semantic segmentation

Labeling every pixel in an image by class, with no notion of separate object instances.

Semantic segmentation assigns every pixel in an image a class label, with no distinction between separate objects of the same class -- three adjacent cars are all labeled "car" and merge into one region. Contrast with instance segmentation, which additionally separates each object.

Examples

A street-scene model labels every 'car' pixel with the same class, merging three parked cars into one undivided region.