Use cases
47 detectors, by model and industry.
41 detector templates and 6 engines from the Atlas repository. Filter by industry, model family or sensor, and each card shows what goes in, what comes out, and the classes it names.
A template is a configuration and its class prompts. Trained weights are not claimed for every one, and no accuracy figures are shown here. The engines are described on Platforms and Research.
Model family by industry.
Each number is a count of detectors. Pick a cell to filter the cards below.
| Model family | Mining | Agriculture | Forestry | Wildfire | Oil and gas | Infrastructure | Maritime | Municipal | Defence |
|---|---|---|---|---|---|---|---|---|---|
| Change detection | |||||||||
| Segmentation | |||||||||
| Object detection | |||||||||
| Classification | |||||||||
| Counting | |||||||||
| Anomaly detection | |||||||||
| Forecasting | |||||||||
| Multi-sensor fusion | |||||||||
| Georeferencing | |||||||||
| Embeddings | |||||||||
| Zero-shot features (DINOv2) | |||||||||
| Zero-shot segmentation (SAM 2) |
Detectors
Showing 47 of 47
Engine · Georeferencing
Automatic georeferencing
Places a scanned map on the satellite frame by matching a parcel graph to a road graph, scores the fit by leave-one-out error, and refuses a poor fit.
- Inputs
- scanned map raster, optical basemap
- Output
- GeoTIFF + GCPs + RMSE report
- Sensors
Georeferencing
Engine · Forecasting
Compliance forecaster
Seasonal forecasts on sensor and index series, with the date each series is expected to cross its limit.
- Inputs
- time series of indices or sensor readings
- Output
- forecast series + threshold-breach dates
- Sensors
Forecasting
Engine · Embeddings
Embedding layers
Per-pixel embeddings from radar and optical time series, for similarity search and clustering without labels.
- Inputs
- S1+S2 time series (TESSERA 128-d), AlphaEarth 64-band
- Output
- embedding rasters for similarity / clustering
- Sensors
Embeddings
Engine · Forecasting
Harvest window forecast
Season curves, growing degree days and the El Nino index shift the expected harvest window and flag the fields at risk.
- Inputs
- Sentinel-2 NDVI/NDRE time series, ERA5/temperature GDD, ENSO index, Elevation
- Output
- harvest window shift (days) + at-risk fields
- Sensors
ForecastingEmbeddings
Engine · Multi-sensor fusion
Hotspot to incident, with spread forecast
Hotspots, cameras, ground nodes and weather open an incident only when two signals agree, with a spread forecast attached.
- Inputs
- satellite thermal hotspots, camera, ground node, weather
- Output
- incident alert + spread forecast
- Sensors
Multi-sensor fusionForecasting
Engine · Multi-sensor fusion
Multi-sensor attention fusion
Per-sensor encoders feed one masked attention model. Any sensor can drop out, and every decision comes with the weight each sensor carried.
- Inputs
- optical (DINOv2), SAR ViT, thermal CNN, RF spectrogram ViT, AIS/sequence, audio/acoustic
- Output
- detect score + anomaly score + per-modality attention explanation
- Sensors
Multi-sensor fusionAnomaly detection
Template · Classification
Activity & OOB Monitor
Temporal activity index and order-of-battle proxy clustering over airfield aprons: class-level counts only (fighter, transport, support), not tail-number ID
- Inputs
- Optical
- Output
- Heatmap
- Sensors
- Resolution
- 10m Sentinel stacks
- Classes
- fighter, transport, support, helo, uav
ClassificationZero-shot features (DINOv2)Anomaly detection
Template · Segmentation
Agricultural Bund Digitization
Automated delineation of agricultural bunds and field embankments for irrigation management and plot mapping
- Inputs
- Optical, Multispectral
- Output
- Polygons
- Sensors
- Resolution
- 0.1-0.5m per pixel
- Classes
- bund, field interior, water channel
SegmentationZero-shot segmentation (SAM 2)Multi-sensor fusion
Template · Segmentation
Agricultural Plot Subdivision
