A free image labeling tool for creating custom training datasets from satellite imagery

Build your own machine learning training data from satellite, drone, and aerial imagery.

Sign up for free
Depiction of GroundWork UI
Geospatial annotation visualization

GroundWork is designed for annotating and labeling geospatial data like satellite imagery

Groundwork converts .tif files that you upload into Cloud-Optimized GeoTiffs (COGs) and stores vector annotations as GeoJSON. You can export your annotated training dataset as a SpatioTemporal Asset Catalog (STAC) at any time, making it compatible with a growing ecosystem of Python libraries, like PySTAC and Franklin.

Get started for free
Geospatial annotation visualization with houses over top of a labeled neighborhood.

Semantic segmentation labeling

Groundwork supports satellite and aerial feature extraction for deep learning. Annotate imagery using our easy freehand lasso and polygon labeling tools.

Stylized illustration of a list overtop of simplified geospatial aerial image.

Object detection in aerial imagery

Object detection in aerial imagery presents unique challenges around resolution, orientation, and noise. Groundwork supports freehand contours and bounding for labeling complex training data.

Stylized illustration of a list overtop of simplified geospatial aerial image.

Image classification

Groundwork breaks large geospatial imagery into smaller pieces, to make creating remote sense training data from satellite and other aerial imagery easy.

Illustration of cloud with a map marker overtop of an image of labeled cars in a parking lot.

Cloud-native architecture

All imagery is stored as cloud optimized GeoTiffs and labels are stored as GeoJSON. Exported data is represented as a SpatioTemporal Asset Catalog.

We offer managed annotation services for satellite, drone, and other aerial imagery. Our R&D team can also work with you to train custom computer vision models that meet your needs. Learn more about our services.

Contact us

Geospatial imagery is not an edge case

Supervised machine learning always starts with a high-quality training dataset, but image annotation tools have always treated geospatial data like an afterthought. Not anymore.

Get Started
Chart of images labeled in GroundWork over the time of the competition.

Labeling Competition Recap

We ran a STAC labeling competition as part of the Cloud Native Geospatial Sprint. Learn about the competition and hear directly from some of the contestants!

Check it out!