer arrives while it is still useful.
Queue management
Counts people in line, predicts the wait and signals for a till to open before the queue builds rather than after a customer has already walked out.
Shelf & stock intelligence
On-shelf availability, planogram compliance and gap detection, so a replenishment task is raised from what the shelf actually looks like.
Warehouse asset tracking
RFID for bulk scanning, BLE for real-time location and NFC for a smartphone read, chosen per asset class rather than standardised into the wrong technology.
Yard & dock management
Plate recognition at the gate, trailer and dock occupancy, and dwell time per carrier, which is what a yard argument usually turns on.
Cold chain & condition
Temperature, humidity and shock monitored across storage and transit, with the excursion recorded against the consignment rather than the sensor.
Loss prevention
Restricted-zone movement, unusual-movement patterns and after-hours activity flagged as events, with the clip attached for the investigation.
04 Modern Workplace Solutions
The building as one operating system
Facility management usually means six systems that do not speak to each other and a spreadsheet holding them together. We put lighting, HVAC, security, assets, access and parking behind one control layer and one digital twin, so a facilities team runs the building from a single view.
Building IoT & automation
Lighting, HVAC and security under one control layer. Occupancy and temperature sensors trim energy automatically instead of on a timer.
AI inventory kiosk
Vision identifies the person, RFID and barcode read the asset, and a spoken line covers anything untagged, leaving one accountable record of who moved what.
AI reception kiosk
On-device voice interaction through a human or holographic avatar. Multilingual, context-aware, and it keeps working when the line goes down.
Facility traffic management
Vehicle flow across gates, internal roads and loading areas, with signal and barrier logic decided on site rather than in a distant server.
Parking management
Bay-level occupancy, plate recognition on entry and exit, live guidance to a free space and enforcement that runs off the same record.
Asset management & monitoring
RFID, NFC and BLE tags read by pocket, handheld and wall-mounted readers, online or offline, feeding one auditable asset history.
Access control
Camera-verified identity at the door, gate logic by role and time, and visitor or contractor flows that reconcile against the same log.
AV rooms for enterprises
Meeting-room AV, room occupancy, booking integration and device health, so a room that is not working raises a ticket before the meeting starts.
Digital twin & ticketing
Dashboards, reports, alerts and maintenance workflows running on a live model of the building rather than a floor plan from handover.
21,646
RFID-tagged assets under live tracking
20
Levels: two basements, ground and 17 floors
A single tower running the full loop: connect the tag, scan with the right reader for the asset class, update status in real time, and keep a paperless verification history. Online and offline modes, with web dashboard and spreadsheet reporting for the audit.
05 Modern Workforce Solutions
Hands-free work, guided and evidenced
Field technicians cannot hold a tablet and a torque wrench at the same time. AR glasses put the procedure in their line of sight, bring a remote expert into the same view, and capture proof of every step as the work happens rather than from memory afterwards.
SmartInspect, remote collaboration
A collaboration platform built for wearables rather than adapted to them. Field technicians run inspections, maintenance, troubleshooting and audits with an expert seeing exactly what they see: live AR annotation on the real object, 4K streaming, multi-user sessions and a report generated from the session.
SmartFlo, guided workflows
Turns written SOPs into step-by-step flows with checklists, validations and approvals. The technician follows the steps on glasses or a phone; photos, video, voice notes, scans and location are captured at each step and sync to the dashboard in real time for audit and compliance.
The glasses we deploy on
Ruggedised, enterprise-managed and rated for the environments our clients actually work in, including cold stores, plant floors and outdoor sites, not a demo room.
- Platform
- Android 11, OTA upgradeable, MDM support from multiple partners
- Memory
- 6 GB LPDDR4 RAM, 64 GB internal flash
- Durability
- IP67, −22 °C to 50 °C operating, drop tested to 2 m
- Display
- OLED HD colour, 16:9, 16.8° diagonal field of view, above 2,000 nits, contrast above 10,000:1
- Camera
- Up to 13 MP stills, up to 4K 60 fps video
- Controls
- Three buttons, two-axis multi-finger touchpad, customisable multilingual voice control
- Connectivity
- Wi-Fi 2.4 / 5 GHz, Bluetooth 5.0 BR/EDR/LE, USB 3.1 Gen 2 Type-C
We customise the device, not just the software on it
Every solution above runs on hardware we specify and build. That is the difference between an integrator fitting a use case to an off-the-shelf camera, and a team choosing the sensor, the illumination, the compute and the enclosure for the job in front of it.
CAMERAS
Purpose-built AI cameras
Sensor, optics and integrated illumination chosen for the scene (backlit hook areas, wet processing lines, dark corridors), with inference running inside the housing.
COMPUTE
Configurable AI compute boxes
Modular units scaled to the workload, from entry-level inference to high-performance edge, supporting multi-model deployment across vision, speech and language on one device.
ACCELERATION
NPU-class acceleration
Built on efficient neural processing platforms including DeepX accelerators and AMD AI Halo, matched to throughput and power budget rather than to whatever was in the reference design.
TAGS & READERS
RFID, NFC and BLE
Tag and reader selection per asset class: RFID for bulk scanning, BLE for real-time location, NFC for a smartphone read, with pocket, handheld and wall-mounted readers.
CONNECTIVITY
Resilient links
5G, LTE, Wi-Fi, LoRa and wired options, with event traffic and evidence traffic separated so alerts keep moving when the main link is stressed.
DEPLOYMENT
On-premises by default
Local databases, offline operation and signed OTA updates. Data residency is a design decision made at the start, not a policy applied afterwards.
Start with one site.
Tell us the environment, the decision you need made and how fast it has to be made. We will scope a pilot on a single floor, line or gate, and show you the numbers before you commit to a rollout.