Automated delineation of agricultural plot boundaries for cadastral mapping and land administration
- Inputs
- Optical, Multispectral
- Output
- Polygons
- Sensors
- Resolution
- 0.3-1m per pixel
- Classes
- plot boundary, plot interior, access path
SegmentationZero-shot segmentation (SAM 2)Multi-sensor fusion
Template · Counting
Agricultural Tree Counter
Automated individual tree detection and counting for orchard management, agroforestry, and plantation inventory
- Inputs
- Optical, Multispectral
- Output
- Points
- Sensors
- Resolution
- 0.1-0.5m per pixel
- Classes
- mature tree, young tree, dead tree
CountingZero-shot segmentation (SAM 2)Multi-sensor fusion
Template · Change detection
Airfield Activity Monitor
Track apron occupancy changes and 14-day activity index over revetments and parking bays: infrastructure-level, not tail-number identification
- Inputs
- Optical
- Output
- Polygons
- Sensors
- Resolution
- 10m Sentinel stacks
- Classes
- occupied bay, new occupancy, empty bay, activity spike
Change detectionZero-shot features (DINOv2)
Template · Object detection
AIS Spoof Anomaly Monitor
Flag AIS tracks with geometry contradictions vs SAR-observed positions: duplicate MMSI, implausible jumps, and claimed-vs-observed discrepancies
- Inputs
- Ship reports, Radar
- Output
- Points
- Sensors
- Resolution
- AIS kinematics + S1 SAR
- Classes
- geometry contradiction, duplicate mmsi, implausible jump, cooperative
Object detectionMulti-sensor fusionAnomaly detection
Template · Change detection
Border Intrusion Monitor
Detect fence-line disturbances and elevated boundary activity along international borders using change detection and temporal anomaly indexing
- Inputs
- Optical, Radar
- Output
- Polygons
- Sensors
- Resolution
- 10m optical/SAR
- Classes
- fence disturbance, activity anomaly, stable boundary
Change detectionZero-shot features (DINOv2)Zero-shot segmentation (SAM 2)Multi-sensor fusion
Template · Segmentation
Carbon Sequestration Monitor
Assess forest carbon storage capacity and sequestration rates
- Inputs
- Multispectral, LiDAR, Hyperspectral
- Output
- Polygons
- Sensors
- Resolution
- 1-10m per pixel
- Classes
- old growth, young forest, degraded forest
SegmentationZero-shot segmentation (SAM 2)Multi-sensor fusion
Template · Segmentation
Crop Damage Assessor
Rapid quantification of crop losses from floods, hailstorms, pest outbreaks, and drought for insurance and government relief programs
- Inputs
- Optical, Multispectral, Radar
- Output
- Polygons
- Sensors
- Resolution
- 1-10m per pixel
- Classes
- no damage, partial damage, total damage, flood waterlogged, lodging
SegmentationZero-shot segmentation (SAM 2)Multi-sensor fusion
Template · Segmentation
Crop Field Boundary Detector
Precisely delineate crop field boundaries and classify crop types for precision agriculture
- Inputs
- Multispectral, Radar
- Output
- Polygons
- Sensors
- Resolution
- 1-3m per pixel
- Classes
- corn, wheat, soybean, rice, fallow, pasture
SegmentationZero-shot segmentation (SAM 2)Multi-sensor fusion
Template · Classification
Crop Health Assessment
Monitor crop health using NDVI and multispectral analysis to detect stress, disease, and nutrient deficiency
- Inputs
- Multispectral, NDVI
- Output
- Heatmap
- Sensors
- Resolution
- 1-5m per pixel
- Classes
- healthy, stressed, diseased, severe damage
ClassificationZero-shot segmentation (SAM 2)Multi-sensor fusion
Template · Change detection
Crop Health Trends Detector
Monitor crop health changes, stress progression, and disease spread over growing season using change detection
- Inputs
- Multispectral, NDVI, Image pairs
- Output
- Polygons
- Sensors
- Resolution
- 1-10m per pixel
- Classes
- health improvement, health decline, disease spread, recovery area
Change detectionZero-shot features (DINOv2)Zero-shot segmentation (SAM 2)Multi-sensor fusion
Template · Object detection
Crop Pest & Disease Monitor
Early detection of crop pests and diseases for timely intervention
