US 12,456,277 B2
Determining object orientation from an image with machine learning
Jonathan Tremblay, Redmond, WA (US); Ming-Yu Liu, San Jose, CA (US); Dieter Fox, Seattle, WA (US); and Philip Ammirato, Chapel Hill, NC (US)
Assigned to NVIDIA Corporation, Santa Clara, CA (US)
Filed by NVIDIA Corporation, Santa Clara, CA (US)
Filed on Oct. 29, 2019, as Appl. No. 16/667,708.
Prior Publication US 2021/0125036 A1, Apr. 29, 2021
Int. Cl. G06N 3/08 (2023.01); G05D 1/00 (2006.01); G06N 3/045 (2023.01); G06V 10/24 (2022.01); G10L 15/16 (2006.01)
CPC G06V 10/242 (2022.01) [G05D 1/0088 (2013.01); G06N 3/045 (2023.01); G06N 3/08 (2013.01); G10L 15/16 (2013.01); G06T 2207/20084 (2013.01)] 24 Claims
OG exemplary drawing
 
15. A system, comprising:
one or more processors use a first neural network to identify an orientation of one or more objects within one or more images, based at least in part, on a three-dimensional model of the one or more objects and one or more memories to store the first neural network, wherein:
a first orientation of the one or more objects is determined by the first neural network using a first image of the one or more objects;
the first orientation is adjusted until it is observably equivalent to a second orientation, based at least in part on a difference between the first image and a second image; and
the second image is generated using a mathematical model of the one or more objects at the first orientation.