Satellite imagery segmentation
Satellite imagery segmentation is the process of separating pixels, regions, or objects into meaningful classes, such as water, vegetation, roads, buildings, clouds, burn scars, crop areas, or solar sites.
RasterForge approaches segmentation as a project workflow. The source raster, AOI, render profile, prompts, examples, mask layers, review state, object edits, and exports are kept together so users can repeat and inspect the process.
Why rendering matters
Many satellite rasters are not simple RGB images. They can be reflectance products, multispectral stacks, 8-bit visual products, or derived exports from another GIS workflow. The segmentation model sees the rendered image, so band selection, stretch, gamma, nodata handling, and derived views can change the output.
A waterline workflow, vegetation workflow, roof workflow, and cloud workflow may each need different render choices even when they use the same source image.
Why profiles matter
Segmentation is rarely just one prompt. A reusable RasterForge profile captures target name, class metadata, prompt text, examples, thresholds, max instances, and postprocessing choices. That makes it possible to test on one image, run selected images, and then review the project consistently.
Shared profiles can speed up setup for simple repeated targets, but they still need to be checked against the render and imagery in the current project.
Why review matters
Segmentation models can produce false positives, missing areas, over-tight masks, noisy boundaries, or plausible-looking vectors that fail downstream requirements. RasterForge keeps review and editing close to the project so users can inspect layers before export.
Review masters, image overviews, coverage summaries, edit drafts, and handoff exports help turn model output into a selected project result.
Where RasterForge fits
Use RasterForge when you need promptable segmentation, repeatable profiles, multi-image runs, review decisions, and GIS-ready exports. It is not a guarantee of ground truth; it is a workflow for preparing, running, checking, and exporting segmentation results.