AI & IntelligenceNow available: AI cadastre reconciliation

Spatial intelligence.
Trained on the real world.

ML-powered feature extraction, predictive land valuation, and automated encroachment detection — trained on curated global spatial datasets and built to scale to national parcel sets.

Request demo Read research →

Production AI models: 3
Explainable predictions: SHAP
Vegetation classes: 8
Inference per parcel: < 2s

Core AI capabilities

Three production-grade models, each purpose-built for land administration workflows.

Automated Feature Extraction

Extract building footprints, roads, vegetation boundaries, and infrastructure from any raster imagery. Works on aerial, satellite, and drone sources at any resolution above 30 cm/px.

InputAny raster (GeoTIFF, ECW, JP2, COG)

OutputGeoJSON or GML with confidence scores

AccuracySurvey-grade, confidence-scored output

ScaleBatch processing on cloud GPU workers

Building footprints with height estimation
Road centreline and width extraction
Vegetation canopy delineation (8 classes)
Impervious surface mapping
Confidence score per extracted feature
Change delta against prior epoch

Predictive Land Valuation

Spatial ML model combining comparable sales, zoning designation, proximity to services, environmental overlays, and market trend signals to predict land and improvement values at parcel level.

AccuracyConfidence interval on every estimate

ModelXGBoost ensemble + spatial lag features

ExplainabilitySHAP value breakdown per prediction

RefreshWeekly re-train on new sales evidence

Comparable sales matching within spatial radius
Zoning, flood, and bushfire overlay weighting
Proximity scoring (schools, transit, CBD)
Market trend adjustment (rolling 12-month)
Confidence interval output per parcel
Bulk portfolio valuation via API

Encroachment Detection

Automated comparison of survey-registered boundaries against extracted building footprints and fences. Flags probable encroachments and ranks them by severity for surveyor triage.

MethodBoundary overlay + spatial difference

ThresholdConfigurable (default: 0.2 m tolerance)

OutputRanked alert queue with geometry diff

WorkflowHuman-in-the-loop surveyor review UI

Compares registered survey bounds to imagery
Detects fence, building, and vegetation encroachments
Severity scoring (minor / moderate / significant)
Exports PDF report per parcel for legal teams
Integrates with cadastre workflow queue
Continuous monitoring mode for large portfolios

How we train

Rigorous data curation, domain-native architecture, and continuous expert feedback — applied across every model we ship.

Curated spatial datasets
We train on curated global cadastre records, proprietary survey control networks, government land-use datasets, and multi-epoch satellite imagery — architected to scale to national and continental parcel datasets.
Domain-specific architecture
Our models use spatial transformers that encode geographic context, graph neural networks for parcel topology, and XGBoost ensembles for tabular valuation signals — purpose-built for land administration.
Human-in-the-loop validation
Licensed surveyors review edge cases and model disagreements on a continuous basis. Their corrections feed back into the training loop every sprint, progressively sharpening boundary accuracy and valuation precision.

Technical specifications

Everything you need to evaluate model fit for your use case.

Attribute	Detail
Model types	Spatial transformer, GNN, XGBoost ensemble
Training data	Global cadastre + proprietary survey corpus
Inference latency	< 2 s per parcel (p95, cloud API)
Explainability	SHAP values per prediction, feature importance report
Deployment	Cloud API (REST + streaming) · On-premise (Enterprise)
Custom fine-tuning	Enterprise plan — bring your own corpus or labels
Data residency	Region-locked inference (AU, EU, US, SG)
Model versioning	Immutable model IDs; pin to a specific version via API

Ready to run AI on your parcels?

Book a demo and our spatial ML team will walk through your dataset, discuss accuracy requirements, and show you live inference on your region.

Request a personalised demo Read the API docs →

Attribute

Detail

Model types

Spatial transformer, GNN, XGBoost ensemble

Training data

Global cadastre + proprietary survey corpus

Inference latency

< 2 s per parcel (p95, cloud API)

Explainability

SHAP values per prediction, feature importance report

Deployment

Cloud API (REST + streaming) · On-premise (Enterprise)

Custom fine-tuning

Enterprise plan — bring your own corpus or labels

Data residency

Region-locked inference (AU, EU, US, SG)

Model versioning

Immutable model IDs; pin to a specific version via API

Spatial intelligence.Trained on the real world.

Core AI capabilities

Automated Feature Extraction

Predictive Land Valuation

Encroachment Detection

How we train

Curated spatial datasets

Domain-specific architecture

Human-in-the-loop validation

Technical specifications

Ready to run AI on your parcels?

Spatial intelligence.Trained on the real world.

Core AI capabilities

Automated Feature Extraction

Predictive Land Valuation

Encroachment Detection

How we train

Curated spatial datasets

Domain-specific architecture

Human-in-the-loop validation

Technical specifications

Ready to run AI on your parcels?

Spatial intelligence.
Trained on the real world.

Spatial intelligence.
Trained on the real world.