promptable segmentation

Segmenting an object from a point, box, or text prompt supplied at inference time, rather than a fixed trained label set.

Promptable segmentation produces a mask for whatever a point, box, or text prompt indicates at inference time, instead of a network trained to output masks for one fixed set of labels -- letting one trained model segment an object class it was never explicitly trained to recognize by name.

Examples

Clicking once on a cat in a photo is enough of a prompt for a promptable segmentation model to return a mask outlining the whole cat.