Import airborne LiDAR
LAS, LAZ, COPC, E57, PLY from any airborne or UAV sensor. No per-project setup.
Topographic mapping · Airborne LiDAR
From raw airborne LiDAR to a finished classified deliverable, in one pass.
The Aerial Mapping AI model classifies a full airborne-LiDAR capture automatically: noise, ground, above-ground features and the power network, then out to DTM / DSM, hillshades and contours. Less manual labelling, faster turnaround, proven at national scale.
50-99%
processing time saved versus manual point-cloud editing
27
classes classified in a single automated pass
1000s
of km² processed per national mapping campaign
200+
customers across 50+ countries, from mapping firms to national agencies
The workflow
The aerial mapping workflow, end to end.
One automated run takes a raw capture to a finished deliverable. You step in only for the final review, not to classify every tile by hand.
Noise filtering
Atmospheric returns, birds, multipath and scan-edge artefacts removed before anything else runs.
Flight-line matching
Align overlapping strips and clean flight-line seams for a consistent, gap-free surface.
Ground & above-ground
Ground and water through buildings, vegetation, infrastructure and the power network, in one run.
Manual QC
Review and correct with annotation tools and keyboard shortcuts. Touch-ups, not labelling from scratch.
Finished outputs
Classified point cloud plus DTM / DSM, hillshades, contours and rasters, ready to ship.
Capabilities
What Flai does for aerial mapping teams.
Automated full-scene classification
Pre-trained AI classifies ground, water, buildings, vegetation, infrastructure and the power-line network straight from raw LAS / LAZ.
Less manual labelling
The model does the heavy lifting. Your team reviews and corrects instead of classifying every tile from scratch.
Faster, simpler processing
50-99% less editing time. Batch hundreds of tiles unattended, with no bespoke pipeline to build or maintain.
Works with your sensors
Sensor- and platform-agnostic: airborne and UAV LiDAR from all major sensors, ingested in the formats you already fly.
Custom retraining for new classes
Extend the model with your own classes and terrain types (AI Learning Point), on Flai cloud or fully on-prem.
Lower project risk
Consistent, repeatable, auditable results across operators, regions and epochs, with per-class QC metrics and reporting.
Manual QC & cleanup tools
Annotation tools, measurements and keyboard shortcuts to speed up the final review, all in one workspace.
Open formats & API
LAS / LAZ / COPC out, plus DTM / DSM, hillshades, contours and rasters, a REST API and a QGIS plugin for direct GIS integration.
Deliverables
Not just a classified cloud.
The classified point cloud is the starting point. From it, Flai generates the surface models and map products your clients actually ask for.
Additional service
Custom retraining for new classes.
Need a class the pre-trained model does not ship with? Extend it with your own. Provide examples through the AI Learning Point and Flai retrains the model on your data and terrain, on Flai cloud or fully on-prem, so no data leaves your network.
- Add objects you care aboutCustom classes on top of the standard schema, tuned to your specifications.
- Adapt to local terrainFine-tune on your national or regional terrain types for higher accuracy.
- Own the modelRetrain in-house and keep improving it on every new dataset.
National mapping & cadastre agencies
Built for national programmes.
Around 150 national mapping and cadastre agencies run country-wide LiDAR campaigns. The bottleneck is no longer collection, it is classification. Deploy FlaiCortex on your own infrastructure and process petabytes consistently, reducing the risk of an ageing workforce and an ever-growing backlog.
- On-premise, fully air-gapped deployment on your own servers
- Batch processing at national scale, thousands of km² per campaign
- Auditable results with per-class QC metrics and reporting
- Support for country-specific classification schemas
- Custom retraining on your national terrain types
Classes the model resolves
27 classes, out of the box.
Every point is assigned a typed class, then vectorised into the objects your workflow consumes. Highlighted classes are the ones downstream extraction depends on most.
01Other (man-made)02Ground03Vegetation06Buildings (roof)07Low noise09Water17Bridge decks18High noise20Roof objects21Vehicles22Walls / facades23Points below ground24Low-voltage wires25High-voltage wires26Railroad wires27Low-voltage towers28High-voltage towers29Railroad towers30Fences35Impervious surfaces37Large vehicle38Containers39Street lamps66Roof solar panels67Solar panels68Wind turbines70Antennas
Colours match the Flai aerial mapping classifier. Extend the schema with custom classes via retraining.
First pass · noise filtering
Noise patterns, across sensors
Before aerial classification, Flai isolates low and high noise in a single automated pass, consistently across sensors, so ground and surface models fit cleanly at national scale.
Low isolated noisebelow ground
High isolated noisedust, birds, cloud
Unclassifiedvalid returns kept
"With the use of Flai's state-of-the-art AI for automatic classification of LiDAR point clouds we have reduced the cost and complexity of data processing."
Deployment & integration
Run it wherever your data must live.
Flai-managed Cloud
AWS EU-Central (Frankfurt) by default. Production-grade, monitored, onboarding in days.
Customer Cloud / Self-hosted
Dedicated deployment in your own GCP, AWS or Azure. Full data sovereignty.
On-premise / Air-gapped
The Flai CLI runs entirely behind your firewall, with only metadata for licence validation.
Classify aerial LiDAR faster.
From raw capture to finished deliverables, at national scale. Cut manual work and deliver consistent, accurate results.