US 12,228,497 B2
Determining material properties based on machine learning models
Nasim Souly, San Mateo, CA (US); Melanie Senn, Mountain View, CA (US); and Gianina Alina Negoita, Belmont, CA (US)
Assigned to VOLKSWAGEN AKTIENGESELLSCHAFT, Wolfsburg (DE)
Filed by VOLKSWAGEN AKTIENGESELLSCHAFT, Wolfsburg (DE)
Filed on Aug. 10, 2021, as Appl. No. 17/398,764.
Prior Publication US 2023/0051237 A1, Feb. 16, 2023
Int. Cl. G06T 7/00 (2017.01); G01N 21/17 (2006.01); G06N 3/08 (2023.01)
CPC G01N 21/17 (2013.01) [G06N 3/08 (2013.01); G06T 7/0004 (2013.01); G01N 2021/1765 (2013.01); G06T 2207/10016 (2013.01); G06T 2207/10056 (2013.01); G06T 2207/20081 (2013.01); G06T 2207/20084 (2013.01)] 18 Claims
OG exemplary drawing
 
1. A method, comprising:
obtaining a sequence of images of a three-dimensional volume of a material;
determining a set of features based on the sequence of images and a first neural network, wherein the set of features indicate microstructure features of the material; and
determining a set of material properties of the three-dimensional volume of the material based on the set of features a first transformer network, a second transformer network, a first crossmodal transformer network, and a second crossmodal transformer network, wherein a first set of transformed features from the first transformer network are provided as input to the first crossmodal transform network and the second crossmodal transformer network, and wherein a second set of transformed features from the second transformer network are provided as input to the first crossmodal transform network and the second crossmodal transformer network.