Method and device for learning a depth prediction model from images filtered by vertical bands.

Abstract

A method or device for learning a depth prediction model associated with a vehicle vision system. The method comprises receiving (31) a batch of images acquired by a camera of the vision system, the images being acquired at distinct time points, dividing (32) these images into vertical bands, comparing the vertical bands of a first image to those of a second image preceding the first, and deleting (34) the first image when a colorimetric difference (33) associated with one of its vertical bands is less than a threshold. A local photometric error is then determined (36) to compare each vertical band of the first image to the vertical bands of the second image. The depth prediction model is then learned (38) by minimizing the largest photometric errors associated with the vertical bands of each first image.Figure for the abbreviation: Figure 3.

Images