Methods and systems for learning from the air for autonomous driving.

Abstract

The invention relates to an aerial learning method for autonomous driving. This method involves acquiring aerial imagery containing aerial data of an area with numerous ground vehicles and road objects. The aerial data is analyzed in a machine learning process from the corresponding viewpoints of multiple ground vehicles. Based on this analysis, corresponding ground vehicle viewpoints are generated, each viewpoint from a different ground vehicle viewpoint and including road object information related to the corresponding viewpoint of each different ground vehicle, for use in training an Artificial Intelligence (AI) model used in controlling autonomous vehicles. The invention also relates to an aerial learning system for autonomous driving.

Images