- Inputs
- RGB, Multispectral, Hyperspectral
- Output
- Boxes
- Sensors
- Resolution
- 0.1-1m per pixel
- Classes
- aphid infestation, fungal disease, nutrient deficiency, viral infection
Object detectionZero-shot segmentation (SAM 2)Multi-sensor fusion
Template · Classification
Crop Yield Predictor
Predict crop yields based on growth patterns and environmental factors
- Inputs
- Multispectral, Weather, Past yields
- Output
- Heatmap
- Sensors
- Resolution
- 1-10m per pixel
- Classes
- high yield, medium yield, low yield, crop failure
ClassificationZero-shot segmentation (SAM 2)Multi-sensor fusionForecasting
Template · Object detection
Dark Vessel Fusion Monitor
Correlate SAR vessel detections with AIS tracks to flag dark transits: vessels in choke points with AIS disabled or no cooperative reporting
- Inputs
- Radar, Ship reports
- Output
- Points
- Sensors
- Resolution
- S1 10m + AIS kinematics
- Classes
- dark candidate, cooperative track, matched fusion
Object detectionMulti-sensor fusion
Template · Object detection
Deforestation Alert System
Real-time detection of illegal logging and forest clearing activities
- Inputs
- Optical, Radar
- Output
- Polygons
- Sensors
- Resolution
- 0.5-3m per pixel
- Classes
- fresh clearcut, logging equipment, access roads, log piles
Object detectionZero-shot segmentation (SAM 2)Multi-sensor fusion
Template · Change detection
Deforestation Change Detector
Detect forest loss, degradation, and deforestation between time periods using ML-based change detection
- Inputs
- Multispectral, Radar, Image pairs
- Output
- Polygons
- Sensors
- Resolution
- 10-30m per pixel
- Classes
- forest loss, forest gain
Change detectionZero-shot features (DINOv2)Zero-shot segmentation (SAM 2)Multi-sensor fusion
Template · Change detection
Environmental Compliance Monitor
Monitor environmental compliance violations and restoration progress using temporal change detection
- Inputs
- Multispectral, Radar, Image pairs
- Output
- Polygons
- Sensors
- Resolution
- 1-10m per pixel
- Classes
- violation appearing, violation resolved, restoration progress, degradation detected
Change detectionZero-shot features (DINOv2)Zero-shot segmentation (SAM 2)Multi-sensor fusion
Template · Change detection
FOB Perimeter Security Monitor
Monitor forward operating base perimeters for SAR change polygons, thermal anomalies, and fused intrusion alerts along fence lines
- Inputs
- Radar, Thermal, Optical
- Output
- Polygons
- Sensors
- Resolution
- 10m SAR + optional TIR
- Classes
- perimeter stable, fence line change, thermal anomaly, intrusion alert
Change detectionZero-shot features (DINOv2)Zero-shot segmentation (SAM 2)Multi-sensor fusion
Template · Segmentation
Forest Canopy Health Monitor
Semantic segmentation of forest canopy to assess health, coverage, and detect deforestation
- Inputs
- Multispectral, RGB, Near infrared
- Output
- Polygons
- Sensors
- Resolution
- 0.5-2m per pixel
- Classes
- healthy canopy, stressed canopy, dead trees, clearcut, regrowth
SegmentationZero-shot segmentation (SAM 2)Multi-sensor fusion
Template · Object detection
Illegal Mining Activity Detector
Detect unauthorized mining activities using multi-temporal satellite imagery and change detection
- Inputs
- Optical, Thermal, Night lights
- Output
- Boxes
- Sensors
- Resolution
- 0.3-1m per pixel
- Classes
- excavation, equipment, access roads, tailings, water pollution
Object detectionZero-shot features (DINOv2)Zero-shot segmentation (SAM 2)Multi-sensor fusion
Template · Change detection
Illegal Mining Change Detector
Detect new mining activities, land disturbance, and environmental impacts using temporal change detection
- Inputs
- Optical, Radar, Night lights, Image pairs
- Output
- Boxes
- Sensors
- Resolution
- 0.5-3m per pixel
- Classes
- new excavation, new access roads, vegetation removal, water pollution, equipment deployment
