Plates: Vehicle intelligence
at the speed of operations.

Multi-stage automatic license plate recognition with vehicle classification, presence tracking, and fleet visibility. From logistics gates to premise security: structured vehicle data feeding the Canvas intelligence layer.

A plate read is the start of an operational record.

A traditional ALPR system gives you a log: timestamp, plate, lane. Plates does more, because the vehicle is rarely the question. The shift is. The route is. The gate that should have been closed is. Plates classifies the vehicle, tracks its time on premise, and pushes the structured event into Canvas, where it correlates with the systems that actually run your operation.

A logistics gate becomes a yard report. A municipal road camera becomes a dwell-time analysis. A residential entry becomes a tenant access record. The plate is the trigger; the operational picture is the product.

Two pipelines, one stream: parking-audit telemetry for occupancy, premise-tracking events for entry and exit. Both flow into the same Canvas vehicle ontology.

Vehicles passing through a busy urban intersection at night, Plates treats every read as the start of an operational record
What ships in every deployment

Five capabilities. One vehicle stream.

Each capability builds on the previous. By the fifth, the plate is no longer a number in a log: it is a vehicle, in context, with evidence attached, on the right side of an operational rule.

  1. Multi-stage recognition.

    YOLO vehicle detection, RetinaPlate localisation, LPRNet character recognition. Three models in series read the plates that single-stage systems miss: angled approaches, partial occlusion, sun glare, mixed lighting, and traffic moving fast enough that one frame is all you get.

  2. Vehicle classification.

    Type, colour, and make recorded alongside every read. A whitelisted plate on the wrong vehicle is no longer a clean entry; a fleet truck that arrived as a sedan triggers the right exception. Classification turns the plate from an identifier into a verifiable claim.

  3. Presence tracking.

    Per-camera mode selection: zoned occupancy snapshots for parking and yard audits, or directional entry-exit pairs for premise tracking with auto-exit for vehicles that arrive but never leave. Both modes write into the same Canvas table, with the lifecycle resolved at the source.

  4. Fleet and logistics visibility.

    Once vehicle events land in Canvas, they correlate with delivery schedules in your ERP, shift patterns in your workforce system, dock assignments in your WMS, and access policies in your security model. The vehicle stops being a data island and starts being part of the operational picture.

  5. Video-linked evidence.

    Every read carries the frame it came from and a pointer back to the source feed (Luxriot EVO and other VMS supported). When a dispute reaches the operations desk, the evidence is already attached to the record. No second tool, no manual lookup, no missing context.

Where Plates is deployed today

Coverage and integration surface.

  • Plate formatsThailand and Malaysia in production. New formats trained from operational footage.
  • Recognition stackYOLO + RetinaPlate + LPRNet, FastAPI service, Socket.IO event stream.
  • Operating modesPer-camera selection between parking-audit zones and premise entry-exit tracking.
  • Video sourcesRTSP ingestion with auto-reconnect; Luxriot EVO integration for evidence playback.
  • Canvas pipelinesOccupancy snapshots, individual reads, premise lifecycle events.
  • DeploymentSingle-binary install (PyInstaller), models bundled, no Python environment to maintain.

Vehicle data that connects to the bigger picture.

Request Demo See what Plates ships