Book a call
Earth at night from the International Space Station: city lights under an aurora and the rising Moon.

Satellite & drone
data, labeled.

Geospatial annotation for machine-learning teams: we label satellite and drone imagery, from Copernicus and high-resolution scenes to solar, power line, orchard and construction surveys, with the accuracy measured per class.

Book a call
01 // Industries

Who we label for.

02 // Data

What we label.

Satellite image annotation on your imagery, free Copernicus data, or commercial imagery we buy for your project. Drag across each example to see it before and after labeling, browse all label types or read the satellite image annotation guide.

Astronaut photograph of The Pearl-Qatar: artificial islands with villas, marinas and roads.
RawLabeled
  • Building28 boxes
  • Road1 line
  • Marina water2 polygons

Optical & aerial

Sub-metre satellite, aerial and drone RGB

  • Building footprints
  • Roads
  • Vehicles
  • Damage

Source: The Pearl-Qatar, ISS photograph ISS053-E-127736 (NASA JSC)

Landsat 8 scene of irrigated farmland, a lake and a town in Montana.
RawLabeled
  • Irrigated field19 instances
  • Water body1 polygon
  • Built-up area1 polygon
  • Road1 line

Multispectral

Sentinel-2, Landsat 8/9 and other VNIR–SWIR sensors

  • Land cover
  • Crop type
  • Water
  • Burn scars

Source: Montana farmland, Landsat 8 OLI (NASA Earth Observatory / USGS)

Airborne radar image of a river and flooded Amazon forest in Peru.
RawLabeled
  • Open waterriver + 4 lakes
  • Flooded forest1 polygon
  • Floating vegetation3 polygons

Radar (SAR)

Sentinel-1 and commercial SAR, day or night, through cloud

  • Flood extent
  • Change
  • Vessels

Source: Pacaya-Samaria, Peru, UAVSAR airborne radar (NASA/JPL-Caltech)

Top-down lidar point cloud of a forest plot in Alaska, coloured by height.
RawLabeled
  • Crown > 15 m8 trees
  • Canopy gap6 polygons
  • Plot boundary1 polygon

Elevation & LiDAR

Airborne point clouds, DSM and DTM

  • Canopy height
  • Structures
  • 3D objects

Source: Tanana Valley, Alaska, G-LiHT airborne lidar (NASA Goddard)

  • Segmentation masks
  • Polygons
  • Bounding boxes
  • Time-series labels
  • GeoJSON · GeoPackage · COG · COCO
03 // Drone

Drone image annotation.

Inspection and survey imagery from your drone flights: RGB, thermal and multispectral frames or orthomosaics. Open a label type to see an example before and after labeling, or read the drone image annotation guide.

04 // Questions

Satellite and drone image annotation, explained.

What is satellite and drone image annotation?

Also called data labeling or geospatial annotation, it means marking what is in satellite, aerial or drone images (field boundaries, crop types, floodwater, buildings, forest loss, solar panels, power lines, trees) as polygons, masks, boxes, points or lines. Machine-learning models learn from these labels, and the same labels can be published as map layers. IKT Spīdolas does this work for teams in agriculture, insurance, forestry, energy, construction and government.

Which imagery do you work with?

Your own imagery; free Copernicus Sentinel-1, Sentinel-2 and Landsat scenes; or commercial very-high-resolution satellite imagery we buy for your project. We also label drone imagery (RGB, thermal and multispectral), aerial photos, radar (SAR) and airborne LiDAR.

What do I receive at the end?

The labels in the format your pipeline reads (GeoJSON, GeoPackage, Cloud-Optimized GeoTIFF masks or COCO) in your coordinate system, together with the written labeling guideline and the accuracy per class measured on a reviewed sample.

Do you fly the drones?

No. We label imagery from your own drone flights or your survey provider: single frames, orthomosaics, thermal data or point clouds, in the formats your software exports.

How is a project priced?

Per project. The price depends on the area, the number of classes, how many dates are needed and whether imagery has to be bought. You get a quote after a short call where we agree the class list.

Can you label something that is not on your list?

Yes. The 20 label types on this site are common requests, not a limit. Rooftop solar, ships in port, greenhouses, mining change or your own class scheme: we write a guideline for it and label to that guideline.

Who is IKT Spīdolas?

SIA IKT Spīdolas is a company registered in Rīga, Latvia (registration number 40203763365). Projects are run remotely, for clients in any country.

05 // Custom cases

Your case isn’t on the list? That’s normal.

Most projects need their own classes, sensor mix or target. Tell us what your model has to learn and we design the labeling around it.

Book a call
  • Rooftop solar panels
  • Ships in port
  • Illegal dumping sites
  • Quarry & mining change
  • Snow and ice cover
  • Greenhouses
  • Road surface condition
  • Your own class scheme