Change detectionZero-shot features (DINOv2)Zero-shot segmentation (SAM 2)Multi-sensor fusion
Template · Change detection
Infrastructure Change Detector
Detect new structures, demolition, and resurface changes at airfields, ports, and strategic facilities using bi-temporal optical / SAR change detection
- Inputs
- Optical, Radar
- Output
- Polygons
- Sensors
- Resolution
- 10m per pixel (Sentinel-2/S1)
- Classes
- new structure, demolition, resurface, no change
Change detectionZero-shot features (DINOv2)Zero-shot segmentation (SAM 2)Multi-sensor fusion
Template · Segmentation
Land Disturbance Monitor
Monitor and quantify land disturbance from mining operations for environmental compliance
- Inputs
- Multispectral
- Output
- Mask
- Sensors
- Resolution
- 1-5m per pixel
- Classes
- undisturbed land, active mining, reclaimed land, infrastructure, water bodies
SegmentationZero-shot segmentation (SAM 2)
Template · Change detection
Land Encroachment Change Detector
Monitor unauthorized land use changes and encroachment into protected or restricted mining areas
- Inputs
- Multispectral, Radar, Image pairs
- Output
- Polygons
- Sensors
- Resolution
- 1-5m per pixel
- Classes
- unauthorized development, vegetation encroachment, boundary violation, infrastructure expansion
Change detectionZero-shot features (DINOv2)Zero-shot segmentation (SAM 2)Multi-sensor fusion
Template · Object detection
Mine Bench Slope Safety Monitor
Detection of geotechnical hazards including slope failures, tension cracks, and instability indicators on mine bench faces for safety management
- Inputs
- Optical, LiDAR, Radar
- Output
- Polygons
- Sensors
- Resolution
- 0.05-0.3m per pixel
- Classes
- tension crack, slope failure, ravelling zone, bench face, displacement zone
Object detectionZero-shot segmentation (SAM 2)Multi-sensor fusion
Template · Change detection
Mine Pit Progression Monitor
Temporal monitoring of open-cut mine pit expansion, waste dump growth, and stockpile changes for production tracking and regulatory compliance
- Inputs
- Optical, Radar, Image pairs
- Output
- Polygons
- Sensors
- Resolution
- 0.5-5m per pixel
- Classes
- active pit, waste dump, ore stockpile, disturbed land, rehabilitation
Change detectionZero-shot segmentation (SAM 2)Multi-sensor fusion
Template · Segmentation
Mine Rehabilitation Tracker
Track post-mining land rehabilitation and vegetation recovery
- Inputs
- Multispectral, RGB, LiDAR
- Output
- Polygons
- Sensors
- Resolution
- 1-5m per pixel
- Classes
- fully rehabilitated, partial recovery, early rehabilitation, bare disturbed
SegmentationZero-shot segmentation (SAM 2)Multi-sensor fusion
Template · Object detection
Mining Haul Road Detector
Automated detection and mapping of haul roads, access tracks, and mining infrastructure for operational planning and compliance monitoring
- Inputs
- Optical, Radar
- Output
- Polygons
- Sensors
- Resolution
- 0.3-2m per pixel
- Classes
- primary haul road, secondary road, drill pattern, infrastructure
Object detectionZero-shot segmentation (SAM 2)Multi-sensor fusion
Template · Segmentation
Mining Tailings Monitor
Monitor tailings dam stability and environmental impact
- Inputs
- Optical, Radar, Thermal
- Output
- Polygons
- Sensors
- Resolution
- 0.5-3m per pixel
- Classes
- stable tailings, seepage risk, structural concern, environmental impact
SegmentationZero-shot segmentation (SAM 2)Multi-sensor fusion
Template · Classification
Ore Exploration Assistant
Identify potential ore deposits and geological formations
- Inputs
- Multispectral, Hyperspectral, Magnetic survey, Radiometric survey
- Output
- Heatmap
- Sensors
- Resolution
- 1-10m per pixel
- Classes
- high potential, moderate potential, low potential, barren rock
ClassificationZero-shot segmentation (SAM 2)Multi-sensor fusion
Template · Counting
Plantation Plant Counter
Automated individual plant counting and survival rate assessment for forestry plantations, nurseries, and agroforestry systems
