GT-AiVision Live Monitoring

GT-AiVision Live Monitoring

Add Real-Time Edge AI to Critical Cameras and Zones

Help customers add configurable event detection, visual evidence, alerts, and analytics to compatible CCTV systems. Real-time inference and event-rule evaluation run on NVIDIA Jetson hardware at the customer site.

The partner opportunity

Add intelligence without replacing the complete camera estate

Live Monitoring can be applied to selected cameras where prompt event awareness matters. Partners can combine the GT-AiVision appliance, platform configuration, annual real-time channel subscription, installation, and first-line service within a broader CCTV or security solution.

Discuss the Partner Program

Services partners can package

  • Configured AI event rules for selected cameras.
  • Edge-based object detection and tracking.
  • Visual snapshots and searchable event history.
  • Telegram alerts for configured events.
  • Dashboard statistics and scheduled reports.
  • Integration and customisation services where agreed.

How it works

From camera stream to reviewed event

1

Connect

Configure a compatible RTSP stream, inference resolution, and processing settings.

2

Define Area

Draw ROI polygons or virtual tripwire lines through the web interface.

3

Configure Rule

Set class, direction, severity, threshold, dwell time, cooldown, and alerts as applicable.

4

Process at Edge

Run DeepStream, YOLOv11, TensorRT, hardware decoding, and tracking on the Jetson device.

5

Create Evidence

Record the event and capture a timestamped JPEG snapshot with the configured overlay.

6

Notify and Review

Use Telegram alerts, dashboard filters, snapshots, statistics, and reports.

7

Escalate

Link relevant events to the Incident Management workflow when required.

8

Improve

Review event history and tune scene-specific rules, thresholds, and operating procedures.

Standard real-time events

Configure supported rules around the customer’s operational risks

Events are enabled and configured by camera according to the supported options and qualified scene.

Tripwire

Virtual line crossing

  • Person crossing a configured line.
  • Vehicle crossing a configured line.
  • Dog or cat crossing a configured line.

Options can include direction, sensitivity, strict direction, object class, severity, cooldown, Telegram notification, and optional VLM analysis.

ROI presence and count

Zone monitoring

  • Person entering a configured polygon.
  • Supported unauthorised-access workflow.
  • Loitering beyond configured dwell time.
  • Overcrowding above a configured threshold.
ROI duration

Objects and vehicles

  • Vehicle remaining in a zone beyond the configured time.
  • Backpack, suitcase, or handbag remaining beyond the configured threshold.
  • Dog or cat activity inside a monitored zone.

The model recognises additional object classes, but they should not be presented as ready-to-use business rules until configured and tested for the required workflow.

Configuration controls

Adapt event rules to each qualified camera and scene

Operators can enable rules by camera and adjust geometry, object class, direction, thresholds, severity, cooldown, notification, VLM description, inference resolution, and frame interval according to the event type.

Key controls

  • ROI polygon or tripwire geometry.
  • Object class and direction rule.
  • Count or dwell-time threshold.
  • Low, medium, or high severity.
  • Cooldown between repeated alerts.
  • Telegram notification.
  • Optional VLM analysis and description.
  • Camera resolution and frame interval.

Operator dashboard

Review cameras, events, evidence, and system health

Live operations

  • Live camera card grid and worker status.
  • Multi-camera snapshot view.
  • Per-camera settings and service controls.

Event review

  • Camera, event-type, and date filters.
  • Snapshot preview and event details.
  • Statistics and trend charts.

Configuration and health

  • ROI and tripwire editors.
  • GPU, memory, and system-health views.
  • Worker and service status.

Alerts and reporting

Provide timestamped evidence and structured review

Configured events can create event records, JPEG snapshots with ROI or tripwire overlays, Telegram alerts, optional VLM descriptions, dashboard history, and scheduled event-summary PDF reports through email.

WhatsApp alerting is not presented as a standard documented GT-AiVision feature.

Typical operator flow

Alert

Receive the configured event and snapshot.

Verify

Review camera, rule, timestamp, and visual evidence.

Respond

Apply the customer’s agreed operating procedure.

Document

Retain event history or escalate into Incident Management.

Real-time channel packages

Match edge hardware to the qualified processing requirement

These are listed package maxima, not unconditional performance guarantees. Actual suitability depends on the complete configuration.

Up to 8 channels

Jetson Orin Nano

Entry real-time tier for selected cameras and qualified event workloads.

Up to 16 channels

Jetson Orin NX

Expanded real-time tier for larger qualified processing requirements.

Up to 24 channels

Jetson AGX Orin

Highest listed real-time tier for demanding qualified deployments.

Actual capacity depends on model, resolution, frame interval, tracking, enabled rules, VLM use, scene complexity, stream stability, and system configuration. Real-time channel packages are separate from Virtual Patrol managed-camera tiers.

Camera and site qualification

Qualify the scene before quotation and deployment

Reliable real-time analytics depends on more than the camera count. Partners should confirm the stream, scene, rule, network, and expected response workflow.

Review Deployment Architecture

Qualification checklist

  • Stable RTSP access, suitable resolution, and codec.
  • Reliable LAN, power, and camera addressing.
  • Camera angle, lighting, object size, and scene visibility.
  • Clear monitored zone and event definition.
  • Expected operator response workflow.
  • Customer authorisation and retention requirements.
  • Required internet-dependent services.

Performance boundaries

Set responsible expectations during the sales process

Detection performance can be affected by camera position, lighting, shadows, reflections, rain, camera shake, occlusion, object size and distance, crowding, stream quality, model configuration, event thresholds, and scene complexity.

GT-AiVision does not claim

  • 100% detection accuracy.
  • Zero false positives.
  • Guaranteed incident prevention.
  • Replacement of human review or operational judgement.
  • Identical performance across unqualified camera scenes.

Partner delivery model

Lead the local solution with Grep Tech technical support

Partners can lead customer discovery, camera and site qualification, installation coordination, first-line support, and service packaging.

Grep Tech provides product training, solution guidance, platform configuration, software updates, and L2/L3 technical escalation.

Combine Live Monitoring with

  • Existing CCTV installation and maintenance services.
  • Virtual Patrol across the wider camera estate.
  • Operator review and managed alert services.
  • Incident Management workflows.
  • Customer-specific integration where agreed.

Partner demonstration

Demonstrate real-time AI on representative camera scenarios

Review event configuration, ROI and tripwire setup, snapshots, alerts, dashboards, and the combined Live Monitoring and Incident Management workflow.