US 11,715,251 B2
Neural network model trained using generated synthetic images
Jonathan Tremblay, Redmond, WA (US); Aayush Prakash, Toronto (CA); Mark A. Brophy, Toronto (CA); Varun Jampani, Nashua, NH (US); Cem Anil, Toronto (CA); Stanley Thomas Birchfield, Sammamish, WA (US); Thang Hong To, Redmond, WA (US); and David Jesus Acuna Marrero, Toronto (CA)
Assigned to NVIDIA Corporation, Santa Clara, CA (US)
Filed by NVIDIA Corporation, Santa Clara, CA (US)
Filed on Oct. 21, 2021, as Appl. No. 17/507,620.
Application 17/507,620 is a continuation of application No. 17/119,971, filed on Dec. 11, 2020, granted, now 11,182,649.
Application 17/119,971 is a continuation of application No. 16/256,820, filed on Jan. 24, 2019, granted, now 10,867,214, issued on Dec. 15, 2020.
Claims priority of provisional application 62/630,722, filed on Feb. 14, 2018.
Prior Publication US 2022/0044075 A1, Feb. 10, 2022
This patent is subject to a terminal disclaimer.
Int. Cl. G06T 15/00 (2011.01); G06T 15/04 (2011.01); G06T 15/50 (2011.01); G06T 15/20 (2011.01); G06F 18/214 (2023.01); G06F 18/211 (2023.01); G06V 10/774 (2022.01); G06V 10/82 (2022.01); G06N 3/04 (2023.01); G06N 3/084 (2023.01)
CPC G06T 15/00 (2013.01) [G06F 18/211 (2023.01); G06F 18/2148 (2023.01); G06T 15/04 (2013.01); G06T 15/20 (2013.01); G06T 15/50 (2013.01); G06V 10/7747 (2022.01); G06V 10/82 (2022.01); G06N 3/04 (2013.01); G06N 3/084 (2013.01); G06T 2210/12 (2013.01); G06V 2201/07 (2022.01)] 22 Claims
OG exemplary drawing
 
1. A processor, comprising:
one or more processing units to receive, using a neural network model, an input image comprising at least one object of interest and an additional object, wherein the neural network is trained to ignore the additional object and to produce task-specific output data corresponding to the object of interest by using a training dataset that includes synthetic input images rendered according to rendering parameters that define visibility of a ground plane in a three-dimensional (3D) scene, and at least one of:
one or more texture maps applied to the object of interest and the additional object, or
one or more colors of the object of interest and the additional object.