determining, by a computing system, lane detection data and object detection data associated with an environment;

selecting, by the computing system, a lane template from a set of predefined lane templates based on a number of lane markings described by the lane detection data, wherein the set of predefined lane templates correspond with lane layouts for roads that are encountered in the environment and the selected lane template includes a road with a number of lanes that corresponds with the number of lane markings described by the lane detection data, wherein the lane template is selected based on a machine learning model, and wherein the machine learning model is trained with training data that includes a first instance of training data that is based on an application of an offset to the lane template as a modification of a lane width of the lane template; and

generating, by the computing system, localization data that identifies a location of an object in the environment based on the lane template and the object detection data.