Represent land-cover areas with labeled boundaries or segmentation regions in the raster’s original coordinate space.
Use lines for roads and points for discrete targets when the mapping task requires those geometries.
It labels the areas and features in geospatial imagery so models can learn classes such as water, vegetation, buildings, roads, or other task-defined categories.
Land cover describes visible surface classes, while land use adds a purpose defined by the task. Your guidelines should clarify which evidence supports each label.
Yes. Define the classes and properties required by your dataset and apply them consistently across satellite or aerial scenes.
Yes. Use polygons or masks for areas and line annotations for roads or other linear features.
The geospatial viewer uses tiled navigation to inspect large images at different zoom levels while retaining full-resolution annotation geometry.
Define whether mixed areas receive a separate class or are divided into constituent regions. Use the same boundary rules across annotators and review difficult cases.
Unitlab supports GeoTIFF, Cloud Optimized GeoTIFF, JPEG 2000, ERDAS IMG, and NITF raster inputs. Tiled navigation helps teams inspect large satellite and aerial images.
Reviewers inspect mapped regions and boundaries in their source raster, add comments, and return corrections through the workflow. A shared ontology keeps the map classes and properties consistent.
Yes. GeoJSON supports geographic output for valid georeferenced rasters. COCO, YOLO, and Unitlab Unified Export Format serve other training needs while retaining the appropriate source geometry.
We use cookies to improve your experience. See our Privacy Policy and Cookie Policy.