Abstract: A method is provided for simplifying a takeover of control of a motor vehicle by a vehicle-external operator. In the method, images of the surroundings of the vehicle are captured from the vehicle and semantically segmented. Errors in a corresponding segmentation model are predicted on the basis of at least one such image each. If a corresponding error prediction triggering a request for the takeover of control is made, an image-based visualization is automatically generated in which exactly one region corresponding to the error prediction is visually highlighted. The request and the visualization are then sent to the vehicle-external operator.