Methods for training a perception system for a vehicle

Abstract

The invention relates to a computer-implemented method (100) for training a perception system (200) for a vehicle, comprising: providing (101) a first training data set (201), wherein the first training data set (201) comprises unlabeled environmental sensor data (203); performing (103) a first training step of a map prediction module (205) of the perception system (200) to generate a map representation (207) of the vehicle's environment based on the environmental sensor data (203) of the first training data set (201); and providing (105) a second training data set (209), wherein the second training data set (209) comprises labeled environmental sensor data (211);and execution (107) of a second training step of the map prediction module (205) of the perception system (200) pre-trained in the first training step to generate the map representation (207) of the environment based on the labeled environment sensor data (211) of the second training data set (209), wherein the generation of the map representation (207) is designed as an onboard map generation.;

Images