US 11,995,854 B2
Mesh reconstruction using data-driven priors
Orazio Gallo, Santa Cruz, CA (US); and Abhishek Badki, Goleta, CA (US)
Assigned to NVIDIA Corporation, Santa Clara, CA (US)
Filed by NVIDIA CORPORATION, Santa Clara, CA (US)
Filed on Dec. 19, 2018, as Appl. No. 16/226,329.
Prior Publication US 2020/0202622 A1, Jun. 25, 2020
Int. Cl. G06T 7/55 (2017.01); G06F 17/16 (2006.01); G06F 18/214 (2023.01); G06N 20/00 (2019.01); G06T 15/10 (2011.01); G06T 17/20 (2006.01)
CPC G06T 7/55 (2017.01) [G06F 17/16 (2013.01); G06F 18/214 (2023.01); G06N 20/00 (2019.01); G06T 15/10 (2013.01); G06T 17/205 (2013.01)] 2 Claims
OG exemplary drawing
 
1. A processor, comprising:
logic to predict one or more three-dimensional (3D) meshes based on a plurality of digital images, wherein the one or more 3D meshes are refined using a trained decoder based on differences between the one or more 3D meshes and the plurality of digital images;
wherein predicting the one or more 3D meshes based on the plurality of digital images comprises:
executing the decoder to produce a mesh of an object from a first value in a latent space; and
refining the mesh of the object using the decoder, wherein refining the mesh of the object comprises selecting a second value in the latent space based on one or more geometric constraints associated with the plurality of digital images of the object;
wherein refining the mesh of the object comprises:
dividing the mesh into a set of meshlets;
for each meshlet in the set of meshlets, selecting the second value in the latent space to learn a prior for a portion of the mesh represented by the meshlet; and
reconstructing the mesh from the set of meshlets; and
wherein refining the mesh of the object further comprises iteratively increasing a resolution of the set of meshlets to meet the one or more geometric constraints.