An autonomous vehicle company trains a vision model that assigns every individual pixel in a camera frame to a category such as road, sidewalk, vehicle, or pedestrian, producing a colored map of the entire scene rather than boxes drawn around individual objects. Which computer vision technique is being described?

Semantic segmentation classifies every individual pixel in an image into a category, producing a dense, per-pixel map of the scene, such as marking which pixels belong to the road, the sidewalk, a vehicle, or a pedestrian, which precisely matches the colored scene map the autonomous vehicle company is generating. The object-detection approach is incorrect because it locates discrete objects and marks each one with a bounding box around its approximate location, producing boxes rather than the pixel-by-pixel category map described in the scenario. The image-classification approach is incorrect because it assigns one overall label to an entire image, such as 'street scene,' and cannot distinguish or label the many different regions within a single frame the way the described model does. The optical-character-recognition approach is incorrect because it extracts and reads printed or handwritten text characters from an image, such as a street sign, which has nothing to do with categorizing every pixel of a scene by object type. Because the scenario explicitly describes pixel-level categorization across the whole frame rather than boxes, a whole-image label, or text extraction, semantic segmentation is the only technique that matches.