Method, apparatus, and computer program product for correcting lane geometry in map data

Abstract

A method is provided to using a machine learning model to predict lane geometry where incorrect or missing lane line geometry is detected. Methods may include: receiving a representation of lane line geometry including a masked area representing an area of broken lane line geometry; processing the representation using an inpainting model including a generator to produce a corrected representation, where the masked area is inpainted to correct the broken lane line geometry, where the generator processes the representation through one or more convolution layers and at least eight dilated convolution layers to generate inpainting of corrected lane line geometry in the corrected representation; processing the corrected representation through a discriminator of the inpainting model, where the discriminator discerns a score reflecting a quality measure of the corrected representations; and updating a map database based on the corrected representation in response to the score satisfying a predetermined threshold.