On your devices
Edge AI Companion
Software running on your device’s onboard computer. Run detection models, retain observations and telemetry while disconnected, and synchronise with the ground node when links return.
APPLICATION
Tactical computer vision on Scaleout Edge.
Keep detection models relevant as field conditions change.
Run detection on live video for counter-UAS, ISR and range surveillance. Improve models locally, evaluate candidate updates and share approved improvements when connectivity allows, keeping sensitive footage within each site.
The field workflow
Turn field observations into model improvements.
New terrain, sensor conditions and objects can expose gaps in detection performance. Monitor model health to identify changes that warrant review, and use field observations to improve models within your site.
Evaluate before deployment.
Your team controls which updates are approved. Raw footage stays within the site.
Unlabelled imagery can also support model pre-training across participating sites through federated learning. This provides a starting point for task-specific fine-tuning, with labelled data and evaluation still part of the workflow.
-
Detect
Run models on video from cameras, vehicles or drones. Onboard detection continues when connections drop.
-
Select
Automatically select frames where models are uncertain or encounter unfamiliar conditions, focusing human review on useful training material.
-
Annotate
Review a prioritised queue of selected frames, focusing annotation on observations that are useful for model improvement. Add or correct labels using agreed labelling guidelines.
Train and evaluate
Train candidate models on local data. Run model versions side by side on the same sensor feeds to compare predictions, and assess performance on labelled evaluation data before approving an update.
-
Distribute
Deploy approved models to connected devices. Coordinate shared improvements across participating sites through Scaleout Edge when connectivity allows.
Where the software runs
On your devices.
At your site.
Across your network.
Three connected software roles support the workflow, from detection on your devices to local training and coordination across sites.
At your site
Vision Ground Node
A software workbench on local GPU-equipped infrastructure. Run detection, annotate observations, and train and evaluate models even while disconnected. Synchronise updates when connectivity returns.
Across your sites
Scaleout Edge
Coordinate federated learning and distribute approved updates. Monitor reported model health and maintain a record of which model ran where, with what results.
Devices ↔ Local siteSelected observations move to the customer-controlled site when links are available. Approved models return to devices for local detection.
Local site ↔ Scaleout EdgeEach site trains on local footage. Federated learning combines model updates to improve a shared model without transferring raw footage between sites.
The software runs on hardware supplied by you or your chosen partners. Scaleout Edge is deployed within your chosen infrastructure. Scaleout provides the software infrastructure and engineering support for this workflow, integrated with your sensor feeds, models, computing infrastructure and operational systems.
How it fits your operation
Connect with the systems you already use.
Integrate tactical computer vision with your sensor feeds, models and operational applications. Scaleout Edge provides the model deployment, learning and coordination layer alongside your existing systems, including central model-development tools. Compatibility and compute requirements are assessed during evaluation.
Your sensor feeds
Connect video from cameras, vehicles or drones through supported interfaces. Run detection on onboard compute or local infrastructure, depending on your setup.
Your models
Bring detection models from your team or chosen suppliers, or start with reference models such as YOLO. Adapt them using local observations and evaluate their suitability for your task and target hardware.
Your operational systems
Make detections available to operator tools, including TAK through OpenTAKServer. Additional C2 connections are scoped around the interfaces and information your team needs.
Combine this workflow with onboard drone AI, integration and engineering support. Explore the Tactical AI Network solution.
For onboard inference, retained observations and model deployment to drones, explore AI for drones.
Your decisions.
Your models. Your data.
Your models. Your data.
Your team retains control of operational decisions and model approvals. Your operational data and trained models remain yours.
Get started
Evaluate it in your environment.
Work with Scaleout engineers to assess tactical computer vision using your sensor feeds, models and computing infrastructure against agreed performance and operational requirements.
Define the scope
Choose a detection task, identify data and equipment, and agree on success criteria: detection performance, the update workflow and operation with limited connectivity.
Test the workflow
Connect selected feeds, run detection and evaluate a local model update. Train your team or chosen partner in data review, annotation and model improvement.
Plan the next step
Review results and plan the move from an initial workflow to field validation or broader deployment, including integration, hardware and ongoing support.
Scaleout provides software, engineering support and knowledge transfer within the agreed scope. Security requirements and accreditation support are scoped around your deployment environment. Hardware is supplied by you or your chosen partners.
Discuss an evaluation