
Land cover, agriculture, flood, burned area, and infrastructure mapping models are only as good as the segmentation data behind them. Producing that data usually calls for GIS expertise and painstaking attention across every tile of a large image. The SAM GeoTIFF service in AI-Builder-Suite puts that capability directly in your team’s hands, turning raw satellite imagery into reviewed, export-ready training data through a guided, interactive workflow.
Earth observation teams often need labelled segmentation datasets to train models for land cover, agriculture, floods, burned areas, or infrastructure mapping. Creating these labelled masks is one of the slowest and most expensive parts of model development.
Traditionally, this meant manually drawing segmentation masks in GIS or annotation tools, a process that is slow and repetitive, especially across large GeoTIFF images.
The SAM GeoTIFF service in AI-Builder-Suite lets users upload GeoTIFF imagery and interactively generate masks for target objects or land-cover classes. Instead of manually tracing every object, the user provides a few interactive prompts and the model generates the mask automatically. The expert then reviews and adjusts the result rather than creating it from scratch. Masks can be reviewed, corrected, and exported directly as labelled training data for downstream AI models, making label creation more consistent and easier to scale.
Create segmentation masks with a few prompts instead of tracing objects manually, pixel by pixel.
Rely on model-generated masks reviewed and adjusted by experts, reducing the variability of fully manual annotation.
Move from a slow, repetitive process to one that scales across large GeoTIFF images and multiple land-cover classes.
Turn reviewed masks directly into labelled datasets for downstream AI model development.
View generated masks as an overlay on the satellite image as soon as they’re created, making quality easy to check on the spot.