US 11,948,272 B2
Computer-implemented method to improve scale consistency and/or scale awareness in a model of self-supervised depth and ego-motion prediction neural networks
Hemang Chawla, Eindhoven (NL); Arnav Varma, Eindhoven (NL); Elahe Arani, Eindhoven (NL); and Bahram Zonooz, Eindhoven (NL)
Assigned to NAVINFO EUROPE B.V., Eindhoven (NL)
Filed by NavInfo Europe B.V., Eindhoven (NL)
Filed on Aug. 13, 2021, as Appl. No. 17/402,349.
Claims priority of application No. 20207576 (EP), filed on Nov. 13, 2020.
Prior Publication US 2022/0156882 A1, May 19, 2022
Int. Cl. G06T 3/40 (2006.01); G06F 18/214 (2023.01); G06N 3/045 (2023.01); G06N 3/08 (2023.01); G06T 3/4038 (2024.01); G06T 3/4046 (2024.01); G06T 7/50 (2017.01); G06V 20/40 (2022.01)
CPC G06T 3/4046 (2013.01) [G06F 18/2148 (2023.01); G06N 3/045 (2023.01); G06N 3/08 (2013.01); G06T 3/4038 (2013.01); G06T 7/50 (2017.01); G06V 20/41 (2022.01)] 15 Claims
OG exemplary drawing
 
1. A computer-implemented method to improve scale consistency and/or scale awareness in a model of self-supervised depth and ego-motion prediction neural networks processing a video stream of monocular images, the method comprising using complementary GPS coordinates synchronized with the images to calculate a GPS to scale loss to enforce the scale-consistency and/or -awareness on the monocular self-supervised ego-motion and depth estimation, wherein a relative weight assigned to the GPS to scale loss exponentially increases as training progresses.