Semantic segmentation is a computer vision task that assigns a class label to each pixel or region of an image. It enables pixel-level understanding of scene content, such as distinguishing roads, people, objects, or background areas.