Definition
Plain language
An AI that outlines and labels the individual objects in an image, like separating each bowl on a table.
As stated in the literature
A vision model that partitions an image into per-object regions; in robot code-as-policy pipelines it returns object masks without inherent ordering, motivating induced skills that sort returned instances by position for disambiguation.
Also called: image segmentation
Why it matters: It lets a system act on specific objects rather than a whole image, which is essential for a robot to grab the right item among several.
For example, given a photo of a set dinner table, it outlines each plate, fork, and glass separately so software knows where one object ends and the next begins.
Heard on the show
“And a disambiguation skill — you say "the front bowl," the segmentation model hands back every bowl with no ordering, so the routine sorts them by position and picks the actual front one.”Episode 194 — How a Robot Builds a Debugging Notebook It Can Read, Edit, and Hand to Another Robot