Extract Objects
The Extract Objects command reads a labelled instance map and creates one segmentation-class object per labelled region. That map can come from Connected Components (or Watershed), or directly from an instance-segmentation command such as AI Cellpose Segmentation or AI Stardist Segmentation, which already separate touching objects on their own. These objects are the raw candidates passed to Classify Objects for final filtering and class assignment.
This is the boundary between “pixels” and “objects” in the pipeline: everything before this step operates on whole images, and everything after operates on individually addressable regions with their own metrics, mask, and object ID. In the command picker it belongs to the Extract category, together with Fill Object Holes.
While extracting, the step measures every object’s intensities on each image channel: sum, mean, minimum and maximum, and the spread of its pixel values - standard deviation, median, lower and upper quartile and median absolute deviation (see Intensity metrics).
Parameters
Section titled “Parameters”| Parameter | Description |
|---|---|
| Max objects before fail | Hard upper limit on the number of regions extracted. If more regions are found the pipeline step fails, preventing runaway processing on misconfigured segmentation (default: 100000) |
| Class Mapping | Which object class the objects of a segmentation class get - see Class mapping. Empty by default |
Class mapping
Section titled “Class mapping”Without a mapping, the objects of a segmentation class get the project class with the same number - that’s what the class choice in Threshold means.
Class Mapping is a list of entries, each with a Segmentation Class and the Object Class its objects get. Segmentation classes without an entry keep the default. Inside a module, map to one of the module’s own classes to keep the objects internal to the module - this is what Create module sets up automatically when a segmentation class is only used by the bundled steps.
A mapping to a project class that doesn’t exist is shown as a warning.
Pipeline position
Section titled “Pipeline position”Threshold → Connected Components → [Watershed] ─┐ ├─→ Extract Objects → Classify Objects AI Cellpose Segmentation / AI Stardist ────┘Extract Objects is always followed by Classify Objects. The segmentation class assigned here is the intermediate class used internally; the final named object class is assigned by Classify Objects.