Semantic Segmentation with U-Net: Labeling Every Pixel on CARLA's Streets

Object detection draws a box around what it finds. Semantic segmentation goes a level deeper: every single pixel gets its own class label. Not "there's a car over there," but "this pixel belongs to a