- Inputs
- Optical, Multispectral
- Output
- Points
- Sensors
- Resolution
- 0.1-0.5m per pixel
- Classes
- healthy plant, stressed plant, dead plant, gap
CountingZero-shot segmentation (SAM 2)Multi-sensor fusion
Template · Segmentation
Precision Agriculture Optimizer
Variable rate application mapping for fertilizers, pesticides, and irrigation
- Inputs
- Multispectral, Soil data, Elevation
- Output
- Heatmap
- Sensors
- Resolution
- 1-5m per pixel
- Classes
- high fertility, medium fertility, low fertility, waterlogged
SegmentationZero-shot segmentation (SAM 2)Multi-sensor fusion
Template · Counting
Reforestation Pit Counter
Automated detection and counting of planting pits for reforestation progress monitoring and survival rate assessment
- Inputs
- Optical, LiDAR
- Output
- Points
- Sensors
- Resolution
- 0.05-0.3m per pixel
- Classes
- empty pit, planted pit, failed pit
CountingZero-shot segmentation (SAM 2)Multi-sensor fusion
Template · Object detection
SAR Vessel Detector
Detect vessels in Sentinel-1 SAR imagery using CFAR prescreen and learned detection head: feeds dark-vessel fusion and maritime MDA
- Inputs
- Radar
- Output
- Boxes
- Sensors
- Resolution
- 10m Sentinel-1 GRD
- Classes
- vessel, fixed structure, clutter
Object detection
Template · Segmentation
Stubble Burnt Area Mapper
Detection and mapping of crop residue burning events for air quality monitoring, carbon accounting, and regulatory compliance
- Inputs
- Optical, Thermal, Multispectral
- Output
- Polygons
- Sensors
- Resolution
- 3-30m per pixel
- Classes
- burnt stubble, active fire, unburnt stubble, cleared field
SegmentationZero-shot segmentation (SAM 2)Multi-sensor fusion
Template · Change detection
Tactical SAR Change Detector
Sentinel-1 SAR bi-temporal change for tactical ISR: vehicle convoy staging, surface disturbance, and emergent targets in denied-weather conditions
- Inputs
- Radar
- Output
- Polygons
- Sensors
- Resolution
- 10m Sentinel-1 GRD
- Classes
- surface disturbance, staging area, emergent target, stable ground
Change detectionZero-shot features (DINOv2)Zero-shot segmentation (SAM 2)
Template · Segmentation
Wildfire Burnt Area Assessor
Precise mapping and quantification of wildfire burnt areas for forest damage assessment, carbon emission estimation, and recovery monitoring
- Inputs
- Optical, Radar, Thermal, Multispectral
- Output
- Polygons
- Sensors
- Resolution
- 10-30m per pixel
- Classes
- high severity burn, moderate severity burn, low severity burn, unburnt, active fire
SegmentationZero-shot segmentation (SAM 2)Multi-sensor fusion
Template · Change detection
Wildfire Damage Change Detector
Assess fire damage extent and post-fire vegetation recovery using temporal change detection
- Inputs
- Multispectral, Thermal, Image pairs
- Output
- Polygons
- Sensors
- Resolution
- 10-30m per pixel
- Classes
- burned area, vegetation loss, recovery area
Change detectionZero-shot features (DINOv2)Zero-shot segmentation (SAM 2)Multi-sensor fusion
Template · Classification
Wildfire Risk Detector
Detect and classify areas at high risk for wildfire based on vegetation density and dryness
- Inputs
- Multispectral, Thermal, Weather
- Output
- Heatmap
- Sensors
- Resolution
- 10-30m per pixel
- Classes
- low risk, moderate risk, high risk, extreme risk
ClassificationZero-shot segmentation (SAM 2)Multi-sensor fusionForecasting
Template · Change detection
WorldMonitor COP
Global COP hotspot / nightlight / LST activity index over a near-global viewport: VIIRS DNB, MODIS LST, FIRMS points, ASTER catalog. OSINT / COP overlay (op_poc); not theatre fine-tune.
- Inputs
- Global open feeds
- Output
- Polygons
- Sensors
- Resolution
- 375m-1km (VIIRS/MODIS)
- Classes
- hotspot, nightlight anomaly, context
Change detection
Use cases · Next step
Need one that is not here?
Most new detectors are a new set of class prompts on the same engines. Tell me the ground and the decision.