What We Detect

Common capabilities below — anything not listed can be customised. If you can describe it, we can train it

📦

Bin Empty / Full Detection

Real-time empty/full state of storage bins, ASRS cells and buffer positions, replacing manual rounds and photoelectric sensors. Results posted straight back to WMS.

🚶

Intrusion & Safety Zone

Virtual fences alarm and interlock the moment a person enters a charging bay, robot cell or ASRS aisle, with optional helmet and hi-vis vest compliance checks.

🧯

Early Fire Detection

Visual pre-detection of flame, smoke and abnormal hot spots — redundant with smoke and heat sensors, and earlier than their thresholds.

🎯

AGV/AMR Docking & Pose

Arrival, offset, skew and pallet centring judged in one shot, handed to the fleet manager for correction or re-localisation.

🏷

Pallets, Bins & Barcodes

Pallet type, presence, stack height and bin counts, plus high-speed 1D/2D/DMC code reading tolerant of damaged, distorted and reflective labels.

🚧

Aisle Occupancy & Safety

Blocked aisles, over-crowding, mixed forklift and pedestrian traffic and missing PPE detected and logged, exportable per shift.

⚙️

Equipment & Indicator Reading

Panel lamps, gauge readings, gate and door states read by camera, bringing “dumb” equipment into your information system.

🎥

Video Structuring & Evidence

Automatic clips and stills around every event, searchable by time, camera and workstation for review and accountability.

Model Training Workflow

From on-site data capture to a live model — five steps in a closed loop, with accuracy improving the longer it runs

1

Data Collection

Streams from site cameras, phones or archive footage, covering varied light, angle, glare and occlusion.

2

Annotation

Scene-specific label schema covering bounding boxes, instance segmentation, keypoints and classification.

3

Model Training

Fine-tuning of pretrained models plus hard-example mining, yielding lightweight models that run on edge devices.

4

On-site Deployment

Edge box, industrial PC or server, with incremental model updates and remote push.

5

Feedback Loop

Hard examples are collected automatically on site; models iterate weekly and accuracy keeps climbing.

Platform Integration

Recognition output is delivered on two channels — structured data and events — built to your interface specification, not shipped half-finished

Camera & Algorithm Side

Structured data (class, count, coordinates, confidence, timestamp) and events (enter/leave, empty→full, alarm) over REST/HTTPS callback, WebSocket push, MQTT subscribe, TCP Socket, Modbus TCP, OPC UA, direct database access and vendor SDKs.

→

Upstream System Side

WMS inventory and bin status, WCS/WES task verification, MES process poka-yoke, RCS/AGV dispatch commands, ASRS control, PLC and host-computer interlock, SCADA dashboards and data feeds into MES/ERP reporting.

WMS WCS / WES MES ERP RCS Dispatch ASRS PLC SCADA OPC UA Modbus TCP MQTT WebSocket REST API Direct DB

Delivery Options

Choose by site conditions and budget — algorithm only, or a complete turnkey package

🖥️

Edge Appliance

Algorithms pre-installed and ready at power-on; one unit covers one to many cameras — fastest route for a single workstation.

🧠

Algorithm SDK Licence

Keep your existing cameras and servers and license the algorithm per camera point; shortest lead time.

🔌

Integration Development

Built to your interface document and commissioned until live WMS/WCS data flows end to end.

🛠️

Turnkey Project

Camera, lens and lighting selection, mounting, algorithm, platform integration and operator training delivered as one package.

Something you want to detect?

Send us site photos or video and we will come back with a feasibility judgement and a training plan; the first demo model usually lands within two weeks.

Contact Us